arXiv Daily Digest

282

Papers

FailureSpot: Label-Efficient Timestamp-Level Failure Detection for Vision-Language-Action Models

FailureSpot：面向视觉-语言-动作模型的标签高效时间戳级故障检测

Ma, Jie, Liu, Zongxi, Zhu, Yi

Abstract

Vision-language-action (VLA) policies have shown strong potential for general-purpose robotic manipulation, but they can still fail unpredictably during long-horizon execution, making reliable failure detection essential for safe deployment. Existing methods either rely on visual models that typically detect failures only after erroneous actions have occurred, or use lightweight proactive detectors trained on VLA internal representations. However, these proactive methods are often supervised with trajectory-level labels, causing normal pre-failure behavior in unsuccessful trajectories to be incorrectly labeled as failure. This supervision mismatch introduces label noise and limits both trajectory-level detection accuracy and precise timestamp-level failure localization. In this work, we study fine-grained timestamp-level VLA failure detection while addressing the cost of dense annotation. We propose a data-efficient framework that first leverages unlabeled VLA action chunks to construct action-derived weak supervision signals, capturing abnormal patterns such as inconsistent consecutive chunks, frozen or idle actions, and aggressive random motions. We then use active learning to select only the most uncertain trajectories for timestamp-level annotation and fine-tune the detector with these informative labels. Experiments across multiple VLA policies show that our method improves both timestamp-level and trajectory-level failure detection performance.

Chinese Translation

视觉-语言-动作（VLA）策略在通用机器人操作中展现出强大的潜力，但在长时间执行过程中仍可能出现不可预测的故障，因此可靠的故障检测对于安全部署至关重要。现有方法要么依赖于视觉模型，这些模型通常在错误动作发生后才检测到故障，要么使用在VLA内部表示上训练的轻量级主动检测器。然而，这些主动方法通常以轨迹级标签进行监督，导致在不成功的轨迹中正常的故障前行为被错误标记为故障。这种监督不匹配引入了标签噪声，限制了轨迹级检测的准确性和精确的时间戳级故障定位。在本研究中，我们研究了细粒度的时间戳级VLA故障检测，同时解决了密集注释的成本问题。我们提出了一种数据高效的框架，首先利用未标记的VLA动作片段构建基于动作的弱监督信号，捕捉异常模式，例如不一致的连续片段、冻结或闲置的动作以及激进的随机运动。然后，我们使用主动学习仅选择最不确定的轨迹进行时间戳级注释，并用这些信息丰富的标签微调检测器。在多个VLA策略上的实验表明，我们的方法提高了时间戳级和轨迹级故障检测性能。

View on arXiv Download PDF AI Translation

cs.RO / 2 / 2609.04355

VLA-Precision: Asymmetric Co-Bootstrapping for Efficient Real-World Online RL of Vision-Language-Action Models

VLA-Precision：用于高效现实世界在线强化学习的非对称协同自助法在视觉-语言-动作模型中的应用

Su, Chenyu, Shen, Zhaolong, Qian, Yuan, Qian, Chen, Zhang, Rui, Yan, Feng, Chen, Weixing, Zhang, Fei, Wang, Jiamin, Cong, Shuang, Shang, Weiwei

Abstract

Pretrained vision-language-action (VLA) models enable broad manipulation but remain unreliable in tasks demanding precision and repeatability. Applying real-world online reinforcement learning (RL) to VLA post-training enables autonomous trial-and-error improvement beyond demonstrations alone, but exposes two bottlenecks: 1) unreliable value signals can induce policy drift; 2) large-VLA overhead constrains throughput and sample efficiency. To address these challenges, we present VLA-Precision, an efficient real-world online RL framework featuring the Asymmetric Co-Bootstrapping (ACoB) algorithm and the ACoB-Stream architecture. Specifically, ACoB establishes asymmetric co-bootstrapping across timescales: early intervention-guided behavioral learning rapidly improves policy performance while enhancing online experience quality. As autonomous experience accumulates, global return propagation and local preference ranking progressively calibrate value estimates, yielding relative action advantages for reference-regularized policy improvement while suppressing drift. To enable ACoB on large VLAs, we develop ACoB-Stream, a closed-loop experience--policy architecture that establishes invariant-state decoupling and on-demand streaming as design principles, delivering up to 10.9$\times$ improvements in throughput and computational efficiency. Extensive evaluations on nine high-precision chemistry tasks across four categories and four robot embodiments show that VLA-Precision achieves 98.3\% mean success rate in 45.8 min/task, with 27.6 s episodes running at 1.2$\times$ and 1.8$\times$ the speeds of VLA and RL baselines. Resources are available at https://vla-precision.github.io.

Chinese Translation

预训练的视觉-语言-动作（VLA）模型能够广泛应用于操作，但在需要精确性和可重复性的任务中仍然不可靠。将现实世界在线强化学习（RL）应用于VLA后训练，可以实现超越单纯演示的自主试错改进，但也暴露出两个瓶颈：1）不可靠的价值信号可能导致策略漂移；2）大规模VLA的开销限制了吞吐量和样本效率。为了解决这些挑战，我们提出了VLA-Precision，这是一种高效的现实世界在线RL框架，采用了非对称协同自助法（Asymmetric Co-Bootstrapping, ACoB）算法和ACoB-Stream架构。具体而言，ACoB在时间尺度上建立了非对称协同自助：早期干预引导的行为学习快速提高策略性能，同时提升在线体验质量。随着自主经验的积累，全球回报传播和局部偏好排序逐步校准价值估计，为参考正则化策略改进提供相对行动优势，同时抑制漂移。为了在大型VLA上实现ACoB，我们开发了ACoB-Stream，这是一种闭环经验-策略架构，以不变状态解耦和按需流式传输作为设计原则，提供高达10.9倍的吞吐量和计算效率提升。在四个类别和四种机器人实施的九个高精度化学任务上的广泛评估表明，VLA-Precision在每个任务中以45.8分钟达到了98.3%的平均成功率，27.6秒的剧集以1.2倍和1.8倍的速度运行，超越了VLA和RL基线。相关资源可在 https://vla-precision.github.io 获取。

View on arXiv Download PDF AI Translation

cs.RO / 3 / 2609.04364

Scalable Edge-assisted Fusion and Path Prediction for Connected Autonomous Vehicles

可扩展的边缘辅助融合与路径预测用于连接自主车辆

Landle, Tyler, Isenberg, Jackson, Chatterjee, Abhijit, Daglis, Alexandros, Ramachandran, Umakishore

Abstract

The planning algorithms inside an Autonomous Vehicle (AV) rely on information from on-board sensors whose line of sight is limited by emerging traffic conditions and occlusions. Edge-assisted creation of a unified world model fusing information from AVs and Road Side Units (RSUs) in a geographical locale, and the prediction of AVs' future trajectories, can enhance the planning algorithms inside AVs to improve quality metrics, such as better traffic flow and collision prevention. AVs participating in such enhancements are called Connected Autonomous Vehicles (CAVs). However, such information generated by the edge (world model and motion predictions) must reach the planners within a tight Age of Information (AoI) time budget to be useful. The state of the art fuses per-CAV information: each AV fuses inputs from other actors locally, which limits both scalability with actor count and quality of results. We present Conductor, an edge-based solution for creating a unified world model from the perspective of a fixed anchor (e.g., an RSU) in a locale and predicting future trajectories of AVs in that locale. Our solution adheres to the AoI time budget by dynamically limiting the number of AVs that would lead to the best quality of results. Specifically, we introduce an occlusion-aware selector that favors information contribution by AVs that detect objects in the locale not covered by RSUs. We pair this selector with a runtime controller that adapts both the number of AV inputs to fuse and the amount of trajectory predictions in each cycle to stay within the AoI time budget. Evaluation on CAV simulation infrastructure shows our joint selector-controller meets the AoI safety bound across traffic scenarios with up to 31 CAVs, with fusion fidelity close to an Oracle and much better than a random selector under the same AoI constraint.

Chinese Translation

自主车辆（AV）内部的规划算法依赖于车载传感器提供的信息，而这些传感器的视线受限于不断变化的交通状况和遮挡物。边缘辅助的统一世界模型创建，通过融合来自自主车辆（AVs）和路边单元（RSUs）在特定地理区域的信息，以及对AV未来轨迹的预测，可以增强AV内部的规划算法，从而改善质量指标，如更好的交通流和碰撞预防。参与此类增强的AV被称为连接自主车辆（CAVs）。然而，边缘生成的信息（世界模型和运动预测）必须在严格的信息年龄（AoI）时间预算内到达规划者，才能发挥作用。现有技术融合每个CAV的信息：每个AV在本地融合来自其他参与者的输入，这限制了参与者数量的可扩展性和结果的质量。我们提出了Conductor，这是一个基于边缘的解决方案，从固定锚点（例如RSU）的视角创建统一的世界模型，并预测该区域内AV的未来轨迹。我们的解决方案通过动态限制AV的数量来满足最佳结果质量，从而遵循AoI时间预算。具体而言，我们引入了一种考虑遮挡的选择器，优先考虑那些能够检测到RSU未覆盖区域内物体的AV的信息贡献。我们将该选择器与运行时控制器配对，后者在每个周期内调整融合的AV输入数量和轨迹预测的数量，以保持在AoI时间预算内。在CAV仿真基础设施上的评估表明，我们的联合选择器-控制器在多达31个CAV的交通场景中满足AoI安全界限，其融合精度接近于Oracle，并且在相同的AoI约束下远优于随机选择器。

View on arXiv Download PDF AI Translation

cs.RO / 4 / 2609.04411

AquaBEV: Monocular Underwater BEV Occupancy with 3D Sonar Supervision

AquaBEV：单目水下鸟瞰视图占用预测与3D声纳监督

Dong, Trung Tien, Jin, Shengji, Chen, Chen, Sheng, Yi, Lin, Xiaomin

Abstract

Autonomous underwater robots are widely used for exploration, monitoring, and inspection, where safe navigation depends on understanding the surrounding free and occupied space. Bird's eye view (BEV) occupancy provides such a representation, but predicting it from a single underwater RGB image is difficult due to limited, unreliable geometric cues from appearance alone. 3D imaging sonar offers complementary geometric measurements to supervise this task. We introduce AquaBEV, a monocular underwater occupancy model that predicts local BEV occupancy from a single RGB image, using paired 3D imaging sonar as geometric supervision during training. AquaBEV maps visual features into a calibration free polar representation and applies causal decoding along the range dimension before reconstructing the prediction in Cartesian BEV coordinates. A controlled underwater occupancy benchmark was established, adapting representative occupancy methods to the same RGB to sonar task under a unified protocol. AquaBEV achieves 31.4 Visible IoU and 38.6 Observed IoU, 4.0% and 4.3% relative improvements over the strongest transferred baseline.

Chinese Translation

自主水下机器人广泛应用于探索、监测和检查，其安全导航依赖于对周围自由空间和被占用空间的理解。鸟瞰视图（BEV）占用提供了这样的表示，但由于仅凭外观提供的几何线索有限且不可靠，从单张水下RGB图像中预测BEV占用是困难的。3D成像声纳提供了补充的几何测量，以监督这一任务。我们提出了AquaBEV，这是一种单目水下占用模型，它利用配对的3D成像声纳作为训练期间的几何监督，从单张RGB图像中预测局部BEV占用。AquaBEV将视觉特征映射到无校准的极坐标表示，并在范围维度上应用因果解码，然后在笛卡尔BEV坐标中重建预测。我们建立了一个受控的水下占用基准，适应了代表性的占用方法，以统一协议将相同的RGB到声纳任务进行比较。AquaBEV实现了31.4的可见IoU和38.6的观察IoU，相较于最强的迁移基线分别提高了4.0%和4.3%。

View on arXiv Download PDF AI Translation

cs.RO / 5 / 2609.04464

Achieving Asymptotic Near-Optimality Without $\delta$-Similarity

在没有 $ ext{δ}$-相似性的情况下实现渐近近似最优性

Moncton, Michael, Frew, Eric

Abstract

Sampling-based motion planning algorithms are a popular class of trajectory planning algorithm due to their speed in complex, high-dimensional environments and ability to handle kinodynamic constraints, specifically through the use of forward dynamics propagation. Many such planners claim to achieve asymptotic near-optimality by proving the almost sure sampling of trajectories that are close to an optimal trajectory in the state space, known as $\delta$-similar trajectories. This paper shows that the proof behind asymptotic $\delta$-similarity relies on an unstated assumption that $\delta$-similar trajectory segments will always be kept once sampled. This assumption does not hold in general. A problematic case, referred to as ``crowding out,'' is described, where locally low-cost paths prevent trajectories that are $\delta$-similar to the optimal trajectory from being added to the tree. It is shown, however, that asymptotic near-optimality guarantees can still be achieved without guarantees of $\delta$-similar solution trajectories when crowding out is properly accounted for. An example environment and system are provided where crowding out is shown to occur, demonstrating a scenario where inductively sampling a $\delta$-similar solution trajectory is impossible.

Chinese Translation

基于采样的运动规划算法是一类受欢迎的轨迹规划算法，因其在复杂的高维环境中速度快且能够处理运动动力学约束，特别是通过使用前向动力学传播。许多此类规划者声称通过证明在状态空间中接近最优轨迹的轨迹（称为 $ ext{δ}$-相似轨迹）的几乎确定采样，来实现渐近近似最优性。本文表明，渐近 $ ext{δ}$-相似性的证明依赖于一个未明言的假设，即一旦采样，$ ext{δ}$-相似的轨迹段将始终被保留。这个假设在一般情况下并不成立。文中描述了一个问题案例，称为“拥挤排斥”，在该案例中，局部低成本路径阻止了与最优轨迹 $ ext{δ}$-相似的轨迹被添加到树中。然而，本文表明，当适当考虑拥挤排斥时，仍然可以在没有 $ ext{δ}$-相似解轨迹的保证下实现渐近近似最优性。提供了一个示例环境和系统，展示了拥挤排斥的发生，演示了一个归纳采样 $ ext{δ}$-相似解轨迹不可能的场景。

View on arXiv Download PDF AI Translation

cs.RO / 6 / 2609.04545

SocioGesture: Real-Time and Adaptive Social Gesture Perception for Human-Robot Interaction

SocioGesture：用于人机交互的实时自适应社交手势感知

Fu, Wenjin, Wu, Li-Fan, Peter, Jerin, Huyen, Chip, Chen, Boyuan, Liphardt, Jan

Abstract

Robots interacting with people must recognize not only explicit commands, but also social cues such as invitations, refusals, and unavailability. In real deployments, these cues must be inferred from noisy onboard perception under partial occlusion, changing viewpoints, and strict latency constraints. We present SocioGesture, a real-time adaptive social gesture perception system for human-robot interaction (HRI). SocioGesture uses a compact confidence-aware body-hand skeleton representation and a lightweight dual-stream model that fuses body motion with hand articulation for low-latency onboard recognition. To improve deployment robustness, we train the model with occlusion-aware skeleton corruption, exposing it to missing hands, occluded arms, and temporally unstable keypoints without increasing the inference cost. On a social gesture dataset collected in mixed indoor-outdoor HRI scenarios, SocioGesture achieves strong held-out-subject recognition, substantially improves robustness under structured joint occlusion, and runs in real time on a robot-mounted edge device. During deployment, uncertain interaction segments are saved for offline labeling and adaptation, enabling SocioGesture to expand its gesture vocabulary while preserving performance in the original classes. These results demonstrate a practical path toward robust, efficient, and adaptive social perception for interactive robots.

Chinese Translation

与人类互动的机器人不仅需要识别明确的指令，还必须理解社交线索，如邀请、拒绝和不可用性。在实际应用中，这些线索必须在部分遮挡、视角变化和严格的延迟限制下，从嘈杂的机载感知中推断出来。我们提出了SocioGesture，一个用于人机交互（HRI）的实时自适应社交手势感知系统。SocioGesture采用了一种紧凑的自信度感知身体-手骨架表示和一种轻量级双流模型，将身体运动与手部动作融合，以实现低延迟的机载识别。为了提高部署的鲁棒性，我们通过遮挡感知的骨架损坏训练模型，使其能够应对缺失的手、被遮挡的手臂和时间上不稳定的关键点，而不增加推理成本。在一个在混合室内外HRI场景中收集的社交手势数据集上，SocioGesture实现了强大的被保留主体识别，显著提高了在结构化关节遮挡下的鲁棒性，并在机器人安装的边缘设备上实时运行。在部署过程中，不确定的交互片段被保存以供离线标注和适应，使SocioGesture能够扩展其手势词汇，同时保持原有类别的性能。这些结果展示了实现互动机器人强大、高效和自适应社交感知的实用路径。

View on arXiv Download PDF AI Translation

cs.RO / 7 / 2609.04552

Continual Field-Adaptive Models (CFAMs) for Post-Deployment Physical AI

持续场适应模型（CFAM）用于部署后的物理人工智能

Singh, Amarjot, Pancholi, Tanmay R., Kothari, Jainam, Mahajan, Shrirang, Bansal, Ketan, Erickson, Zackory, Loianno, Giuseppe, Bayen, Alexandre M., Schneider, Jeff, Nakayama, Vince

Abstract

Unattended interactive autonomy - machines that step into danger in place of humans and complete tasks with human tools - remains a missing capability in mission-critical operations. These domains offer scarce training data and only onboard compute, yet deployed systems must face novelty without erasing prior competence. We introduce Continual Field-Adaptive Models (CFAMs), which learn efficiently in the lab and continue learning after deployment through autonomous, gradient-free, on-device updates. CFAM uses a complementary learning architecture with a frozen slow-learning component and a fast-learning Capsule Field. The slow component contains three cortices: Sensor, which maps multimodal input into 3D-grounded geometry; Reasoning, which decomposes tasks into skills and evaluates outcomes; and Action, which executes geometric skills. The Capsule Field stores field learning one-shot and gradient-free as Competence Capsules. Skill installation is few-shot in the lab and continual in the field; open-world novelty is outside scope. We evaluate CFAM across five embodiments: manipulator, quadruped, humanoid, quadrotor, and off-road vehicle. Baselines (pi0, CogACT, SpatialVLA) use the same in-house multi-embodiment dataset for physical-platform comparisons. CFAM reaches the operating point of a standard policy trained on the full prior-training dataset using 40% of the data, or 2.5x fewer trajectories. At test time, autonomous capture of verified near-OOD cases improves action success by 13.9 percentage points. In sequential simulation, backward transfer is -0.5 percentage points versus -11.4 for LoRA. CFAM therefore provides a bounded form of post-deployment physical intelligence: few-shot skill learning, autonomous field growth from verified near-OOD experience, and retention of prior competence.

Chinese Translation

无人值守的互动自主性——机器代替人类进入危险环境并使用人类工具完成任务——仍然是关键任务操作中缺失的一项能力。这些领域提供稀缺的训练数据且仅具备机载计算能力，但已部署的系统必须在不抹去先前能力的情况下面对新颖性。我们提出了持续场适应模型（CFAM），该模型能够在实验室中高效学习，并通过自主、无梯度的设备内更新在部署后继续学习。CFAM采用了一种互补的学习架构，包含一个冻结的慢学习组件和一个快速学习的胶囊场（Capsule Field）。慢组件包含三个皮层：传感器（Sensor），将多模态输入映射到三维基础几何；推理（Reasoning），将任务分解为技能并评估结果；以及行动（Action），执行几何技能。胶囊场以一次性和无梯度的方式存储场学习作为能力胶囊（Competence Capsules）。技能安装在实验室中是少样本的，而在现场是持续的；开放世界的新颖性超出范围。我们在五种实现上评估CFAM：操纵器、四足机器人、人形机器人、四旋翼和越野车。基准（pi0、CogACT、SpatialVLA）使用相同的内部多实现数据集进行物理平台比较。CFAM在使用40%的数据或2.5倍更少的轨迹时，达到了在完整先前训练数据集上训练的标准策略的操作点。在测试时，自动捕获经过验证的近OOD案例使得行动成功率提高了13.9个百分点。在顺序模拟中，反向转移为-0.5个百分点，而LoRA为-11.4个百分点。因此，CFAM提供了一种有限形式的部署后物理智能：少样本技能学习、基于经过验证的近OOD经验的自主场增长，以及对先前能力的保留。

View on arXiv Download PDF AI Translation

cs.RO / 8 / 2609.04602

NavArena: Automated Construction of Goal-Oriented Navigation Benchmarks from 3D Gaussian Splatting Reconstructions

NavArena：从3D高斯点云重建自动构建目标导向导航基准

Wang, Junhui, Yang, Wei, Li, Xinyao, Fan, Ningjing, Yin, Yuehao, Chen, Xuecheng, Gao, Chao

Abstract

Fixed 3D Gaussian Splatting (3DGS) reconstructions provide realistic novel views but lack the traversability constraints, valid goals, and closed-loop protocols required for navigation evaluation. We introduce NavArena, an automated framework that transforms fixed 3DGS reconstructions into benchmarks for goal-oriented visual navigation. NavArena integrates a frozen 3DGS model for egocentric RGB-D rendering, an occupancy costmap derived from Gaussian density and height statistics for reachability and collision queries, and semantic goal candidates lifted from multi-view open-vocabulary masks. These components support the automatic generation and unified closed-loop evaluation of goal-oriented navigation episodes. Across more than 2{,}000 scenes, NavArena generates 22.2 million expert trajectories. Spatial and semantic evaluations assess the derived navigation representations, while policy rollouts demonstrate the diagnostic value of the unified evaluation protocol. NavArena enables scalable and reproducible navigation evaluation on large-scale 3DGS reconstructions, and all benchmark-generation tools, evaluation protocols, and derived assets will be released publicly.

Chinese Translation

固定的3D高斯点云（3DGS）重建提供了逼真的新视图，但缺乏进行导航评估所需的可 traversability 约束、有效目标和闭环协议。我们提出了NavArena，一个自动化框架，将固定的3DGS重建转化为目标导向视觉导航的基准。NavArena集成了一个用于自我中心RGB-D渲染的冻结3DGS模型、一个基于高斯密度和高度统计的占用成本图，用于可达性和碰撞查询，以及从多视角开放词汇掩码中提取的语义目标候选。这些组件支持目标导向导航情节的自动生成和统一的闭环评估。在超过2,000个场景中，NavArena生成了2220万个专家轨迹。空间和语义评估评估了所得到的导航表示，而策略展开则展示了统一评估协议的诊断价值。NavArena使得在大规模3DGS重建上进行可扩展和可重复的导航评估成为可能，所有基准生成工具、评估协议和派生资产将公开发布。

View on arXiv Download PDF AI Translation

cs.RO / 9 / 2609.04607

Open-Set 3D Scene Graphs for Field Robotics: An Outdoor Case Study

开放集3D场景图在现场机器人中的应用：一个户外案例研究

Samuelson, Chad R., Slade, Gabriel R., Mangelson, Joshua G.

Abstract

Three-dimensional scene graphs (3DSGs) have emerged as a promising approach for building geometrically grounded, semantically informed, hierarchical general-purpose maps to support high-level robotic reasoning. However, the behavior of 3DSGs in real-world outdoor deployments remains poorly understood, particularly when combined with open-set vision-language models (VLMs). In this field report, we analyze the components common to most 3DSG representations across five outdoor robotic datasets to characterize challenges that arise in complex outdoor environments. Using the recently proposed Terra 3DSG as a case study, we investigate semantic point embeddings, place-node graph navigation, region-level understanding, and memory size across the five diverse datasets. We additionally introduce novel consistency metrics to evaluate whether semantic and structural graph properties remain stable across repeated traversals of the same environment. Our analysis reveals that outliers and multiple modes are common in VLM point embeddings across all tested datasets with outlier ratios above $0.1$ for around $30\%$ of points. We demonstrate the feasibility of outdoor 3DSGs for navigation-based object retrieval, achieving success rates near $70\%$, though performance is limited by traversability failures and inefficient routing, with trajectories averaging approximately $66\%$ suboptimal path efficiency. Region-level understanding remains challenging in complex natural environments, with low average F1 scores around $0.359$. Overall, our results show that outdoor 3DSGs can maintain compact (less than $600$MB for multi-kilometer trajectories) and relatively consistent large-scale environment representations, while highlighting open challenges in handling multiple semantic modes, incorporating traversability into graph structures, and improving higher-level region understanding.

Chinese Translation

三维场景图（3DSGs）作为一种有前景的方法，已被提出用于构建几何基础、语义信息丰富的层次化通用地图，以支持高级机器人推理。然而，3DSGs在实际户外部署中的表现仍然不够清楚，特别是在与开放集视觉-语言模型（VLMs）结合时。在本次实地报告中，我们分析了五个户外机器人数据集中大多数3DSG表示的共同组成部分，以表征在复杂户外环境中出现的挑战。以最近提出的Terra 3DSG为案例，我们研究了语义点嵌入、地点节点图导航、区域级理解和五个不同数据集的内存大小。此外，我们引入了新的一致性度量，以评估语义和结构图属性在同一环境的重复遍历中是否保持稳定。我们的分析显示，在所有测试数据集中，VLM点嵌入中常见异常值和多模态现象，异常值比例超过$0.1$的点约占$30\%$。我们展示了户外3DSGs在基于导航的物体检索中的可行性，成功率接近$70 ext{ extperthousand}$，尽管性能受到可遍历性失败和低效路径规划的限制，轨迹的平均次优路径效率约为$66 ext{ extperthousand}$。在复杂自然环境中，区域级理解仍然具有挑战性，平均F1分数低至$0.359$。总体而言，我们的结果表明，户外3DSGs能够保持紧凑（多公里轨迹小于$600$MB）且相对一致的大规模环境表示，同时突显了处理多种语义模式、将可遍历性纳入图结构以及改善高级区域理解等开放挑战。

View on arXiv Download PDF AI Translation

cs.RO / 10 / 2609.04620

Pack It My Way: Triadic Human-Robot Collaboration for Personalized Autonomous Packing

按我的方式打包：三方人机协作实现个性化自主打包

Kotapati, Sandeep Chowdary, Gao, Yanxin, Lin, Tsung-Chi

Abstract

Personalized autonomous packing requires robots to account for resident preferences that cannot be inferred from scene geometry alone. Expert teleoperators can interpret these preferences and translate them into feasible robot actions, but continuous expert involvement limits scalable deployment. In this paper, we investigate triadic human-robot collaboration among a resident, a correction mediator, and a robot by comparing human-expert and voice-agent mediation. We evaluate the two conditions in a user study across Protection, Compactness, andGrouping tasks, using a Show-Correct-Generalize process to assess preference correction and subsequent generalization after the surrounding objects are rearranged. Results show that voice-agent mediation achieves outcomes comparable to human-expert mediation in two of the three preference categories, despite receiving shorter and less detailed instructions. Both mediators are similarly easy to use, although the human expert is perceived as more reliable. These findings demonstrate the potential of voice agents to reduce expert involvement while identifying perceived reliability and preference generalization as remaining challenges.

Chinese Translation

个性化自主打包要求机器人考虑居民的偏好，而这些偏好无法仅通过场景几何推断。专家远程操作员能够解读这些偏好并将其转化为可行的机器人行动，但持续的专家参与限制了可扩展部署。在本文中，我们通过比较人类专家和语音代理的调解，研究居民、纠正中介和机器人之间的三方人机协作。我们在用户研究中评估了这两种条件，涉及保护性、紧凑性和分组任务，采用展示-纠正-概括的过程来评估偏差纠正及在周围物体重新排列后的后续概括。结果表明，尽管语音代理接受的指令较短且不够详细，但在三类偏好中的两类中，语音代理调解的结果与人类专家调解相当。两种调解方式的使用难易程度相似，尽管人类专家被认为更可靠。这些发现表明，语音代理在减少专家参与方面具有潜力，同时识别感知可靠性和偏好概括仍然是挑战。

View on arXiv Download PDF AI Translation

cs.RO / 11 / 2609.04759

Dressing in Motion: A Human Motion-Aware Diffusion Policy for Robot-Assisted Dressing

动态穿衣：一种基于人类运动感知的机器人辅助穿衣扩散策略

Sun, Haoxiang, Wang, Fangyuan, Huang, Songhao, Liu, Justina Y. W., Zhu, Jihong, Zhou, Peng, Navarro-Alarcon, David

Abstract

Robotic dressing assistance is a promising solution for supporting older adults with physical impairments in daily living. However, dressing under human motion remains challenging, as complex garment--human contact and occlusions make it difficult to generate actions aligned with arm movements. In this letter, we propose a visuomotor policy that learns dressing skills from static expert demonstrations and generalizes to dynamic user-motion scenarios. A diffusion policy tailored to garment--human interaction geometry learns from partially observed point clouds with varied arm postures. We then introduce an object-centric representation based on PDE diffusion to capture the axial distribution of the arm. By sampling motion-relevant regions and registering them across consecutive observations, the proposed method approximates arm motion and reactively adapts the executed trajectory. We evaluate our method in simulation and a real-world human study involving nine participants, three garment types, and six arm-motion patterns. Results show that our method outperforms baselines in dressing progress, freedom of movement, and user comfort. The project website is https://anonymous.4open.science/w/dressing-in-motion.

Chinese Translation

机器人穿衣辅助是支持身体残疾老年人日常生活的一个有前景的解决方案。然而，在人类运动下穿衣仍然具有挑战性，因为复杂的服装与人体接触以及遮挡使得生成与手臂运动一致的动作变得困难。在本文中，我们提出了一种视觉运动策略，该策略从静态专家演示中学习穿衣技能，并能够推广到动态用户运动场景。针对服装与人体交互几何的扩散策略，从部分观察到的点云中学习，适应不同的手臂姿势。然后，我们引入了一种基于偏微分方程（PDE）扩散的以物体为中心的表示，以捕捉手臂的轴向分布。通过对运动相关区域进行采样并在连续观察中进行配准，所提出的方法近似手臂运动并实时调整执行轨迹。我们在模拟和涉及九名参与者、三种服装类型和六种手臂运动模式的真实人类研究中评估了我们的方法。结果表明，我们的方法在穿衣进展、运动自由度和用户舒适度方面优于基线。项目网站为 https://anonymous.4open.science/w/dressing-in-motion。

View on arXiv Download PDF AI Translation

cs.RO / 12 / 2609.04770

Continuous Cognitive Coverage for Autonomous Robots via Event-Dependent Cognitive Treatment and Learning

通过事件依赖的认知处理与学习实现自主机器人连续认知覆盖

Su, Hong

Abstract

Autonomous robots continuously encounter objects, changes, and situations, and every event admitted into cognition should receive an appropriate cognitive treatment rather than remain untreated until an explicit task requires attention. However, existing task-driven, reactive, or fixed-reasoning approaches generally process only selected events or apply predefined reasoning procedures, making it difficult to provide continuous cognitive coverage with differentiated treatment. This paper proposes a continuous cognitive coverage framework in which every cognitively admitted event is assigned an event-dependent cognitive treatment according to its state, context, and history. Different events may therefore invoke description, memory, risk prediction, planning, diagnosis, analogy, or other learned treatments. Familiar events can be processed automatically by learned mechanisms, whereas unfamiliar or uncertain events invoke explicit deliberation or fallback reasoning. Multiple cognitive processes can be suspended, resumed, and interleaved so that cognitive processing continues as new events arrive or existing events await evidence. Validated experiences are continuously learned to automate, refine, and revise event-specific treatments. Experiments achieve 96.76% structured treatment accuracy with 93.66% automatic processing, 92.64% cognitive coverage under bursty-delayed workloads, and 79.53% continual-learning joint accuracy, with novel-event reuse reaching 100% automatic processing.

Chinese Translation

自主机器人不断遇到物体、变化和情况，每一个被认知接受的事件都应当获得适当的认知处理，而不是在明确的任务需要关注时才被处理。然而，现有的任务驱动、反应式或固定推理方法通常只处理选定的事件或应用预定义的推理程序，这使得提供具有差异化处理的连续认知覆盖变得困难。本文提出了一种连续认知覆盖框架，其中每个被认知接受的事件根据其状态、上下文和历史被分配事件依赖的认知处理。因此，不同的事件可能会引发描述、记忆、风险预测、规划、诊断、类比或其他学习到的处理。熟悉的事件可以通过学习机制自动处理，而不熟悉或不确定的事件则需要明确的深思熟虑或后备推理。多个认知过程可以被暂停、恢复和交错，以便在新事件到达或现有事件等待证据时继续认知处理。经过验证的经验被持续学习，以自动化、精炼和修订事件特定的处理。实验结果显示，在突发延迟工作负载下，结构化处理准确率达到96.76%，自动处理率为93.66%，认知覆盖率为92.64%，持续学习的联合准确率为79.53%，新事件重用实现100%的自动处理。

View on arXiv Download PDF AI Translation

cs.RO / 13 / 2609.04799

HaptiNet: Networked Haptic Robots Enable Physical Co-presence in Geographically-Unconstrained Rehabilitation

HaptiNet：网络化触觉机器人实现地理无约束康复中的身体共存

Sun, Chenyang, Dong, Mingjie, Deng, Haodong, Liu, Yudong, Chen, Yi-Feng, Lin, Jun, Huang, Changlong, Guo, Jie, Liu, Yantong, Liu, Yang, Lin, Yuzhou, Long, Jianjun, Xing, Zheng, Zhao, Sining, Zhang, Xuemin, Wang, Zhiyong, Li, Zhenhong, Wu, Dongrui, Liu, Honghai, Dai, Jian S., Zhang, Mingming

Abstract

Cooperative rehabilitation enhances engagement, task performance, and social-motor interaction, yet it demands physical co-presence: users must transmit forces, coordinate movements, and infer intent through haptic contact. Telerehabilitation promises to expand access for patients constrained by distance, mobility, or clinical disparities, yet current techniques remain predominantly audiovisual while leaving users haptically and physically isolated. Here, we introduce HaptiNet, a networked haptic robotic system enabling physical co-presence for geographically distributed users via force-mediated interaction. Each robotic terminal features a low-inertia, long-stroke design with high force-feedback capacity, tailored for haptic rendering in upper-limb training. Building on these terminals, HaptiNet creates a distributed haptic network with an imitation-learning-based delay compensator, enabling users to physically perceive and coordinate with one another over distance. We validated HaptiNet in 284 healthy participants and 111 patients with neurological impairments across progressively realistic settings, including laboratory tests, cross-city deployments, and clinical applications. HaptiNet preserved task-level force rendering consistency across single-user and multi-user scenarios. Compared with solo and visual cooperative training, haptic cooperation improved task performance by 24% and 22%, respectively, while also boosting engagement and interpersonal motor synchrony. Across three intercity links totaling approximately 4,000 km, HaptiNet maintained stable haptic interaction among patients with neurological impairments, producing a 3.87-fold greater baseline-to-training score improvement and a 106% higher patient-applied effort over the solo condition.

Chinese Translation

合作康复增强了参与感、任务表现和社会运动互动，但这需要身体共存：用户必须通过触觉接触传递力量、协调动作和推断意图。远程康复承诺为因距离、行动能力或临床差异受限的患者扩大获取机会，但目前的技术仍主要依赖视听方式，使用户在触觉和身体上处于孤立状态。在此，我们介绍了HaptiNet，一个网络化触觉机器人系统，通过力介导的互动使地理分布的用户能够实现身体共存。每个机器人终端具有低惯量、长行程的设计，具备高力反馈能力，专为上肢训练中的触觉呈现而定制。在这些终端的基础上，HaptiNet创建了一个分布式触觉网络，配备模仿学习基础的延迟补偿器，使用户能够在距离上身体感知和协调彼此。我们在284名健康参与者和111名神经损伤患者中验证了HaptiNet，涵盖了逐步逼真的环境，包括实验室测试、跨城市部署和临床应用。HaptiNet在单用户和多用户场景中保持了任务级别的力呈现一致性。与单独和视觉合作训练相比，触觉合作分别提高了任务表现24%和22%，同时也增强了参与感和人际运动同步性。在三条总长约4000公里的城市间链接中，HaptiNet在神经损伤患者之间维持了稳定的触觉互动，产生了3.87倍的基线到训练分数改善和106%的患者施加努力高于单独条件。

View on arXiv Download PDF AI Translation

cs.RO / 14 / 2609.04807

CoLMIN: LLM-based Multi-Decision Path Negotiation for Cooperative Autonomous Driving

CoLMIN：基于大型语言模型的合作自主驾驶多决策路径协商

Huang, Zhe, Fan, Zhaoxin, Wang, Shuo, Wu, Wenjun, Zhao, Xuan, Liu, Min

Abstract

Multi-vehicle cooperative autonomous driving enhances the safety and reliability of autonomous driving systems through information sharing among connected vehicles, demonstrating significant potential for improving traffic safety. LLM-based approaches leverage strong reasoning capabilities of LLMs to enable effective inter-vehicle negotiation and improve cooperative driving performance. However, driving decisions in complex traffic scenarios are inherently multi-solution in nature. As a result, existing negotiation-based methods often converge prematurely to suboptimal solutions, hindering consensus formation and limiting the practical deployment of cooperative autonomous driving systems. To address this challenge, we propose CoLMIN, the LLM-based multi-decision path negotiation framework for cooperative autonomous driving, achieving stable decision consensus through multi-decision path negotiation and reflective reasoning. To achieve stable and high-quality consensus in cooperative autonomous driving, CoLMIN consists of three key components: (i) an LLM-based Multi-Intent Negotiation module (LMin), which adopts a Negotiator-Evaluator paradigm and generates multiple candidate driving intentions for joint evaluation; (ii) an Evaluation-based Shallow Reflection Module (ESRM), which analyzes negotiation outcomes and provides feedback to guide subsequent negotiations, thereby accelerating consensus formation; and (iii) an LLM-based Deep Reflection Module (LDRM), which performs long-term reflection over negotiation histories to mitigate cognitive fixation and prevent the system from converging to suboptimal solutions. Experimental results in the CARLA simulation environment demonstrate that CoLMIN significantly outperforms existing methods in challenging interactive driving scenarios.

Chinese Translation

多车辆合作自主驾驶通过连接车辆之间的信息共享增强了自主驾驶系统的安全性和可靠性，展现了显著提升交通安全的潜力。基于大型语言模型（LLM）的方法利用LLM强大的推理能力，促进有效的车际协商并改善合作驾驶性能。然而，复杂交通场景中的驾驶决策本质上具有多解性。因此，现有的基于协商的方法往往过早收敛到次优解，阻碍了共识的形成，限制了合作自主驾驶系统的实际部署。为了解决这一挑战，我们提出了CoLMIN，一种基于LLM的合作自主驾驶多决策路径协商框架，通过多决策路径协商和反思推理实现稳定的决策共识。为了在合作自主驾驶中实现稳定和高质量的共识，CoLMIN由三个关键组件组成：（i）基于LLM的多意图协商模块（LMin），采用协商者-评估者范式，生成多个候选驾驶意图以供联合评估；（ii）基于评估的浅层反思模块（ESRM），分析协商结果并提供反馈以指导后续协商，从而加速共识形成；（iii）基于LLM的深层反思模块（LDRM），对协商历史进行长期反思，以减轻认知固定并防止系统收敛到次优解。在CARLA仿真环境中的实验结果表明，CoLMIN在具有挑战性的互动驾驶场景中显著优于现有方法。

View on arXiv Download PDF AI Translation

cs.RO / 15 / 2609.04817

Strict Modes Everywhere - Bringing Order Into Dynamics of Mechanical Systems by a Potential Compatible With the Geodesic Flow

严格模式无处不在 - 通过与测地线流相容的势能为机械系统的动力学带来秩序

Sachtler, Arne, Albu-Schäffer, Alin

Abstract

Strict nonlinear normal modes provide very regular families of oscillations within conservative mechanical systems. However, a strict normal mode will generally be an isolated curve within the configuration space of the system. In this letter, we design a potential that will densely fill the configuration space with strict normal modes such that each configuration belongs to one mode and each mode passes through a common point, the equilibrium. As the potential can be realized by (nonlinear) elastic elements it can be used to execute a variety of periodic trajectories very efficiently. Most of the required torques will come from the elastic elements in the system and not from the actuators. We also design a controller stabilizing the system to a desired target mode and a controller performing swing-up and compensating dissipated energy. Finally, we showcase the approach for a two DoF manipulator. The experiments show that the approach performed well for the example system.

Chinese Translation

严格的非线性正常模式在保守机械系统中提供了非常规则的振荡族。然而，严格的正常模式通常在系统的配置空间中是一个孤立的曲线。在本文中，我们设计了一种势能，能够在配置空间中密集填充严格的正常模式，使得每个配置都属于一个模式，并且每个模式都通过一个共同的点，即平衡点。由于该势能可以通过（非线性）弹性元件实现，因此可以非常高效地执行多种周期性轨迹。所需的大多数扭矩将来自系统中的弹性元件，而不是来自执行器。我们还设计了一个控制器，使系统稳定到期望的目标模式，以及一个控制器用于摆动提升和补偿耗散能量。最后，我们展示了该方法在一个两自由度操纵器上的应用。实验表明，该方法在示例系统中表现良好。

View on arXiv Download PDF AI Translation

cs.RO / 16 / 2609.04851

Coupled Control and Wireless World Models for Resilient Remote Robotic Control

用于弹性远程机器人控制的耦合控制与无线世界模型

Madushanka, H. P., Samarakoon, Sumudu, Bennis, Mehdi

Abstract

Remote robotic systems operating over wireless networks must maintain reliable control despite limited communication resources, changing channel conditions, and environmental disturbances.However, continuously transmitting high-dimensional sensory observations, such as camera images, increases communication overhead and energy consumption while reducing robustness under unreliable connectivity.To address these challenges, this paper proposes a resilient communication-aware remote robotic control framework based on coupled control and wireless Joint Embedding Predictive Architecture (JEPA) world models that jointly capture robot dynamics and wireless channel evolution from visual observations and a combination of raw and structured radio frequency (RF) representations based on spectrograms and Persistence Images(PIs).The learned latent representations enable predictive communication scheduling by jointly forecasting future robot states and wireless conditions, thereby reducing unnecessary uplink transmissions while maintaining reliable control performance.Furthermore, an adaptive resilience mechanism detects latent prediction discrepancies and efficiently adapts perception embeddings to accommodate wireless and visual environmental changes without retraining the complete control policy.The proposed framework is evaluated in a synchronized Gazebo-Robot Operating System (ROS)-Sionna robot-wireless simulation environment under diverse wireless propagation and perception perturbations.Experimental results demonstrate significant improvements in communication efficiency, robustness, and resilience while maintaining navigation performance compared with conventional Proportional Integral Derivative (PID), model-free Deep Q-Network (DQN), and predictive approaches based on Vision Transformers(ViTs).

Chinese Translation

在无线网络上运行的远程机器人系统必须在有限的通信资源、变化的信道条件和环境干扰下保持可靠的控制。然而，持续传输高维感知观测（如摄像头图像）会增加通信开销和能耗，同时在不可靠的连接下降低鲁棒性。为了解决这些挑战，本文提出了一种基于耦合控制和无线联合嵌入预测架构（Joint Embedding Predictive Architecture, JEPA）世界模型的弹性通信感知远程机器人控制框架，该框架共同捕捉来自视觉观测的机器人动态和无线信道演变，以及基于声谱图和持久图像（Persistence Images, PIs）的原始和结构化射频（Radio Frequency, RF）表示的组合。所学习的潜在表示通过共同预测未来的机器人状态和无线条件，实现预测通信调度，从而减少不必要的上行传输，同时保持可靠的控制性能。此外，自适应弹性机制能够检测潜在预测差异，并有效调整感知嵌入，以适应无线和视觉环境的变化，而无需重新训练完整的控制策略。所提出的框架在多种无线传播和感知扰动下的同步Gazebo-机器人操作系统（Robot Operating System, ROS）-Sionna机器人无线仿真环境中进行了评估。实验结果表明，与传统的比例-积分-微分控制（Proportional Integral Derivative, PID）、无模型深度Q网络（Deep Q-Network, DQN）和基于视觉变换器（Vision Transformers, ViTs）的预测方法相比，通信效率、鲁棒性和弹性显著提高，同时保持导航性能。

View on arXiv Download PDF AI Translation

cs.RO / 17 / 2609.04893

Reasoning Without Inference Cost: Latent Semantic Scaffolding for Robot VLA Policies

无推理成本的推理：机器人视觉-语言-动作政策的潜在语义支架

Li, Andrew Ting Yan, Li, Zhuo, Yang, Zhelin, Dong, Zhipeng, Rouxel, Quentin, Chen, Fei

Abstract

Vision-language-action (VLA) models are trained by imitation and capture what action to take but not why; adding causal reasoning improves manipulation, but current methods pay for it at inference time - generating reasoning tokens or rolling out predicted future states at every step, a cost that compounds over long horizons. We ask whether this benefit can instead be captured during training and discarded before deployment. We introduce Latent Semantic Scaffolding (LSS), an auxiliary loss applied during human-demonstration pretraining that aligns a VLA's action-token representations to text embeddings of physical-reasoning rationales through a small projection head. The head is dropped at inference, leaving the unmodified base policy with zero added cost. Our central finding concerns alignment granularity: aligning each action token to the rationale of its own manipulation phase (Dense LSS) rather than to a single pooled episode-level embedding (Pooled LSS) yields representations that transfer markedly better to held-out tasks. Dense LSS attains both the best in-distribution success and the best transfer to tasks unseen during alignment, whereas pooled alignment over-specializes to the training task. A representational probe shows Dense LSS induces roughly twice the per-phase separability in the backbone, supporting that phase-local alignment is the operative mechanism.

Chinese Translation

视觉-语言-动作（VLA）模型通过模仿训练，捕捉到应该采取的行动，但并未解释原因；增加因果推理可以改善操作，但当前方法在推理时需要付出代价——在每一步生成推理令牌或展开预测的未来状态，这种成本在长时间跨度中会累积。我们探讨是否可以在训练过程中捕捉这一好处，并在部署前丢弃。我们引入了潜在语义支架（Latent Semantic Scaffolding, LSS），这是一种在人工示范预训练期间应用的辅助损失，通过一个小的投影头将VLA的动作令牌表示与物理推理理由的文本嵌入对齐。该投影头在推理时被丢弃，从而使未修改的基础策略没有额外成本。我们的核心发现涉及对齐的粒度：将每个动作令牌与其自身操作阶段的理由对齐（密集LSS）而不是与单一汇总的情节级嵌入对齐（汇总LSS），会产生在保留任务中显著更好的迁移表示。密集LSS在分布内成功率和对未在对齐期间见过的任务的最佳迁移方面均表现最佳，而汇总对齐则过度专注于训练任务。一个表示探针显示，密集LSS在主干网络中引入了大约两倍的每阶段可分离性，支持阶段局部对齐是有效机制的观点。

View on arXiv Download PDF AI Translation

cs.RO / 18 / 2609.05084

ToPos: Automated Optimal Positioning on Topographic Manifolds using Constrained Geodesic Voronoi Decomposition

ToPos：基于约束测地Voronoi分解的地形流形自动最优定位

Raveendran, Rajesh, Vanhamaa, Akseli, Suutala, Jaakko, Tikanmäki, Antti, Röning, Juha

Abstract

Reliable autonomous mapping, environmental sampling, last-mile logistics, and infrastructure deployment depend on the optimal surface area-balanced distribution of Spatial Reference Sites (SRS). Conventional 2D Euclidean methods often fail in high-relief environments by neglecting topographic variations and physical obstructions. This leads to significant planimetric distortion, spatial clustering, and the placement of targets in inaccessible or shadowed regions, compromising both data integrity and operational safety. This paper introduces ToPos, an automated framework for TOPography-aware Optimal Sampling on topographic manifolds. We treat the terrain as a discrete 2-dimensional manifold embedded in 3D Euclidean space and replace standard flat-map distances with non-Euclidean geodesic distances that follow the actual surface geometry. The point distribution is formulated as an optimization problem using a Constrained Geodesic Voronoi Decomposition, solved via a Riemannian Nesterov Accelerated Gradient (NAG) engine. Our approach restricts target locations to a feasible "safe zone," accounting for non-traversable slopes, vegetation, environmental occlusions, etc. Through evaluations on non-convex sinusoidal manifolds, we show that ToPos mitigates planimetric distortion by utilizing geodesic metrics. This approach results in a $\sim$74% improvement in optimal surface area-balanced distribution, as measured by the coefficient of variation (CV) of the Voronoi cell areas. The framework is architected as a Geographic Information System (GIS)-ready micro-service to bolster the mentioned applications. Index Terms: Topographic Manifolds, Geodesic Voronoi Decomposition, Infrastructure Deployment, 3D Mapping, Spatial Sampling, and Non-Euclidean Optimization.

Chinese Translation

可靠的自主映射、环境采样、最后一公里物流和基础设施部署依赖于空间参考点（SRS）的最优表面积平衡分布。传统的二维欧几里得方法在高地形环境中常常失效，因为它们忽视了地形变化和物理障碍。这导致了显著的平面畸变、空间聚类以及目标放置在不可达或阴影区域，从而危及数据完整性和操作安全性。本文介绍了ToPos，一个针对地形流形的地形感知最优采样的自动化框架。我们将地形视为嵌入在三维欧几里得空间中的离散二维流形，并用遵循实际表面几何的非欧几里得测地距离替代标准的平面图距离。点分布被构造为一个优化问题，使用约束测地Voronoi分解，通过黎曼Nesterov加速梯度（NAG）引擎求解。我们的方法将目标位置限制在可行的“安全区”内，考虑了不可通行的坡度、植被、环境遮挡等因素。通过对非凸正弦流形的评估，我们展示了ToPos通过利用测地度量减轻了平面畸变。这种方法在Voronoi单元面积的变异系数（CV）测量下，实现了约74%的最优表面积平衡分布的改善。该框架被设计为一个地理信息系统（GIS）就绪的微服务，以增强上述应用。关键词：地形流形、测地Voronoi分解、基础设施部署、三维映射、空间采样和非欧几里得优化。

View on arXiv Download PDF AI Translation

cs.RO / 19 / 2609.05133

A Schema Bounded Language Model for Refining Robot Policies Without Destabilizing Local Learning

一种模式约束语言模型用于优化机器人策略而不破坏局部学习

Dong, Chongwen, Saint-Germain, Mithun Paul, Asif, Pinjari, daCunha, Carlo R.

Abstract

This paper addresses navigation by composite heterogeneous robots in a decentralized system when policy reasoning and local control operate at different update levels. In a NetLogo--Python implementation, three robots share motion dynamics but use different LLM backends. Each robot independently combines a large language model (LLM) policy agent, an Upper Confidence Bound (UCB) bandit, and a Double Deep Q-Network (Double DQN) controller; no central LLM generates team actions. LLM inference is confined to round-level policy generation and refinement rather than tick-level action selection. The robots perform cross-LLM communication through a shared round summary containing policies, outcomes, and learning feedback. UCB performs refinement-mode selection, and the policy-conditioned Double DQN performs tick-level action selection from navigation variables, active policy parameters, and the LLM action prior. Each of the four configurations was evaluated over 30 rounds. In the fixed simulation, the complete configuration reached the goal in all 90 correlated robot--round records and achieved the lowest median completion time (42 ticks) and P90 (73.2 ticks); its median was 25.0--39.1\% lower than those of the other configurations. These observations provide descriptive, configuration-level evidence from the evaluated configurations.

Chinese Translation

本文探讨了在去中心化系统中，由复合异构机器人进行导航时，策略推理和局部控制在不同更新级别下的操作。在NetLogo--Python的实现中，三台机器人共享运动动态，但使用不同的LLM（大语言模型）后端。每台机器人独立结合了一个大语言模型（LLM）策略代理、一个上置信界（UCB）赌博机和一个双深度Q网络（Double DQN）控制器；没有中央LLM生成团队行动。LLM推理仅限于轮次级别的策略生成和优化，而不是时钟级别的动作选择。机器人通过包含策略、结果和学习反馈的共享轮次摘要进行跨LLM通信。UCB执行优化模式选择，而基于策略的Double DQN则从导航变量、活动策略参数和LLM动作先验中执行时钟级别的动作选择。对四种配置进行了30轮的评估。在固定仿真中，完整配置在所有90个相关的机器人-轮次记录中达到了目标，并实现了最低的中位完成时间（42个时钟周期）和P90（73.2个时钟周期）；其中位数比其他配置低25.0--39.1\%。这些观察为评估配置提供了描述性的配置级别证据。

View on arXiv Download PDF AI Translation

cs.RO / 20 / 2609.05178

LIBERO-RECOVER: Beyond Task Success Towards Failure Recovery in Robotic Manipulation Models

LIBERO-RECOVER：超越任务成功，迈向机器人操作模型中的失败恢复

Liu, Lin, Bao, Zhicheng, Zhang, Lu, Song, Ziying, Yang, Wu, Tao, Shuai, Liu, Wulong, Lu, Huchuan

Abstract

Vision-Language-Action (VLA) or World Action (WAM) models have recently demonstrated remarkable performance in robotic manipulation. On LIBERO, SOTA method have achieved nearly 100\% success rates, seemingly suggesting that the models are ready for deployment in real world. However, near perfect performance on existing benchmarks can be misleading: success under ideal conditions does not imply real world robustness. Existing benchmarks primarily evaluate task completion from predefined initial states, while real world interactions inevitably involve failures such as failed grasps, collisions, and unintended object movements. A robot must therefore not only execute tasks successfully, but also recognize and recover from failures to continue the task. Yet this capability remains largely unmeasured, revealing a critical gap between benchmark performance and real world reliability. To address this gap, we introduce LIBERO-Recover Benchmark, a large scale benchmark for failure recovery in robotic manipulation. Built upon LIBERO, we collect real execution failures from SOTA embodied models and construct 1,000+ scenarios across four recovery levels: (1) Action Retry, (2) Action Adaptation, (3) Object State Recovery, and (4) Environmental Recovery. We evaluate four core capabilities: spatial understanding, object structure reasoning, interaction understanding, and topological reasoning. As the first large-scale benchmark for embodied failure recovery, LIBERO-Recover shifts evaluation from \emph{Can the robot succeed?''} to \emph{Can the robot recover after failure?''}, promoting robust and generalizable embodied agents. The project will be avaible in \textcolor{blue}{https://liulin815.github.io/LIBERO-Recovery/}.

Chinese Translation

视觉-语言-动作（VLA）或世界动作（WAM）模型最近在机器人操作中展现了显著的性能。在LIBERO上，最先进的方法（SOTA）已实现近100%的成功率，这似乎表明这些模型已准备好在现实世界中部署。然而，在现有基准测试中接近完美的表现可能具有误导性：在理想条件下的成功并不意味着在现实世界中的鲁棒性。现有基准主要评估从预定义初始状态的任务完成情况，而现实世界的交互不可避免地涉及诸如抓取失败、碰撞和意外物体移动等失败。因此，机器人不仅必须成功执行任务，还必须识别并从失败中恢复，以继续任务。然而，这种能力在很大程度上仍未被测量，揭示了基准性能与现实世界可靠性之间的关键差距。为了解决这一差距，我们引入了LIBERO-Recover基准，这是一个针对机器人操作中失败恢复的大规模基准。基于LIBERO，我们收集了来自最先进具身模型的真实执行失败，并构建了1000多个场景，涵盖四个恢复级别：（1）动作重试，（2）动作适应，（3）物体状态恢复，以及（4）环境恢复。我们评估了四个核心能力：空间理解、物体结构推理、交互理解和拓扑推理。作为首个大规模的具身失败恢复基准，LIBERO-Recover将评估从“机器人能否成功？”转变为“机器人能否在失败后恢复？”，促进鲁棒且可推广的具身智能体。该项目将可在 https://liulin815.github.io/LIBERO-Recovery/ 上获取。

View on arXiv Download PDF AI Translation

cs.RO / 21 / 2609.05206

Morphology and actuation as inductive biases in robotic hand manipulation

形态与驱动作为机器人手部操作中的诱导偏置

Tari, Zalán, Birtalan, Eszter, Polcz, Péter, Koller, Miklós

Abstract

Robotic hands vary widely in anatomical fidelity and mechanical complexity, and these structural choices influence the coordination of joint motions and the difficulty of controlling the system. A unified framework is presented in which the kinematic and actuation stages are analysed separately and in composition, through the conditioning of the task Jacobian, the actuation matrix, and their product. It is applied to two hands representing opposing design philosophies, the Shadow Dexterous Hand and the Anatomically Correct, Biomechatronic Hand, along four morphological aspects: joint axis geometry, actuator-to-DOF ratio, coupling architecture, and authority distribution. All parameters are derived from the hands' canonical digital representations. Anatomical fidelity carries no uniform advantage: oblique axes improve thumb conditioning but leave the long fingers worse conditioned than the orthogonal-axis design, while the branching tendon network improves the effective control mapping at every long finger and worsens it significantly at the thumb, where actuator authority is concentrated on thumb opposition. Predictions derived from these metrics are evaluated against reinforcement learning experiments using PPO, DDPG+HER, and TQC+HER, across three different tasks.

Chinese Translation

机器人手在解剖学逼真度和机械复杂性上差异显著，这些结构选择影响关节运动的协调性以及系统控制的难度。本文提出了一个统一框架，在该框架中，运动学和驱动阶段通过任务雅可比矩阵、驱动矩阵及其乘积的条件化进行单独和组合分析。该框架应用于两种代表对立设计理念的手部，分别是Shadow Dexterous Hand和Anatomically Correct, Biomechatronic Hand，从四个形态学方面进行分析：关节轴几何、驱动器与自由度（DOF）比率、耦合架构和权威分配。所有参数均来源于手部的标准数字表示。解剖学逼真度并不具有统一的优势：斜轴改善了拇指的条件化，但使长指的条件化比正交轴设计更差，而分支肌腱网络在每个长指的有效控制映射中有所改善，但在拇指处显著恶化，因为驱动器的权威集中在拇指对抗上。基于这些指标的预测结果与使用PPO、DDPG+HER和TQC+HER的强化学习实验进行了评估，涵盖了三种不同的任务。

View on arXiv Download PDF AI Translation

cs.RO / 22 / 2609.05260

One Word, Different Action: A Real-Robot Benchmark for Language-Conditioned Embodied Reasoning

一个词，不同的动作：基于语言条件的具身推理真实机器人基准

Liu, Yiwei, Yang, Luwei, Lei, Shunbo

Abstract

Natural-language instruction changes can directly alter robot behavior. A reliable embodied system should preserve its action when the task is unchanged and update it correctly when the task itself changes. We introduce One Word, Different Action, a real-robot benchmark built on physical decision states and executable actions, using task-preserving and task-changing instruction pairs to jointly evaluate Decision Invariance and Decision Sensitivity, with further evaluation under multi-constraint reasoning and real-RGB grounding. Experiments show that modern models are near saturation on single-constraint instruction changes, yet several models degrade noticeably when multiple task constraints must be integrated into one executable decision. These results suggest that the more salient remaining challenge is no longer recognizing an isolated instruction change, but reliably composing multiple task requirements into a correct robot action decision.

Chinese Translation

自然语言指令的变化可以直接改变机器人的行为。一个可靠的具身系统应该在任务不变时保持其动作，并在任务本身变化时正确更新。我们介绍了‘一个词，不同的动作’（One Word, Different Action），这是一个基于物理决策状态和可执行动作的真实机器人基准，使用保持任务和改变任务的指令对共同评估决策不变性（Decision Invariance）和决策敏感性（Decision Sensitivity），并在多约束推理和真实RGB基础上进行进一步评估。实验表明，现代模型在单约束指令变化上接近饱和，但当多个任务约束必须整合成一个可执行决策时，多个模型明显退化。这些结果表明，当前更显著的挑战不再是识别孤立的指令变化，而是可靠地将多个任务要求组合成正确的机器人动作决策。

View on arXiv Download PDF AI Translation

cs.RO / 23 / 2609.05266

TacPAC: Tactile Prediction and Real-Time Action Correction in World-Action Models for Contact-Rich Manipulation

TacPAC：用于接触丰富操作的世界行动模型中的触觉预测与实时动作修正

Ma, Zipei, Wei, Xiaofei, Jiang, Junzhe, Lu, Shunlin, Zhang, Li

Abstract

World-action models guide action generation with predicted future observations, but vision-centric predictions miss the local contact cues that decide contact-rich manipulation. However, naively predicting future tactile observations as additional views recovers only a third of the achievable gain in our experiments. This gap reflects a timing mismatch: predictions precede execution, while tactile feedback arrives during it. We introduce TacPAC, which turns tactile prediction into real-time action correction. Once the base model has planned an action chunk, TacPAC caches the predicted contact that plan was conditioned on together with the plan's own representation, and a tactile expert reads each newly observed tactile image against that cache to correct the actions not yet executed. Feedback is thus interpreted against what the plan anticipated rather than in isolation, and one correction is a single pass over that cache, $20.7\times$ cheaper than regenerating the chunk. On five real-robot tasks spanning precision insertion, fragile-object handling, object reorientation, and long-horizon manipulation, TacPAC leads every task and raises the average from 22% for its vision-only base model to 64%. Code is available at https://github.com/LogosRoboticsGroup/TacPAC.

Chinese Translation

世界行动模型通过预测未来观察来指导动作生成，但以视觉为中心的预测忽视了决定接触丰富操作的局部接触线索。然而，简单地将未来触觉观察作为额外视图进行预测，仅在我们的实验中恢复了可实现增益的三分之一。这个差距反映了时间上的不匹配：预测在执行之前，而触觉反馈在执行过程中到达。我们引入了TacPAC，它将触觉预测转化为实时动作修正。一旦基础模型规划了一个动作块，TacPAC会缓存该计划所依赖的预测接触以及该计划自身的表示，触觉专家会根据该缓存读取每个新观察到的触觉图像，以修正尚未执行的动作。因此，反馈是相对于计划的预期进行解释，而不是孤立地进行修正，并且一次修正只需对该缓存进行一次遍历，成本比重新生成该块低$20.7 imes$。在涵盖精确插入、易碎物体处理、物体重新定向和长时间操作的五个真实机器人任务中，TacPAC在每个任务中都表现出色，并将其视觉仅基础模型的平均表现从22%提升至64%。代码可在 https://github.com/LogosRoboticsGroup/TacPAC 获取。

View on arXiv Download PDF AI Translation

cs.RO / 24 / 2609.05282

Temporal Tactile Encoding and Compliance for Intent-Aware Robot-to-Human Bimanual Handover

意图感知机器人与人类双手交接的时间触觉编码与顺应性

Marra, Pasquale, Berti, Stefano, Caddeo, Gabriele Mario, Natale, Lorenzo

Abstract

Reliable robot-to-human handover requires the robot to infer when the person is ready to receive the object, and release it safely, comfortably, and at the right time. This is challenging because visual observations alone may not disambiguate clear taking intent from accidental contact, weak grasping, wrong-direction forces, or transient interactions. In this work we treat human-robot handover as an intrinsically multimodal problem. Our approach couples a VLA model with a compliance controller that reduces interaction forces during object transfer. We finetune the VLA model with human demonstrations using RGB observation, temporally encoded tactile feedback and proprioception. We evaluate the complete system in a human-subject study against two baselines: one without tactile feedback and one using tactile feedback without compliance control. We hypothesize that combining compliance and temporal tactile encoding yields the most reliable and comfortable handovers, as compliance facilitates physical interaction while tactile history captures sustained taking intent. Performance is measured through objective metrics and an ad-hoc questionnaire. The results show that the two components provide complementary benefits and substantially outperform the baselines. Code and data will be released upon acceptance.

Chinese Translation

可靠的机器人与人类交接要求机器人能够推断出何时人准备好接收物体，并在安全、舒适和恰当的时机释放物体。这一过程具有挑战性，因为仅依靠视觉观察可能无法清晰地区分明确的接收意图与偶然接触、弱抓握、错误方向的力或瞬时交互。在本研究中，我们将人机交接视为一个内在的多模态问题。我们的方法将 VLA 模型与顺应性控制器相结合，以减少物体转移过程中的交互力。我们使用 RGB 观察、时间编码的触觉反馈和本体感觉对 VLA 模型进行微调，基于人类示范。我们在一项人类受试者研究中评估完整系统，并与两个基线进行比较：一个没有触觉反馈，另一个使用触觉反馈但没有顺应性控制。我们假设结合顺应性和时间触觉编码能够实现最可靠和舒适的交接，因为顺应性促进了物理交互，而触觉历史则捕捉了持续的接收意图。通过客观指标和临时问卷来衡量性能。结果表明，这两个组件提供了互补的好处，并显著优于基线。代码和数据将在接受后发布。

View on arXiv Download PDF AI Translation

cs.RO / 25 / 2609.05300

Human-Human & Human-Robot Interaction Transformer (H2INT) for Robot Navigation in Dense and Uncertain Crowds

人际与人机交互变换器（H2INT）在密集和不确定人群中的机器人导航

Shen, Ao, Chen, Kaixi, Liu, Shiwei, Deng, Fang, Chen, Chen

Abstract

Safe robot navigation in dense crowds requires reasoning about pedestrian motion and how it may change in response to a robot. However, many learning-based approaches generate pedestrian motion independently of the robot or assume uniform reciprocity, omitting an important source of interaction uncertainty. This paper presents a Human-Human & Human-Robot Interaction Transformer (H2INT), a reinforcement learning framework that retains robot-conditioned changes in pedestrian motion during policy learning while allowing responsiveness to vary across pedestrians. Responsiveness affects the crowd dynamics when the robot is visible but is not supplied as a policy input; the policy must instead infer its consequences from robot-centered relative positions. A two-stage gated Transformer progressively encodes human-human and human-robot relations, while a recurrent policy captures their temporal evolution. A curriculum gradually reduces pedestrian responsiveness to increase interaction difficulty. Simulation experiments demonstrate improved navigation safety and robustness over representative baselines across response conditions and crowd densities, and show transfer without retraining to structurally distinct crowd-flow layouts. Ablations support the hierarchical relational encoding and gated updates. Real-robot deployment further verifies that the learned policy can operate with sparse observations in a physical environment.

Chinese Translation

在密集人群中安全导航的机器人需要推理行人运动及其如何响应机器人的变化。然而，许多基于学习的方法独立于机器人生成行人运动，或假设均匀的互惠性，从而忽略了一个重要的交互不确定性来源。本文提出了一种人际与人机交互变换器（H2INT），这是一种强化学习框架，在策略学习过程中保留了行人运动在机器人的条件变化，同时允许不同的行人对机器人的响应性有所不同。响应性在机器人可见时影响人群动态，但并未作为策略输入提供；策略必须从以机器人为中心的相对位置推断其后果。一个两阶段的门控变换器逐步编码人际和人机关系，而一个递归策略则捕捉它们的时间演变。课程学习逐渐降低行人的响应性，以增加交互难度。仿真实验表明，在不同响应条件和人群密度下，相较于代表性基线，导航的安全性和鲁棒性得到了改善，并且在结构上不同的人群流布局中无需重新训练即可实现迁移。消融实验支持了层次关系编码和门控更新。真实机器人部署进一步验证了所学习的策略能够在物理环境中以稀疏观察进行操作。

View on arXiv Download PDF AI Translation

cs.RO / 26 / 2609.05324

RoboSPA: Can VLA Models Go Beyond Simple Scenes and Short-Horizon Tasks?

RoboSPA：VLA模型能否超越简单场景和短期任务？

Fan, Zhenxuan, Zhang, Bo, Lin, Yutong, Yuan, Yuqian, Lin, Juekai, Liang, Liang, Huang, Zhuoyi, Zhang, Wenqiao, Li, Juncheng, Tang, Siliang, Xiao, Jun, Zhuang, Yueting

Abstract

Vision-Language-Action (VLA) models have shown promising progress in language-conditioned robotic manipulation. However, existing datasets and benchmarks mainly evaluate task completion under predefined settings, offering limited insight into model reasoning under increasing spatial and procedural complexity. We introduce \textbf{RoboSPA} (\textbf{Robo}t \textbf{S}patial-\textbf{P}rocedural \textbf{A}ssessment), a large-scale robotic manipulation dataset and benchmark for diagnosing embodied reasoning in VLA models. \texttt{RoboSPA} focuses on two core dimensions, Fine-Grained Spatial Reasoning and Long-Horizon Procedural Planning, covering 10 task categories and 56 base tasks. Each task is instantiated across five difficulty levels, yielding 280 variants with increasing spatial ambiguity and procedural complexity. We collect 527K trajectories across multiple embodiments and diverse scenes. Beyond binary success rate, \texttt{RoboSPA} introduces diagnostic metrics for more detailed evaluation. Experiments on representative VLA models show that current systems still struggle with complex spatial relations, precise low-level execution, and memory-intensive planning. These results establish \texttt{RoboSPA} as a challenging diagnostic benchmark for developing more capable, reliable, and generalizable embodied agents. Our data and code are available at https://github.com/fanzhenxuan/RoboSPA.

Chinese Translation

视觉-语言-动作（VLA）模型在语言条件下的机器人操控方面显示出令人鼓舞的进展。然而，现有的数据集和基准主要在预定义设置下评估任务完成情况，提供的见解有限，无法深入了解模型在日益复杂的空间和程序环境下的推理能力。我们引入了 extbf{RoboSPA}（ extbf{Robo}t extbf{S}patial- extbf{P}rocedural extbf{A}ssessment），这是一个大规模的机器人操控数据集和基准，用于诊断VLA模型的具身推理能力。 exttt{RoboSPA}专注于两个核心维度：细粒度空间推理和长时间程序规划，涵盖10个任务类别和56个基础任务。每个任务在五个难度级别上进行实例化，产生280个变体，具有逐渐增加的空间模糊性和程序复杂性。我们收集了527K条轨迹，涵盖多个具身实例和多样化场景。除了二元成功率外， exttt{RoboSPA}还引入了诊断指标，以便进行更详细的评估。对代表性VLA模型的实验表明，当前系统在复杂空间关系、精确的低级执行和内存密集型规划方面仍然面临挑战。这些结果确立了 exttt{RoboSPA}作为一个具有挑战性的诊断基准，以促进更强大、可靠和可推广的具身智能体的发展。我们的数据和代码可在https://github.com/fanzhenxuan/RoboSPA获取。

View on arXiv Download PDF AI Translation

cs.RO / 27 / 2609.05325

FIRE-LIVWO: Robust LiDAR-Inertial-Visual-Wheel Odometry via Failure-Immune mmWave Radar Enhancement

FIRE-LIVWO：通过故障免疫毫米波雷达增强实现稳健的激光雷达-惯性-视觉-轮式里程计

Hu, Kun, Li, Menggang, Wu, Kaidi, Jin, Zhiwen, Zhao, Yingjie, Tang, Chaoquan, Hu, Eryi, Zhou, Gongbo

Abstract

Achieving robust SLAM in large-scale underground coal mines with complex structures and severe degeneracies remains highly challenging. Dense smoke and dust cause substantial loss of visual information and degrade LiDAR point-cloud features, while long, self-similar corridors induce geometric degeneration, leading to pronounced odometry drift. To address these issues, we propose FIRE-LIVWO: Failure-Immune mmWave Radar-Enhanced LiDAR-Inertial-Visual-Wheel Odometry, a tightly coupled multi-modal odometry framework based on an iterated error-state Kalman filter (IESKF). The framework fuses 4D mmWave radar, LiDAR, and visual features within a unified VoxelMap and jointly constructs LiDAR-radar point-to-plane residuals and sparse visual photometric residuals. In smoke-filled environments, we exploit the strong penetration of 4D mmWave radar and introduce pointwise Doppler velocity constraints to preserve state observability. In geometrically degenerate corridors, we tightly couple wheel odometry using non-holonomic constraints (NHC) and online lever-arm compensation to reduce drift. Our central contribution is a degeneration detection and adaptive fusion model switching strategy grounded in geometric and visual observability analysis, which quantifies observability online and dynamically adjusts modality weights. Real-world experiments in underground coal mines demonstrate that FIRE-LIVWO accurately identifies failure boundaries, enabling reliable modality switching under extreme conditions. Compared with baselines, it achieves superior accuracy and robustness (average localization error of 5.677m). We open source our code on Github to benefit the robotics community.

Chinese Translation

在复杂结构和严重退化的地下煤矿中实现稳健的SLAM仍然具有很高的挑战性。浓密的烟雾和灰尘导致视觉信息的显著丢失，并降低激光雷达点云特征，而长且自相似的走廊则引发几何退化，导致明显的里程计漂移。为了解决这些问题，我们提出了FIRE-LIVWO：故障免疫毫米波雷达增强的激光雷达-惯性-视觉-轮式里程计，这是一个基于迭代误差状态卡尔曼滤波器（IESKF）的紧耦合多模态里程计框架。该框架在统一的体素地图中融合4D毫米波雷达、激光雷达和视觉特征，并共同构建激光雷达-雷达点到平面的残差和稀疏视觉光度残差。在充满烟雾的环境中，我们利用4D毫米波雷达的强穿透能力，并引入逐点多普勒速度约束以保持状态可观测性。在几何退化的走廊中，我们通过非完整约束（NHC）和在线杠杆臂补偿紧密耦合轮式里程计，以减少漂移。我们的核心贡献是基于几何和视觉可观测性分析的退化检测与自适应融合模型切换策略，该策略在线量化可观测性并动态调整模态权重。在地下煤矿的实际实验中，FIRE-LIVWO准确识别故障边界，使得在极端条件下能够可靠地进行模态切换。与基线相比，它实现了更优的准确性和鲁棒性（平均定位误差为5.677米）。我们在Github上开源了我们的代码，以造福机器人社区。

View on arXiv Download PDF AI Translation

cs.RO / 28 / 2609.05331

Adaptation Needs in Robotic Systems: Assessing Behavior Trees and Their Enhancement

机器人系统中的适应需求：行为树及其增强的评估

Rostamnia, Mehran, Filippone, Gianluca, Caldas, Ricardo, Pelliccione, Patrizio

Abstract

Robotic systems increasingly operate in dynamic, uncertain, and open-ended environments, where design-time assumptions may no longer hold, and adaptation becomes necessary to maintain effective and safe operation. Behavior Trees (BTs) are widely used in robotic control architectures due to their modularity, readability, and reactivity. This raises a central question: are BTs sufficient to meet the adaptation needs of modern robotic systems? This paper investigates this question through a literature-driven study complemented by empirical validation. First, we derive a classification of robotic adaptation needs from the literature, organizing them into six categories: Knowledge, Perception, Actuation, System, Mission, and Environment. Then, we analyze the capabilities and limitations of classical BTs with respect to these needs. Then, we characterize BT-based approaches for adaptation from the existing literature and organize them into four primary families, i.e., generation, extension, evolution, and refinement, including approaches that combine multiple families. Our analysis shows that the modularity, flexibility, and reactivity of classical BTs are insufficient for adaptation needs involving runtime restructuring, reasoning under uncertainty, mission reinterpretation, learning, or integration with external knowledge and planning mechanisms. Enhanced BT approaches address several of these limitations, but to different extents and often with limitations of their own. Our findings relate adaptation needs to both the capabilities and limitations of classical and enhanced BTs, providing guidance on when classical BTs are sufficient, when enhanced mechanisms are needed, and which challenges remain or emerge for adaptive robotic control architectures.

Chinese Translation

机器人系统越来越多地在动态、不确定和开放的环境中运行，在这些环境中，设计时的假设可能不再成立，因此适应变得必要，以维持有效和安全的操作。行为树（Behavior Trees, BTs）因其模块化、可读性和反应性而广泛应用于机器人控制架构。这引出了一个核心问题：行为树是否足以满足现代机器人系统的适应需求？本文通过文献驱动的研究并辅以实证验证来探讨这一问题。首先，我们从文献中推导出机器人适应需求的分类，将其组织为六个类别：知识、感知、驱动、系统、任务和环境。然后，我们分析了经典行为树在满足这些需求方面的能力和局限性。接着，我们对现有文献中基于行为树的适应方法进行了特征化，并将其组织为四个主要类别，即生成、扩展、演化和细化，包括结合多个类别的方法。我们的分析表明，经典行为树的模块化、灵活性和反应性不足以满足涉及运行时重构、不确定性推理、任务重新解释、学习或与外部知识和规划机制集成的适应需求。增强型行为树方法解决了其中一些局限性，但程度不同，且通常自身也存在局限性。我们的研究结果将适应需求与经典和增强型行为树的能力和局限性联系起来，为何时使用经典行为树、何时需要增强机制以及适应性机器人控制架构中仍然存在或新出现的挑战提供了指导。

View on arXiv Download PDF AI Translation

cs.RO / 29 / 2609.05361

Development of a Humanoid Robot Prototype for Multimodal Human-Robot Interaction

多模态人机交互的人形机器人原型开发

Viet, Thang Tran, Canh, Thanh Nguyen, Gia, Huy Uong, Van, Phuc Dinh, Duc, Son Tran, Do, Ngoc Minh, HoangVan, Xiem

Abstract

Human-robot interaction (HRI) enables intuitive and intelligent collaboration between humans and robots in real-world environments. This paper introduces a humanoid robot prototype designed as a flexible testbed for developing and integrating artificial intelligence (AI) modules in HRI tasks. The system features a 12 degree-of-freedom (DOFs) dual-arm mechanism and a 2 DOFs head with an expressive LCD screen to express facial emotions. All hardware components are controlled by a custom-designed controller board with real-time AI processing supported by an onboard Jetson module. The system incorporates three AI modules: (1) gesture recognition using MediaPipe Pose and an LSTM classifier, (2) object detection with YOLO and 3D localization, and (3) voice-command processing through speech recognition and large language model(LLM)-based semantic parsing. The platform is validated through experiments on positioning accuracy, with results showing average manipulation errors of approximately 1.83 cm. To demonstrate its versatility, experimental results show over 90% task accuracy, with gesture recognition reaching 96%, speech recognition reaching 92%. The results confirm the effectiveness of the proposed system as a reproducible and accessible humanoid platform for research and prototyping in HRI.

Chinese Translation

人机交互（HRI）使人类与机器人在现实环境中能够进行直观和智能的协作。本文介绍了一种人形机器人原型，旨在作为一个灵活的测试平台，用于开发和集成人工智能（AI）模块以执行HRI任务。该系统具有12个自由度（DOFs）的双臂机制和一个具有表现力的LCD屏幕的2个自由度的头部，用于表达面部情感。所有硬件组件由一个定制设计的控制板控制，配备了支持实时AI处理的嵌入式Jetson模块。该系统集成了三个AI模块：（1）使用MediaPipe Pose和LSTM分类器的手势识别，（2）使用YOLO进行物体检测和3D定位，以及（3）通过语音识别和基于大型语言模型（LLM）的语义解析进行语音命令处理。通过对定位精度的实验验证该平台，结果显示平均操作误差约为1.83厘米。为了展示其多功能性，实验结果显示任务准确率超过90%，手势识别达到96%，语音识别达到92%。结果确认了所提出系统作为一个可重复和可访问的人形平台在HRI研究和原型开发中的有效性。

View on arXiv Download PDF AI Translation

cs.RO / 30 / 2609.05369

Towards Neuro-Symbolic Procedural Reasoning for Long-Horizon Vision-Language-Action Manipulation

面向长时间视语言行动操控的神经符号程序推理

Chavan, Vivek, Shi, Yahuan, Heimann, Oliver, Haninger, Kevin, Krüger, Jörg

Abstract

Vision-language-action (VLA) models can execute short manipulation skills, but remain brittle in long-horizon procedures requiring persistent task state, dependency-aware reasoning, conditional decisions, and reliable grounding. We investigate a neuro-symbolic framework that combines learned VLA control with explicit task graphs and multimodal procedural memory. Task graphs encode action dependencies, valid transitions, and branch conditions, while memory maintains the active step, completed actions, textual context, and task-relevant visual evidence. Together, these structures guide object selection, destination grounding, subgoal dispatch, and verification of expected state transitions. Human demonstrations provide additional spatial and temporal guidance through gaze or saliency cues. To isolate their effect on policy learning, our initial study bypasses cross-view gaze transfer and directly annotates pseudo-gaze in robot-view teleoperation videos. The resulting guidance is used during VLA fine-tuning and inference. We study two long-horizon manipulation domains, workspace clearing and surgical-instrument handling, which require ordered execution, visually grounded decisions, and conditional branching. We evaluate correct-object and destination selection, subtask completion, task progress, step-order consistency, complete-task success, and procedural or execution mistakes. This work positions structured symbolic reasoning and demonstration-derived visual guidance as complementary mechanisms for reliable long-horizon VLA manipulation.

Chinese Translation

视语言行动（VLA）模型能够执行短期操控技能，但在需要持久任务状态、依赖感知推理、条件决策和可靠基础的长时间程序中仍然脆弱。我们研究了一种神经符号框架，该框架将学习到的VLA控制与显式任务图和多模态程序记忆相结合。任务图编码了动作依赖关系、有效转移和分支条件，而记忆则维护当前步骤、已完成的动作、文本上下文和与任务相关的视觉证据。这些结构共同指导对象选择、目标定位、子目标分派以及预期状态转移的验证。人类示范通过注视或显著性线索提供额外的空间和时间指导。为了隔离其对策略学习的影响，我们的初步研究绕过了跨视角注视转移，直接在机器人视角的遥操作视频中标注伪注视。生成的指导在VLA微调和推理过程中使用。我们研究了两个长时间操控领域：工作空间清理和外科仪器处理，这些领域需要有序执行、视觉基础决策和条件分支。我们评估了正确对象和目标选择、子任务完成、任务进展、步骤顺序一致性、完整任务成功率以及程序或执行错误。本研究将结构化符号推理和基于示范的视觉指导定位为可靠的长时间VLA操控的互补机制。

View on arXiv Download PDF AI Translation

cs.RO / 31 / 2609.05376

What Matters, When? Diagnosing and Improving Conditional Visual Grounding in Visuomotor Imitation Policies

什么重要，何时重要？诊断和改善视觉运动模仿策略中的条件视觉定位

Chavan, Vivek, Xie, Pengtao, Shi, Yahuan, Heimann, Oliver, Haninger, Kevin, Krüger, Jörg

Abstract

Visuomotor imitation policies can achieve high performance under in-distribution visual conditions yet fail when visually similar objects or receptacles are introduced. We study this behavior as a problem of conditional visual grounding: the visual target required for successful control changes with the manipulation phase and, in more complex tasks, with the observed task state. Using Action Chunking with Transformers (ACT), we systematically introduce distractor objects and receptacles with controlled color and shape similarity and localize failures to picking and placement. We find that distractor sensitivity is specific to both the type of visual similarity and the manipulation stage. Guided by this diagnosis, we evaluate distractor augmentation, phase-dependent attention regularization, and appearance-based visual prompting as complementary interventions for improving target selection while preserving spatial information required for control. These interventions substantially improve robustness in simulation and on a physical UR3e. We further examine the same failure pattern in a pretrained vision-language-action policy on a state-conditioned instrument-handling task, where the observed state of a medical instrument determines the correct destination. Together, the results show that visual distractors can cause incorrect object or destination selection even when the underlying manipulation skill remains intact, and that explicitly improving target selection can substantially recover performance across distinct visuomotor policy-learning regimes.

Chinese Translation

视觉运动模仿策略在分布内视觉条件下可以达到高性能，但在引入视觉相似的物体或容器时却会失败。我们将这种行为研究为条件视觉定位的问题：成功控制所需的视觉目标随着操作阶段的变化而变化，并且在更复杂的任务中，随着观察到的任务状态而变化。通过使用带有变换器的动作分块（Action Chunking with Transformers, ACT），我们系统地引入具有控制的颜色和形状相似性的干扰物体和容器，并将失败定位于抓取和放置。我们发现，干扰敏感性特定于视觉相似性的类型和操作阶段。在这一诊断的指导下，我们评估了干扰物增强、阶段依赖的注意力正则化和基于外观的视觉提示作为改善目标选择的互补干预措施，同时保留控制所需的空间信息。这些干预措施在仿真和物理UR3e上显著提高了鲁棒性。我们进一步在一个预训练的视觉-语言-动作策略中检查了相同的失败模式，该策略在一个状态条件的工具处理任务中，其中观察到的医疗工具状态决定了正确的目的地。综合来看，结果表明，视觉干扰物可以导致错误的物体或目的地选择，即使基础的操作技能保持完好，并且明确改善目标选择可以在不同的视觉运动策略学习模式中显著恢复性能。

View on arXiv Download PDF AI Translation

cs.RO / 32 / 2609.05401

Same Trajectory, Contradictory Rewards (ROBORMBENCH): Paraphrase Fragility in Vision Language Reward Models

相同轨迹，矛盾奖励（ROBORMBENCH）：视觉语言奖励模型中的释义脆弱性

Jeung, Wonje, Yoon, Sangyeon, Hong, Hyesoo, Cho, Yoonjun, Jeon, Dongjae, Kim, Bumjun, Oh, Jean, Yu, Youngjae, No, Albert

Abstract

Vision-language models are increasingly used as reward functions for robotic learning, but this role requires paraphrase invariance: the same trajectory should receive the same reward under semantically equivalent goal descriptions. We show that current VLM reward models often violate this property. Paraphrasing the instruction alone can substantially change predicted progress scores, and can even flip identical robot behavior between failure and success. To measure this failure mode, we introduce ROBORMBENCH, a benchmark with 2,390 real-robot trajectories, ground-truth progress labels, and 21,673 verified paraphrases spanning lexical, syntactic, and action-goal rewrites. Across proprietary and open-source VLMs, paraphrase-induced instability is widespread and severe, grows under more divergent rewrites, and is not reliably reduced by scale or explicit reasoning. Dedicated reward models trained with trajectory-grounded supervision are substantially more stable. These results show that paraphrase robustness is a core requirement for reliable VLM-based reward modeling in robotics.

Chinese Translation

视觉语言模型越来越多地被用作机器人学习的奖励函数，但这一角色要求释义不变性：在语义上等价的目标描述下，相同的轨迹应获得相同的奖励。我们展示了当前的视觉语言模型（VLM）奖励模型常常违反这一属性。仅仅对指令进行释义就可以显著改变预测的进展评分，甚至可以将相同的机器人行为在失败和成功之间翻转。为了衡量这种失败模式，我们引入了ROBORMBENCH，这是一个包含2390个真实机器人轨迹、真实进展标签和21673个经过验证的释义（涵盖词汇、句法和动作-目标重写）的基准。在专有和开源的VLM中，释义引起的不稳定性普遍且严重，并且在更具差异性的重写下会加剧，而通过规模或明确推理并不能可靠地减少这种不稳定性。经过轨迹基础监督训练的专用奖励模型则显著更稳定。这些结果表明，释义鲁棒性是基于VLM的机器人奖励建模的核心要求。

View on arXiv Download PDF AI Translation

计算机视觉 (Computer Vision)

cs.CV / 1 / 2609.04276

FailSAE: Towards Interpretable Failure Prediction for Vision-Language Models via Sparse Autoencoders

FailSAE：通过稀疏自编码器实现可解释的视觉语言模型故障预测

Ma, Jie, Liu, Zongxi, Zhu, Yi

Abstract

Vision-language models (VLMs), such as CLIP, have achieved strong performance across multimodal tasks by aligning visual and textual representations in a shared embedding space. As VLMs are increasingly used for high-stakes domains, failure prediction becomes critical for risk-aware deployment and human intervention. Existing failure prediction methods typically rely on confidence scores or auxiliary classifiers. Although these methods are effective on predicting VLM failures, they provide limited interpretability. In this work, we investigate the use of Sparse Autoencoders (SAEs) for interpretable failure prediction in VLMs. We formulate failure prediction as a classification task over sparse SAE latent activations and introduce a three-stage failure-aware training pipeline that encourages the learned latent directions to remain interpretable while becoming more informative for failure prediction. Our experiments show that the resulting framework outperforms the evaluated baselines in failure prediction. Further analysis suggests that failure-aware training encourages SAE latent directions to capture more class-specific concepts. We also use the SAE to provide a concept-level analysis of how model representations change during failures, revealing a shift from class-specific concepts toward more ambiguous or style-related concepts. Finally, we explore how the learned SAE latent directions can support runtime failure recovery.

Chinese Translation

视觉语言模型（VLMs），如CLIP，在多模态任务中通过在共享嵌入空间中对齐视觉和文本表示，取得了强大的性能。随着VLMs在高风险领域的日益应用，故障预测对于风险意识部署和人类干预变得至关重要。现有的故障预测方法通常依赖于置信度评分或辅助分类器。尽管这些方法在预测VLM故障方面有效，但它们提供的可解释性有限。在本研究中，我们探讨了使用稀疏自编码器（SAEs）进行VLM可解释故障预测的方法。我们将故障预测形式化为对稀疏SAE潜在激活的分类任务，并引入一个三阶段的故障感知训练流程，鼓励学习到的潜在方向在变得更具信息性的同时保持可解释性。我们的实验表明，所提出的框架在故障预测方面优于评估的基线。此外，进一步分析表明，故障感知训练促使SAE潜在方向捕捉更多类特定概念。我们还利用SAE提供了一个概念级分析，揭示模型表示在故障期间如何变化，显示出从类特定概念向更模糊或与风格相关的概念的转变。最后，我们探讨了学习到的SAE潜在方向如何支持运行时故障恢复。

View on arXiv Download PDF AI Translation

cs.CV / 2 / 2609.04281

When Seeing Overrides Knowing: Visual Dominance and Deferral-Based Method for Personalized Safety in VLMs

视觉优先于知识：基于延迟的方法用于视觉语言模型中的个性化安全

Sun, Edward, Wu, Yuchen, Ma, Zixian, Jiang, Eric Hanchen, Xiao, Yijia, Yi, Xiaoyuan, Krishna, Ranjay, Wang, Wei, Wang, Jindong, Caliskan, Aylin

Abstract

Vision-language models (VLMs) are increasingly deployed in high-stakes settings, where a response that is reasonable in general may still be unsafe for a particular user whose medical, emotional, or situational context is unknown to the model. We study this problem of personalized safety in multimodal systems and introduce MPS-Bench, a benchmark of 5,181 scenarios from 584 real-world images across 12 high-risk domains, each paired with a hidden user profile. Evaluating eight frontier VLMs, we find that they almost always respond directly (86-99%) rather than seek missing context, and none exceeds 2.6/5 on personalized safety. To understand why these failures arise, we analyze multimodal interactions and identify visual dominance: visual information enters text representations early and suppresses textual risk signals during multimodal fusion. Causal interventions reveal a two-stage mechanism in which visual affect is first transferred into the text stream in early layers and then shapes the final decision through this altered text representation, making late-stage internal remediation unreliable. Motivated by this mechanism, we propose PRISM, a lightweight input monitor that uses bidirectional cross-modal modulation to predict when a query is likely to require deferral. PRISM achieves 0.978 AUC and strictly dominates the safety-utility Pareto frontier across all tested models.

Chinese Translation

视觉语言模型（VLMs）越来越多地应用于高风险环境中，在这些环境中，通常合理的响应可能对特定用户而言仍然是不安全的，因为模型对用户的医疗、情感或情境背景并不知情。我们研究了多模态系统中个性化安全的问题，并引入了MPS-Bench，这是一个包含来自584张真实世界图像的5,181个场景的基准，涵盖12个高风险领域，每个场景都配有一个隐藏的用户档案。对八个前沿VLM的评估发现，它们几乎总是直接响应（86-99%），而不是寻求缺失的上下文，并且没有一个模型在个性化安全性上超过2.6/5。为了理解这些失败的原因，我们分析了多模态交互并识别出视觉优先性：视觉信息在早期阶段进入文本表示，并在多模态融合过程中抑制文本风险信号。因果干预揭示了一种两阶段机制，其中视觉影响首先在早期层中转移到文本流中，然后通过这种改变的文本表示塑造最终决策，使得后期内部补救变得不可靠。基于这一机制，我们提出了PRISM，这是一种轻量级输入监控器，利用双向跨模态调制来预测查询何时可能需要延迟。PRISM在所有测试模型中实现了0.978的AUC，并严格主导了安全性-效用的帕累托前沿。

View on arXiv Download PDF AI Translation

cs.CV / 3 / 2609.04282

Step Back to Move Forward: Reflection-Aware Preference Optimization for Visual Generation

后退以前进：面向反思的视觉生成偏好优化

Wu, Junlong, Lin, Jiuzhou, Sun, Jia, Zhang, Boheng, Wang, Huaiqing, Fan, Dewen, Liu, Houde, Gan, Qianqian, Yang, Fan, Gao, Tingting

Abstract

Diffusion models have become the mainstream paradigm for modern visual generation and have substantially advanced multimedia content synthesis, especially in text-to-image and text-to-video tasks. To further align such generative models with human preferences, reinforcement learning (RL) has recently shown strong potential as a post-training strategy. Nevertheless, existing policy gradient-based methods often explore inefficiently, making them vulnerable to local optima that may degrade semantic faithfulness and visual realism. To address these challenges, we present Reflection-Aware GRPO (RA-GRPO), a new RL-based preference alignment framework for diffusion generative models. The core idea is to improve "forward" generation by incorporating "backward" reflection during optimization. We first introduce Diffusion Reflection, which rectifies intermediate sampling trajectories by inverting the diffusion process with a weak estimator, guiding latent states toward higher-probability regions of the true data manifold. Furthermore, we introduce Counterfactual Path Synthesis to implicitly distill these rectified trajectories into the policy, enabling the model to internalize the benefits of search-based exploration without incurring inference-time overhead. Extensive experiments on T2I and T2V models demonstrate that RA-GRPO significantly outperforms existing methods, particularly in mitigating reward hacking and improving generalization. The method remains architecture-agnostic and integrates seamlessly with standard pipelines, suggesting a promising direction for stable preference alignment.

Chinese Translation

扩散模型已成为现代视觉生成的主流范式，并在多媒体内容合成方面取得了显著进展，尤其是在文本到图像（text-to-image）和文本到视频（text-to-video）任务中。为了进一步使这些生成模型与人类偏好对齐，强化学习（Reinforcement Learning, RL）最近作为一种后训练策略显示出强大的潜力。然而，现有的基于策略梯度的方法往往探索效率低下，使其容易陷入局部最优，从而可能降低语义忠实性和视觉真实感。为了解决这些挑战，我们提出了反思感知的广义强化偏好优化（Reflection-Aware GRPO, RA-GRPO），这是一个基于RL的偏好对齐框架，适用于扩散生成模型。其核心思想是在优化过程中通过引入“向后”反思来改善“向前”生成。我们首先介绍了扩散反思（Diffusion Reflection），该方法通过使用弱估计器反转扩散过程来修正中间采样轨迹，引导潜在状态朝向真实数据流形的高概率区域。此外，我们引入了反事实路径合成（Counterfactual Path Synthesis），以隐式地将这些修正的轨迹提炼到策略中，使模型能够内化基于搜索的探索的好处，而不增加推理时间的开销。在文本到图像和文本到视频模型上的大量实验表明，RA-GRPO显著优于现有方法，特别是在减轻奖励黑客行为和改善泛化能力方面。该方法保持架构无关性，并与标准流程无缝集成，显示出稳定偏好对齐的有希望方向。

View on arXiv Download PDF AI Translation

cs.CV / 4 / 2609.04283

Joint Alignment and Distillation for Video Generation via Sample-Guided Distribution Matching

通过样本引导分布匹配实现视频生成的联合对齐与蒸馏

Lin, Jiuzhou, Wu, Junlong, Zuo, Fei, Ouyang, Huan, Fan, Dewen, Zhang, Boheng, Wang, Huaiqing, Sun, Jia, Yang, Fan, Liu, Houde, Chen, Kehai, Zhang, Min, Gao, Tingting, Li, Han

Abstract

Aligning video generative models to human preferences heavily relies on Reinforcement Learning (RL), which suffers from extensive computational overhead. Existing workflows typically treat RL and distillation as disconnected stages: applying RL before distillation incurs prohibitive computational costs, whereas applying RL after distillation frequently leads to model collapse. To overcome these limitations, we propose a unified, single-stage optimization framework grounded in Distribution Matching (DM). In the standard DM framework, distillation updates the model via a gradient direction that minimizes the gap between the real and fake models, guiding generations toward clarity and high fidelity. Building upon this, we introduce DM-Align, which derives a complementary gradient direction to guide the model toward human-preferred samples. Inspired by DPO and GRPO, our method leverages the distributional gap -- formulated from either preference pairs or intra-group exploration -- to directly construct this preference-guided gradient. By synergizing these two gradient directions, our approach eliminates the need for multi-step reward evaluation and complex ODE-SDE conversions inherent in traditional RL. Comprehensive experiments across multiple foundational video models demonstrate that this sample-guided framework robustly enhances both distillation quality and preference alignment, consistently outperforming both standalone variants and sequential two-stage pipelines.

Chinese Translation

将视频生成模型与人类偏好对齐在很大程度上依赖于强化学习（Reinforcement Learning, RL），但这会导致巨大的计算开销。现有的工作流程通常将RL和蒸馏视为不相干的阶段：在蒸馏之前应用RL会产生高昂的计算成本，而在蒸馏之后应用RL则常常导致模型崩溃。为了克服这些局限性，我们提出了一种基于分布匹配（Distribution Matching, DM）的统一单阶段优化框架。在标准的DM框架中，蒸馏通过一个梯度方向更新模型，该方向最小化真实模型与虚假模型之间的差距，从而引导生成结果朝向清晰和高保真度。基于此，我们引入了DM-Align，它推导出一个互补的梯度方向，以引导模型朝向人类偏好的样本。受到DPO和GRPO的启发，我们的方法利用分布差距——可以通过偏好对或组内探索来构建——直接构造这种偏好引导的梯度。通过协同这两个梯度方向，我们的方法消除了传统RL中多步骤奖励评估和复杂的常微分方程-随机微分方程（ODE-SDE）转换的需求。在多个基础视频模型上的全面实验表明，这种样本引导框架稳健地提升了蒸馏质量和偏好对齐，始终优于单独变体和顺序两阶段管道。

View on arXiv Download PDF AI Translation

cs.CV / 5 / 2609.04325

The microscope is the mask: privileged views and labels from a cryo-ET forward model

显微镜是面具：来自冷冻电子断层成像前向模型的特权视图和标签

Toader, Bogdan, Jamali, Kiarash, Bharat, Tanmay A. M., Scheres, Sjors H. W.

Abstract

We explore the use of simulated data for training a model for protein annotation in crowded cryo-electron tomography volumes reconstructed from images collected at limited tilt angles and severely corrupted by the measurement operator. Firstly, we leverage the corruptions imposed by the forward model to generate domain-specific augmented paired views of the exact same scene for an invariance objective integrated into the LeJEPA self-supervised training framework. Secondly, we use additional information from the simulation pipeline such as the positions and identity of proteins in the simulated volumes to inform the architecture of the model and the loss function, so that semantic information is localised at protein positions in the resulting dense feature volume. The resulting model, CARNIVAL, is evaluated without finetuning on classification and detection tasks in real tomograms, using a benchmark dataset containing multiple protein types and two tomogram processing types. We show that CARNIVAL outperforms a state-of-the-art model trained using a contrastive objective on simulated data but without forward model-based paired views or privileged information.

Chinese Translation

我们探讨了使用模拟数据来训练蛋白质注释模型在拥挤的冷冻电子断层成像体积中，这些体积是从在有限倾斜角度下收集的图像重建而成，并受到测量算子的严重干扰。首先，我们利用前向模型施加的干扰生成特定领域的增强配对视图，以实现集成在LeJEPA自监督训练框架中的不变性目标。其次，我们使用来自模拟管道的额外信息，例如模拟体积中蛋白质的位置和身份，以指导模型的架构和损失函数，从而使语义信息在生成的密集特征体积中的蛋白质位置上得以定位。最终模型CARNIVAL在真实断层图的分类和检测任务中进行了无微调评估，使用的基准数据集包含多种蛋白质类型和两种断层图处理类型。我们展示了CARNIVAL在使用对比目标训练的最先进模型上表现优越，后者使用了模拟数据但没有基于前向模型的配对视图或特权信息。

View on arXiv Download PDF AI Translation

cs.CV / 6 / 2609.04348

Object Concepts Emerge from Motion

运动中物体概念的出现

Li, Boshi, Wang, Xiaohui, Wu, Xiaoyang, Li, Zhichao, Yang, Ya, Wang, Naiyan

Abstract

Object-centric visual representations are important for physical-world perception, but existing visual pretraining methods often capture semantic categories without preserving the identity and coherence of individual instances. We present a biologically inspired framework that learns object-centric representations for single images from raw videos. Our approach uses motion boundaries as a source of object-level grouping: off-the-shelf optical flow and clustering produce pseudo-instance masks, which supervise a single-image encoder with pixel-level pairwise metric learning. The framework requires neither human annotations nor camera calibration. We first obtain 195 million pseudo-labeled frames from 7,163 hours of driving and web videos, then expand the supervision to 421 million frames with Motion-Verified Self-Training, which combines model proposals with motion evidence. We train encoders up to Swin-H and distill the learned representations into a family of Swin backbones. Across monocular depth estimation, 3D object detection, 3D occupancy prediction, and end-to-end planning, the resulting models achieve competitive or superior performance relative to supervised and self-supervised pretraining baselines, with particularly strong transfer on geometry- and instance-sensitive tasks. These results show that motion-derived supervision can teach static image encoders to represent visual instances, providing a complementary direction for scalable visual pretraining.

Chinese Translation

以物体为中心的视觉表征对于物理世界的感知至关重要，但现有的视觉预训练方法往往捕捉语义类别，而未能保留单个实例的身份和一致性。我们提出了一种生物启发的框架，该框架从原始视频中学习单幅图像的以物体为中心的表征。我们的方法利用运动边界作为物体级分组的来源：现成的光流和聚类生成伪实例掩码，这些掩码用于监督单幅图像编码器的像素级成对度量学习。该框架不需要人工标注或相机校准。我们首先从7163小时的驾驶和网络视频中获得1.95亿个伪标记帧，然后通过运动验证自我训练（Motion-Verified Self-Training）将监督扩展到4.21亿帧，该方法结合了模型提议和运动证据。我们训练了高达Swin-H的编码器，并将学习到的表征提炼成一系列Swin主干网络。在单目深度估计、3D物体检测、3D占用预测和端到端规划等任务中，所得到的模型相较于监督和自监督预训练基线表现出竞争力或优越的性能，尤其在几何和实例敏感任务上具有强大的迁移能力。这些结果表明，基于运动的监督可以教会静态图像编码器表示视觉实例，为可扩展的视觉预训练提供了一种互补的方向。

View on arXiv Download PDF AI Translation

cs.CV / 7 / 2609.04369

AdaptVPR: Route-Aware Hard Positive Generation for Robust Visual Place Recognition

AdaptVPR：基于路线的困难正样本生成以增强视觉地点识别的鲁棒性

Chen, Shunpeng, Zhang, Jingyi, Wang, Changwei, Xu, Shengpeng, Song, Yukun, Pei, Xingtian, Lin, Jinzhou, Guo, Li, Xu, Shibiao

Abstract

Visual Place Recognition (VPR) localizes a query image by retrieving database images of the same or nearby place, yet its robustness is often degraded by domain shifts arising from illumination, weather, seasonal changes, and dynamic occlusions. One contributing factor is the limited appearance diversity of the same place in existing training data. To address this issue, we propose AdaptVPR, a route-aware generative augmentation framework that constructs same-place hard positives for robust VPR training. AdaptVPR first uses a vision language model to parse scene attributes and estimate editing feasibility, while a rule-based scheduler determines the generation route according to editability scores and risk constraints. The generation process is decomposed into three complementary routes: the Global Appearance Route introduces global scene changes in weather, illumination, and time of day; the Local Occlusion Route inserts plausible dynamic occluders; and the Dual Route combines both types of perturbations to produce more challenging appearance shifts. Each generated candidate is evaluated using a VPR-oriented verification scheme based on geometric consistency and appearance diversity, reducing the risk of structural drift while ensuring sufficient appearance variation. Global candidates are generated once and rejected if verification fails, while Local Occlusion and Dual candidates use verification feedback for limited prompt refinement and regeneration. Using this framework, we construct AdaptCities, containing 160K verified synthetic same-place hard positives. Experiments across multiple VPR baselines and vision foundation backbones show consistent gains on standard benchmarks and substantial improvements under challenging domain shifts, with R@1 gains of up to 9.2%. The source code and data resources are publicly available at https://github.com/chenshunpeng/AdaptVPR.

Chinese Translation

视觉地点识别（VPR）通过检索同一地点或附近地点的数据库图像来定位查询图像，但其鲁棒性常常受到照明、天气、季节变化和动态遮挡等领域转移的影响。造成这一问题的一个因素是现有训练数据中同一地点的外观多样性有限。为了解决这一问题，我们提出了AdaptVPR，一个基于路线的生成增强框架，旨在为鲁棒的VPR训练构建同一地点的困难正样本。AdaptVPR首先使用视觉语言模型解析场景属性并评估编辑可行性，同时基于可编辑性评分和风险约束，采用基于规则的调度器确定生成路线。生成过程被分解为三条互补路线：全球外观路线引入天气、照明和时间变化的全球场景变化；局部遮挡路线插入合理的动态遮挡物；双重路线结合这两种扰动，以产生更具挑战性的外观变化。每个生成的候选样本都使用基于几何一致性和外观多样性的VPR导向验证方案进行评估，从而降低结构漂移的风险，同时确保足够的外观变化。全球候选样本生成一次，如果验证失败则被拒绝，而局部遮挡和双重候选样本则利用验证反馈进行有限的提示优化和再生成。通过该框架，我们构建了AdaptCities，包含160K经过验证的合成同一地点困难正样本。在多个VPR基线和视觉基础模型上的实验显示，在标准基准测试中取得了一致的提升，并在具有挑战性的领域转移下取得了显著改善，R@1的提升高达9.2%。源代码和数据资源可在https://github.com/chenshunpeng/AdaptVPR上公开获取。

View on arXiv Download PDF AI Translation

cs.CV / 8 / 2609.04381

Where Appearance Fails, Geometry Recognizes: A CAD-Free 3D Shape Prior That Complements Vision Foundation Models

外观失效时，几何识别：一种无需计算机辅助设计（CAD）的三维形状先验，补充视觉基础模型

Tao, Chenxi, Choi, Seung-Kyum

Abstract

Recognizing specific objects onboarded without a labeled training set recurs across manufacturing and service robotics, yet the conventional renderable prior, a computer-aided-design (CAD) model, is often unavailable. Two-dimensional capture supplies no shape prior, and frozen foundation features fail on geometrically similar, low-texture industrial parts. We ask what a short object-centric scan buys for recognition beyond the captured images themselves: each object is reconstructed with 3D Gaussian Splatting (3DGS), summarized into a per-class shape prototype, and fused with frozen DINOv2 image features. First, the scan recovers the recognition value of CAD without CAD: geometry from RGB-D depth (on T-LESS), 3DGS, and CAD gives comparable recognition (tied on HOPE, within 1.6 points on T-LESS); 3DGS is only a convenient route to a point cloud. Second, the payoff is governed by how recognizable the shape is: on shape-distinctive household objects (HOPE) geometry alone reaches 0.920 versus image-only 0.832, a ceiling below which fixed-weight fusion (0.872) sits. On shape-confusable textureless industrial parts (T-LESS) the gain is modest but consistent (0.560 to 0.591 fused, above both single signals). Third, the prior is complementary, not uniformly additive: it rescues far more image failures than it breaks successes, and its benefit grows under partial occlusion. Finally, the worth lies in geometry, not rendered pixels: 3DGS renderings do not help the image side, and frozen-feature recognition is nearly lighting-invariant (within 2.5 points). The study is scoped to recognition, not the BOP pose benchmark.

Chinese Translation

在制造和服务机器人领域，识别未标记训练集中的特定物体是一个反复出现的问题，然而传统的可渲染先验，即计算机辅助设计（CAD）模型，通常不可用。二维捕捉无法提供形状先验，而冻结的基础特征在几何相似、低纹理的工业部件上表现不佳。我们探讨短暂的物体中心扫描在识别上能带来什么超出捕获图像本身的价值：每个物体通过三维高斯点云（3D Gaussian Splatting, 3DGS）重建，汇总为每类形状原型，并与冻结的DINOv2图像特征融合。首先，扫描恢复了CAD的识别价值，而无需CAD：来自RGB-D深度（在T-LESS上）、3DGS和CAD的几何形状提供了可比的识别效果（在HOPE上持平，在T-LESS上相差1.6分）；3DGS仅是获取点云的便捷途径。其次，收益取决于形状的可识别性：在形状独特的家用物体（HOPE）上，仅凭几何形状的识别率达到0.920，而仅依赖图像的识别率为0.832，固定权重融合（0.872）则低于此上限。在形状易混淆的无纹理工业部件（T-LESS）上，增益虽小但稳定（融合后从0.560提升至0.591，均高于单一信号）。第三，先验是互补的，而非均匀叠加的：它挽救了更多的图像失败，而非打破成功，其益处在部分遮挡下更为显著。最后，价值在于几何，而非渲染像素：3DGS渲染对图像侧没有帮助，冻结特征的识别几乎不受光照影响（相差不超过2.5分）。本研究的范围限于识别，而非边界物体姿态基准（BOP pose benchmark）。

View on arXiv Download PDF AI Translation

cs.CV / 9 / 2609.04383

What Moves? Localized Motion Representations for Compositional Scene Control

什么在移动？用于组合场景控制的局部运动表示

Fundel, Frank, Alaya, Malek Ben, Ressler-Antal, Thomas, Baumann, Stefan Andreas, Ommer, Björn

Abstract

Real-world dynamics are inherently compositional: multiple entities move simultaneously within a shared scene, each exhibiting distinct motion patterns. Yet most existing video representations encode motion globally, without explicitly capturing localized motion for individual entities. Crucially, motion is defined relative to a global reference frame, including camera motion and scene layout. However, localized embeddings are often computed from cropped images or obtained by masking features after encoding, discarding the context needed to interpret motion. To address this, we introduce a promptable localized motion representation that produces persistent embeddings for user-specified regions defined by spatial masks. Rather than cropping the input or masking features, our model processes the full video and conditions motion encoding directly on the queried region. This yields temporally consistent, region-addressable embeddings that isolate local dynamics while retaining the global context required for disambiguation. We demonstrate object-level motion transfer, enabling controlled composition of dynamic scenes. Beyond generative control, our embeddings support localized action classification in multi-actor videos. Across both tasks, our approach improves controllability and outperforms global representations localized through cropping or post-hoc masking. Project Page: https://compvis.github.io/WhatMoves

Chinese Translation

现实世界的动态本质上是组合性的：多个实体在共享场景中同时移动，每个实体表现出不同的运动模式。然而，大多数现有的视频表示是全局编码运动，而没有明确捕捉单个实体的局部运动。关键是，运动是相对于全局参考框架定义的，包括相机运动和场景布局。然而，局部嵌入通常是从裁剪的图像计算得出的，或者通过在编码后对特征进行掩蔽来获得，这样就丢弃了理解运动所需的上下文。为了解决这个问题，我们引入了一种可提示的局部运动表示，该表示为用户指定的由空间掩蔽定义的区域生成持久的嵌入。我们的模型处理完整视频，而不是裁剪输入或掩蔽特征，并直接在查询区域上对运动编码进行条件处理。这产生了时间一致、区域可寻址的嵌入，能够隔离局部动态，同时保留了解释所需的全局上下文。我们展示了对象级运动转移，使动态场景的控制组合成为可能。除了生成控制外，我们的嵌入还支持多演员视频中的局部动作分类。在这两项任务中，我们的方法提高了可控性，并优于通过裁剪或事后掩蔽局部化的全局表示。项目页面：https://compvis.github.io/WhatMoves

View on arXiv Download PDF AI Translation

cs.CV / 10 / 2609.04392

FAVE: Foveated Adaptive Visual Encoding for Efficient Fine-Grained Visual Understanding

FAVE：用于高效细粒度视觉理解的注视适应视觉编码

Mukherjee, Amitangshu, Roy, Kaushik

Abstract

Fine-grained visual understanding depends on local detail, yet visual encoders face a trade-off between costly full-image high-resolution processing and compact global encoding that can weaken such evidence. Inspired by human active vision, we separate where to look from what to encode. We focus on the latter and introduce FAVE (Foveated Adaptive Visual Encoding), a lightweight variable-resolution ViT that encodes externally selected regions at high acuity while preserving native geometry. We first isolate this encoding problem using oracle ground-truth crops in a controlled small-object regime. On ImageNet objects with a native maximum side of 96 pixels, FAVE improves Top-1 by 9.4 points over a fixed-resolution ViT on the same crop window with 12.7 times lower FLOPs. Increasing global resolution or backbone capacity does not recover the same operating point. We then integrate FAVE as a complementary local branch in FastVLM. Its local tokens are combined with FastVLM's global visual tokens, while the original global pathway and language model remain frozen. With at most 16 additional local tokens, FAVE improves TextVQA by 1.60 points and achieves a 3.3 times controlled TTFT speedup over SmolVLM2-2.2B. On GQA attribute questions, it improves FastVLM-1.5B by 1.31 points, extending the benefit beyond text while narrowing the gap to FastVLM-7B. Together, these results show that selectively allocating high-acuity local capacity provides an efficient complement to broader global representations and model scaling for fine-grained understanding of small objects, text, and attributes.

Chinese Translation

细粒度视觉理解依赖于局部细节，但视觉编码器面临着昂贵的全图高分辨率处理与可能削弱此类证据的紧凑全局编码之间的权衡。受到人类主动视觉的启发，我们将注视位置与编码内容分开。我们专注于后者，提出了FAVE（Foveated Adaptive Visual Encoding），这是一种轻量级的可变分辨率视觉变换器（ViT），能够以高精度编码外部选择的区域，同时保持原生几何形状。我们首先在一个受控的小物体环境中，利用真实的基准裁剪来隔离这个编码问题。在具有最大边长为96像素的ImageNet对象上，FAVE在相同裁剪窗口下比固定分辨率的ViT提高了9.4个点，同时FLOPs降低了12.7倍。增加全局分辨率或主干网络容量并不能恢复相同的操作点。随后，我们将FAVE作为FastVLM的一个补充局部分支进行整合。其局部标记与FastVLM的全局视觉标记相结合，而原始的全局路径和语言模型保持不变。最多增加16个局部标记后，FAVE将TextVQA提高了1.60个点，并在SmolVLM2-2.2B上实现了3.3倍的控制TTFT加速。在GQA属性问题上，它将FastVLM-1.5B提高了1.31个点，扩展了文本之外的收益，同时缩小了与FastVLM-7B之间的差距。综合来看，这些结果表明，选择性地分配高精度局部容量为细粒度理解小物体、文本和属性提供了对更广泛全局表示和模型扩展的高效补充。

View on arXiv Download PDF AI Translation

cs.CV / 11 / 2609.04438

ICM-Bench: Person-Level Identity Reasoning in Multimodal Agents with Long-Term Memory

ICM-Bench：具有长期记忆的多模态智能体中的个体级身份推理

Ren, Shidu, Liu, Yunze, Liu, Xing, Wu, Chi-Hao, Zhou, Enmin, Shen, Junxiao

Abstract

Long-horizon multimodal agents should remember not only what happened but also who participated. This capability depends on linking recurring faces, voices, names, person-associated objects, events, and social relations to consistent identities over time. Existing long-video and multimodal-agent benchmarks measure broad memory question answering, but they do not isolate the ability to maintain recurring person identities and reason over their cross-time relations. We introduce ICM-Bench (Identity-Centric Memory Benchmark), which, to the best of our knowledge, is the first benchmark specifically designed to evaluate identity-centric reasoning over long video memories in multimodal agents. The benchmark contains 839 synthetic clips spanning 141 minutes and 1,217 open-ended questions about six recurring adults in a one-year life album. A theme-configurable pipeline generates the video collection and associates each question with its target identities and traceable supporting evidence. We compare direct caption-memory baselines, memory-augmented agents, and graph-retrieval systems. Gemini 3.1 Pro achieves the highest overall accuracy of 74.0%, yet its score falls to 60.3% on questions that require long-term identity profiles. The results show that current systems recover many event-level memories but remain less reliable when evidence must be accumulated around a stable person.

Chinese Translation

长时间跨度的多模态智能体不仅应记住发生了什么，还应记住谁参与了其中。这一能力依赖于将重复出现的面孔、声音、名字、与个人相关的物品、事件和社会关系链接到一致的身份上。现有的长视频和多模态智能体基准测试衡量广泛的记忆问答能力，但并未单独评估维持重复个体身份及推理其跨时间关系的能力。我们提出了ICM-Bench（以身份为中心的记忆基准），据我们所知，这是第一个专门设计用于评估多模态智能体在长视频记忆中进行身份中心推理的基准。该基准包含839个合成片段，时长141分钟，以及关于六位重复出现的成年人在一年生活相册中的1,217个开放式问题。一个主题可配置的管道生成视频集合，并将每个问题与其目标身份及可追溯的支持证据关联。我们比较了直接字幕记忆基线、增强记忆的智能体和图检索系统。Gemini 3.1 Pro在整体准确率上达到了74.0%的最高分，但在需要长期身份档案的问题上，其得分降至60.3%。结果表明，当前系统能够恢复许多事件级别的记忆，但在需要围绕稳定个体积累证据时，可靠性较低。

View on arXiv Download PDF AI Translation

cs.CV / 12 / 2609.04439

Segmentation of the aorta in 4D flow MRI using 4D convolutional kernels and learning from sparse annotations

基于4D卷积核和稀疏注释学习的4D流动MRI主动脉分割

Rahlfs, Hinrich, Garcia, Julio, Manini, Chiara, Hüllebrand, Markus, Schmitter, Sebastian, Nordmeyer, Sarah, Kühne, Titus, Stern, Heiko, Meierhofer, Christian, Harloff, Andreas, Kelle, Sebastian, Lenz, Alexander, Bannas, Peter, Schulz-Menger, Jeanette, Trauzeddel, Ralf F, Hennemuth, Anja

Abstract

Automated aortic segmentation in 4D flow MRI is essential for reproducible hemodynamic assessment but is limited by scarce dense annotations and high computational demands. We developed a fully automated 4D (3D+time) U-Net for segmenting the ascending aorta, arch, and proximal descending aorta, using a parameter-efficient hybrid 4D kernel to capture temporal context and sparse 4D labels derived from existing 2D expert contours and centerlines, thereby avoiding the need for dense 4D annotations. Training comprised 268 scans from 8 centers and 2 vendors, with evaluation on an internal test set (32 scans) and an external post-contrast set (30 scans; different site, protocol, and annotator), compared against frame-wise 3D networks and two semi-automatic references. Against time-resolved annotations, the 4D U-Net achieved Dice scores of 0.927 (internal) and 0.911 (external), versus 0.919/0.847 for the 3D U-Net, 0.893 for static PC-MRA, and 0.808 for registration-based propagation; differences were small in systole but pronounced in diastole. Agreement with expert contours for peak velocity, net flow, axial and circumferential wall shear stress, and diameters was excellent (ICC >=0.954 internal, >=0.980 external), while semi-automatic references performed worse. The method thus provides reproducible, time-resolved aortic segmentation for automated hemodynamic analysis and generalizes across multicenter, multivendor, and independent post-contrast data. The model is publicly available.

Chinese Translation

在4D流动MRI中，自动化的主动脉分割对于可重复的血流动力学评估至关重要，但受到稀缺的密集注释和高计算需求的限制。我们开发了一种完全自动化的4D（3D+时间）U-Net，用于分割升主动脉、主动脉弓和近端降主动脉，采用了一种参数高效的混合4D核以捕捉时间上下文，并利用从现有2D专家轮廓和中心线派生的稀疏4D标签，从而避免了对密集4D注释的需求。训练数据包括来自8个中心和2个供应商的268个扫描，评估在一个内部测试集（32个扫描）和一个外部后对比集（30个扫描；不同地点、协议和注释者）上进行，并与逐帧3D网络和两个半自动参考进行比较。与时间分辨注释相比，4D U-Net在内部和外部分别达到了0.927和0.911的Dice分数，而3D U-Net为0.919/0.847，静态PC-MRA为0.893，基于配准的传播为0.808；在收缩期的差异较小，但在舒张期则明显。与专家轮廓在峰值速度、净流量、轴向和周向壁剪切应力以及直径方面的符合度极佳（ICC >=0.954 内部，>=0.980 外部），而半自动参考的表现较差。因此，该方法为自动化血流动力学分析提供了可重复的、时间分辨的主动脉分割，并在多中心、多供应商和独立后对比数据中具有良好的泛化能力。该模型已公开发布。

View on arXiv Download PDF AI Translation

cs.CV / 13 / 2609.04454

Topology-Aware Training and Spatial Diagnostics for Fiber Bundle Segmentation in Tracer Histology

基于拓扑的训练和空间诊断用于追踪组织学中的纤维束分割

Avila, Joselyn Romero, Bintsi, Kyriaki-Margarita, Habte, Ermias, Lehman, Julia F., Haber, Suzanne N., Yendiki, Anastasia

Abstract

Anatomic tracer studies reveal how axon bundles project from an injection site, branch into smaller groups of axons, and course through the brain to reach their destinations. Histological data from such studies provide anatomical reference information for validating diffusion MRI tractography. However, manual annotation of the histological data is very labor-intensive, and although automated segmentation methods have been proposed, they rely mainly on pixel-overlap losses such as BCE and Dice; topology-aware loss functions have not been studied for this task. We compare BCE-Dice, clDice, Betti matching, and Topograph for fiber bundle segmentation in macaque tracer histology using a frozen DINOv3 backbone. To our knowledge, this is the first exploration of foundation-model features for this task. BCE-Dice achieved the highest Dice, while clDice achieved the highest bundle recall but poor mask overlap. Topograph had similar Dice to BCE-Dice, the lowest $\beta_0$ error, and fewer false positives than BCE-Dice and Betti matching. Fiber bundle segmentation methods are typically evaluated with a permissive rule that counts a bundle as detected given any overlap with the prediction. We show this rule does not capture oversegmentation, and that per-section TPR can be inflated by empty sections assigned perfect recall. To quantify this, we introduce Excess32, a spatial diagnostic measuring predicted pixels outside a 32-pixel tolerance band around annotated bundles. In validation, a Betti-Topograph union raises sparse-bundle TPR from 0.818 to 0.933, but worsens FDR from 0.296 to 0.509, Excess32 from 0.108 to 0.466, and area ratio from 0.94 to 3.34. These results show detection metrics alone are insufficient to characterize segmentation quality.

Chinese Translation

解剖追踪研究揭示了轴突束如何从注射点投射、分支成更小的轴突组，并穿过大脑到达其目的地。这类研究的组织学数据为验证扩散MRI轨迹成像提供了解剖参考信息。然而，手动注释组织学数据非常耗时，尽管已经提出了自动分割方法，但它们主要依赖于像BCE和Dice这样的像素重叠损失；针对该任务的拓扑感知损失函数尚未得到研究。我们比较了BCE-Dice、clDice、Betti匹配和Topograph在猕猴追踪组织学中的纤维束分割效果，使用了冻结的DINOv3骨干网络。据我们所知，这是首次探索基础模型特征在该任务中的应用。BCE-Dice达到了最高的Dice值，而clDice实现了最高的束召回率但掩膜重叠较差。Topograph的Dice值与BCE-Dice相似，$eta_0$错误率最低，且假阳性数量少于BCE-Dice和Betti匹配。纤维束分割方法通常使用一种宽松规则来评估，即只要与预测有任何重叠就算检测到一个束。我们展示了这一规则未能捕捉到过度分割，并且每个部分的真正阳性率（TPR）可能因被分配完美召回的空部分而被夸大。为量化这一点，我们引入了Excess32，一种空间诊断方法，用于测量标注束周围32像素容差带外的预测像素。在验证中，Betti-Topograph联合将稀疏束的TPR从0.818提高到0.933，但FDR从0.296恶化到0.509，Excess32从0.108增加到0.466，面积比从0.94增加到3.34。这些结果表明，仅依靠检测指标不足以表征分割质量。

View on arXiv Download PDF AI Translation

cs.CV / 14 / 2609.04550

VISTA: Dense Multi-Label Classroom Coding with Vision-Language Models

VISTA：基于视觉-语言模型的密集多标签课堂编码

Franck, Andrew, Ng, Brendan, Fitzgerald, Ben, Derrod, Zane, Cianci, Chris, Craney, Chris

Abstract

Video-language benchmarks are usually constructed by the dataset authors without published reliability statistics, leaving the noise floor of the construct unknown. We argue that multimodal benchmarking benefits from methods taken from research communities that have already invested in strategies to ensure reliability. We illustrate the case with the Classroom Observation Protocol for Undergraduate STEM (COPUS): a 24-code multi-label observation instrument with a decade of peer-reviewed reliability literature. We recast COPUS as a video benchmark for multimodal foundation models, where it provides a dense set of structured labels (a 24-dimensional binary vector every 2 minutes across a 50-90 minute lecture), an externally validated vocabulary, and established literature that provides a per-code reliability target based on human evaluators. Annotations in our evaluation corpus are produced by a 5-person human-evaluator panel whose consensus matrix is our reference. We propose VISTA, a baseline that runs MiniCPM-V-4.5 over a dense sliding window, refines its per-window outputs with a lightweight multi-layer perceptron (MLP) head trained on top of the frozen backbone, and max-pools the resulting predictions onto the 2-minute COPUS grid. On three held-out chemistry lectures, VISTA reaches 80.1% restricted macro accuracy versus 74.9% for the zero-shot variant, with the largest residual errors on visually similar instructor codes and on rare audio-dependent codes. We characterize three systematic failure modes (audio-partial observability, fine-grained group-work discrimination, long-tail recall) and release the benchmark tooling, prompts and baseline code at https://github.com/ajfranck/VISTA.

Chinese Translation

视频-语言基准通常由数据集作者构建，但未发布可靠性统计数据，导致构建的噪声水平未知。我们认为，多模态基准测试受益于那些已经在确保可靠性策略上投入研究的社区的方法。我们以本科STEM课堂观察协议（COPUS）为例：这是一个具有十年同行评审可靠性文献的24代码多标签观察工具。我们将COPUS重新构建为多模态基础模型的视频基准，其中提供了一组密集的结构化标签（在50-90分钟的讲座中每2分钟一个24维二进制向量）、一个经过外部验证的词汇，以及基于人类评估者提供的每个代码可靠性目标的已建立文献。我们评估语料库中的注释由一个5人评估小组生成，其共识矩阵是我们的参考。我们提出了VISTA，一个在密集滑动窗口上运行MiniCPM-V-4.5的基线，通过在冻结的主干上训练的轻量级多层感知器（MLP）头来细化每个窗口的输出，并将结果预测最大池化到2分钟的COPUS网格上。在三场保留的化学讲座中，VISTA达到了80.1%的限制宏观准确率，而零-shot变体为74.9%，最大残差错误出现在视觉上相似的讲师代码和稀有的音频依赖代码上。我们描述了三种系统性失败模式（音频部分可观察性、细粒度小组工作区分、长尾召回），并在https://github.com/ajfranck/VISTA发布基准工具、提示和基线代码。

View on arXiv Download PDF AI Translation

cs.CV / 15 / 2609.04555

DART: Depth-as-Target Pretraining for Surgical Vision Foundation Models

DART：用于外科视觉基础模型的深度作为目标预训练

Han, John J., Schmidt, Adam, Jamal, Muhammad Abdullah, Wu, Jie Ying, Mohareri, Omid

Abstract

Vision foundation models (VFMs) are valuable in data-scarce domains such as surgery, where a single pretrained backbone can provide rich representations for many downstream tasks. Yet the dominant self-supervised pretraining paradigm uses only RGB images, leaving readily available complementary signals, such as depth maps, unused. This is a particular missed opportunity in surgery, where natural-image VFMs transfer poorly while the scene geometry is rich and informative. With strong off-the-shelf models now able to produce pseudo-labeled dense depth for any image corpus, we hypothesize that such signals can be folded into pretraining to learn better representations. We present DART, an RGB-D pretraining recipe that builds on DINOv2 with a simple modification: a pixel-space depth reconstruction objective applied to masked iBOT patches, supervised by pseudo-labeled depth. Depth is used only during pretraining, so fine-tuning and inference remain RGB-only. We find that this pixel-level reconstruction head improves representation quality rather than disrupting it. We further show that depth, which encodes scene geometry, is more effective as a target than alternative dense signals such as Canny edges, confirming that the gains stem from depth rather than added supervision alone. Across eight surgical benchmarks spanning segmentation, depth estimation, and image-level recognition, DART outperforms both natural-image and in-domain baselines, including a vanilla DINOv2 trained on identical data, improving dense prediction while also strengthening image-level understanding. More broadly, DART shows that freely available geometric pseudo-labels can strengthen foundation model pretraining without extra labels or added inference cost, pointing toward stronger backbones for surgery.

Chinese Translation

视觉基础模型（VFMs）在数据稀缺的领域（如外科手术）中具有重要价值，其中单一的预训练骨干网络可以为许多下游任务提供丰富的表示。然而，主流的自监督预训练范式仅使用RGB图像，未能利用可用的补充信号，如深度图。这在外科手术中尤其是一个错失的机会，因为自然图像的VFMs转移效果较差，而场景几何信息丰富且具有信息量。随着强大的现成模型现在能够为任何图像数据集生成伪标记的稠密深度，我们假设这些信号可以被纳入预训练中，以学习更好的表示。我们提出了DART，这是一种基于DINOv2的RGB-D预训练方案，简单地修改为：对被遮挡的iBOT补丁应用像素空间深度重建目标，并由伪标记的深度进行监督。深度仅在预训练期间使用，因此微调和推理仍然保持RGB模式。我们发现，这种像素级重建头提高了表示质量，而不是破坏它。我们进一步表明，编码场景几何的深度作为目标比其他稠密信号（如Canny边缘）更有效，确认收益源于深度而不仅仅是增加的监督。在涵盖分割、深度估计和图像级识别的八个外科基准测试中，DART超越了自然图像和领域内基准，包括在相同数据上训练的普通DINOv2，改善了稠密预测，同时也增强了图像级理解。更广泛地说，DART表明，免费提供的几何伪标签可以在没有额外标签或增加推理成本的情况下增强基础模型的预训练，为外科手术提供更强的骨干网络指明了方向。

View on arXiv Download PDF AI Translation

cs.CV / 16 / 2609.04590

Dual-Part Multi-Lateral Branched Network for Multi-Class Segmentation in Cardiovascular Catheterization Angiograms

用于心血管导管造影多类分割的双部分多边分支网络

Omisore, Olatunji, Elazab, Ahmed, Shahidinejad, Ali, Sabrina, Fariza

Abstract

Catheterisation image processing requires segmentation models that are fast, accurate and explainable. While most of the existing studies usually focus on binary segmentation, there is a recent demand for simultaneous segmentation of multiple structures found in catheterization scenes. In this study, a dual-part MLBNet architecture is designed with multi-lateral encoder blocks and multi-head decoder branches for class-aware segmentation in cardiovascular catheterization scenes. Lateral branches in the encoder enables repeated feature extraction to learn diverse shared representations, while multiple decoder heads are used to introduce class-skewed branches that specialize in different structural properties in catheterization scenes. To analyze the performances of the dual-part MLBNet architecture, several multi-class segmentation angiogram data obtained during cardiovascular catheterization in phantom models, synthetic human-simulated aorta, and animal model are used for model training and evaluation. Results obtained showed the dual-part models could effectively separate guidewire, catheter, vessels and background pixels to their classes of memberships with high probability. The results demonstrate that all models were able to distinguish the dominant background class from foreground structures with high overall accuracy.

Chinese Translation

导管成像处理需要快速、准确且可解释的分割模型。尽管现有的大多数研究通常集中于二元分割，但最近对同时分割导管成像场景中多种结构的需求日益增加。在本研究中，设计了一种双部分的 MLBNet（Multi-Lateral Branched Network）架构，采用多边编码器块和多头解码器分支，以实现心血管导管成像场景中的类别感知分割。编码器中的侧支允许重复特征提取，以学习多样的共享表示，而多个解码器头则用于引入类别偏斜的分支，专门针对导管成像场景中不同的结构特性。为了分析双部分 MLBNet 架构的性能，使用了在幻影模型、合成的人体模拟主动脉和动物模型中获得的多类分割造影数据进行模型训练和评估。结果表明，双部分模型能够有效地将导丝、导管、血管和背景像素分离到其所属类别，并具有高概率。结果显示，所有模型都能够以高整体准确率将主导背景类别与前景结构区分开来。

View on arXiv Download PDF AI Translation

cs.CV / 17 / 2609.04592

Hidden In Plain Gaze: Gaze Representations as Privacy Controls for Utility and Re-identification Risk in XR

隐匿于显现之中：注视表示作为扩展现实中效用与再识别风险的隐私控制

Ilo, Cory, John, Brendan-David, Bowman, Doug A.

Abstract

Intelligent extended reality (XR) systems increasingly use eye and head tracking to infer user intent, task, and attention, but the same signals can also reveal biometric identity. We study whether gaze data representation choice can serve as a lightweight privacy control at feature extraction, before adding perturbation or formal privacy mechanisms. Using the egocentric HoloAssist dataset, we compare three gaze representations under matched model capacity: raw gaze, spatial attention heatmaps, and engineered eye-movement features. We evaluate each representation on action recognition as task utility and closed-set user re-identification as privacy leakage. Representation choice substantially changes the privacy-utility tradeoff. Engineered features retain roughly 85% of raw gaze's action-recognition accuracy while reducing re-identification by about an order of magnitude, to roughly four times the chance rate across 206 identities. This reduction attenuates rather than eliminates identity leakage, and the differences across representations show that abstraction alone does not guarantee privacy. Engineered features expose interpretable and auditable structure, giving designers a transparent privacy lever that complements mechanisms such as differential privacy.

Chinese Translation

智能扩展现实（XR）系统越来越多地使用眼动和头部追踪来推断用户意图、任务和注意力，但这些信号也可能揭示生物特征身份。我们研究注视数据表示选择是否可以作为一种轻量级的隐私控制，在特征提取阶段进行，而不是在添加扰动或正式隐私机制之后。使用自我中心的 HoloAssist 数据集，我们在匹配模型容量的情况下比较了三种注视表示：原始注视、空间注意热图和工程化眼动特征。我们评估每种表示在任务效用方面的动作识别和在隐私泄露方面的闭集用户再识别。表示选择显著改变了隐私与效用之间的权衡。工程化特征在保留约 85% 原始注视的动作识别准确率的同时，将再识别降低了大约一个数量级，约为 206 个身份的机会率的四倍。这种减少减轻了身份泄露，而不是消除了身份泄露，且不同表示之间的差异表明，仅仅抽象化并不能保证隐私。工程化特征暴露了可解释和可审计的结构，为设计者提供了一个透明的隐私杠杆，补充了差分隐私等机制。

View on arXiv Download PDF AI Translation

cs.CV / 18 / 2609.04603

An Evaluation Framework for Generating Multi-View Images of a Person in a Scene

生成场景中人物多视角图像的评估框架

Majid, Mahir, Kim, Young Kyung, Sapiro, Guillermo

Abstract

Recent generative image-editing Diffusion Transformers (DiTs) demonstrate impressive semantic editing capabilities but still struggle with spatially consistent camera angle changes. A primary bottleneck in training foundation models to execute free-form, promptable camera angle changes is the lack of specialized training data. While multi-view datasets exist for generic 3D environments and objects, there remains an absence of paired, multi-view datasets featuring human subjects at fixed locations in natural scenes, including frontal and side-profile views. Capturing such multi-camera data in unconstrained environments is logistically challenging and unscalable. In this paper, we first experiment with multiple state-of-the-art image editing models to create this data synthetically, but find that the outputs are frequently prone to hallucinations involving how much the subject's head turns relative to the background, often producing inconsistent environments. To address this issue, we propose the Head Scene Rotation Difference (HSRD) metric to quantitatively evaluate camera movements around a person. The proposed metric operates by decoupling camera movement from localized head pose manipulation. As demonstrated by the extensive experimentation, HSRD provides the pipeline necessary to evaluate 3D spatial parallax for a person in a scene, paving the way to reliably construct high-quality multi-view synthetic datasets.

Chinese Translation

最近的生成图像编辑扩散变换器（Diffusion Transformers, DiTs）展示了令人印象深刻的语义编辑能力，但在空间一致的摄像机角度变化方面仍然存在困难。训练基础模型以执行自由形式、可提示的摄像机角度变化的主要瓶颈是缺乏专门的训练数据。尽管存在针对通用3D环境和物体的多视角数据集，但在自然场景中固定位置的人物配对多视角数据集仍然缺乏，包括正面和侧面视图。在不受限制的环境中捕捉这种多摄像机数据在后勤上具有挑战性且难以扩展。在本文中，我们首先尝试使用多种最先进的图像编辑模型合成这些数据，但发现输出结果常常容易出现幻觉，涉及主体头部相对于背景的转动程度，通常产生不一致的环境。为了解决这个问题，我们提出了头部场景旋转差异（Head Scene Rotation Difference, HSRD）指标，以定量评估围绕人物的摄像机运动。所提指标通过将摄像机运动与局部头部姿态操控解耦来运作。通过广泛的实验验证，HSRD提供了评估场景中人物3D空间视差所需的管道，为可靠构建高质量的多视角合成数据集铺平了道路。

View on arXiv Download PDF AI Translation

cs.CV / 19 / 2609.04646

Importance-Aware Low-Rank Distillation of Diffusion Transformers

重视重要性的低秩蒸馏扩散变换器

Zavadski, Denis, Heid, Sebastian, Kalšan, Damjan, Roth, Stefan, Rother, Carsten

Abstract

Diffusion Transformers (DiTs) have emerged as a dominant architecture for high-quality text-to-image generation, yet their scale poses challenges for efficient deployment. While truncated singular value decomposition (SVD) is a principled tool for parameter reduction, evidence from large language models (LLMs) suggests that naive low-rank approximation can cause catastrophic failure. In contrast, we find that truncated SVD in DiTs produces smooth degradation even under substantial global compression, with redundancy distributed across projection matrices throughout the whole network rather than concentrated in a few transformer blocks. Building on these insights, we introduce SVDtrunc, a two-step block-level compression scheme, first allocating ranks across blocks and compressing the least important ones via truncated SVD under a global parameter budget, and then fine-tuning all blocks with modular knowledge distillation and a rectified-flow objective. We apply SVDtrunc to FLUX.dev across compression levels ranging from 40-90% of the original parameter count. Across three benchmarks, GenEval, HPSv2, and DPG, we outperform all competing approaches. Notably, and in contrast to prior work, we retain near-full performance at 68% and remain competitive even at 57% of the original parameter budget. Furthermore, we show that SVDtrunc complements step distillation and achieves strong results even without fine-tuning, positioning it as a practical continuation of efficiency improvements beyond diffusion step reduction for large-scale generative models. Project page: https://vislearn.github.io/SVDtrunc/

Chinese Translation

扩散变换器（Diffusion Transformers, DiTs）已成为高质量文本到图像生成的主导架构，但其规模给高效部署带来了挑战。尽管截断奇异值分解（truncated singular value decomposition, SVD）是一种原则性的参数减少工具，但来自大型语言模型（Large Language Models, LLMs）的证据表明，简单的低秩近似可能导致灾难性失败。相反，我们发现，在DiTs中，截断SVD在大规模全局压缩下仍能产生平滑的性能下降，冗余分布在整个网络的投影矩阵中，而不是集中在少数变换器块中。基于这些见解，我们提出了SVDtrunc，一种两步块级压缩方案，首先在块之间分配秩，并在全局参数预算下通过截断SVD压缩最不重要的块，然后使用模块化知识蒸馏和修正流目标对所有块进行微调。我们将SVDtrunc应用于FLUX.dev，压缩水平范围从原始参数数量的40%到90%。在三个基准测试中，GenEval、HPSv2和DPG，我们的表现超越了所有竞争方法。值得注意的是，与先前的工作相比，我们在68%的参数预算下保持接近满性能，甚至在57%的原始参数预算下仍具竞争力。此外，我们展示了SVDtrunc与步骤蒸馏相辅相成，即使在不进行微调的情况下也能取得良好结果，使其成为大型生成模型在效率改进方面超越扩散步骤减少的实际延续。项目页面：https://vislearn.github.io/SVDtrunc/

View on arXiv Download PDF AI Translation

cs.CV / 20 / 2609.04649

ReaDiT Guidance: Control for Image and Video Generation using Diffusion Transformer Features

ReaDiT 指导：基于扩散变换器特征的图像和视频生成控制

Mahajan, Jay, Liu, Chang, Makharov, Rauf, Shah, Viraj, Schwing, Alexander, Lazebnik, Svetlana

Abstract

We present DiT Readout (ReaDiT) Guidance, a lightweight framework for controlling generation with Diffusion Transformer (DiT) models via their internal feature representations. ReaDiT Guidance uses features from a single DiT block to steer the generative process according to spatial targets - like depth, pose, or edge maps - provided at test time. Furthermore, since modern text-to-video models are largely built on DiT backbones, ReaDiT Guidance naturally extends to video generation, enabling camera and motion control. Experimental results demonstrate that our approach achieves competitive or improved results compared to existing feature-based and off-the-shelf adapter-based approaches while requiring fewer parameters.

Chinese Translation

我们提出了 DiT 读出（ReaDiT）指导，这是一个轻量级框架，用于通过扩散变换器（Diffusion Transformer, DiT）模型的内部特征表示控制生成过程。ReaDiT 指导利用来自单个 DiT 块的特征，根据在测试时提供的空间目标（如深度、姿态或边缘图）引导生成过程。此外，由于现代文本到视频模型主要基于 DiT 骨干网络，ReaDiT 指导自然扩展到视频生成，实现了相机和运动控制。实验结果表明，我们的方法在参数更少的情况下，能够实现与现有基于特征和现成适配器方法相竞争或更优的结果。

View on arXiv Download PDF AI Translation

cs.CV / 21 / 2609.04689

Retinal OCTA Phenotyping with LLM Reporting for Alzheimer's Disease

阿尔茨海默病的视网膜OCTA表型分析与LLM报告

Dutta, Progga Paromita, Maliha, Jeba, Kabir, Md Rafiul

Abstract

Early identification of Alzheimer's disease (AD) remains challenging because established assessment methods can be costly, resource-intensive, or unsuitable for population-scale screening. Optical coherence tomography angiography (OCTA) provides non-invasive visualization of retinal microvasculature, but existing approaches often require diagnostic labels and provide limited measurement-level interpretation. We present an explainable OCTA pipeline that integrates annotation-aware vessel segmentation, layer-specific vascular biomarker extraction, label-free phenotyping, and measurement-grounded LLM reporting. Using 117 ROSE-1 images from 39 subjects, we apply annotation-matched segmentation models to superficial vascular complex (SVC), deep vascular complex (DVC), and combined SVC+DVC representations. The models achieve ROC-AUC values of 0.916-0.970 and Dice scores of 0.695-0.781. Six density and fractal-dimension biomarkers form subject-level profiles for exploratory clustering. Analysis of nine held-out subjects identifies an internally consistent lower-density, lower-fractal-dimension phenotype, although the absence of diagnostic labels prevents clinical interpretation. Reports generated using GPT, Gemini, and Llama are evaluated for measurement grounding, citation faithfulness, and diagnostic caution. Overall, the framework provides a transparent, non-diagnostic connection between retinal vascular measurements, exploratory phenotyping, and evidence-linked interpretation for Alzheimer's research.

Chinese Translation

阿尔茨海默病（AD）的早期识别仍然面临挑战，因为现有的评估方法可能成本高、资源密集或不适合大规模筛查。光学相干断层扫描血管成像（OCTA）提供了视网膜微血管的非侵入性可视化，但现有方法通常需要诊断标签，并且提供的测量级别解释有限。我们提出了一种可解释的OCTA流程，集成了注释感知的血管分割、特定层次的血管生物标志物提取、无标签表型分析和基于测量的LLM报告。使用来自39名受试者的117幅ROSE-1图像，我们将匹配注释的分割模型应用于表层血管复合体（SVC）、深层血管复合体（DVC）和组合SVC+DVC表示。模型的ROC-AUC值为0.916-0.970，Dice得分为0.695-0.781。六个密度和分形维度生物标志物形成了受试者级别的特征轮廓，以便进行探索性聚类。对九名保留受试者的分析识别出一种内部一致的低密度、低分形维度表型，尽管缺乏诊断标签阻碍了临床解释。使用GPT、Gemini和Llama生成的报告在测量基础、引用准确性和诊断谨慎性方面进行了评估。总体而言，该框架提供了视网膜血管测量、探索性表型分析和与证据相关的解释之间的透明、非诊断性连接，为阿尔茨海默病研究提供支持。

View on arXiv Download PDF AI Translation

cs.CV / 22 / 2609.04690

Enhancing Multimodal Emotion Recognition via Multi-Feature Encoding and Attention-Based Fusion

通过多特征编码和基于注意力的融合增强多模态情感识别

Lin, Xu, Wang, Ke, Kang, Hui, Wang, Xinying

Abstract

Multimodal emotion recognition has attracted growing interest due to its importance in human-computer interaction, remote education, and healthcare. This paper proposes a novel multimodal emotion recognition framework that integrates rich audio and visual feature extraction with an attention-based fusion strategy. For audio, we extract three complementary feature types: semantic embeddings from Wav2Vec2, MFCC features, and statistical acoustic descriptors such as pitch, energy, and rhythm. These are aligned and fused via a BiLSTM to capture temporal dependencies. For video, we propose a ResNet50-BiLSTM architecture that combines deep residual learning and sequential modeling to extract expressive spatiotemporal features from facial sequences. To enhance multimodal synergy, we introduce a feature-level fusion mechanism based on multi-head attention, allowing the model to adaptively weigh contributions across modalities. Experiments conducted on the MELD and IEMOCAP datasets demonstrate that our model significantly outperforms baselines in both accuracy and robustness. Furthermore, ablation studies show that the attention-based fusion strategy significantly improves performance in unbalanced data settings. Our findings suggest that the proposed framework effectively captures diverse emotional cues from speech and visual expressions, and offers a practical and generalizable approach for real-world multimodal emotion recognition tasks.

Chinese Translation

多模态情感识别因其在人与计算机交互、远程教育和医疗保健中的重要性而受到越来越多的关注。本文提出了一种新颖的多模态情感识别框架，该框架将丰富的音频和视觉特征提取与基于注意力的融合策略相结合。在音频方面，我们提取了三种互补的特征类型：来自Wav2Vec2的语义嵌入、MFCC特征以及音高、能量和节奏等统计声学描述符。这些特征通过双向长短期记忆网络（BiLSTM）对齐并融合，以捕捉时间依赖性。在视频方面，我们提出了一种ResNet50-BiLSTM架构，该架构结合了深度残差学习和序列建模，从面部序列中提取富有表现力的时空特征。为了增强多模态协同，我们引入了一种基于多头注意力的特征级融合机制，使模型能够自适应地加权不同模态的贡献。在MELD和IEMOCAP数据集上进行的实验表明，我们的模型在准确性和鲁棒性方面显著优于基线。此外，消融研究表明，基于注意力的融合策略在不平衡数据设置中显著提高了性能。我们的研究结果表明，所提出的框架有效捕捉了来自语音和视觉表达的多样情感线索，并为现实世界中的多模态情感识别任务提供了一种实用且可推广的方法。

View on arXiv Download PDF AI Translation

cs.CV / 23 / 2609.04698

LookThere! Sparse Vision by Reinforced Selection

看这里！通过强化选择实现稀疏视觉

Rammohan, Sreehari, Yassin, Yousef, Fuller, Anthony, Wen, Junfeng, Vondrick, Carl, Shelhamer, Evan

Abstract

Vision transformers typically treat every image token as equally important, yet for most tasks in computer vision only a fraction are needed. Adaptive computation methods accelerate inference by choosing which tokens to process, but existing methods struggle at extreme sparsity and require heuristics that may not generalize like token diversity and attention scores. We address these limitations with LookThere, achieving a new pareto frontier in performance-compute trade-offs through an end-to-end reinforcement learning framework that jointly trains a shallow input selector and a deep representation extractor. The selector learns where to look and the extractor learns what to see, together saving computation by selecting only what is worth processing for a given task without relying on auxiliary signals. We show that LookThere only selects the task-specific input, excelling at sparse recognition in high-resolution settings (traffic signs, billiards), and maintaining accuracy with as little as 0.2% of the input. It generalizes across tasks and models, including global recognition (ImageNet classification), local recognition (ADE20K segmentation), zero-shot classification (by distillation), and regression (counting). Across all settings, LookThere surpasses state-of-the-art selection to provide a general and scalable framework for specialized and efficient adaptive computation.

Chinese Translation

视觉变换器通常将每个图像标记视为同等重要，然而在计算机视觉的大多数任务中，仅需处理一小部分标记。自适应计算方法通过选择处理哪些标记来加速推理，但现有方法在极端稀疏情况下表现不佳，并且需要一些可能无法泛化的启发式方法，如标记多样性和注意力分数。我们通过LookThere解决了这些限制，通过一个端到端的强化学习框架实现了性能与计算开销之间的新帕累托前沿，该框架共同训练一个浅层输入选择器和一个深层表示提取器。选择器学习在哪里观察，提取器学习观察什么，从而通过仅选择对特定任务值得处理的内容来节省计算，而无需依赖辅助信号。我们展示了LookThere仅选择任务特定的输入，在高分辨率设置下（如交通标志、台球）在稀疏识别方面表现出色，并且在输入量仅为0.2%的情况下仍能保持准确性。它在任务和模型之间具有良好的泛化能力，包括全局识别（ImageNet分类）、局部识别（ADE20K分割）、零样本分类（通过蒸馏）和回归（计数）。在所有设置中，LookThere超越了最先进的选择方法，为专门和高效的自适应计算提供了一个通用且可扩展的框架。

View on arXiv Download PDF AI Translation

cs.CV / 24 / 2609.04709

AngelFingerprint: A Traceable, Explainable, and White-Box Stealthy Watermark for Text-Guided Image Editing

AngelFingerprint：一种可追踪、可解释且白盒隐蔽的文本引导图像编辑水印

Kung, Bo-Han, Waseda, Futa, Chang, Ching-Chun, Echizen, Isao, Chen, Shang-Tse

Abstract

Text-guided diffusion editing raises disinformation concerns, making reliable image provenance essential. While watermarks are commonly used for this purpose, most methods carry a fixed ID that cannot explain what was changed and which prompt produced it. Furthermore, under open-source white-box access, attackers can easily locate and remove watermarks added as separate modules. Targeting this setting, we propose AngelFingerprint, a novel watermarking framework ensuring edit traceability, explainability, and white-box stealthiness. It integrates a LoRA into the diffusion model to embed the editing prompt's CLIP text embedding directly into the model's weights. An extractor then recovers this embedding from the image pixels alone. This semantic payload explains the edit, while the weight-integrated design makes it hard to detect and isolate even under full white-box access. Two techniques make this possible: a velocity-alignment anchor that preserves edit quality, and a specially designed frequency filter that keeps the watermark imperceptible yet recoverable and robust. On the MagicBrush dataset, our extractor achieves $86\%$ top-1 accuracy in a 200-way prompt retrieval, versus $20\%$ for prompt inversion.

Chinese Translation

文本引导的扩散编辑引发了虚假信息的担忧，使得可靠的图像来源变得至关重要。虽然水印通常用于此目的，但大多数方法携带固定的ID，无法解释所做的更改以及产生这些更改的提示。此外，在开放源代码的白盒访问下，攻击者可以轻松定位并移除作为独立模块添加的水印。针对这一设置，我们提出了AngelFingerprint，一种新颖的水印框架，确保编辑的可追踪性、可解释性和白盒隐蔽性。它将LoRA集成到扩散模型中，直接将编辑提示的CLIP文本嵌入嵌入到模型的权重中。然后，提取器仅通过图像像素恢复该嵌入。这个语义负载解释了编辑，而权重集成设计使得即使在完全的白盒访问下也难以检测和隔离。两种技术使这一切成为可能：一种保持编辑质量的速度对齐锚点，以及一种特别设计的频率滤波器，使水印在不可感知的情况下仍然可恢复且具有鲁棒性。在MagicBrush数据集上，我们的提取器在200种提示检索中实现了86%的顶级准确率，而提示反演的准确率仅为20%。

View on arXiv Download PDF AI Translation

cs.CV / 25 / 2609.04716

Counting Beyond Instances: A Benchmark for Group-Individual Object Counting

超越实例计数：群体-个体物体计数基准

Wang, Rui, Huang, Junyi, Li, Jiahui, Yu, Qiao, Hao, Yixue, Hu, Long, Huang, Baoru

Abstract

Visual counting is commonly formulated at the instance level, aiming to estimate how many objects of a queried category appear in an image. However, real-world counting often involves higher-level semantic units formed by multiple instances, such as a bunch of grapes, a stack of plates, or a pair of shoes. This exposes a key limitation of existing counting formulations, which mainly focus on what to count, while largely overlooking at which semantic unit to count. We introduce Group-Individual Object Counting (GIC), a new setting that requires models to count both individual objects and semantic groups within a unified framework. To support this new task, we present BunchCount, a real-world benchmark with 1,330 images, 89,254 individual annotations, and 11,065 group annotations. BunchCount provides paired individual-group annotations within the same image and explicitly records containment relations between each group and its constituent individuals. Experiments on BunchCount show that current advanced counting models perform well on individual instances but fail to count semantic groups more accurately. To mitigate semantic granularity conflict, we propose a counting-unit guided relational counting framework, which exploits group-individual containment relations to regularize cross-granularity representations during training. Our method substantially improves group-level counting while better preserving individual-level counting ability, establishing a strong baseline for counting beyond instances.

Chinese Translation

视觉计数通常是在实例层面进行的，旨在估计图像中出现的特定类别物体的数量。然而，现实世界中的计数往往涉及由多个实例形成的更高层次的语义单元，例如一串葡萄、一堆盘子或一双鞋子。这暴露了现有计数公式的一个关键局限性，即主要关注计数的对象，而在很大程度上忽视了计数的语义单元。我们引入了群体-个体物体计数（Group-Individual Object Counting, GIC），这是一个新的设置，要求模型在统一框架内同时计数个体物体和语义群体。为了支持这一新任务，我们提出了BunchCount，这是一个包含1,330幅图像、89,254个个体注释和11,065个群体注释的真实世界基准。BunchCount在同一图像中提供配对的个体-群体注释，并明确记录每个群体与其组成个体之间的包含关系。在BunchCount上的实验表明，当前先进的计数模型在个体实例上表现良好，但在更准确地计数语义群体时却存在不足。为了缓解语义粒度冲突，我们提出了一种计数单元引导的关系计数框架，该框架利用群体-个体包含关系在训练过程中对跨粒度表示进行规范化。我们的方法显著提高了群体层面的计数能力，同时更好地保留了个体层面的计数能力，为超越实例的计数建立了一个强有力的基准。

View on arXiv Download PDF AI Translation

cs.CV / 26 / 2609.04718

HiSfM: Disambiguating Structure-from-Motion via Scaffold-Anchored Hierarchical Reconstruction

HiSfM：通过支架锚定的层次重建消歧运动结构

Zhao, Ziding, Cui, Hainan, Tao, Peilin, Shen, Shuhan

Abstract

Structure-from-Motion (SfM) is a fundamental tool for sparse 3D reconstruction with broad impact in robotics and vision, supporting mapping, localization, and large-scale scene modeling. However, conventional pipelines often fail under hard visual ambiguity caused by repeated or symmetric structures, and incur heavy computational cost due to redundant cameras and constraints. We present HiSfM, a hierarchical coarse-to-fine SfM framework that improves robustness and efficiency through scaffold construction. HiSfM first forms strong local communities using geometrical induced heuristics, then connects communities with a compact yet strong skeleton by packing edge-disjoint spanning trees (EDST) while verifying skeletal edges with a two-view disambiguator. We reconstruct a stable scaffold on this verified skeleton, serving as an anchor to capture the essence of the scene, and subsequently absorb remaining images via efficient registration and triangulation for further refinements. Experiments on ambiguity-focused benchmarks and general datasets show that HiSfM prevents ambiguity-induced failures while substantially reducing runtime compared to previous methods, and improves completeness over aggressive sparsification methods. Code is available at https://github.com/3dv-casia/HiSfM.

Chinese Translation

运动结构（Structure-from-Motion, SfM）是稀疏三维重建的基础工具，在机器人技术和计算机视觉领域具有广泛影响，支持地图构建、定位和大规模场景建模。然而，传统的处理流程在面对重复或对称结构造成的视觉歧义时常常失败，并且由于冗余的相机和约束而产生较高的计算成本。我们提出了HiSfM，一种层次化的粗到细的SfM框架，通过支架构建提高了鲁棒性和效率。HiSfM首先利用几何诱导启发式方法形成强大的局部社区，然后通过打包边不相交的生成树（Edge-Disjoint Spanning Trees, EDST）连接社区，并通过双视图消歧器验证骨架边。我们在这个经过验证的骨架上重建一个稳定的支架，作为捕捉场景本质的锚点，随后通过高效的配准和三角测量吸收剩余图像以进行进一步的优化。在针对歧义的基准测试和一般数据集上的实验表明，HiSfM能够防止因歧义引起的失败，同时与之前的方法相比显著减少运行时间，并在激进稀疏化方法上提高了完整性。代码可在 https://github.com/3dv-casia/HiSfM 获取。

View on arXiv Download PDF AI Translation

cs.CV / 27 / 2609.04726

Bridging Modalities and Tasks: A Unified Hierarchical ViT for SAR-to-Optical Translation and Semantic Segmentation

桥接模态与任务：一种统一的层次化ViT用于SAR到光学图像转换和语义分割

Liu, Siyuan, Zhang, Xuze, Wang, Yongshun, Pan, Licong, Liu, Hang, Li, Huihui

Abstract

Synthetic Aperture Radar (SAR) images have all-weather, day-and-night observation capabilities. However, compared with optical images, their speckle noise and non-intuitive scattering mechanism limit the interpretability of the images. Generative models for SAR-to-optical (S2O) conversion can improve visual interpretability, but existing methods often ignore the constraints on semantic structure, which are necessary for downstream tasks, for the sake of visual effects. We propose a unified collaborative dual-task learning framework, termed BMT (Bridging Modalities and Tasks), that jointly optimizes S2O image translation and semantic segmentation through a shared hierarchical Vision Transformer. The framework integrates: (1) a LocalViTBlock that fuses global self-attention with spatial depthwise convolution through a learnable gating mechanism; (2) an enhanced output module combining multi-scale refinement processing, color correction and anti-aliasing, which calibrates channel-level color statistics through feature fusion; (3) a ControlNet-style conditional injection mechanism that encodes SAR wavelet features and segmentation labels into a multi-scale feature pyramid and injects them at each encoder layer through zero-initialized convolution; (4) a bounded Kendall uncertainty weighting scheme that prevents either task from dominating the shared representation. We evaluate the framework under both paired and unpaired translation settings, on the public WHU-OPT-SAR paired dataset and a self-constructed unpaired ship dataset built from HRSID and DIOR, respectively. The experimental results show that the proposed method achieves competitive S2O translation quality and semantic segmentation performance. The dataset and source code have been publicly released at https://github.com/Lewisyuaner/BMT-S2O-main.

Chinese Translation

合成孔径雷达（SAR）图像具有全天候、昼夜观察能力。然而，与光学图像相比，其斑点噪声和非直观的散射机制限制了图像的可解释性。用于SAR到光学（S2O）转换的生成模型可以提高视觉可解释性，但现有方法往往为了视觉效果而忽视了对语义结构的约束，而这些约束对于下游任务是必要的。我们提出了一种统一的协作双任务学习框架，称为BMT（Bridging Modalities and Tasks），该框架通过共享的层次化视觉变换器（Vision Transformer）联合优化S2O图像翻译和语义分割。该框架集成了：（1）一个LocalViTBlock，通过可学习的门控机制将全局自注意力与空间深度卷积融合；（2）一个增强的输出模块，结合多尺度细化处理、颜色校正和抗锯齿，通过特征融合校准通道级颜色统计；（3）一种ControlNet风格的条件注入机制，将SAR小波特征和分割标签编码到多尺度特征金字塔中，并通过零初始化卷积在每个编码器层注入它们；（4）一种有界Kendall不确定性加权方案，防止任何一个任务主导共享表示。我们在配对和非配对翻译设置下评估该框架，使用公共的WHU-OPT-SAR配对数据集和自构建的非配对船舶数据集，该数据集分别基于HRSID和DIOR构建。实验结果表明，所提出的方法在S2O翻译质量和语义分割性能上均表现出竞争力。数据集和源代码已公开发布在https://github.com/Lewisyuaner/BMT-S2O-main。

View on arXiv Download PDF AI Translation

cs.CV / 28 / 2609.04741

Where to Look Matters: Learning Influential Views for VLM-based 3D Visual Grounding

视角选择的重要性：基于VLM的3D视觉定位中的影响视角学习

Chiang, Tsung-Chih, Yang, Hsuan-Kung, Liu, Jou-Min, Liu, Ting-Ru, Huang, Chun-Wei, Kong, Quan, Lee, Chun-Yi

Abstract

Recent zero-shot 3D visual grounding methods leverage vision-language models (VLMs) to localize objects in 3D scenes from natural language queries. However, these methods typically rely on heuristic rules to select which camera views are provided to the VLM, often prioritizing object visibility rather than grounding relevance. We present IVSGround, a framework that learns Influential View Selection for VLM-based 3D visual grounding. Instead of using fixed heuristics, a lightweight view selector is trained to identify views that provide discriminative evidence for grounding. To obtain supervision signals, we generate training signals using feedback from a reasoning VLM through a two-stage rejection sampling process. During inference, the learned selector predicts query-conditioned influential views for each candidate object, which are then evaluated by a frozen reasoning VLM through comparative grounding. Experiments on ScanRefer and NR3D show that IVSGround consistently improves grounding accuracy over existing zero-shot pipelines, demonstrating that selecting where to look is crucial for effective 3D visual grounding. Project page: https://ivsground.github.io/

Chinese Translation

近期的零样本3D视觉定位方法利用视觉-语言模型（VLM）从自然语言查询中定位3D场景中的物体。然而，这些方法通常依赖启发式规则来选择提供给VLM的相机视角，往往优先考虑物体的可见性而非定位的相关性。我们提出了IVSGround，一个学习基于VLM的3D视觉定位的影响视角选择的框架。与使用固定启发式规则不同，我们训练了一个轻量级的视角选择器，以识别提供区分性证据的视角。为了获得监督信号，我们通过两阶段拒绝采样过程，利用推理VLM的反馈生成训练信号。在推理过程中，学习到的选择器为每个候选物体预测查询条件下的影响视角，这些视角随后通过冻结的推理VLM进行比较定位评估。在ScanRefer和NR3D上的实验表明，IVSGround在现有的零样本管道上始终提高了定位准确性，证明了选择观察位置对有效的3D视觉定位至关重要。项目页面：https://ivsground.github.io/

View on arXiv Download PDF AI Translation

cs.CV / 29 / 2609.04751

SeamFlow: Structure-Aware Flow Matching on Edge Probabilities for Artist-Like UV Unwrapping

SeamFlow：基于边缘概率的结构感知流匹配用于艺术家风格的 UV 展开

Zhao, Yuming, Xian, Zangyueyang, Zhang, Qijian, Liang, Rendong, Jia, Qin, He, Ying, Hou, Junhui

Abstract

3D surface cutting and UV unwrapping are fundamental problems in computer graphics. Traditional geometric optimization methods mainly focus on reducing parameterization distortion, but they often overlook visual semantic coherence in seam layouts. Recent autoregressive generative methods improve semantic coherence, yet limited perception of mesh topology often causes inaccurate local cuts. To address these limitations, we introduce SeamFlow, a novel generative framework for 3D surface cutting. We reformulate the discrete mesh-cutting problem as continuous flow matching in a high-dimensional edge-probability space. Through continuous relaxation, SeamFlow learns a deterministic mapping from a Gaussian prior to a target seam-probability distribution. An evolution network couples local topological tokens with global shape priors and guides smooth probability flow through Ordinary Differential Equation solving. Compared with existing autoregressive generative frameworks, SeamFlow improves topology awareness through edge tokenization while eliminating both 3D spatial projection errors and artificial sequential-order bias. Extensive experiments demonstrate that SeamFlow achieves exceptional semantic coherence and remarkably low parameterization distortion. The project page is https://meshy-dev.github.io/seamflow.

Chinese Translation

3D 表面切割和 UV 展开是计算机图形学中的基本问题。传统的几何优化方法主要关注减少参数化失真，但往往忽视了缝合布局中的视觉语义一致性。近期的自回归生成方法改善了语义一致性，但对网格拓扑的有限感知常导致局部切割不准确。为了解决这些局限性，我们提出了 SeamFlow，一种用于 3D 表面切割的新型生成框架。我们将离散网格切割问题重新表述为高维边缘概率空间中的连续流匹配。通过连续松弛，SeamFlow 学习从高斯先验到目标缝合概率分布的确定性映射。演化网络将局部拓扑标记与全局形状先验耦合，并通过常微分方程求解引导平滑的概率流。与现有的自回归生成框架相比，SeamFlow 通过边缘标记化提高了拓扑感知，同时消除了 3D 空间投影误差和人为顺序偏差。大量实验表明，SeamFlow 实现了卓越的语义一致性和显著较低的参数化失真。项目页面为 https://meshy-dev.github.io/seamflow。

View on arXiv Download PDF AI Translation

cs.CV / 30 / 2609.04775

LUMIN: Lightweight Universal Manufacturing Inspection Network for Anomaly Detection

LUMIN：轻量级通用制造检测网络用于异常检测

Yang, Pengfei

Abstract

Industrial anomaly detection faces two engineering bottlenecks: memory bank construction latency and inference efficiency. Traditional sampling algorithms (Farthest Point Sampling, K-Means, etc.) rely on numerous backbone forward passes and iterative distance computations, with construction times ranging from minutes to hours; heavy computation components such as multi-scale feature extraction struggle to meet the millisecond-level real-time requirements of production lines. This paper focuses on sampling efficiency and inference optimization for industrial deployment with two core contributions: (1) PSP (Plugin Sampler Pipeline)---a four-stage adaptive memory bank sampling pipeline based on 18-dimensional pixel metadata and five complementary visual plugins. PSP completes all sampling with zero backbone forward passes; coarse filtering is sub-second numerical sorting, and metadata extraction is a one-time offline cost. PSP supports progressive deployment and incremental updates. (2) Two engineering optimization strategies---parallel memory bank similarity computation (reducing inference memory and latency by over 95\%) and stratified pixel sampling for large-scale evaluation (reducing computation time by 20$\times$ while keeping metrics stable). As a vehicle for validation, we introduce LUMIN (Lightweight Universal Manufacturing Inspection Network) with extreme segmentation-head compression, systematically exploring the accuracy-efficiency frontier against strong baselines. Experiments on five benchmarks demonstrate that PSP matches state-of-the-art sampling accuracy at near-random construction cost (341$\times$ faster than FPS), while inference optimizations reduce evaluation time by 20$\times$ with negligible accuracy loss.

Chinese Translation

工业异常检测面临两个工程瓶颈：内存库构建延迟和推理效率。传统的采样算法（最远点采样、K均值等）依赖于大量的骨干网络前向传播和迭代距离计算，构建时间从几分钟到几小时不等；多尺度特征提取等重计算组件难以满足生产线毫秒级实时要求。本文聚焦于工业部署的采样效率和推理优化，提出两个核心贡献：（1）PSP（插件采样管道）——一个基于18维像素元数据和五个互补视觉插件的四阶段自适应内存库采样管道。PSP在零骨干前向传播的情况下完成所有采样；粗过滤为亚秒级数值排序，元数据提取为一次性离线成本。PSP支持渐进部署和增量更新。（2）两种工程优化策略——并行内存库相似性计算（将推理内存和延迟减少超过95%）和分层像素采样用于大规模评估（在保持指标稳定的同时将计算时间减少20倍）。作为验证工具，我们引入LUMIN（轻量级通用制造检测网络），通过极端的分割头压缩，系统性地探索准确性与效率的边界，针对强基线进行比较。五个基准实验表明，PSP在接近随机构建成本下匹配了最先进的采样准确性（比最远点采样快341倍），而推理优化将评估时间减少了20倍，且准确性损失微乎其微。

View on arXiv Download PDF AI Translation

cs.CV / 31 / 2609.04781

CoMLP: Cooperatively-Gated MLPs for Fine-Grained Cross-Modal Information Fusion in Medical Image Segmentation

CoMLP：用于医学图像分割的协同门控多层感知器在细粒度跨模态信息融合中的应用

Meng, Mingyuan, Ye, Shuchang, Li, Mingjian, Zhao, Zhenyu, Kim, Jinman, Bi, Lei

Abstract

Multi-modal medical images and clinical reports provide complementary anatomical, functional, and semantic information for medical image segmentation. Effectively exploiting these heterogeneous sources requires fine-grained cross-modal information fusion that preserves subtle spatial details while capturing semantic dependencies across modalities. Existing fusion approaches frequently rely on cross-attention, whose computational burden increases rapidly with spatial resolution, making dense cross-modal interaction difficult on high-resolution feature maps, particularly for volumetric medical images. In this work, we propose CoMLP, a cooperatively-gated MLP module for fine-grained cross-modal information fusion in medical image segmentation. CoMLP models cross-modal dependencies through cooperative cross-gating, built upon complementary regional and dilated MLP interactions, to capture local and global cross-modal dependencies. We further develop a multi-source fusion architecture in which CoMLP performs both inter-image fusion across imaging modalities and vision-language fusion between visual features and textual reports, enabling heterogeneous information to be integrated without relying on dense cross-attention. Extensive experiments on five medical segmentation benchmarks, covering 2D/3D images, clinical reports, multiple imaging modalities, and diverse anatomical regions, demonstrate consistent improvements over state-of-the-art multi-modal and language-guided segmentation methods. Ablation studies further show that fine-grained interaction at high spatial resolutions and complementary local-global fusion are critical to the performance gains. These results demonstrate the potential of MLP-based interaction as an effective alternative for fine-grained cross-modal information fusion in medical image segmentation.

Chinese Translation

多模态医学图像和临床报告为医学图像分割提供了互补的解剖、功能和语义信息。有效利用这些异构来源需要细粒度的跨模态信息融合，以保留微妙的空间细节，同时捕捉跨模态的语义依赖关系。现有的融合方法通常依赖于交叉注意力，但其计算负担随着空间分辨率的增加而迅速增加，使得在高分辨率特征图上进行密集的跨模态交互变得困难，尤其是对于体积医学图像。在本研究中，我们提出了CoMLP，一种用于医学图像分割的协同门控多层感知器模块，以实现细粒度的跨模态信息融合。CoMLP通过协同交叉门控建模跨模态依赖关系，基于互补的区域和扩张多层感知器交互，捕捉局部和全局的跨模态依赖关系。我们进一步开发了一种多源融合架构，其中CoMLP在成像模态之间进行图像间融合，并在视觉特征与文本报告之间进行视觉-语言融合，从而实现异构信息的整合，而无需依赖密集的交叉注意力。在五个医学分割基准上的大量实验，涵盖了2D/3D图像、临床报告、多种成像模态和多样的解剖区域，均显示出相对于最先进的多模态和语言引导分割方法的一致性提升。消融研究进一步表明，在高空间分辨率下的细粒度交互和互补的局部-全局融合对性能提升至关重要。这些结果展示了基于多层感知器的交互作为医学图像分割中细粒度跨模态信息融合的有效替代方案的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 32 / 2609.04784

CLON: Cue-Calibrated Linguistic Object Onboarding for Zero-Shot 6D Pose Front-Ends

CLON：用于零样本6D姿态前端的线索校准语言对象引导

Ji, Seojin, Kwon, Yoojin, Kim, Hyung-Sin

Abstract

Zero-shot 6D pose estimation pipelines increasingly rely on strong downstream pose solvers, but their performance is often limited by the front-end: object proposals must preserve partially visible true positives while rejecting semantically plausible distractors. We introduce Cue-Calibrated Linguistic Object Onboarding (CLON), a front-end requiring no task-specific training for new objects. Given rendered templates of the onboarded object set, CLON constructs a linguistic semantic memory for top-down proposal generation and object-set cue weights for calibrated proposal scoring. The linguistic memory guides SAM 3 toward high-recall proposals for onboarded objects, while cue weights are computed once from the onboarded object set before scene inference and kept fixed during online scoring. On seven BOP-Classic-Core datasets, CLON improves detection AP by 8.1 percentage points (pp), segmentation AP by 6.2 pp, and downstream 6D pose AR by up to 4.1 pp over CNOS and SAM-6D front-ends.

Chinese Translation

零样本6D姿态估计管道越来越依赖于强大的下游姿态求解器，但其性能往往受到前端的限制：对象提议必须保留部分可见的真实正样本，同时拒绝语义上合理的干扰项。我们提出了线索校准语言对象引导（CLON），这是一种不需要针对新对象进行特定任务训练的前端。给定已引导对象集的渲染模板，CLON 构建了一个用于自上而下提议生成的语言语义记忆和用于校准提议评分的对象集线索权重。语言记忆引导 SAM 3 朝向高召回率的已引导对象提议，而线索权重在场景推理之前从已引导对象集中计算一次，并在在线评分期间保持固定。在七个 BOP-Classic-Core 数据集上，CLON 将检测 AP 提高了 8.1 个百分点（pp），分割 AP 提高了 6.2 pp，下游 6D 姿态 AR 提高了最多 4.1 pp，相较于 CNOS 和 SAM-6D 前端。

View on arXiv Download PDF AI Translation

cs.CV / 33 / 2609.04791

An Attention-Guided Global and Local Fusion Framework for Lesion-Focused Image Classification

一种基于注意力引导的全局与局部融合框架用于病变聚焦图像分类

Tasnima, Mst Shafia, Elaheea, Md Samaun, Aurpab, Tanjim Taharat, Anwar, Md Musfique

Abstract

Lesion-focused image classification presents a core analytical challenge, as discriminative signals are often sparse, spatially dispersed, and easily obscured by background noise, while conventional convolutional neural networks (CNNs) process entire images uniformly and may dilute signal relevance. This study hypothesizes that adaptive fusion of global contextual information and lesion-focused local information can improve classification performance compared with using either representation independently. We propose a three-branch, attention-guided deep learning framework built on Densely Connected Convolutional Network-121 (DenseNet-121) to improve feature attribution, interpretability, and classification reliability. The architecture consists of a global branch that learns representations from full images, followed by Gradient-weighted Class Activation Mapping (Grad-CAM) to generate attention maps that highlight prediction-relevant regions and produce masked inputs, and a local branch enhanced with a Convolutional Block Attention Module (CBAM) to extract refined spatial and channel-wise features from these focused regions. An adaptive fusion branch integrates global and local representations by learning instance-specific weights, allowing dynamic prioritization between contextual and localized information. The framework is evaluated on a synthetic Spot Pattern Dataset (SSPD) and three benchmark datasets, including skin lesion, guava leaf, and grape leaf image datasets, where the fusion branch outperformed the individual global and local branches, reaching 97.75% accuracy on the skin lesion dataset and 99.64% on the guava leaf dataset. The results highlight the value of attention-guided architectures in healthcare analytics by improving model transparency, strengthening feature relevance, and supporting more reliable data-driven decision-making in medical image analysis.

Chinese Translation

病变聚焦图像分类呈现出核心分析挑战，因为判别信号通常稀疏、空间分散，并且容易被背景噪声遮蔽，而传统的卷积神经网络（CNN）对整个图像进行均匀处理，可能会稀释信号的相关性。本研究假设，全球上下文信息与病变聚焦局部信息的自适应融合可以提高分类性能，相较于独立使用任一表示。我们提出了一种基于密集连接卷积网络-121（DenseNet-121）的三分支注意力引导深度学习框架，以改善特征归因、可解释性和分类可靠性。该架构由一个全局分支组成，该分支从完整图像中学习表示，随后通过梯度加权类激活映射（Grad-CAM）生成注意力图，突出与预测相关的区域并产生掩蔽输入；局部分支则通过卷积块注意力模块（CBAM）增强，从这些聚焦区域提取精细的空间和通道特征。自适应融合分支通过学习实例特定的权重，将全局和局部表示进行整合，从而允许在上下文信息和局部化信息之间进行动态优先级排序。该框架在合成斑点模式数据集（SSPD）和三个基准数据集上进行了评估，包括皮肤病变、番石榴叶和葡萄叶图像数据集，其中融合分支的表现优于单独的全局和局部分支，在皮肤病变数据集上达到97.75%的准确率，在番石榴叶数据集上达到99.64%的准确率。结果突显了注意力引导架构在医疗分析中的价值，通过提高模型透明度、增强特征相关性，并支持更可靠的数据驱动决策制定在医学图像分析中。

View on arXiv Download PDF AI Translation

cs.CV / 34 / 2609.04800

Intrinsic Temporal Adaptation of CLIP for Partially Relevant Video Retrieval

CLIP的内在时间适应用于部分相关视频检索

Seong, Hyun Seok, Jun, Woojin, Lee, SuBeen, Heo, Jae-Pil

Abstract

Partially Relevant Video Retrieval (PRVR) aims to retrieve untrimmed videos that contain moments relevant to a text query. Since the target moment occupies only a portion of the video, PRVR requires retrieval based on fine-grained understanding beyond coarse video-level matching. However, existing methods often rely on frozen CLIP frame features, which lack temporal understanding. Even with recent progress in parameter-efficient CLIP adaptation, video-level predictions can still be supported by imprecise frame-level evidence. In this paper, we propose an Intrinsic Temporal Adaptation (ITA) framework for PRVR. First, our Backbone-Internal Temporal Adaptation allows the last few visual transformer layers to attend over groups of neighboring frames. This provides temporally aware frame embeddings while keeping CLIP frozen and training only adaptation parameters. Second, we introduce Affinity-Weighted Gradient Propagation to address the weakly supervised nature of PRVR, softly aggregating top-$k$ frames based on text-frame affinities and propagating learning signals to multiple query-relevant frames. Our method achieves state-of-the-art performance on PRVR benchmarks, demonstrates robust cross-dataset transfer, and retrieves substantially more accurate frame-level evidence within ground-truth query-relevant moments. Our code is available at github.com/hynnsk/ITA.

Chinese Translation

部分相关视频检索（PRVR）旨在检索包含与文本查询相关时刻的未剪辑视频。由于目标时刻仅占视频的一部分，PRVR需要基于细粒度理解进行检索，而不仅仅是粗略的视频级匹配。然而，现有方法往往依赖于静态的CLIP帧特征，这缺乏时间理解。即使在参数高效的CLIP适应方面取得了近期进展，视频级预测仍然可能受到不精确的帧级证据的支持。本文提出了一种用于PRVR的内在时间适应（ITA）框架。首先，我们的主干内部时间适应允许最后几层视觉变换器对相邻帧组进行关注。这在保持CLIP静态并仅训练适应参数的同时，提供了时间感知的帧嵌入。其次，我们引入了亲和力加权梯度传播，以应对PRVR的弱监督特性，基于文本-帧亲和度柔和地聚合前$k$帧，并将学习信号传播到多个与查询相关的帧。我们的方法在PRVR基准上实现了最先进的性能，展示了强大的跨数据集迁移能力，并在真实查询相关时刻内检索到显著更准确的帧级证据。我们的代码可在github.com/hynnsk/ITA获取。

View on arXiv Download PDF AI Translation

cs.CV / 35 / 2609.04802

Linguistic Trajectory Encoding for Efficient Long-Horizon Spatial Memory in Embodied Agents

用于高效长时间空间记忆的语言轨迹编码在具身智能体中的应用

Xie, Tianyidan, Wang, Shenyi, Tang, Qiang, Wang, Mingjie, Qiu, Zhicheng, Li, Xuanfu, Xu, Zhan, Yang, Jian, Wang, Lanjun, Yi, Zili

Abstract

Embodied agents performing long-horizon tasks require a memory representation in which the state transitions of dynamic objects remain queryable in natural language across hours-to-days observation horizons. Existing systems either drop fine-grained motion (clip-level video-language embeddings), keep it only as raw coordinates (geometric SLAM), or organise it around immediate task context (agent working memories). None of them gives the agent a per-object timeline whose state transitions are themselves queryable in language. Our key contribution is \textbf{Linguistic Trajectory Encoding} (LTE), which compresses dynamic object motion histories via a hybrid representation combining natural language descriptions, sparse spatial anchors, and visual anchors. LTE adapts compression to motion complexity by anchoring periods without reliable observations to the last seen location, while representing motion with geometric waypoints and linguistic descriptions to preserve accuracy. To evaluate these capabilities across extended time horizons, we construct the \textbf{Spatial Memory Benchmark} (SMB) from EgoLife multi-day recordings, targeting capabilities absent in existing benchmarks: semantic trajectory retrieval and long-horizon object retrieval. On SMB, the LTE-based system achieves $45.3\%$ success in semantic trajectory retrieval and $48.7\%$ in long-horizon object retrieval, outperforming structured-memory and VLM baselines (best prior: $31.9\%$ and $34.4\%$). LTE achieves trajectory compression by factors of $8.7\times$ to $26.1\times$ with sub-second query latency on $24$\,h video. On Ego4D natural-language queries, the system reaches $28.75\%$ / $55.10\%$ R@1/R@5, $+15.80$ / $+31.30$ pts over EgoVLPv2.

Chinese Translation

执行长时间任务的具身智能体需要一种记忆表示，使得动态对象的状态转变在数小时到数天的观察范围内能够用自然语言进行查询。现有系统要么忽略细粒度的运动（剪辑级视频-语言嵌入），要么仅将其保留为原始坐标（几何SLAM），或围绕即时任务上下文进行组织（智能体工作记忆）。这些系统都没有为智能体提供一个每个对象的时间线，其状态转变本身可以用语言查询。我们的关键贡献是 extbf{语言轨迹编码}（Linguistic Trajectory Encoding, LTE），它通过结合自然语言描述、稀疏空间锚点和视觉锚点的混合表示来压缩动态对象的运动历史。LTE通过将没有可靠观察的时间段锚定到最后一次看到的位置，适应运动复杂性的压缩，同时用几何路径点和语言描述来表示运动，以保持准确性。为了评估这些能力在扩展时间范围内的表现，我们从EgoLife多日录音构建了 extbf{空间记忆基准}（Spatial Memory Benchmark, SMB），目标是填补现有基准中缺失的能力：语义轨迹检索和长时间对象检索。在SMB上，基于LTE的系统在语义轨迹检索中取得了$45.3\%$的成功率，在长时间对象检索中取得了$48.7\\%$，超越了结构化记忆和VLM基线（最佳先前结果：$31.9\\%$和$34.4\\%$）。LTE实现了$8.7 imes$到$26.1 imes$的轨迹压缩，并在$24$小时视频上实现了亚秒级查询延迟。在Ego4D自然语言查询中，该系统达到了$28.75\\%$ / $55.10\\%$的R@1/R@5，相较于EgoVLPv2提高了$+15.80$ / $+31.30$个百分点。

View on arXiv Download PDF AI Translation

cs.CV / 36 / 2609.04827

Weather-Conditioned Depth Anything

天气条件下的深度估计模型

Xu, Zhaoming, Hu, Chan-Wei, Huang, Kuan-Ru, Zhu, Zihao, Li, Renjie, Zhou, Yang, Tu, Zhengzhong

Abstract

Monocular depth estimation foundation models, such as the Depth Anything series, have achieved remarkable performance across diverse domains. However, they still suffer from critical failures under adverse weather conditions, such as fog, rain, snow, or at night. To address this, we present Weather-Conditioned Depth Anything (DA-W), a framework that explicitly disentangles style from content for weather-robust depth estimation. Specifically, we introduce a Style Filter trained on a curated mix of real and synthetic degradation datasets to extract content-independent, degradation-aware weather embeddings. This style embedding is then injected into the Depth Anything backbone using a parameter-efficient, zero-initialized adapter. Such a lightweight modulation allows a single unified model to robustly adapt to diverse conditions, including fog, rain, snow, and low-light, while avoiding catastrophic forgetting of its core generalization abilities in normal conditions. We train the adapter using a pseudo-label distillation and alignment strategy. Our comprehensive experiments demonstrate that our proposed DA-W achieves state-of-the-art robust depth estimation, improving AbsRel by an average of 3.7% on our curated weather benchmarks, while matching or slightly outperforming performance on standard clean benchmarks. Our project page is available at https://zhaoming-tamu.github.io/WCDA/.

Chinese Translation

单目深度估计基础模型，如Depth Anything系列，在多个领域取得了显著的性能。然而，它们在恶劣天气条件下（如雾、雨、雪或夜间）仍然面临严重的失效问题。为了解决这个问题，我们提出了天气条件下的深度估计模型（Weather-Conditioned Depth Anything, DA-W），该框架明确地将风格与内容分离，以实现对天气的鲁棒深度估计。具体而言，我们引入了一种风格过滤器，该过滤器在经过精心挑选的真实和合成降质数据集上进行训练，以提取内容无关、降质感知的天气嵌入。然后，这种风格嵌入通过一种参数高效、零初始化的适配器注入到Depth Anything主干网络中。这种轻量级的调制使得单一统一模型能够稳健地适应多种条件，包括雾、雨、雪和低光照，同时避免在正常条件下对其核心泛化能力的灾难性遗忘。我们使用伪标签蒸馏和对齐策略训练适配器。我们的全面实验表明，所提出的DA-W在鲁棒深度估计方面达到了最先进的水平，在我们精心挑选的天气基准上，AbsRel平均提高了3.7%，同时在标准清晰基准上表现相当或略有超越。我们的项目页面可访问 https://zhaoming-tamu.github.io/WCDA/。

View on arXiv Download PDF AI Translation

cs.CV / 37 / 2609.04837

PAPT++: Risk-Aware Adversarial Tuning and Generation for Single Domain Generalization

PAPT++：面向单域泛化的风险感知对抗调优与生成

Xu, Zhipeng, Cheng, De, Jiang, Xinyang, He, Lingfeng, Wang, Huaijie, Li, Dongsheng, Wang, Nannan, Gao, Xinbo

Abstract

Single domain generalization (SDG) aims to learn a model from one labeled source domain that generalizes to unseen target domains. A common strategy is to enrich the source distribution with augmented or generated samples, and recent text-to-image (T2I) diffusion models provide a strong generative prior for this purpose. However, diversity alone is insufficient for robust generalization, because useful generated samples should also capture variations that the current classifier finds difficult. Motivated by distributionally robust optimization (DRO), we define a semantic ambiguity set in the class-conditional generative space of a pretrained T2I model and search it for samples with high classification loss under the current classifier. To this end, we introduce PAPT++, a risk-aware adversarial generation-training framework for SDG. PAPT++ first learns diverse semantic reference images for each class through image-text alignment and intra-class diversity regularization. These references then serve as denoising targets during classifier-guided diffusion synthesis, reducing semantic drift while guiding generation toward challenging variations. The generated samples are combined with the source data to update the classifier, and the updated classifier guides the next synthesis round in return. In this way, PAPT++ progressively exposes the classifier to challenging yet semantically consistent variations. Extensive experiments on standard SDG benchmarks demonstrate the superiority of the proposed PAPT++ method and the effectiveness of its main components.

Chinese Translation

单域泛化（SDG）旨在从一个标记的源域学习一个模型，使其能够泛化到未见过的目标域。一种常见策略是通过增强或生成样本来丰富源分布，而最近的文本到图像（T2I）扩散模型为此提供了强大的生成先验。然而，仅仅依靠多样性不足以实现稳健的泛化，因为有用的生成样本还应捕捉当前分类器难以处理的变异。受到分布鲁棒优化（DRO）的启发，我们在预训练的T2I模型的类别条件生成空间中定义了一个语义模糊集，并在其中搜索在当前分类器下具有高分类损失的样本。为此，我们提出了PAPT++，一个面向SDG的风险感知对抗生成训练框架。PAPT++首先通过图像-文本对齐和类内多样性正则化为每个类别学习多样的语义参考图像。这些参考图像随后作为去噪目标，在分类器引导的扩散合成过程中使用，减少语义漂移，同时引导生成朝向具有挑战性的变异。生成的样本与源数据结合，用于更新分类器，而更新后的分类器则反过来引导下一轮合成。通过这种方式，PAPT++逐步使分类器接触到具有挑战性但语义一致的变异。在标准SDG基准上的大量实验表明，所提出的PAPT++方法及其主要组件的有效性和优越性。

View on arXiv Download PDF AI Translation

cs.CV / 38 / 2609.04846

LetOccVote: Learning Weakly Supervised 3D Occupancy through Consensus

LetOccVote：通过共识学习弱监督的3D占用预测

Zhang, Chi, Song, Qi, Li, Feifei, Li, Jie, Huang, Rui

Abstract

Weakly supervised 3D occupancy prediction reduces the reliance on costly 3D annotations by learning from 2D pseudo-labels generated by vision foundation models. However, existing methods typically use these imperfect pseudo-labels directly as supervision, making occupancy learning vulnerable to erroneous geometric and semantic targets. We observe that agreement across repeated observations provides an inexpensive and reliable cue for assessing pseudo-label reliability. Based on this observation, we propose \textbf{LetOccVote}, a weakly supervised Gaussian-based occupancy framework that leverages cross-frame voting to improve both geometric and semantic supervision. For geometry, Depth Vote exploits cross-frame geometric agreement to refine supported pseudo depth and reject contradictory estimates before volumetric lifting and depth supervision. For semantics, Semantic Vote aggregates pseudo-semantic observations in a shared 3D space to identify reliable and contested evidence, strengthening reliable semantic supervision while filtering unreliable pseudo-label segments. The entire framework is trained solely with 2D pseudo-label supervision without requiring 3D occupancy annotations. On Occ3D-nuScenes, LetOccVote achieves 53.27 IoU and 20.39 mIoU, establishing state-of-the-art performance among methods with 2D pseudo-label supervision.

Chinese Translation

弱监督的3D占用预测通过利用视觉基础模型生成的2D伪标签，减少了对昂贵的3D标注的依赖。然而，现有方法通常直接将这些不完美的伪标签用作监督，这使得占用学习容易受到错误的几何和语义目标的影响。我们观察到，重复观察之间的一致性为评估伪标签的可靠性提供了一种廉价且可靠的线索。基于这一观察，我们提出了 extbf{LetOccVote}，一个基于高斯的弱监督占用框架，利用跨帧投票来改善几何和语义监督。在几何方面，Depth Vote利用跨帧几何一致性来精炼支持的伪深度，并在体积提升和深度监督之前拒绝矛盾的估计。在语义方面，Semantic Vote在共享的3D空间中聚合伪语义观察，以识别可靠和有争议的证据，增强可靠的语义监督，同时过滤不可靠的伪标签片段。整个框架仅使用2D伪标签监督进行训练，而无需3D占用标注。在Occ3D-nuScenes数据集上，LetOccVote达到了53.27的IoU和20.39的mIoU，在使用2D伪标签监督的方法中建立了最先进的性能。

View on arXiv Download PDF AI Translation

cs.CV / 39 / 2609.04860

Mitigating Performance Discrepancy in Cross-Domain 3D Class-Incremental Learning

缓解跨域3D类增量学习中的性能差异

Ma, Jinge, Vinod, Gautham, Coburn, Bruce, Chi, Jui-Feng, Raghavan, Siddeshwar, Zhu, Fengqing

Abstract

3D perception plays a crucial role in real-world applications such as autonomous driving, robotics, and AR/VR. In practical scenarios, 3D perception models need to continually adapt to newly emerging 3D object categories, making class-incremental learning (CIL) particularly important. However, unlike 2D images, 3D point clouds are inherently heterogeneous: objects from the same class may not only come from the clean CAD domain, but also from RGB-D camera scans of varying quality, video reconstructions, or even corrupted observations. We discover that such heterogeneity introduces a new challenge beyond catastrophic forgetting: the degree of performance degradation can vary substantially across domains, a phenomenon we term performance discrepancy. To investigate this problem, we establish the Domain3D-CIL training and evaluation protocol, which contains point cloud categories from heterogeneous domains. We further adapt a wide range of mainstream CIL methods to the 3D modality. The results demonstrate that this performance discrepancy consistently appears across these baselines. To mitigate this issue, we introduce PolyMem, an exemplar-free approach that implicitly models rich high-order statistics of the feature distribution to enhance cross-domain robustness. Experiments demonstrate that our method effectively alleviates the performance discrepancy while improving the model's performance across domains. Code will be made publicly available upon acceptance.

Chinese Translation

3D感知在自动驾驶、机器人技术以及增强现实/虚拟现实等现实应用中发挥着至关重要的作用。在实际场景中，3D感知模型需要不断适应新出现的3D物体类别，这使得类增量学习（Class-Incremental Learning, CIL）变得尤为重要。然而，与2D图像不同，3D点云本质上是异质的：同一类别的物体不仅可能来自干净的CAD域，还可能来自不同质量的RGB-D相机扫描、视频重建，甚至是损坏的观测。我们发现，这种异质性引入了一个超出灾难性遗忘的新挑战：性能下降的程度在不同域之间可能存在显著差异，这一现象我们称之为性能差异。为了解决这个问题，我们建立了Domain3D-CIL训练和评估协议，该协议包含来自异质域的点云类别。我们进一步将一系列主流CIL方法适配到3D模态。结果表明，这种性能差异在这些基线方法中始终存在。为了缓解这一问题，我们引入了PolyMem，一种无示例的方法，隐式建模特征分布的丰富高阶统计量，以增强跨域鲁棒性。实验表明，我们的方法有效缓解了性能差异，同时提高了模型在不同域之间的性能。代码将在接受后公开发布。

View on arXiv Download PDF AI Translation

cs.CV / 40 / 2609.04886

SimFuse3D: Source-Guided Target Simulation and Confidence-Guided Multi-Stage Localization Reweighting for Cross-Platform 3D Object Detection

SimFuse3D：源引导目标模拟与信心引导的多阶段定位重加权用于跨平台3D物体检测

Lin, Yongchun, Zhang, Xinliang, Zou, Yun, Xiao, Zhixuan, Lei, Liang, Guo, Jianya, Zhai, Yuqiang, Wang, Xiaofeng, Xu, HaiKuo, Li, Haoang

Abstract

Changes in sensor height and viewpoint alter object-level point distributions, making cross-platform LiDAR unsupervised domain adaptation (UDA) difficult. Self-training uses labeled source scans and unlabeled target scans, yet a retained prediction may provide a useful target location while enclosing sparse foreground returns, background clutter, or points inconsistent with the predicted box. We refer to this mismatch as box-point inconsistency. We introduce SimFuse3D, which preserves the target placement and repairs the associated pseudo-object using measured geometry from labeled source scans. Object Memory retrieves a compatible labeled source instance. Target Simulation places its ground-truth box at the target location, aligns its points with the target viewing geometry, and filters the aligned crop to approximate the target observation. Confidence-Guided Multi-Stage Localization Reweighting (CMLR) maps each target pseudo-object confidence score to a bounded weight shared by RPN localization and R-CNN box regression. All components operate only during adaptation, leaving the detector architecture and inference graph unchanged. Across six cross-platform transfers, SimFuse3D exceeds Pi3DET-Net on every reported AP metric and ranks first among the compared adaptation methods on nearly all metrics. On nuScenes-to-KITTI, it ranks first among the compared adaptation methods with both evaluated detectors.

Chinese Translation

传感器高度和视角的变化会改变物体级点分布，使得跨平台激光雷达无监督领域适应（UDA）变得困难。自我训练利用标记的源扫描和未标记的目标扫描，然而保留的预测可能提供有用的目标位置，同时包围稀疏的前景返回、背景杂波或与预测框不一致的点。我们将这种不匹配称为框-点不一致。我们引入SimFuse3D，它保留目标放置并利用来自标记源扫描的测量几何修复相关的伪物体。对象记忆检索兼容的标记源实例。目标模拟将其真实框放置在目标位置，调整其点与目标视角几何对齐，并过滤对齐的裁剪以近似目标观察。信心引导的多阶段定位重加权（CMLR）将每个目标伪物体的信心得分映射到一个由RPN定位和R-CNN框回归共享的有界权重。所有组件仅在适应期间运行，保持检测器架构和推理图不变。在六个跨平台转移中，SimFuse3D在每个报告的AP指标上均超过Pi3DET-Net，并在几乎所有指标上在比较的适应方法中排名第一。在nuScenes到KITTI的转移中，它在比较的适应方法中排名第一，且评估的检测器均表现优异。

View on arXiv Download PDF AI Translation

cs.CV / 41 / 2609.04902

Sound-based Multi-Person 3D Pose Estimation

基于声音的多人人体三维姿态估计

Oumi, Yusuke, Shibata, Yuto, Irie, Go, Kimura, Akisato, Aoki, Yoshimitsu, Isogawa, Mariko

Abstract

Can we recover the 3D poses of multiple people using only sound? This paper presents the first attempt to estimate multi-person 3D poses solely from acoustic signals. Estimating the poses of multiple individuals using acoustic signals is inherently challenging due to the superposition of motion-dependent signal variations. Unlike single-person scenarios, the presence of multiple subjects leads to overlapping acoustic signatures, making it difficult to attribute specific signal changes to an individual's pose. Furthermore, the complexity is compounded by inter-person reflections, which introduce intricate propagation delays that obscure the temporal motion-acoustic relationship. To address these issues, we propose SoundMHPE (Sound-based Multi-person Human Pose Estimator), a novel encoder-decoder framework consisting of two key components. First, the Acoustic Multi-scale Encoder captures diverse temporal and fine-grained frequency features to isolate subtle acoustic signatures from complex, overlapping signals. Second, the Temporal Pose Decoder employs an attention mechanism to disentangle multi-person information across successive frames. By jointly accounting for temporal dynamics and inter-person dependencies, this component precisely reconstructs frame-wise individual poses. To validate our approach, we constructed the 6-hour Acoustic Multi-person Pose (AMP) dataset consisting of 432K synchronized frames of multi-person pose and acoustic data, and demonstrated that our SoundMHPE outperforms baseline models. Project page: https://oumi03.github.io/sound-mhpe/

Chinese Translation

我们能否仅通过声音恢复多人的三维姿态？本文首次尝试仅基于声学信号估计多人的三维姿态。使用声学信号估计多个人体的姿态本质上具有挑战性，因为运动依赖的信号变化会叠加在一起。与单人场景不同，多主体的存在导致声学特征重叠，使得难以将特定的信号变化归因于个体的姿态。此外，复杂性还因个体间的反射而加剧，这引入了复杂的传播延迟，模糊了时间上的运动与声学之间的关系。为了解决这些问题，我们提出了SoundMHPE（基于声音的多人人体姿态估计器），这是一种新颖的编码-解码框架，由两个关键组件组成。首先，声学多尺度编码器捕捉多样的时间和细粒度频率特征，以从复杂的重叠信号中隔离微妙的声学特征。其次，时间姿态解码器采用注意力机制，解开连续帧中的多人人体信息。通过共同考虑时间动态和个体间的依赖关系，该组件精确重建逐帧的个体姿态。为了验证我们的方法，我们构建了一个包含432K同步的多人人体姿态和声学数据的6小时声学多人人体姿态（AMP）数据集，并证明我们的SoundMHPE优于基线模型。项目页面：https://oumi03.github.io/sound-mhpe/

View on arXiv Download PDF AI Translation

cs.CV / 42 / 2609.04903

InterSing: Explicit Interaction Dynamics for 3D Duet Singing Animation and Beyond

InterSing：用于三维二重唱动画及其扩展的显式互动动态

Zhou, Yihan, Huang, Zikai, Yu, Yuyang, Xu, Xuemiao, Xu, Cheng, He, Shengfeng

Abstract

We present InterSing, a framework for generating realistic 3D head animations for duet singing performances. Unlike solo singing, duet performance requires each singer to balance individual expressiveness with intermittent interaction at musically salient moments, such as phrase boundaries, synchronized rhythms, and call-and-response passages. Because these interactions are sparse and rhythm-dependent, existing audio-driven animation methods and conversational interaction models do not adequately capture their structure. Our key insight is that duet coordination can be represented as a time-varying signal that reflects how strongly performers engage with one another throughout a song. Based on this observation, we introduce interaction logits, an interpretable latent representation that models the degree of cross-performer engagement at each time step. We learn these logits using weak supervision and use them to condition an interaction-aware diffusion model jointly driven by audio features and interaction dynamics. This formulation enables unified multi-mode generation, spanning independent motion, coordinated behavior, and smooth transitions between them. Experiments show that InterSing generates realistic and expressive singing head animations with stronger coordination and musical alignment than existing methods, while preserving each performer's characteristic motion style. We further demonstrate that the same formulation generalizes to multi-singer performances and provides intuitive control over when and how performers engage.

Chinese Translation

我们提出了InterSing，一个用于生成二重唱表演的逼真三维头部动画的框架。与独唱不同，二重唱表演要求每位歌手在音乐上显著的时刻（如短语边界、同步节奏和呼应段落）之间平衡个人表现力与间歇性互动。由于这些互动是稀疏且依赖于节奏的，现有的音频驱动动画方法和对话互动模型无法充分捕捉其结构。我们的关键见解是，二重唱的协调可以表示为一个随时间变化的信号，反映表演者在整首歌曲中彼此互动的强度。基于这一观察，我们引入了互动对数（interaction logits），这是一种可解释的潜在表示，建模每个时间步跨表演者的参与程度。我们使用弱监督学习这些对数，并利用它们来调节一个同时受音频特征和互动动态驱动的互动感知扩散模型。这种表述使得统一的多模式生成成为可能，涵盖独立运动、协调行为以及它们之间的平滑过渡。实验表明，InterSing生成的二重唱头部动画在协调性和音乐对齐方面优于现有方法，同时保留了每位表演者的特征运动风格。我们进一步证明，相同的表述可以推广到多歌手表演，并提供直观的控制，决定表演者何时以及如何互动。

View on arXiv Download PDF AI Translation

cs.CV / 43 / 2609.04906

Methane Detection On Board Satellites from Unorthorectified Imagery

基于未正射校正影像的卫星甲烷检测

Marini, Luca, Chen, Maggie, Lamdouar, Hala, Martínez-Ferrer, Laura, Bridges, Dr C. P., Acciarini, Giacomo

Abstract

As a potent greenhouse gas, methane is a major driver of climate change. Its effective mitigation relies on timely detection. Conventional detection methods rely on orthorectification to correct geometric distortions and matched filters to enhance plume signals, which are steps designed for ground processing and poorly suited to onboard execution. We introduce UnorthoDOS, a dataset and approach for training machine learning models directly on unorthorectified hyperspectral imagery, bypassing both orthorectification and matched-filter products. Our U-Net models trained on unorthorectified data approach the performance of models trained on orthorectified data (IoU 16.91% vs. 18.47% on all plumes), while both substantially outperform the mag1c matched-filter baseline (IoU 4.76%). We further demonstrate the feasibility of onboard deployment: FP16 compression halves model size with under 0.3% output deviation. The trained ML models and two ML-ready datasets -- orthorectified and unorthorectified hyperspectral imagery from the EMIT sensor -- are publicly available at https://huggingface.co/datasets/SpaceML/UnorthoDOS, with code at https://github.com/spaceml-org/plume-hunter.

Chinese Translation

甲烷作为一种强效温室气体，是气候变化的重要驱动因素。其有效减排依赖于及时的检测。传统检测方法依赖于正射校正来修正几何失真，并使用匹配滤波器来增强气流信号，这些步骤旨在地面处理，不适合在卫星上执行。我们提出了UnorthoDOS，一个用于直接在未正射校正高光谱影像上训练机器学习模型的数据集和方法，绕过了正射校正和匹配滤波产品。我们在未正射校正数据上训练的U-Net模型，其性能接近于在正射校正数据上训练的模型（所有气流的IoU为16.91%对比18.47%），而且两者均显著优于mag1c匹配滤波基线（IoU为4.76%）。我们进一步展示了在卫星上部署的可行性：FP16压缩将模型大小减半，输出偏差低于0.3%。训练好的机器学习模型和两个适合机器学习的数据集——来自EMIT传感器的正射校正和未正射校正高光谱影像——已公开发布，链接为https://huggingface.co/datasets/SpaceML/UnorthoDOS，代码可在https://github.com/spaceml-org/plume-hunter获取。

View on arXiv Download PDF AI Translation

cs.CV / 44 / 2609.04911

TourPhysics: Bringing Physics to World Models for Exploration and Manipulation from a Single Image

TourPhysics：将物理学引入世界模型，以便从单幅图像进行探索和操控

Zhang, Xin, Chen, Yabo, Duan, Zixuan, Huang, Haibin, Zhang, Chi, Xu, Feng, Li, Xuelong

Abstract

Interactive visual world models must distinguish observation from physical intervention. Camera motion reveals new surfaces, whereas intervention changes object motion, contact, and deformation. Current video world models are largely driven by appearance priors and often lose physical or spatial consistency over long horizons. We present TourPhysics, an online framework initialized from a single image and a declarative physical configuration. TourPhysics extends PhysOmni, our ACM Multimedia 2026 work, from finite physics-grounded video synthesis to persistent exploration and manipulation. TourPhysics combines deterministic simulation with video generation while assigning separate roles to simulator state, geometric evidence, generator controls, and appearance memory. For each action, the simulator computes a finite physical and camera trajectory before the corresponding observation is generated. Accepted observations publish the terminal state and update the appearance memory and subsequent generator controls, while the committed state and simulator geometry remain fixed throughout synthesis and retry. We further separate the simulator geometry used for projection and visibility from the relative depth used to condition the generator. A reference-anchored memory retrieves accepted static appearance through geometric cross-view correspondence and incorporates it through a bounded residual that reverts to the native path when no valid correspondence exists. On simulator-defined camera tours and object manipulations, TourPhysics follows prescribed camera and object trajectories more closely than the evaluated baselines, preserves the input scene, and reduces appearance drift during long-horizon revisits.

Chinese Translation

交互式视觉世界模型必须区分观察与物理干预。相机运动揭示了新的表面，而干预则改变了物体的运动、接触和变形。目前的视频世界模型主要依赖外观先验，往往在长时间范围内失去物理或空间一致性。我们提出了TourPhysics，这是一个从单幅图像和声明性物理配置初始化的在线框架。TourPhysics将我们在ACM Multimedia 2026上提出的PhysOmni从有限的物理基础视频合成扩展到持久的探索和操控。TourPhysics结合了确定性模拟与视频生成，同时为模拟器状态、几何证据、生成器控制和外观记忆分配了不同的角色。对于每个动作，模拟器在生成相应观察之前计算有限的物理和相机轨迹。被接受的观察发布终态并更新外观记忆和后续生成器控制，而承诺状态和模拟器几何在合成和重试过程中保持不变。我们进一步将用于投影和可见性的模拟器几何与用于条件生成器的相对深度分开。参考锚定记忆通过几何交叉视图对应检索已接受的静态外观，并通过一个有界残差将其纳入，当没有有效对应存在时则恢复到原始路径。在模拟器定义的相机巡游和物体操控中，TourPhysics比评估的基线更紧密地遵循规定的相机和物体轨迹，保持输入场景，并在长时间范围的重访中减少外观漂移。

View on arXiv Download PDF AI Translation

cs.CV / 45 / 2609.04921

One Diffusion Model, Two Roles: Guided Trajectory Planning and Safety-Critical Scenario Generation in Closed-Loop Simulation

一个扩散模型，两个角色：闭环仿真中的引导轨迹规划和安全关键场景生成

Pal, Arka, Kumar, Rajesh, Eriksson, Hannes, Lacombe, Rémi, Ling, Arvid Laveno, Gupta, Ankit, Wozniak, Maciej

Abstract

Diffusion probabilistic models can capture the multi-modal, interaction-rich distribution of joint future trajectories in driving scenes. We show that a single pretrained diffusion traffic model can serve two complementary roles in the autonomous driving development loop: as an ego motion planner, and as a controllable generator of safety-critical scenarios for stress-testing the planners. On the planning side, we introduce a Single-Stream Dual-Stream (SSDS) diffusion-transformer decoder that fuses scene context via joint attention rather than late cross-attention, improving closed-loop performance on nuPlan. We further propose Decoupled Annealing Posterior Sampling with Energy (DAPSE), a training-free guidance scheme that injects arbitrary energy functions at the clean-sample level, avoiding the first-order approximation errors while requiring no auxiliary networks. Beyond planning, we leverage the same diffusion model as a controllable scenario generator to create realistic long-tail driving interactions for closed-loop evaluation. Through inference-time guidance, selected agents are steered toward safety-critical behaviors, including aggressive cut-ins, lead-vehicle braking, and combined longitudinal-lateral interactions, while preserving realistic traffic behaviors. Evaluated in closed-loop nuPlan simulations with independent black-box planners, the generated scenarios expose failure modes that remain hidden under standard benchmarks. Although the SSDS-based planner achieves stronger nominal performance, it experiences larger degradation under these challenging scenarios, demonstrating that benchmark superiority does not necessarily translate to robustness. These results demonstrate that a single learned traffic prior can simultaneously improve motion planning and provide a realistic framework for systematic planner robustness evaluation.

Chinese Translation

扩散概率模型能够捕捉驾驶场景中联合未来轨迹的多模态、丰富交互的分布。我们展示了一个预训练的扩散交通模型可以在自动驾驶开发循环中发挥两个互补的角色：作为自我运动规划器，以及作为可控的安全关键场景生成器，用于对规划器进行压力测试。在规划方面，我们引入了一种单流双流（Single-Stream Dual-Stream, SSDS）扩散变换器解码器，通过联合注意力而非后期交叉注意力融合场景上下文，从而提高了在nuPlan上的闭环性能。我们进一步提出了带能量的解耦退火后验采样（Decoupled Annealing Posterior Sampling with Energy, DAPSE），这是一种无训练的引导方案，在干净样本级别注入任意能量函数，避免了一阶近似误差，同时不需要辅助网络。除了规划，我们利用相同的扩散模型作为可控场景生成器，创建现实的长尾驾驶交互以进行闭环评估。通过推理时引导，选定的代理被引导向安全关键行为，包括激进的插入、前车刹车以及纵向-横向的组合交互，同时保持现实的交通行为。在与独立黑箱规划器的闭环nuPlan仿真中评估，生成的场景暴露出在标准基准下隐藏的失败模式。尽管基于SSDS的规划器在名义性能上表现更强，但在这些具有挑战性的场景下却经历了更大的性能下降，表明基准的优越性并不一定转化为鲁棒性。这些结果表明，单一学习的交通先验可以同时改善运动规划，并为系统的规划器鲁棒性评估提供现实的框架。

View on arXiv Download PDF AI Translation

cs.CV / 46 / 2609.04939

LensStyle: Learning the Optical Aesthetics for Controllable Stylized Lens Effect Rendering

LensStyle：学习可控风格化镜头效果渲染的光学美学

Huang, Yachuan, Xiao, Liwen, Shen, Liao, Wang, Qiwen, Sun, Huiqiang, Pan, Zhiyu, Cao, Zhiguo

Abstract

The visual aesthetics of photographs are deeply influenced by lens characteristics such as aperture shape, optical vignetting and optical diffraction, which together define a camera's unique optical style. Existing lens effect rendering methods primarily focus on accurately simulating the blur transition from small to large apertures but overlook the stylistic aspects of lens effects. As a result, they fail to produce diverse bokeh effects under large apertures or capture distinctive photographic phenomena such as starbursts that emerge under small apertures. In this work, we introduce LensStyle, a unified framework for controllable stylized lens effect rendering that explicitly models lens aesthetics through joint continuous-discrete control. Our model incorporates a Dual-Path Controller that disentangles continuous optical parameter modulation (e.g., focus distance and blur strength) from discrete lens-style conditioning (e.g., circular, polygonal, donut, cat-eye, and starburst effects), enabling fine-grained, interpretable, and physically grounded lens manipulation within a single unified framework. To support model training, we curate a comprehensive MultiLens dataset containing multi-lens image pairs synthesized under real optical constraints. Extensive experiments demonstrate that LensStyle achieves superior realism, controllability, and aesthetic quality compared with existing lens effect rendering approaches and diffusion-based image editing models, advancing computational photography toward multiple-lens-style simulation.

Chinese Translation

照片的视觉美学受到镜头特性如光圈形状、光学渐晕和光学衍射的深刻影响，这些特性共同定义了相机独特的光学风格。现有的镜头效果渲染方法主要关注从小光圈到大光圈的模糊过渡的准确模拟，但忽视了镜头效果的风格化方面。因此，它们未能在大光圈下产生多样的散景效果，或捕捉在小光圈下出现的独特摄影现象，如星芒。在本研究中，我们介绍了LensStyle，一个可控风格化镜头效果渲染的统一框架，通过联合连续-离散控制显式建模镜头美学。我们的模型结合了一个双路径控制器，将连续光学参数调节（例如，焦距和模糊强度）与离散镜头风格条件（例如，圆形、多边形、甜甜圈、猫眼和星芒效果）解耦，从而在一个统一框架内实现细粒度、可解释和物理基础的镜头操作。为了支持模型训练，我们整理了一个全面的MultiLens数据集，包含在真实光学约束下合成的多镜头图像对。大量实验表明，LensStyle在现实感、可控性和美学质量上优于现有的镜头效果渲染方法和基于扩散的图像编辑模型，推动了计算摄影向多镜头风格模拟的发展。

View on arXiv Download PDF AI Translation

cs.CV / 47 / 2609.04942

Learning 3D Editing without Paired Supervision via Generative Prior Distillation

通过生成先验蒸馏实现无配对监督的3D编辑学习

Wen, Hao, Yun, Weibin, Fan, Hongxing, Lu, Haotian, Chen, Rui, Huang, Zehuan, Sheng, Lu

Abstract

Instruction-guided 3D editing is essential for interactive content creation, yet it faces a significant bottleneck: the severe scarcity of high-quality paired training data. Existing approaches attempt to bypass this by either relying on slow test-time optimization or training on pseudo-pairs constructed via complex pipelines, which often introduce structural drift and geometric artifacts. In this paper, we propose a novel framework that learns feed-forward 3D editing without paired 3D supervision via Generative Prior Distillation. Instead of relying on ground-truth 3D pairs, our core idea is to distill visual, semantic, and geometric knowledge from powerful foundation models directly into a 3D editing model. Specifically, through a differentiable rendering pipeline, we supervise the 3D representation using two complementary signals: a 2D visual prior from an image editing model at the main editing view, and a semantic prior from a Vision-Language Model at novel views to ensure strict instruction following and source identity preservation. Crucially, to address the geometric collapse and multi-view inconsistencies inherent in 2D projection supervision, we introduce a 3D-aware Distribution Matching regularization. Acting as a geometric prior, this term operates in the 3D latent space, constraining the edited output to remain within the manifold of realistic 3D assets defined by a pretrained image to 3D teacher model. Extensive experiments demonstrate that our method achieves superior instruction fidelity and cross-view consistency, significantly outperforming state-of-the-art baselines. Our project is available at: https://github.com/thiamine128/PriorEdit3D.

Chinese Translation

指导性3D编辑对于互动内容创作至关重要，但面临着一个显著瓶颈：高质量配对训练数据的严重匮乏。现有方法试图通过依赖缓慢的测试时优化或在复杂管道中构建伪配对来绕过这一问题，这往往会引入结构漂移和几何伪影。本文提出了一种新颖的框架，通过生成先验蒸馏学习无配对3D监督的前馈3D编辑。我们的核心思想是直接从强大的基础模型中提取视觉、语义和几何知识，而不是依赖真实的3D配对。具体而言，通过可微渲染管道，我们使用两个互补信号对3D表示进行监督：来自图像编辑模型的主编辑视图的2D视觉先验，以及来自视觉-语言模型的新视图的语义先验，以确保严格遵循指令和源身份的保留。关键是，为了解决2D投影监督中固有的几何崩溃和多视图不一致性，我们引入了一种3D感知分布匹配正则化。作为几何先验，该项在3D潜在空间中操作，约束编辑输出保持在由预训练的图像到3D教师模型定义的真实3D资产流形内。大量实验表明，我们的方法在指令保真度和跨视图一致性方面表现优越，显著超越了最先进的基线。我们的项目可在以下链接获取：https://github.com/thiamine128/PriorEdit3D。

View on arXiv Download PDF AI Translation

cs.CV / 48 / 2609.04947

MCPO: Modality-Contrastive Preference Optimization for Multimodal Chain-of-Thought Compression

MCPO：用于多模态思维链压缩的模态对比偏好优化

Yang, Guangheng, Ni, Zhenliang, Wu, Zhenkai, Shu, Han, Feng, Juan, Yang, Wenming, Hu, Jie

Abstract

Recently, multimodal large-scale reasoning models have demonstrated remarkable capabilities in solving complex tasks through long Chains-of-Thought (M-CoT). However, excessively long reasoning trajectories incur substantial computational costs and significant KV-cache pressure. Existing CoT compression and alignment paradigms mainly rely on static rules or single-dimensional preferences, lacking fine-grained cross-modal constraints; as a result, they are prone to inducing visual laziness and hallucinatory reasoning. To address these issues, we propose Modality-Contrastive Preference Optimization (MCPO), a highly sample-efficient two-stage length-compression method that requires fewer than 900 training samples. In the compression stage, we introduce a step-level Normalized Cross-Modal Mutual Information (NCMI) pruning algorithm, which automatically identifies and removes visual-independent reasoning steps by comparing the reasoning discrepancies between with-image and no-image contexts. This significantly reduces redundancy and hallucinatory content in the reasoning chains. In the alignment stage, the model first undergoes supervised fine-tuning to achieve domain-adaptive initialization, followed by optimization using an asymmetric multimodal length-controlled preference loss. This objective adopts a highly nonlinear odds-ratio formulation that provides steep gradients in the with-image context to reinforce length constraints for preferred trajectories, while applying a scaled, flat-gradient linear difference in the no-image context to maintain modality consistency, thereby achieving stable cross-modal preference alignment. Extensive experiments on mainstream base models such as Qwen3-VL-Thinking show that our method can reduce CoT length by up to 69.5% and achieve up to 3.34x end-to-end inference speedup while preserving original accuracy.

Chinese Translation

近年来，多模态大规模推理模型在通过长思维链（M-CoT）解决复杂任务方面展现出了显著的能力。然而，过长的推理轨迹会导致巨大的计算成本和显著的KV缓存压力。现有的思维链压缩和对齐范式主要依赖于静态规则或单一维度的偏好，缺乏细粒度的跨模态约束；因此，它们容易导致视觉懒惰和幻觉推理。为了解决这些问题，我们提出了模态对比偏好优化（MCPO），这是一种高样本效率的两阶段长度压缩方法，训练样本少于900个。在压缩阶段，我们引入了一种步级归一化跨模态互信息（NCMI）剪枝算法，通过比较有图像和无图像上下文之间的推理差异，自动识别并移除视觉独立的推理步骤。这显著减少了推理链中的冗余和幻觉内容。在对齐阶段，模型首先经过监督微调以实现领域自适应初始化，然后使用不对称多模态长度控制偏损失进行优化。该目标采用高度非线性的赔率比公式，在有图像上下文中提供陡峭的梯度，以强化偏好的轨迹长度约束，同时在无图像上下文中应用缩放的平坦梯度线性差异，以保持模态一致性，从而实现稳定的跨模态偏好对齐。在主流基础模型如Qwen3-VL-Thinking上的大量实验表明，我们的方法可以将思维链长度减少多达69.5%，并在保持原始准确性的同时实现高达3.34倍的端到端推理加速。

View on arXiv Download PDF AI Translation

cs.CV / 49 / 2609.04948

VICAL: Vicinal Consistency Alignment for Long-Tailed Visual Recognition

VICAL：用于长尾视觉识别的邻域一致性对齐

Zhu, Jiangang, Wang, Zheng, Zhu, Bin, Chen, Yi-Ping Phoebe, Chen, Jingjing

Abstract

Multi-expert models have become the dominant paradigm for long-tailed learning, largely attributed to their presumed ability to benefit from expert diversity. However, we revisit this central assumption and reveal that diversity induced by logit adjustment or explicit regularizers does not guarantee better ensemble accuracy. Our work suggests that multi-expert models benefit more from variance reduction than diversity maximization. We introduce \textbf{VICAL}, a \textbf{VI}cinal \textbf{C}onsistency \textbf{AL}ignment framework that improves long-tailed recognition not by enforcing expert diversity, but by reducing prediction variance. Specifically, our approach comprises two key components: Self-Consistency Learning and Deep Ensemble Distillation. Self-Consistency Learning discourages reliance on unstable high-frequency information, smoothing the local loss landscape and mitigating overfitting, especially for tail classes. Deep Ensemble Distillation promotes cross-expert low-frequency semantic agreement using a low-resolution view, thereby sidestepping optimization conflicts with established knowledge. Extensive experiments on CIFAR-LT, ImageNet-LT, and iNaturalist 2018 show that VICAL consistently outperforms state-of-the-art methods, validating the effectiveness of our consistency-driven design. Our code is available at \href{https://github.com/FlamieZhu/Vicinal-Consistency-Alignment}{VICAL}.

Chinese Translation

多专家模型已成为长尾学习的主流范式，这在很大程度上归因于它们被认为能够从专家多样性中受益。然而，我们重新审视了这一核心假设，揭示了通过对数调整或显式正则化所引起的多样性并不保证更好的集成准确性。我们的研究表明，多专家模型更能从方差减少中获益，而非多样性最大化。我们提出了 extbf{VICAL}，一个 extbf{VI}cinal extbf{C}onsistency extbf{AL}ignment框架，它通过减少预测方差而非强制专家多样性来改善长尾识别。具体而言，我们的方法包含两个关键组件：自一致性学习（Self-Consistency Learning）和深度集成蒸馏（Deep Ensemble Distillation）。自一致性学习抑制对不稳定高频信息的依赖，平滑局部损失景观并减轻过拟合，特别是对于尾部类别。深度集成蒸馏通过低分辨率视图促进跨专家的低频语义一致性，从而避免与既有知识的优化冲突。在CIFAR-LT、ImageNet-LT和iNaturalist 2018上的大量实验表明，VICAL始终优于最先进的方法，验证了我们基于一致性驱动设计的有效性。我们的代码可在 extbf{VICAL}的GitHub页面上获取。

View on arXiv Download PDF AI Translation

cs.CV / 50 / 2609.04958

MINT: A Unified Model for World-Space Camera and Hand Motion Estimation from Scalable Egocentric Pipeline Supervision

MINT：一种统一模型用于从可扩展的自我中心管道监督中估计世界空间相机和手部运动

Zhu, Zijie, Cai, Weiren, Wang, Yizhou, Yang, Zhenjie, Liu, Yide, Chen, Jiahao, He, Guanqi

Abstract

Recovering camera and hand motion in world coordinates from egocentric video is a key capability for activity understanding, robot learning, and augmented reality. Existing systems typically decompose this problem into separate stages for camera motion, depth, hand reconstruction, and trajectory refinement, resulting in substantial computational overhead and preventing the joint modeling of camera and hand motion. We introduce MINT (Minting IN-the-Wild Trajectories), the first foundation model that directly produces complete world-space two-hand trajectories from ego-centric RGB video. From a single shared spatiotemporal video representation, MINT jointly predicts the camera trajectory, camera-frame hand states, and per-frame hand presence, and then produces world-space hand motion via explicit coordinate transformations. Training such a model at scale is challenging, since paired world-space camera and hand annotations are scarce. We therefore develop an open-source labeling EGOPIPELINE that converts large collections of public egocentric videos into structured camera-and-hand trajectory supervision. MINT is first pretrained on these large-scale pseudo-labels and then fine-tuned on a small set of high-quality joint annotations. Across public benchmarks, MINT achieves [xxx] improvement in world-space hand trajectory accuracy, [xxx] improvement in camera trajectory estimation, and [xxx] faster end-to-end trajectory generation than the labeling pipeline, while generalizing zero-shot to unseen egocentric datasets. We release the model, training and inference code, labeling pipeline, and a curated 1,021-hour egocentric trajectory dataset.

Chinese Translation

从自我中心视频中恢复世界坐标系下的相机和手部运动是活动理解、机器人学习和增强现实的关键能力。现有系统通常将此问题分解为相机运动、深度、手部重建和轨迹优化等独立阶段，导致了显著的计算开销，并阻碍了相机和手部运动的联合建模。我们提出了MINT（Minting IN-the-Wild Trajectories），这是第一个直接从自我中心RGB视频生成完整世界空间双手轨迹的基础模型。MINT从单一共享的时空视频表示中，联合预测相机轨迹、相机帧手部状态和每帧手部存在性，然后通过显式坐标变换生成世界空间手部运动。在大规模训练这样的模型是具有挑战性的，因为配对的世界空间相机和手部注释稀缺。因此，我们开发了一个开源标注工具EGOPIPELINE，将大量公共自我中心视频转换为结构化的相机和手部轨迹监督。MINT首先在这些大规模伪标签上进行预训练，然后在一小组高质量的联合注释上进行微调。在公共基准测试中，MINT在世界空间手部轨迹精度上实现了[xxx]的提升，在相机轨迹估计上实现了[xxx]的提升，并且在端到端轨迹生成上比标注管道快[xxx]倍，同时在未见过的自我中心数据集上实现了零样本泛化。我们发布了该模型、训练和推理代码、标注管道以及一个精心策划的1,021小时自我中心轨迹数据集。

View on arXiv Download PDF AI Translation

cs.CV / 51 / 2609.04965

ARC-Loc: Leveraging Azimuthal Ray Convergence as a Geometric Cue for Direct Cross-View Localization

ARC-Loc：利用方位射线收敛作为直接跨视图定位的几何线索

Kim, Hyeongsik, Kim, Mincheol, Moon, Heejoon, Hong, Je Hyeong

Abstract

Cross-view localization (CVL) estimates the pose of a ground image by matching it to a geo-referenced satellite image. To bridge the extreme viewpoint gap, mainstream pipelines rely on Bird's-Eye-View (BEV) transformations or 2D-to-3D lifting. However, deriving 3D structures from a single ground image is fundamentally ill-posed, causing these methods to endure geometric distortions and computational costs during 3D lifting or BEV projection. Furthermore, relying on external depth foundation models to resolve this introduces latency and remains susceptible to noisy predictions. In this work, we present a different approach inspired by a human navigation technique called resection, that can perform direct ground to satellite image matching and localization without relying on external depth foundation models. The key insights of our method are that (i) ground keypoints can be translated into azimuthal rays on the satellite map, and (ii) these rays ideally converge at the user location. Exploiting this geometric constraint through direct line-to-point correspondences, we introduce a minimal Azimuthal Ray Convergence (ARC) solver to identify the intersection, alongside an ARC loss to optimize the matching network. By eliminating dependencies on computationally heavy BEV transformations and external depth foundation models, our approach achieves faster, memory-efficient inference, while its explicit feature matching ensures straightforward compatibility with existing frameworks. Experiments on VIGOR and KITTI demonstrate that ARC-Loc maintains competitive localization accuracy compared to recent approaches, highlighting its practicality.

Chinese Translation

跨视图定位（CVL）通过将地面图像与地理参考的卫星图像进行匹配来估计地面图像的姿态。为了弥补极端视角差距，主流流程依赖于鸟瞰图（BEV）变换或2D到3D的提升。然而，从单一地面图像推导3D结构在本质上是一个病态问题，这导致这些方法在3D提升或BEV投影过程中遭受几何失真和计算成本的增加。此外，依赖外部深度基础模型来解决这一问题引入了延迟，并且仍然容易受到噪声预测的影响。在本研究中，我们提出了一种不同的方法，灵感来源于一种称为重设的人工导航技术，该技术可以在不依赖外部深度基础模型的情况下直接进行地面与卫星图像的匹配和定位。我们方法的关键见解是：（i）地面关键点可以转换为卫星地图上的方位射线，以及（ii）这些射线理想情况下在用户位置收敛。通过直接的点对线对应关系利用这一几何约束，我们引入了一种最小化的方位射线收敛（ARC）求解器来识别交点，同时引入ARC损失以优化匹配网络。通过消除对计算密集型BEV变换和外部深度基础模型的依赖，我们的方法实现了更快、更节省内存的推理，同时其显式特征匹配确保了与现有框架的简单兼容性。在VIGOR和KITTI上的实验表明，ARC-Loc在定位精度上与近期方法保持竞争力，突显了其实用性。

View on arXiv Download PDF AI Translation

cs.CV / 52 / 2609.04976

RefDiT: Local Attribute Guidance in Reference-Based Image Generation

RefDiT：基于参考的图像生成中的局部属性引导

Mishra, Rameshwar, Karanam, Srikrishna, Subramanyam, A V

Abstract

Personalization models generate new images guided by a few subject references, while style transfer methods aim to produce images aligned with a global style derived from a reference image. Recent approaches perform well when the reference image contains a single object, effectively capturing a global style that encompasses all implicit attributes. However, when applied to complex real-world scenes containing multiple objects with distinct attribute characteristics, these methods, due to their global-level guidance, fail to localize relevant elements in the reference image. The global guidance restricts their ability to generate new images based on the local attributes in the reference image. Moreover, existing methods typically employ a single identifier token to capture all details from the reference, resulting in a lack of individual, attribute-level control. Motivated by these limitations, we propose RefDiT, a novel framework for reference-guided image generation. RefDiT takes as input a reference image, a text prompt, and an optional user-provided guidance context. RefDiT employs local region guidance using the attributes of local elements. It constructs an attribute-aware conditioning signal from the reference image by performing attribute-level decomposition of the identifier token and performs context adjustment in the inference prompt to train low-rank adapter (LoRA) blocks of a diffusion transformer (DiT)-based generative model. RefDiT learns the correspondence between identifier tokens and local regions in the reference image, enabling more effective local guidance.

Chinese Translation

个性化模型通过少量主题参考生成新图像，而风格迁移方法旨在生成与源自参考图像的全局风格一致的图像。近期的方法在参考图像包含单一对象时表现良好，有效捕捉到涵盖所有隐含属性的全局风格。然而，当应用于包含多个具有不同属性特征的对象的复杂现实场景时，由于其全局级别的引导，这些方法未能在参考图像中定位相关元素。全局引导限制了它们基于参考图像中的局部属性生成新图像的能力。此外，现有方法通常使用单个标识符令牌来捕捉参考中的所有细节，导致缺乏对个别属性级别的控制。基于这些局限性，我们提出了RefDiT，一种用于参考引导图像生成的新框架。RefDiT以参考图像、文本提示和可选的用户提供的引导上下文作为输入。RefDiT采用局部区域引导，利用局部元素的属性。通过对标识符令牌进行属性级分解，RefDiT从参考图像构建属性感知的条件信号，并在推理提示中进行上下文调整，以训练基于扩散变换器（DiT）的生成模型的低秩适配器（LoRA）块。RefDiT学习标识符令牌与参考图像中局部区域之间的对应关系，从而实现更有效的局部引导。

View on arXiv Download PDF AI Translation

cs.CV / 53 / 2609.04984

Temporal Residual Neural Radiance Fields for Monocular Video Dynamic Human Body Reconstruction

用于单目视频动态人体重建的时序残差神经辐射场

Du, Tianle, Wang, Jie, Xie, Xiaolong, Li, Wei, Su, Pengxiang, Liu, Jie

Abstract

In the field of computer vision and graphics, high-quality reconstruction of the human body in static scenes has been achieved in recent years by a single multilayer perceptron (MLP) in a number of approaches. However, MLPs have capacity limitations, requiring substantial training time and computational resources for dynamic scene reconstruction. And the quality of reconstruction is significantly constrained. This paper proposes a method for effectively processing complex spatiotemporal signals in dynamic scene human 3D modeling. The proposed method uses Temporal Residual Neural Radiance Fields to achieve novel view rendering and new pose synthesis of human bodies.To address the problem of representing temporal signals in video sequences, we construct a temporal residual field which is not related to the MLP architecture. Secondly, to improve reconstruction efficiency, we propose an integrated approach that reduces trainable parameters and accelerates rendering, thereby enhancing the network's feature representation capability. Finally, we design a multi-dimensional loss function to accurately measure the loss between predicted and actual spatial pixel values. The experimental results show that our proposed approach improves the peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) accuracy metrics compared to the latest representative methods. It maintains similar accuracy to Anim-NeRF and Neural Body while achieving a nearly 780-fold increase in time efficiency.

Chinese Translation

在计算机视觉和图形学领域，近年来通过单个多层感知器（MLP）在静态场景中实现了高质量的人体重建。然而，MLP存在容量限制，动态场景重建需要大量的训练时间和计算资源，重建质量受到显著制约。本文提出了一种有效处理动态场景中复杂时空信号的人体三维建模方法。所提方法使用时序残差神经辐射场（Temporal Residual Neural Radiance Fields）实现人体的新视角渲染和新姿态合成。为了解决视频序列中时序信号表示的问题，我们构建了一个与MLP架构无关的时序残差场。其次，为了提高重建效率，我们提出了一种集成方法，减少可训练参数并加速渲染，从而增强网络的特征表示能力。最后，我们设计了一个多维损失函数，以准确测量预测的空间像素值与实际值之间的损失。实验结果表明，与最新的代表性方法相比，我们提出的方法在峰值信噪比（PSNR）和结构相似性指数（SSIM）准确性指标上有所提升。它在保持与Anim-NeRF和Neural Body相似的准确性的同时，实现了近780倍的时间效率提升。

View on arXiv Download PDF AI Translation

cs.CV / 54 / 2609.05028

Compositional Reward Models for Conditional Medical Image Generation

用于条件医学图像生成的组合奖励模型

Tyagi, Aayush Kumar, P., Prathosh A., Mausam

Abstract

Acquiring high quality annotated medical image data is critical for training deep learning models; however, annotation is expensive, time consuming, and requires domain expertise. Conditional diffusion models, such as ControlNet, offer an alternative by generating images conditioned on semantic masks and text. However, existing approaches fail to capture fine grained properties (e.g., intensity and texture), as well as semantic consistency expected by domain experts, limiting their effectiveness for downstream tasks. Recent attempts to address these issues using reinforcement learning fine-tuning remain limited due to the reliance on a single scalar reward, which conflates diverse failure modes and provides weak corrective signals. We propose PRISM, a Compositional Reward Model (CRM) framework for conditional medical image generation. Instead of assigning a single reward, we decompose image quality into verifier grounded stages, each evaluating a distinct aspect of correctness from fine to coarse properties, including low level attributes (intensity and texture), structural alignment with conditioning inputs, and high level semantic fidelity. These stage wise rewards are composed through a Hierarchical Constrained Propagation (HCP) mechanism that enforces a fine to coarse notion of correctness, ensuring that lower level deficiencies are resolved before higher level rewards are accrued, preventing easier objectives from masking critical failures. We evaluate PRISM across three datasets spanning diverse medical imaging tasks: PanNuke (multi-class cell segmentation), CeDeM (villi/crypt detection and measurement), and ISIC (skin lesion classification). Training downstream models with data generated by PRISM yields improvements over closest baselines, including a 2.3% increase in mDice on PanNuke, a 8.5% reduction in Mean Relative Error (MRE) on CeDeM, and increases ISIC F1 by 5.9%.

Chinese Translation

获取高质量的标注医学图像数据对于训练深度学习模型至关重要；然而，标注过程既昂贵又耗时，并且需要领域专业知识。条件扩散模型，如 ControlNet，通过基于语义掩码和文本生成图像提供了一种替代方案。然而，现有方法未能捕捉到细粒度属性（例如，强度和纹理）以及领域专家所期望的语义一致性，限制了其在下游任务中的有效性。最近尝试通过强化学习微调来解决这些问题的努力仍然有限，因为它们依赖于单一的标量奖励，这混淆了多样的失败模式，并提供了较弱的纠正信号。我们提出了 PRISM，一个用于条件医学图像生成的组合奖励模型（CRM）框架。我们不是分配单一奖励，而是将图像质量分解为基于验证者的阶段，每个阶段评估从细到粗的正确性不同方面，包括低级属性（强度和纹理）、与条件输入的结构对齐以及高级语义保真度。这些阶段性奖励通过一种层次约束传播（HCP）机制进行组合，该机制强制实施从细到粗的正确性概念，确保在获得更高级奖励之前解决低级缺陷，防止较简单的目标掩盖关键失败。我们在三个涵盖多样医学成像任务的数据集上评估了 PRISM：PanNuke（多类细胞分割）、CeDeM（绒毛/隐窝检测和测量）和 ISIC（皮肤病变分类）。使用 PRISM 生成的数据训练下游模型在最接近的基线之上取得了改进，包括在 PanNuke 上 mDice 提高了 2.3%，在 CeDeM 上平均相对误差（MRE）降低了 8.5%，以及 ISIC F1 提高了 5.9%。

View on arXiv Download PDF AI Translation

cs.CV / 55 / 2609.05038

PuTR-CouT: Counting-by-Tracking in Camera-Trap Image Sequences

PuTR-CouT：基于跟踪的相机捕捉图像序列计数

Cunha, Fagner, Colonna, Juan G., Santos, Eulanda M. dos

Abstract

Species identification in camera trap images has been widely studied, but key ecological modeling tasks such as species abundance or density estimation also require counting individual animals. However, the lack of counting labels in most datasets and low frame rates (typically ~1 frame per second) make sequence-level tracking and count estimation particularly challenging. In this work, we present PuTR-CouT, a counting-by-tracking framework built on a transformer-based learned association mechanism for sequence-level animal counting in camera trap images. To address the scarcity of annotated tracking data, we generate synthetic training data by exploiting structural priors, such as static backgrounds and short temporal bursts, to heuristically create pseudo-tracking labels in a weakly supervised manner. The resulting tracker associates detections across frames, using these tracks to estimate per-species counts. We also refine the MaxBoxCount heuristic used by the top solutions of the iWildCam 2021 challenge as a strong baseline, setting the highest score reported to date. When evaluated on the iWildCam 2021 benchmark, our framework PuTR-CouT delivers competitive counting results compared to the improved MaxBoxCount, with the added capability of multi-species predictions and track-level verification.

Chinese Translation

相机捕捉图像中的物种识别已得到广泛研究，但关键生态建模任务如物种丰度或密度估计也需要对个体动物进行计数。然而，大多数数据集中缺乏计数标签，且低帧率（通常约为每秒1帧）使得序列级跟踪和计数估计特别具有挑战性。在本研究中，我们提出了PuTR-CouT，一个基于变换器（transformer）学习关联机制的基于跟踪的计数框架，用于相机捕捉图像中的序列级动物计数。为了解决标注跟踪数据稀缺的问题，我们通过利用结构先验（如静态背景和短暂时间突发）生成合成训练数据，以启发式方式在弱监督下创建伪跟踪标签。生成的跟踪器在帧之间关联检测，利用这些轨迹估计每种物种的计数。我们还改进了iWildCam 2021挑战赛顶尖解决方案中使用的MaxBoxCount启发式方法，作为强基线，设定了迄今为止报告的最高分数。在iWildCam 2021基准测试中，我们的框架PuTR-CouT与改进的MaxBoxCount相比，提供了具有竞争力的计数结果，并具备多物种预测和轨迹级验证的附加能力。

View on arXiv Download PDF AI Translation

cs.CV / 56 / 2609.05049

Efficient Multi-Timescale Event Representations for Feed-Forward Object Detection

高效的多时间尺度事件表示用于前馈物体检测

Lundell, Fredrik, Forssen, Per-Erik, Wadenbäck, Mårten, Lundmark, Astrid

Abstract

Autonomous systems require robust low-latency perception under rapidly changing scene dynamics and challenging illumination. In event cameras object detection commonly relies on recurrent architectures to accumulate sparse temporal information over time. This work investigates how temporal information can be encoded directly within the event representation. We propose a confidence-normalized continuous multi-timescale representation based on logarithmic B-spline temporal encoding together with a geometry-aware local confidence mechanism that exploits the spatial structure of event generation. Using a fixed feed-forward EventCenterNet detector, we show that the proposed representations consistently outperform the compact CSTR representation on PEDRo and Gen1 datasets. We further introduce a recursive exponential-polynomial approximation that enables efficient event-by-event updates while largely preserving detection performance. These results demonstrate that carefully designed event representations can capture a substantial portion of the temporal information learned through recurrent temporal modeling, providing a promising foundation for efficient feed-forward, event-driven, and future neuromorphic object detection.

Chinese Translation

自主系统在快速变化的场景动态和挑战性的光照条件下，需要具备稳健的低延迟感知能力。在事件相机中，物体检测通常依赖递归架构来随时间积累稀疏的时间信息。本研究探讨了如何直接在事件表示中编码时间信息。我们提出了一种基于对数B样条时间编码的置信度归一化连续多时间尺度表示，并结合了一种几何感知的局部置信度机制，以利用事件生成的空间结构。使用固定的前馈EventCenterNet检测器，我们展示了所提出的表示在PEDRo和Gen1数据集上始终优于紧凑的CSTR表示。我们进一步引入了一种递归指数多项式近似方法，使得能够高效地逐事件更新，同时在很大程度上保持检测性能。这些结果表明，精心设计的事件表示能够捕捉通过递归时间建模学习到的相当大一部分时间信息，为高效的前馈、事件驱动和未来的神经形态物体检测提供了有希望的基础。

View on arXiv Download PDF AI Translation

cs.CV / 57 / 2609.05066

Adaptive Multi-Granularity Temporal Modeling for Weakly Supervised Video Anomaly Detection

自适应多粒度时间建模用于弱监督视频异常检测

Li, Changyi, Xiao, Yu

Abstract

As the scale of video surveillance data outpaces manual annotation capacities, weakly supervised video anomaly detection (WSVAD) has emerged as a critical research frontier. Most existing approaches formulate WSVAD within a Multiple Instance Learning (MIL) framework that relies on rigid, hand-crafted temporal priors to supervise anomaly scoring. However, such formulations exhibit limited adaptability to the wide variation in anomaly durations and temporal dynamics observed in real-world videos, often leading to unstable or unreliable snippet-level predictions. To address this limitation, we propose an adaptive temporal modeling framework for WSVAD that explicitly accounts for variations in video dynamics across multiple temporal granularities. First, we introduce a Temporal Refinement Module (TRM) that leverages dynamic positional encoding and a learnable class token to model long-range temporal dependencies while distilling a stable global video-level representation. Second, to capture anomalous events with varying frequency and duration, we develop an adaptive Event Segmentation Module (ESM) that identifies event boundaries through temporal discontinuity analysis and aggregates snippet features into discriminative event-level representations. Finally, for snippet-level and event-level predictions, we propose an adaptive similarity-based fusion strategy that dynamically integrates anomaly scores into video-level predictions, replacing fixed top-k aggregation heuristics with global semantic relevance. Extensive experiments on two benchmarks demonstrate that the proposed framework consistently outperforms state-of-the-art methods.

Chinese Translation

随着视频监控数据的规模超出人工标注能力，弱监督视频异常检测（WSVAD）已成为一个关键的研究前沿。现有的大多数方法在多实例学习（MIL）框架内构建WSVAD，依赖于刚性、手工设计的时间先验来监督异常评分。然而，这种构建方式在面对现实视频中异常持续时间和时间动态的广泛变化时表现出有限的适应性，常常导致片段级预测的不稳定或不可靠。为了解决这一限制，我们提出了一种自适应时间建模框架用于WSVAD，明确考虑了跨多个时间粒度的视频动态变化。首先，我们引入了一个时间细化模块（Temporal Refinement Module, TRM），利用动态位置编码和可学习的类别标记来建模长距离时间依赖关系，同时提炼出稳定的全局视频级表示。其次，为了捕捉频率和持续时间各异的异常事件，我们开发了一个自适应事件分割模块（Event Segmentation Module, ESM），通过时间不连续性分析识别事件边界，并将片段特征聚合为具有区分性的事件级表示。最后，对于片段级和事件级预测，我们提出了一种自适应基于相似性的融合策略，动态地将异常评分整合到视频级预测中，用全局语义相关性替代固定的 top-k 聚合启发式方法。在两个基准数据集上的大量实验表明，所提出的框架在性能上始终优于最先进的方法。

View on arXiv Download PDF AI Translation

cs.CV / 58 / 2609.05070

MultiAttenGastro: Multi-Dimensional Attention Augmentation for Gastrointestinal Endoscopy Classification

MultiAttenGastro：用于胃肠内窥镜分类的多维注意力增强

Devarajan, Sadhana, Chandaliya, Praveen Kumar, Shah, Dhruvin Jashvant Kumar, Upla, Kishor, Raja, Kiran

Abstract

Automated gastrointestinal (GI) endoscopy classification requires models that generalize across diverse modalities and class distributions, often far from natural-image pretraining. We propose MultiAttenGastro, a plug-and-play attention framework with parallel 1-D channel, 2-D spatial, and 3-D contextual heads, and present the first systematic cross-dataset evaluation across eight CNN and transformer backbones on five public GI datasets (80 backbone--dataset runs). We find that attention effectiveness is not universal but tracks the representational gap between ImageNet features and the target distribution: MultiAttenGastro improves 6 of 8 backbones on Kvasir-Capsule (14-class WCE, large gap; best macro F1 98.33\%), is uniformly negative on the small-gap Kvasir-v2 benchmark (0/8), and shows mixed outcomes on datasets with intermediate gap. Five-seed ablation on the strongest case (Kvasir-Capsule, ConvNeXt-Tiny) shows this improvement is directionally consistent, but not statistically decisive (paired $t$: $p=0.47$; Wilcoxon: $p=0.63$), and that individual attention heads are not uniformly beneficial in isolation only their combination yields a positive mean effect. Centered Kernel Alignment (CKA) analysis links this pattern to representational redundancy: low inter-head CKA under large domain gaps coincides with the framework's only consistent gains, while high redundancy under small gaps coincides with its losses. We report these results, including the non-significant margins, as evidence for when and why multi-dimensional attention helps GI endoscopy classification, rather than as a claim that MultiAttenGastro is a strictly superior architectural choice.

Chinese Translation

自动化的胃肠（GI）内窥镜分类需要能够在多样化的模态和类别分布中泛化的模型，这通常远离自然图像的预训练。我们提出了MultiAttenGastro，一个即插即用的注意力框架，具有并行的1维通道、2维空间和3维上下文头，并首次在五个公共GI数据集上对八个CNN和变换器骨干网络进行系统的跨数据集评估（80个骨干网络-数据集运行）。我们发现注意力的有效性并非普遍适用，而是与ImageNet特征与目标分布之间的表征差距相关：MultiAttenGastro在Kvasir-Capsule（14类WCE，大差距；最佳宏F1为98.33%）上改善了8个骨干网络中的6个，在小差距的Kvasir-v2基准上表现均为负（0/8），并在具有中等差距的数据集上显示出混合结果。对最强案例（Kvasir-Capsule，ConvNeXt-Tiny）进行的五次种子消融实验表明，这种改善在方向上是一致的，但在统计上并不决定性（配对$t$：$p=0.47$；Wilcoxon：$p=0.63$），并且单个注意力头在孤立状态下并不总是有利，只有它们的组合才能产生正的平均效果。中心核对齐（CKA）分析将这种模式与表征冗余联系起来：在大领域差距下，低的头间CKA与该框架唯一一致的增益相吻合，而在小差距下的高冗余则与其损失相吻合。我们报告这些结果，包括非显著的边际，作为多维注意力何时以及为何有助于GI内窥镜分类的证据，而不是声称MultiAttenGastro是严格优越的架构选择。

View on arXiv Download PDF AI Translation

cs.CV / 59 / 2609.05091

Training-Free Logical and Structural Anomaly Detection via Calibrated Fusion

无训练的逻辑与结构异常检测通过校准融合

Li, Changyi, Yu, Miao, Dong, Kai, Xiao, Yu

Abstract

Industrial anomaly detection must handle two distinct defect families: structural anomalies, which manifest as local texture corruptions, and logical anomalies, which violate global rules on object count, composition, or arrangement. Existing detectors typically favor one family at the expense of the other. In particular, training-free methods effectively exploit frozen representations but lack an explicit notion of object count, while methods that reason about counts usually rely on category-specific component modeling. We show that counting ability can be introduced into training-free anomaly detection without additional training or part-level supervision. Our key idea is a normal-set calibration that aligns heterogeneous anomaly cues using statistics from normal images, enabling their direct fusion within a unified training-free framework. Built upon this calibration, our detector combines complementary frozen cues to address both logical and structural anomalies. On MVTec-LOCO, our method achieves image-level AUROCs of 89.0 and 95.9 on logical and structural anomalies, respectively, yielding a 92.5 average---the best among training-free detectors in our comparison. It remains competitive with methods requiring network training or part annotations, while its structural variant matches PatchCore on MVTec-AD (99.1 image-AUROC), suggesting that the proposed calibration generalizes beyond logical anomaly detection.

Chinese Translation

工业异常检测必须处理两种不同的缺陷类型：结构异常，表现为局部纹理损坏，以及逻辑异常，违反对象数量、组成或排列的全局规则。现有的检测器通常偏向于某一类缺陷，而牺牲另一类。特别是，无训练的方法有效利用冻结的表示，但缺乏对象数量的明确概念，而关于数量推理的方法通常依赖于特定类别的组件建模。我们展示了在无训练的异常检测中可以引入计数能力，而无需额外的训练或部分级别的监督。我们的关键思想是通过使用正常图像的统计数据来校准正常集，进而对齐异质异常线索，使其能够在统一的无训练框架内直接融合。在此校准的基础上，我们的检测器结合互补的冻结线索，以解决逻辑和结构异常。在 MVTec-LOCO 数据集上，我们的方法在逻辑和结构异常上分别达到了89.0和95.9的图像级 AUROC，平均值为92.5——在我们的比较中是所有无训练检测器中表现最好的。它在与需要网络训练或部分注释的方法的竞争中依然保持竞争力，同时其结构变体在 MVTec-AD 上与 PatchCore 的表现相当（99.1 图像 AUROC），这表明所提出的校准方法在逻辑异常检测之外具有广泛的适用性。

View on arXiv Download PDF AI Translation

cs.CV / 60 / 2609.05114

VoxelFix: Post-Hoc Semantic Correction of Completed 3D Voxel Maps

VoxelFix：已完成3D体素地图的事后语义修正

Sundarasami, Sunesh Praveen Raja, Kim, Taehyoung, Scherer, Johannes, Cotič, Tomaž, Subbiah, Sivasubiramaniam, Greiner, Andreas, Spannaus, Paul, Houben, Sebastian

Abstract

Semantic 3D maps are increasingly constructed automatically for aerial robotics by integrating learned semantic predictions into 3D representations. While this avoids costly manual 3D annotation, errors in the perception and mapping pipeline can persist in the resulting map, reducing its reliability for downstream autonomous tasks. Existing 3D semantic map refinement methods either rely on the original observations, treat occupancy as part of the prediction problem, or apply non-learned local regularization to completed maps. Instead, we study post-hoc semantic correction, asking whether semantic accuracy can be recovered directly from the completed map while keeping its geometry and occupancy fixed. We introduce \method, a graph-based model that corrects voxel labels based on local geometry and neighboring semantic information. To obtain training pairs, we corrupt contiguous regions of annotated OccuFly maps according to class confusions observed in upstream maps. We evaluate \method on completed OccuFly maps generated from predictions of four independently trained 2D segmentation models. \method consistently improves mIoU by 4.23--5.00 percentage points, with gains broadly distributed across the evaluated semantic classes and particularly strong improvements for tree, roof, and wall. Results on an independently reconstructed out-of-distribution aerial scene further suggest that the learned correction can transfer beyond the environments seen during training.

Chinese Translation

语义3D地图正越来越多地通过将学习到的语义预测整合到3D表示中，自动构建用于空中机器人。虽然这避免了昂贵的手动3D标注，但感知和映射流程中的错误可能会在生成的地图中持续存在，从而降低其在下游自主任务中的可靠性。现有的3D语义地图精炼方法要么依赖于原始观测，要么将占用率视为预测问题的一部分，或者对已完成的地图应用非学习的局部正则化。相反，我们研究事后语义修正，探讨是否可以在保持几何形状和占用率不变的情况下，从已完成的地图中直接恢复语义准确性。我们引入了 extit{VoxelFix}，一种基于图的模型，依据局部几何形状和邻近的语义信息修正体素标签。为了获得训练对，我们根据上游地图中观察到的类别混淆，破坏标注的OccuFly地图的连续区域。我们在由四个独立训练的2D分割模型生成的已完成OccuFly地图上评估 extit{VoxelFix}。 extit{VoxelFix}始终将mIoU提高了4.23到5.00个百分点，增益在评估的语义类别中广泛分布，尤其在树木、屋顶和墙壁类别上表现出显著的改善。在一个独立重建的分布外空中场景上的结果进一步表明，学习到的修正可以超越训练期间所见的环境。

View on arXiv Download PDF AI Translation

cs.CV / 61 / 2609.05171

WeAgent-MMGenEdit: A Full-Stack Recipe for Multimodal Agentic Image Generation and Editing

WeAgent-MMGenEdit：多模态智能图像生成与编辑的全栈方案

Zhang, Hui, Liu, Zongkai, Niu, Liqiang, Liu, Juntao, Li, Han, Cao, Zhen, Chen, Wenchao, Zhao, Chengduo, Meng, Fandong

Abstract

Image generation and editing models have advanced rapidly, yet remain unreliable when prompts require external world knowledge. Bounded and long-tail parametric knowledge prevents direct or reason-then-generate approaches from recovering the required facts and visual appearances. Existing agentic generation and editing methods mitigate this limitation with retrieval tools, yet remain constrained by insufficient visual verification, overloaded policy models, and weak integration of retrieved textual and visual evidence. To address these limitations, we present WeAgent-MMGenEdit, a full-stack recipe including a multimodal harness, a scalable data construction pipeline, a comprehensive benchmark, and post-training methods for the agent policy and image backend. We first introduce WeAgent-Harness, a multimodal runtime with persistent evidence management and dedicated verification and integration tools that organize retrieved multimodal evidence into a dense carrier. Upon this, we develop a scalable pipeline for prompt synthesis and agentic trajectory collection, yielding 23K supervised trajectories and 14.7K RL tasks with three-layer verifiable checklists. We further introduce WeBench-MMGenEdit, a bilingual benchmark covering both knowledge-intensive image generation and multi-image editing. Finally, a two-sided post-training recipe based on SFT and RL improves the agent policy and image backend. Together, WeAgent-MMGenEdit enables a 30B-total/3B-active policy to outperform similarly sized policy models and approach the performance of a 1T-parameter agent.

Chinese Translation

图像生成和编辑模型发展迅速，但在需要外部世界知识的提示下仍然不可靠。有限的和长尾的参数知识使得直接或推理后生成的方法无法恢复所需的事实和视觉外观。现有的智能生成和编辑方法通过检索工具缓解了这一限制，但仍然受到视觉验证不足、策略模型过载以及检索的文本和视觉证据整合不力的制约。为了解决这些问题，我们提出了WeAgent-MMGenEdit，一个全栈方案，包括多模态工具、可扩展的数据构建管道、全面的基准测试以及用于智能策略和图像后端的后训练方法。我们首先介绍WeAgent-Harness，一个具有持久证据管理和专用验证与整合工具的多模态运行时，它将检索到的多模态证据组织成一个密集载体。在此基础上，我们开发了一个可扩展的提示合成和智能轨迹收集管道，生成了23K个监督轨迹和14.7K个具有三层可验证检查表的强化学习任务。我们进一步介绍WeBench-MMGenEdit，这是一个涵盖知识密集型图像生成和多图像编辑的双语基准。最后，基于SFT和RL的双向后训练方案改善了智能策略和图像后端。综上所述，WeAgent-MMGenEdit使得一个30B总参数/3B活跃策略超越了同等规模的策略模型，并接近1T参数智能体的性能。

View on arXiv Download PDF AI Translation

cs.CV / 62 / 2609.05174

SMILE: Self-Explainable Multimodal Information Bottleneck for Medical Diagnosis

SMILE：自解释的多模态信息瓶颈用于医学诊断

Yang, Yuqing, Schmatz, Alexander, Ma, Zhaozhao, Choi, Changkyu, Jenssen, Robert, Yu, Shujian

Abstract

Explainability is increasingly seen as a crucial requirement in AI-based medical diagnosis, particularly in safety-critical clinical decision-making. Most existing explainability methods in healthcare operate in a post-hoc manner and are predominantly designed for unimodal data, which limits their applicability in increasingly prevalent multimodal diagnostic settings. This paper addresses the problem of self-explainable multimodal diagnosis by formulating it within the information bottleneck (IB) framework. We propose a unified learning paradigm that jointly optimizes predictive performance and modality-specific explainability by identifying the most informative elements inside each modality that contribute to diagnostic decisions. To enable tractable and stable optimization, we employ a matrix-based Renyi's $\alpha$-order entropy functional under the assumption of sufficiently expressive encoders. Extensive experiments on representative medical datasets spanning heterogeneous modalities demonstrate that the proposed method consistently achieves strong diagnostic performance, including an absolute accuracy improvement of 9.1 percentage points on the iCTCF dataset. Moreover, the learned explanations provide transparent and modality-aware insights into feature relevance, thereby improving both the explainability and generalization.

Chinese Translation

可解释性越来越被视为基于人工智能的医学诊断中的一个关键要求，特别是在安全关键的临床决策中。现有的大多数医疗可解释性方法都是事后解释的，且主要针对单模态数据设计，这限制了它们在日益普遍的多模态诊断环境中的适用性。本文通过在信息瓶颈（IB）框架内对自解释的多模态诊断问题进行表述，解决了这一问题。我们提出了一种统一的学习范式，该范式通过识别每种模态中对诊断决策贡献最大的最具信息性的元素，来共同优化预测性能和模态特定的可解释性。为了实现可处理和稳定的优化，我们在假设编码器具有足够表达能力的前提下，采用了基于矩阵的Renyi的$eta$阶熵泛函。对涵盖异构模态的代表性医学数据集进行的广泛实验表明，所提出的方法在诊断性能上始终表现出色，包括在iCTCF数据集上绝对准确率提高了9.1个百分点。此外，学习到的解释提供了透明且考虑模态的特征相关性洞察，从而提高了可解释性和泛化能力。

View on arXiv Download PDF AI Translation

cs.CV / 63 / 2609.05182

Conserved Immune Topology Improves Pathology Foundation Model Generalization for Cross-Cancer MSI-H Prediction

保守免疫拓扑改善病理基础模型在跨癌症MSI-H预测中的泛化能力

Raju, Dasari Naga

Abstract

Pathology foundation models integrated with multiple instance learning achieve competitive accuracy within single-cancer cohorts, yet cross-cancer generalization remains unresolved due to organ-specific histological and architectural differences. In this paper, we propose Conserved Immune Topology (CIT), a lightweight spatial representation for cross-cancer MSI-H prediction that augments foundation-model embeddings with biologically motivated immune descriptors. CIT uses unsupervised clustering to identify immune-associated tiles, then encodes tertiary lymphoid structures, peritumoral immune reactions, multi-scale tumor-infiltrating lymphocyte density, and immune-tumor mixing from frozen foundation-model embeddings and tile coordinates without requiring annotations or target-domain data. The proposed method was evaluated under cross-site and cross-cancer settings using CPTAC-COAD and TCGA-STAD cohorts, which introduce scanner variability, distribution shifts, and organ-specific architectural variations. Zero-shot cross-cancer transfer with CIT increased TransMIL AUC from 0.6627 to 0.7161, an absolute gain of 0.0534 (p=0.003), with consistent improvements across all three MIL aggregators. These results suggest that spatial immune topology provides potentially an organ-invariant representation for MSI-H prediction, supporting cross-cancer generalization of pathology foundation models.

Chinese Translation

整合多实例学习的病理基础模型在单一癌症队列中实现了竞争性的准确性，然而由于器官特异性的组织学和结构差异，跨癌症的泛化问题仍未解决。本文提出了保守免疫拓扑（Conserved Immune Topology, CIT），一种轻量级的空间表示，用于跨癌症MSI-H预测，增强了基础模型嵌入与生物学驱动的免疫描述符。CIT利用无监督聚类识别与免疫相关的图块，然后从冷冻的基础模型嵌入和图块坐标中编码三级淋巴结构、肿瘤周围免疫反应、多尺度肿瘤浸润淋巴细胞密度及免疫-肿瘤混合，而无需注释或目标领域数据。所提方法在CPTAC-COAD和TCGA-STAD队列下进行了跨站点和跨癌症设置的评估，这些队列引入了扫描仪变异性、分布偏移和器官特异性结构变化。使用CIT进行零样本跨癌症迁移，TransMIL AUC从0.6627提高到0.7161，绝对增益为0.0534（p=0.003），在所有三种MIL聚合器中均表现出一致的改善。这些结果表明，空间免疫拓扑为MSI-H预测提供了一种可能的器官不变表示，支持病理基础模型的跨癌症泛化。

View on arXiv Download PDF AI Translation

cs.CV / 64 / 2609.05210

BLASt3R: Bundle Adjustment of Any Image Set with Multi-View Matching and Monocular Priors

BLASt3R：具有多视图匹配和单目先验的任意图像集的束调整

Leroy, Vincent, Weinzaepfel, Philippe, Zust, Lojze, Cabon, Yohann, Revaud, Jérome

Abstract

Recent hybrid Structure-from-Motion (SfM) systems combine the robustness of feed-forward 3D reconstruction with the accuracy of traditional bundle adjustment (BA) with pixel matching. They are usually the best performing methods however their scalability and usability remains limited since estimating dense correspondences between views is prohibitively costly, especially considering time constraints inherent to online applications like Visual SLAM (VSLAM). In this paper, we introduce a regularized BA framework that leverages a fast multi-view matcher and monocular priors for initialization and regularization. In contrast to existing systems, our unified approach seamlessly supports both online VSLAM and offline reconstruction from unordered image collections within the same optimization framework and sharing common hyperparameters for all tasks. Extensive experiments across both domains demonstrate improved performance and speed tradeoffs over traditional, feed-forward, and hybrid baselines. Notably for VSLAM, our uncalibrated method outperforms all previous calibrated approaches.

Chinese Translation

最近的混合结构光束法（SfM）系统结合了前馈三维重建的鲁棒性和传统束调整（BA）与像素匹配的准确性。尽管它们通常是性能最优的方法，但由于在视图之间估计密集对应关系的成本过高，尤其是在考虑到视觉同步定位与地图构建（VSLAM）等在线应用的时间限制时，其可扩展性和可用性仍然有限。本文提出了一种正则化的BA框架，利用快速的多视图匹配器和单目先验进行初始化和正则化。与现有系统相比，我们的统一方法在同一优化框架内无缝支持在线VSLAM和从无序图像集合进行离线重建，并为所有任务共享共同的超参数。在这两个领域的广泛实验表明，与传统的、前馈的和混合基线相比，性能和速度的权衡得到了改善。值得注意的是，对于VSLAM，我们的未校准方法超越了所有先前的校准方法。

View on arXiv Download PDF AI Translation

cs.CV / 65 / 2609.05224

First Things First: Teaching LLM-Based Agents to Prioritize Must-Haves before Nice-to-Haves

优先考虑重要事项：教导基于大型语言模型的智能体在可有可无的需求之前优先满足必须的需求

Ju, Tianjie, Xu, Xinyue, Sun, Wanxuan, Diao, Lingxiao, Liu, Gongshen, Zhang, Zhuosheng, Yang, Cheng

Abstract

Recent progress in multimodal large language models (MLLMs) has fueled significant enthusiasm in their potential to act as autonomous agents for real-world tasks. However, scenarios requiring agents to fulfill users' complex, structured requirements remain largely underexplored. In this work, we examine reasoning tasks under three distinct requirement scenarios: (i) Must-have requirements uniquely determine a unique feasible solution; (ii) Multiple answers satisfy the must-have requirements and are prioritized via the nice-to-have requirements; and (iii) No candidate solution satisfies the must-have requirements, in which case the agent should abstain from generating a response. We evaluate state-of-the-art MLLMs on 3,649 carefully constructed problems that reflect realistic service scenarios, including e-commerce, booking, and map-based or ride-hailing. Our evaluation reveals that existing MLLMs exhibit catastrophic failures in all scenarios. They frequently misinterpret task requirements, violate must-have requirements, and produce invalid solutions. To address this critical gap, we propose First Things First Reinforcement Learning FTF-rl that explicitly optimizes reasoning over multi-priority user requirements. Experimental results show that our method substantially improves the task success rate compared to strong baselines. Moreover, FTF-rl yields general effectiveness on popular logical and mathematical reasoning tasks, including LogicVista, MathVision, and InfoQA. Our findings suggest that enhancing requirement-aware reasoning capability provides a simple yet effective pathway to improve generalization of MLLM agents. Code and dataset are available at https://github.com/claire62/FTF-RL.

Chinese Translation

近期在多模态大型语言模型（MLLMs）方面的进展激发了人们对其作为自主智能体执行现实任务潜力的极大热情。然而，要求智能体满足用户复杂、结构化需求的场景仍然在很大程度上未被探索。在本研究中，我们考察了三种不同需求场景下的推理任务：（i）必须的需求唯一确定一个可行解；（ii）多个答案满足必须的需求，并通过可有可无的需求进行优先排序；（iii）没有候选解满足必须的需求，在这种情况下，智能体应避免生成响应。我们在3649个精心构建的问题上评估了最先进的MLLMs，这些问题反映了现实服务场景，包括电子商务、预订以及基于地图的打车服务。我们的评估显示，现有的MLLMs在所有场景中均表现出灾难性的失败。它们经常误解任务需求，违反必须的需求，并产生无效的解决方案。为了解决这一关键问题，我们提出了优先事项强化学习（First Things First Reinforcement Learning，FTF-rl），该方法明确优化多优先级用户需求的推理。实验结果表明，我们的方法相比强基线显著提高了任务成功率。此外，FTF-rl在流行的逻辑和数学推理任务上也表现出一般有效性，包括LogicVista、MathVision和InfoQA。我们的研究结果表明，增强需求感知推理能力为改善MLLM智能体的泛化提供了一条简单而有效的途径。代码和数据集可在https://github.com/claire62/FTF-RL获取。

View on arXiv Download PDF AI Translation

cs.CV / 66 / 2609.05234

Measured Sliders: Learning Continuous Controls from Differentiable Image Measurements

测量滑块：从可微分图像测量中学习连续控制

Chen, Yijia, Wei, Boyu, Yin, Xuanhua

Abstract

Continuous sliders are useful only when coefficient changes produce predictable image changes. Yet most diffusion sliders derive their axes from text or learned representations, leaving their scales disconnected from observable image properties. Consequently, we cannot tell in advance which attributes are learnable, compare control strengths directly, or anticipate interference when multiple controls are combined. We propose Measured Sliders, a framework that defines continuous controls through closed-form differentiable image measurements. A common measurement space unifies the pipeline. Before training, an observability test identifies usable supervision. During training, a measurement-guided objective learns target movement while suppressing non-target changes. After training, decoded calibration expresses controls in comparable units of realized image change. Multiple LoRA branches are stored in one checkpoint and composed without training on joint activations. Across SDXL and FLUX.1-dev, the resulting controls are ordered, selective, and composable. On 553 prompts, lighting direction reaches rho = 0.995 and 98.9% monotone sweeps. A five-attribute checkpoint achieves average selectivity 2.59, compared with 1.50 for the strongest baseline, and preserves every requested direction in 96.7% of pair and 86.1% of triple compositions. The observability test also separates every subsequently successful measurement from the failed candidate. Overall, image-space measurement provides a common basis for learning, diagnosing, calibrating, and composing continuous generative controls.

Chinese Translation

连续滑块只有在系数变化产生可预测的图像变化时才有用。然而，大多数扩散滑块的轴线来源于文本或学习表示，使得它们的尺度与可观察的图像属性脱节。因此，我们无法提前判断哪些属性是可学习的，无法直接比较控制强度，也无法预测多个控制组合时的干扰。我们提出了测量滑块（Measured Sliders），这是一个通过封闭形式的可微分图像测量定义连续控制的框架。一个共同的测量空间统一了整个流程。在训练之前，观察性测试识别可用的监督信号。在训练过程中，测量引导的目标学习目标运动，同时抑制非目标变化。训练后，解码的校准以可比较的实现图像变化单位表达控制。多个 LoRA 分支存储在一个检查点中，并且可以在不对联合激活进行训练的情况下组合。在 SDXL 和 FLUX.1-dev 上，生成的控制是有序的、选择性的和可组合的。在 553 个提示中，光照方向的相关系数达到 rho = 0.995，单调变化的比例为 98.9%。一个包含五个属性的检查点实现了平均选择性 2.59，而最强基线为 1.50，并在 96.7% 的双重组合和 86.1% 的三重组合中保留了每个请求的方向。观察性测试还将每个随后成功的测量与失败的候选者分开。总体而言，图像空间测量为学习、诊断、校准和组合连续生成控制提供了共同基础。

View on arXiv Download PDF AI Translation

cs.CV / 67 / 2609.05239

Cross-Domain Tracker Adaptation Without Target-Domain Labels via Vision-Language Agents

通过视觉-语言代理实现无目标领域标签的跨域跟踪器适应

Davila, Daniel, Balakrishnan, Ravikumar, Cochran, Mike

Abstract

We present a system that uses a Vision-Language Model (VLM) as a diagnostic agent for adapting a detect-to-track pipeline to a new target domain without access to target-domain labels. Rather than optimizing against annotated metrics, the VLM directly inspects rendered tracking outputs, identifies visual failure modes, and recommends parameter updates through an iterative tuning loop. We first demonstrate that ground-truth-supervised hyperparameter transfer can be brittle. On MOT17->MOT20, applying a source-derived oracle configuration reduces mean HOTA by 0.090, from a target-domain ceiling of 0.357, to 0.267. Without using any target-domain labels, our VLM-based tuner recovers 67.8% of this lost headroom, finishing within 0.029 HOTA of the target ceiling; on the highest-density target sequence, it recovers up to 86.7%. We further show that label-free Bayesian optimization with handcrafted proxy objectives struggles under large domain shifts and can degrade configurations that are already strong. In contrast, the VLM tuner acts selectively: when its visual diagnosis reveals no clear failure mode, it declines to modify the configuration, preserving performance on easy transfers while improving hard ones. Finally, we characterize the conditions under which this approach succeeds, namely, when domain shift manifests through exposed detection-level parameters, versus where it is less effective, such as MOT17->DanceTrack, where the source oracle is already near-optimal.

Chinese Translation

我们提出了一种系统，利用视觉-语言模型（VLM）作为诊断代理，将检测到跟踪的流程适应于新的目标领域，而无需访问目标领域标签。该系统并非针对标注指标进行优化，而是直接检查渲染的跟踪输出，识别视觉失败模式，并通过迭代调优循环推荐参数更新。我们首先展示了基于真实标签的超参数迁移可能会不稳定。在MOT17->MOT20的实验中，应用源导出的oracle配置使得平均HOTA降低了0.090，从目标领域的上限0.357降至0.267。在不使用任何目标领域标签的情况下，我们基于VLM的调优器恢复了67.8%的丢失余量，最终与目标上限的HOTA差距仅为0.029；在最高密度的目标序列中，恢复率高达86.7%。我们进一步表明，使用手工设计的代理目标进行无标签贝叶斯优化在大领域转移下表现不佳，甚至可能降低已经较强的配置。相比之下，VLM调优器采取选择性行动：当其视觉诊断未能揭示明显的失败模式时，它拒绝修改配置，从而在简单迁移中保持性能，同时改善困难迁移。最后，我们描述了该方法成功的条件，即当领域转移通过显露的检测级参数表现出来时，而在某些情况下效果较差，例如MOT17->DanceTrack，在这种情况下源oracle已经接近最优。

View on arXiv Download PDF AI Translation

cs.CV / 68 / 2609.05242

Few-Shot Video Recognition via Hierarchical Metric Learning

通过层次度量学习实现少样本视频识别

Zhang, Jiaxin, Gao, Haoran, Gao, Xizhan, Dong, Zihao, Wang, Tingwei, Niu, Sijie

Abstract

Few-shot action recognition (FSAR) aims to recognize unseen action categories with only a small number of annotated video samples. Recent works typically apply single-prototype supervision at the network output and fail to sufficiently exploit rich cross-frame global spatial information in videos. Even existing multi-level metric schemes only impose parallel prototype constraints on intermediate layers, without progressive supervision along the full feature pipeline, which results in limited generalization ability of the learned class prototypes. Inspired by this, we present a novel method, hierarchical metric learning for few-shot action recognition (HML-FSAR). First, a spatial-enhanced module is developed to capture cross-frame global spatial representations. Combined with temporal MHA, heterogeneous alignment, spatial-temporal feature fusion and dictionary learning modules, it constructs the complete feature processing pipeline. Second, a hierarchical metric learning (HML) strategy is embedded into HML-FSAR. Composed of center metric, alignment metric, contrastive metric, dictionary metric and prototype metric, HML imposes progressive multi-stage complementary constraints from frame-level representations to final class prototypes, so as to jointly optimize feature compactness, heterogeneous spatial-temporal alignment, inter-class discriminability and anti-noise robustness. The proposed HML-FSAR method is validated on five widely-used FSAR datasets, and experimental results fully demonstrate its effectiveness.

Chinese Translation

少样本动作识别（FSAR）旨在仅通过少量标注视频样本识别未见过的动作类别。近期的研究通常在网络输出处应用单原型监督，未能充分利用视频中丰富的跨帧全局空间信息。即使现有的多级度量方案也仅在中间层施加并行原型约束，而未在完整特征处理管道中进行渐进式监督，这导致所学习的类别原型的泛化能力有限。受此启发，我们提出了一种新方法，层次度量学习用于少样本动作识别（HML-FSAR）。首先，开发了一个空间增强模块，以捕捉跨帧全局空间表示。结合时间多头自注意力（temporal MHA）、异构对齐、时空特征融合和字典学习模块，构建完整的特征处理管道。其次，将层次度量学习（HML）策略嵌入HML-FSAR中。HML由中心度量、对齐度量、对比度量、字典度量和原型度量组成，从帧级表示到最终类别原型施加渐进式多阶段互补约束，以共同优化特征紧凑性、异构时空对齐、类别间可分性和抗噪声鲁棒性。所提出的HML-FSAR方法在五个广泛使用的FSAR数据集上进行了验证，实验结果充分证明了其有效性。

View on arXiv Download PDF AI Translation

cs.CV / 69 / 2609.05255

Compact Neural Appearance Models for Efficient Gaussian Splatting

高效高斯点云的紧凑神经外观模型

Hahlbohm, Florian, Condor, Jorge, Franke, Linus, Eisemann, Martin, Magnor, Marcus

Abstract

Explicit primitive-based radiance fields such as 3D Gaussian Splatting typically model view-dependent appearance using low-order spherical harmonics (SH). Although efficient to evaluate, SH coefficients dominate per-primitive storage and memory traffic, while their band-limited basis restricts angular detail. We present a thorough, end-to-end comparison of SH and recent spherical appearance models and introduce an implicit alternative that decodes compact per-primitive latent codes using a tiny shared MLP. We integrate all models into the same optimized pipeline, fusing their forward and backward passes into a differentiable CUDA rasterizer and provide a portable WebGL viewer for laptop and mobile GPUs. Our evaluation across reconstruction quality, memory use, and optimization and rendering performance shows that recent spherical models offer the strongest overall quality-efficiency trade-off. Our neural representation is the most compact model evaluated and, compared to third-degree SH, reduces the per-primitive appearance footprint from 192 to 28 bytes, accelerates optimization by 1.3$\times$, while improving reconstruction quality. We further analyze how appearance parametrization shapes optimization, identifying differences in recovered geometry and the tendency of expressive models to absorb non-static scene content. Together, our framework and analysis provide practical guidance for replacing SH beyond what image metrics alone can capture.

Chinese Translation

显式基于原始的辐射场，如3D高斯点云，通常使用低阶球谐函数（SH）来建模视角依赖的外观。尽管评估效率较高，但SH系数在每个原始的存储和内存流量中占主导地位，而其带限基函数限制了角度细节。我们对SH和近期的球形外观模型进行了全面的端到端比较，并提出了一种隐式替代方案，该方案使用一个小型共享多层感知器（MLP）解码紧凑的每个原始潜在编码。我们将所有模型集成到同一优化管道中，将它们的前向和后向传递融合到一个可微分的CUDA光栅化器中，并提供了一个便携的WebGL查看器，适用于笔记本电脑和移动GPU。我们在重建质量、内存使用、优化和渲染性能方面的评估表明，近期的球形模型提供了最佳的整体质量效率权衡。我们的神经表示是评估中最紧凑的模型，与三阶SH相比，将每个原始的外观占用空间从192字节减少到28字节，优化速度提升了1.3倍，同时改善了重建质量。我们进一步分析了外观参数化如何影响优化，识别了恢复几何形状的差异以及表现力强的模型吸收非静态场景内容的倾向。我们的框架和分析共同为替代SH提供了实用指导，超越了仅依赖图像度量所能捕捉的内容。

View on arXiv Download PDF AI Translation

cs.CV / 70 / 2609.05303

Learning Spatial-Spectral Refinement and Calibrating Complementary Observations for Hyperspectral Image Super-Resolution

学习空间-光谱细化与校准互补观测以实现高光谱图像超分辨率

Yang, Liqian, Chen, Xingchi, Gui, Xinfeng, Cao, Xiangyong, Yi, Qianxin

Abstract

Hyperspectral and multispectral image fusion (HMIF) aims to reconstruct a high-resolution hyperspectral image (HR-HSI) by combining the fine spatial details of a high-resolution multispectral image (HR-MSI) with the rich spectral information of a low-resolution hyperspectral image (LR-HSI). Recent advances in implicit neural representations (INRs) have enabled flexible coordinate-based modeling for HMIF; however, existing INR-based approaches may not fully capture fine-grained spatial structures and rich spectral dependencies. Moreover, the LR-HSI and HR-MSI are primarily incorporated through degradation-consistency constraints, leaving their complementary information underexploited. To address these limitations, we propose Two-Stage Reconstruction with Implicit Tensor Neural Representation (TSR-ITNR), a unified self-supervised framework integrating representation refinement and observation-guided calibration. In Stage 1, TSR-ITNR learns an implicit Tucker representation and refines its low-rank spatial coefficient tensor and spectral basis to better capture fine spatial structures and interband correlations. A fixed pretrained denoiser further provides a deep prior for the preliminary reconstruction. In Stage 2, parameter-free calibration derives complementary and noninterfering corrections from both observations to recover information insufficiently captured in Stage 1. Theoretical analysis establishes the geometry-preserving property of spectral refinement and the orthogonal complementarity of calibration. Extensive experiments on multiple benchmark datasets demonstrate strong quantitative, visual, and spectral reconstruction performance without ground-truth HR-HSI supervision. Beyond conventional reconstruction metrics, we further assess the effectiveness of TSR-ITNR using downstream semantic segmentation accuracy.

Chinese Translation

高光谱与多光谱图像融合（HMIF）旨在通过结合高分辨率多光谱图像（HR-MSI）的精细空间细节与低分辨率高光谱图像（LR-HSI）的丰富光谱信息，重建高分辨率高光谱图像（HR-HSI）。近期在隐式神经表示（INRs）方面的进展使得HMIF的坐标基础建模变得灵活；然而，现有的基于INR的方法可能无法充分捕捉细粒度的空间结构和丰富的光谱依赖关系。此外，LR-HSI和HR-MSI主要通过降解一致性约束进行结合，导致其互补信息未得到充分利用。为了解决这些局限性，我们提出了隐式张量神经表示的两阶段重建（TSR-ITNR），这是一个统一的自监督框架，整合了表示细化和观测引导的校准。在第一阶段，TSR-ITNR学习隐式Tucker表示，并细化其低秩空间系数张量和光谱基，以更好地捕捉细致的空间结构和波段间的相关性。一个固定的预训练去噪器进一步为初步重建提供了深度先验。在第二阶段，无参数校准从两个观测中推导出互补且不干扰的修正，以恢复在第一阶段中未充分捕获的信息。理论分析建立了光谱细化的几何保持特性和校准的正交互补性。在多个基准数据集上的大量实验表明，在没有真实HR-HSI监督的情况下，具有强大的定量、视觉和光谱重建性能。除了传统的重建指标外，我们还通过下游语义分割精度进一步评估了TSR-ITNR的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 71 / 2609.05320

Adaptive Gated Deepfake Detection for Low-Resolution and Resource-Constrained Environments

适应性门控深度伪造检测用于低分辨率和资源受限环境

Sen, Vaishnavi, Laurie, Cody, Hasan, Rashida

Abstract

Deepfake detection models often rely on high-quality inputs, fixed inference paths, and computationally expensive architectures, limiting their use in low-resolution and resource-constrained settings. This paper proposes AdaGate-DF, an adaptive gated deepfake detection framework that uses image-quality cues to route samples through a dual multi-exit system so high-quality images can exit earlier and save compute. We evaluated AdaGate-DF against MaD-CoRN, DefakeHop++, and ShuffleNetV2 on two benchmark datasets (Celeb-DF and FaceForensics++) under multiple configurations to test image resolution dependence and training and inference efficiency. On Celeb-DF, AdaGate-DF achieves an AUC of 0.9370, outperforming MaD-CoRN and DefakeHop++ while maintaining a low inference latency. Resolution-based testing shows consistent improvement as input resolution increases, reaching an AUC of 0.9708 at 384 by 384. The FaceForensics++ results highlight that AdaGate-DF remains effective under class imbalance, following competitive results with evaluated models. Overall, AdaGate-DF demonstrated a practical balance between detection performance, uncertainty-aware prediction, and computational efficiency for variable-quality deepfake detection.

Chinese Translation

深度伪造检测模型通常依赖于高质量输入、固定推理路径和计算开销大的架构，这限制了它们在低分辨率和资源受限环境中的应用。本文提出了AdaGate-DF，一种适应性门控深度伪造检测框架，利用图像质量线索通过双重多出口系统引导样本，使高质量图像能够更早退出，从而节省计算资源。我们在两个基准数据集（Celeb-DF和FaceForensics++）上对AdaGate-DF进行了评估，比较了MaD-CoRN、DefakeHop++和ShuffleNetV2，测试了图像分辨率依赖性以及训练和推理效率。在Celeb-DF上，AdaGate-DF达到了0.9370的AUC，超越了MaD-CoRN和DefakeHop++，同时保持了较低的推理延迟。基于分辨率的测试显示，随着输入分辨率的增加，性能持续改善，在384x384时达到了0.9708的AUC。FaceForensics++的结果强调了AdaGate-DF在类别不平衡情况下仍然有效，并与评估模型的竞争结果相符。总体而言，AdaGate-DF在检测性能、不确定性感知预测和计算效率之间展示了实用的平衡，适用于不同质量的深度伪造检测。

View on arXiv Download PDF AI Translation

cs.CV / 72 / 2609.05323

Scalable Detection of Fossil Palynomorphs in Multifocal Digital Microscopy Images

可扩展的化石花粉体在多焦点数字显微镜图像中的检测

Shaikh, Abbas, Mayor, Praise, Ainlay-Vazquez, Patrick, Viswanathan, Aditya, Golden, Teon, Zhang, Eric, Romero, Ingrid C., White, Alexander E., Wing, Scott, Barman, Arko

Abstract

Palynomorphs (microscopic, organic-walled fossils such as pollen, spores, and dinoflagellates) are important high-resolution records of past climates and are critical to the study of ancient ecosystems. Existing methods rely on manual analysis of high-resolution, multifocal digital microscopy images, which is slow and time-consuming and requires researchers to compromise on the scale of their investigations. To the best of our knowledge, our work proposes the first ever scalable end-to-end pipeline for automated palynomorph detection in whole slide images that addresses this bottleneck through: (1) efficient methods for decomposing and compressing digitized multifocal microscope slide images into tractable 2-dimensional tiles for analysis; (2) benchmarking modern object detection models, including RF-DETR, for the detection of palynomorphs, achieving an AP@50 of 0.879; (3) an efficient algorithm for the synthesis of detection outputs across large-scale, high-resolution images; and (4) an I/O optimization resulting in faster inference time. Our methods drastically reduce the time required for palynomorph detection in a single slide from often days of manual inspection to under one hour of automated analysis, enabling palynological research at a substantially greater scale.

Chinese Translation

花粉体（如花粉、孢子和甲藻等微观有机壁化石）是过去气候的重要高分辨率记录，对于研究古代生态系统至关重要。现有方法依赖于对高分辨率多焦点数字显微镜图像的手动分析，这一过程缓慢且耗时，迫使研究人员在调查规模上做出妥协。根据我们所知，我们的工作提出了首个可扩展的端到端自动化花粉体检测管道，旨在通过以下方式解决这一瓶颈：（1）高效的方法将数字化的多焦点显微镜切片图像分解和压缩为可处理的二维图块进行分析；（2）基准测试现代目标检测模型，包括RF-DETR，以检测花粉体，达到AP@50为0.879；（3）一种高效算法用于在大规模高分辨率图像中合成检测输出；（4）I/O优化以实现更快的推理时间。我们的方法显著减少了在单个切片中进行花粉体检测所需的时间，从通常需要数天的手动检查缩短至不到一小时的自动化分析，使得花粉学研究能够在更大规模上进行。

View on arXiv Download PDF AI Translation

cs.CV / 73 / 2609.05334

Lightweight Vision Transformer Compression for On-Device Plant Disease Detection in Resource-Constrained Agricultural Field Conditions

资源受限农业现场条件下用于设备端植物疾病检测的轻量级视觉变换器压缩

Kumar, Mahadev Sunil, Gondi, Bhavika, Swapnith, Desaisetty Venkata Satya Sai, Jogi, Gangireddy Rahul, Manalil, Sudheesh, Raha, Arnab, Mukherjee, Amitava, Seethapathy, Parthasarathy, Gopakumar, G.

Abstract

Chilli (Capsicum annuum) is one of India's most economically significant crops, yet its productivity is persistently threatened by diseases that are difficult to identify without expert intervention. While Vision Transformers (ViTs) have achieved high classification accuracy, their large computational footprint makes deployment on resource constrained devices challenging. Existing compression approaches typically address pruning, quantization, and knowledge distillation in isolation, leaving the potential benefits and interactions of their combined application insufficiently explored. We propose a unified Vision Transformer compression framework that combines Hessian-Balanced Adaptive Block Pruning (H-BAC), guided by second-order sensitivity estimation, with quantization and attention-based knowledge distillation. To systematically identify the most effective configuration within each compression family, each technique is first evaluated independently through controlled ablation studies, after which the best-performing components are integrated into a sequential deployment pipeline tailored to real-world agricultural constraints. On a chilli 3-class village-split dataset with a genuine cross-village, cross-device out-of-distribution test split, the resulting compressed models match or exceed the 95.13% FP32 baseline's accuracy, alongside 74-98% model size reduction, and the fully integrated compression pipeline achieves a 54.5x size reduction (327.42 MB to 6.01 MB) at 95.13 +/- 2.32% accuracy across four tested configurations. A direct comparison further reveals that, on this dataset, a directly-trained student of the same final size, without pruning or distillation, reaches comparable accuracy of 94.87%, at the same 6.01 MB INT8 size, indicating where H-BAC and knowledge distillation are, and are not yet shown to be, worth their computational cost.

Chinese Translation

辣椒（Capsicum annuum）是印度最具经济意义的作物之一，但其生产力持续受到难以在没有专家干预的情况下识别的疾病威胁。尽管视觉变换器（ViTs）已实现高分类准确率，但其庞大的计算开销使得在资源受限设备上的部署面临挑战。现有的压缩方法通常孤立地解决剪枝、量化和知识蒸馏的问题，未能充分探讨它们联合应用的潜在好处和相互作用。我们提出了一种统一的视觉变换器压缩框架，该框架结合了基于二阶灵敏度估计的Hessian平衡自适应块剪枝（H-BAC）、量化和基于注意力的知识蒸馏。为了系统性地识别每个压缩类别中最有效的配置，首先通过受控消融研究独立评估每种技术，然后将表现最佳的组件集成到一个针对现实农业限制量身定制的顺序部署管道中。在一个具有真实跨村、跨设备的分布外测试集的辣椒三类村庄分割数据集中，得到的压缩模型的准确率与95.13%的FP32基线相匹配或超过，同时模型大小减少了74-98%。完全集成的压缩管道在四种测试配置下实现了54.5倍的大小减少（从327.42 MB降至6.01 MB），准确率为95.13 +/- 2.32%。直接比较进一步表明，在该数据集中，一个经过直接训练的同样最终大小的学生模型，在没有剪枝或蒸馏的情况下，达到了94.87%的可比准确率，且同样为6.01 MB的INT8大小，表明H-BAC和知识蒸馏在计算成本上是值得的地方和尚未显示出价值的地方。

View on arXiv Download PDF AI Translation

cs.CV / 74 / 2609.05351

MEOX: Compact Multimodal Mixture-of-Experts for Earth Observation

MEOX：用于地球观测的紧凑型多模态专家混合模型

Albughdadi, Mohanad

Abstract

Recent advances in Earth Observation representation learning accommodate heterogeneous sensors and missing observations, often through larger architectures. We present MEOX (Multimodal Earth Observation with eXperts), a multimodal masked autoencoder with a 2.939 million-parameter encoder and 3.115 million parameters in total. Sensor-specific adapters, explicit validity signals, and a shared sparse-expert block preserve modality-dependent processing before a learned patch-wise fusion. Four metadata tokens then accompany a single spatial sequence through fourteen further encoder blocks. Shared expert projections with private low-rank residuals constrain parameter growth, while rotary attention supports downstream spatial grids different from pretraining. The model is pretrained on 1.228 million MMEarth64 samples using modality-balanced masked reconstruction and structured sensor dropout. Frozen transfer is evaluated on six GEO-Bench tasks at both 64 and 224 pixels. The model reaches 64.42% mean intersection-over-union on cashew segmentation at 64 pixels and 90.56% average accuracy on EuroSAT at 224 pixels, exceeding the corresponding reported CSMoE results. BigEarthNet finetuning reaches 72.95% micro-average precision. Routing diagnostics distinguish expert participation, spatial dependence, modality association, and functional contribution. A held-out WorldCover probe measures a 0.64-percentage-point benefit from metadata, while retrieval separates same-sensor semantics from cross-sensor alignment. These results demonstrate sensor-flexible representation learning and strong task transfer using a compact parameter budget.

Chinese Translation

近期在地球观测表示学习方面的进展能够处理异构传感器和缺失观测，通常通过更大的架构来实现。我们提出了MEOX（多模态地球观测与专家），这是一种多模态掩蔽自编码器，具有293.9万参数的编码器和总计311.5万参数。传感器特定的适配器、显式有效性信号以及共享稀疏专家块在学习的块级融合之前保持模态依赖的处理。随后，四个元数据标记伴随一个单一的空间序列通过十四个进一步的编码器块。共享专家投影与私有低秩残差限制了参数的增长，而旋转注意力支持与预训练不同的下游空间网格。该模型在122.8万MMEarth64样本上进行预训练，使用模态平衡的掩蔽重建和结构化传感器丢失。在64像素和224像素的六个GEO-Bench任务上评估冻结迁移。该模型在64像素的腰果分割任务上达到64.42%的平均交并比，在224像素的EuroSAT任务上达到90.56%的平均准确率，超越了相应报告的CSMoE结果。BigEarthNet微调达到72.95%的微平均精度。路由诊断区分专家参与、空间依赖、模态关联和功能贡献。一个保留的WorldCover探针测量到元数据带来的0.64个百分点的收益，而检索则将同传感器语义与跨传感器对齐区分开。这些结果展示了传感器灵活的表示学习和在紧凑参数预算下的强任务迁移能力。

View on arXiv Download PDF AI Translation

cs.CV / 75 / 2609.05382

Reflection-aware Generative Novel View Synthesis

反射感知生成新视角合成

Kim, GeonU, Dong-Yeon, Shin, Oh, Tae-Hyun

Abstract

We propose Ref-GeNVS, a training-free, reflection-aware method for generative novel view synthesis (NVS) in mirror scenes. Existing multi-view diffusion models often fail to recognize the mirror in the scene and cannot exploit reflected content for scene generation. To fix this issue without additional training, our key idea is to treat a mirror image as two complementary views. From input images, we estimate the mirror plane and reflect camera poses to form virtual views. Based on this virtual view setup, we propose a two-stage generation method consisting of Mirror-gated attention and Reflection injection, which enables reflection-consistent NVS by explicitly leveraging reflection relationships in a multi-view diffusion model. Ref-GeNVS inherits the strong generalizability of the multi-view diffusion backbone, while it does not require finetuning. On synthetic and real scenes including mirrors, Ref-GeNVS outperforms recent generative NVS methods by generating reflection-consistent and contextually coherent novel views, revealing scene structure visible only through mirrors. Project page: https://kim-geonu.github.io/Ref-GeNVS/

Chinese Translation

我们提出了一种名为 Ref-GeNVS 的无训练、反射感知方法，用于镜面场景中的生成新视角合成（NVS）。现有的多视角扩散模型通常无法识别场景中的镜子，无法利用反射内容进行场景生成。为了解决这个问题而不需要额外的训练，我们的关键思想是将镜像视为两个互补的视角。通过输入图像，我们估计镜面平面并反射相机姿态以形成虚拟视角。基于这种虚拟视角设置，我们提出了一种由镜门控注意力（Mirror-gated attention）和反射注入（Reflection injection）组成的两阶段生成方法，通过明确利用多视角扩散模型中的反射关系，实现反射一致的 NVS。Ref-GeNVS 继承了多视角扩散骨干的强泛化能力，同时不需要微调。在包括镜子的合成和真实场景中，Ref-GeNVS 通过生成反射一致且上下文连贯的新视角，超越了最近的生成 NVS 方法，揭示了仅通过镜子可见的场景结构。项目页面：https://kim-geonu.github.io/Ref-GeNVS/

View on arXiv Download PDF AI Translation

cs.CV / 76 / 2609.05388

Think-Verify-Revise: Neuro-Symbolic Visual Reasoning with Vision-Language Models and Dynamic Logic Tensor Networks

思考-验证-修正：结合视觉-语言模型和动态逻辑张量网络的神经符号视觉推理

Afshari, Homayoun, Basci, Pietro, Russo, Alessandro, Morra, Lia

Abstract

Visual reasoning tasks require a system to jointly perceive visual content and apply formal relational constraints---a combination that neither pure neural nor purely symbolic approaches handle well in isolation. This paper proposes a Neuro-Symbolic (NeSy) framework that closes this gap by tightly coupling a Vision-Language Model (VLM) for automatic First-Order Logic (FOL) rule induction with a Dynamic Logic Tensor Network (D-LTN) for differentiable rule verification, in a closed iterative feedback loop. The VLM receives a small set of labelled visual examples and proposes candidate FOL rules conforming to a strict grammar (Think); the D-LTN is automatically assembled from these rules at runtime and evaluates them grounding on CNN-produced visual embeddings (Verify); and verification failures are fed back to guide the VLM's next hypothesis (Revise). Evaluated on the ViSudo-PC benchmark across four visual domains (MNIST, EMNIST, KMNIST, FMNIST), the system induces valid Sudoku constraint rules using only three training examples as visual context. The proposed method achieves AUC scores matching or outperforming previous methods (NeuPSL, LTN), showing the potential for automatic rule discovery through VLM. Code is available at https://github.com/homayoun-afshari/nesy.

Chinese Translation

视觉推理任务要求系统共同感知视觉内容并应用形式关系约束——这一组合在纯神经或纯符号方法中都难以单独处理。本文提出了一种神经符号（Neuro-Symbolic, NeSy）框架，通过将自动的一阶逻辑（First-Order Logic, FOL）规则归纳的视觉-语言模型（Vision-Language Model, VLM）与可微分规则验证的动态逻辑张量网络（Dynamic Logic Tensor Network, D-LTN）紧密结合，形成一个闭环的迭代反馈机制，从而填补这一空白。VLM接收一小组标记的视觉示例，并提出符合严格语法的候选FOL规则（思考）；D-LTN在运行时根据这些规则自动组装，并基于卷积神经网络（CNN）生成的视觉嵌入进行评估（验证）；验证失败的结果反馈用于指导VLM的下一个假设（修正）。在四个视觉领域（MNIST、EMNIST、KMNIST、FMNIST）上评估的ViSudo-PC基准测试中，该系统仅使用三个训练示例作为视觉上下文即可归纳有效的数独约束规则。所提方法的AUC得分与之前的方法（NeuPSL、LTN）相当或更优，显示了通过VLM进行自动规则发现的潜力。代码可在 https://github.com/homayoun-afshari/nesy 获取。

View on arXiv Download PDF AI Translation

cs.CV / 77 / 2609.05397

CrossDepth: Geometry-Constrained Attention for Generalizable Multi-View Surround Depth Estimation

CrossDepth：几何约束注意力用于可泛化的多视图周围深度估计

Abualhanud, Samer, Mehltretter, Max

Abstract

Reliable 3D understanding of the surrounding environment is a core requirement for autonomous driving. Multi-view surround camera rigs provide broad scene coverage, but the spatially adjacent images typically overlap only minimally. Consequently, the depth of most pixels must be inferred from monocular appearance cues. These cues can appear differently across images and may therefore be interpreted differently by the depth estimation model. We target two main sources of cross-image inconsistency: differences in camera intrinsics and the limited receptive field of each image. We address the former by conditioning the features on per-pixel camera-aware ray embeddings, enabling the network to account for camera-dependent variations in monocular cues. We address the latter by extending each pixel's context beyond its own image through cross-image attention constrained to geometrically plausible regions, derived from the calibrated rig setup. The model is trained in a fully self-supervised manner based on photometric consistency. Evaluations on DDAD and nuScenes show improved overall depth accuracy and cross-image depth consistency over state-of-the-art self-supervised methods under in-domain and cross-domain evaluation. Code is available at https://abualhanud.github.io/CrossDepthPage/.

Chinese Translation

对周围环境的可靠三维理解是自动驾驶的核心需求。多视图周围摄像头设备提供了广泛的场景覆盖，但空间上相邻的图像通常仅有最小的重叠。因此，大多数像素的深度必须通过单目外观线索进行推断。这些线索在不同图像中可能表现不同，因此可能会被深度估计模型以不同的方式解读。我们针对跨图像不一致性的两个主要来源：相机内参的差异和每个图像的有限感受野。我们通过对每个像素的相机感知光线嵌入进行特征条件化来解决前者，使网络能够考虑单目线索中的相机依赖性变化。我们通过扩展每个像素的上下文超出其自身图像，采用几何上合理的区域进行跨图像注意力机制来解决后者，这些区域是从标定的设备设置中推导而来的。该模型以完全自监督的方式基于光度一致性进行训练。在DDAD和nuScenes上的评估显示，在领域内和跨领域评估中，相较于最先进的自监督方法，整体深度准确性和跨图像深度一致性均有所改善。代码可在 https://abualhanud.github.io/CrossDepthPage/ 获取。

View on arXiv Download PDF AI Translation

cs.CV / 78 / 2609.05399

From Interpretability Methods to Interpretable Models

从可解释性方法到可解释模型

Colin, Julien, Oliver, Nuria, Serre, Thomas

Abstract

More than a decade in, explainable AI (XAI) for computer vision has assembled a mature toolbox: attribution, feature visualization, concept-based, and circuit-based methods. Yet almost all of the field's effort has gone into building and comparing these methods, and little into the question they were meant to answer---how interpretable are our models, and are we making progress as they evolve? We argue for shifting the field's focus from methods to models, along two complementary lines. One is already within reach: existing tools let us characterize and compare what different models represent and compute. The other is harder, and largely neglected: whether a model can actually be understood by the humans who rely on it---the independent evaluators on whom trust and certification depend, not the experts confirming what they already expect. It can only be measured, not inferred. We review why the toolbox is mature enough to support both, survey the thin body of work comparing models, draw a parallel to systems neuroscience, and close with a model-centric XAI agenda.

Chinese Translation

经过十多年的发展，计算机视觉领域的可解释人工智能（XAI）已经建立了一个成熟的工具箱：归因、特征可视化、基于概念的方法和基于电路的方法。然而，该领域几乎所有的努力都集中在构建和比较这些方法上，而很少关注它们所要回答的问题——我们的模型有多可解释，随着模型的发展我们是否在取得进展？我们主张将该领域的重点从方法转向模型，沿着两个互补的方向进行探索。一个方向已经触手可及：现有工具使我们能够表征和比较不同模型所代表和计算的内容。另一个方向则更为困难，且在很大程度上被忽视：一个模型是否能够被依赖于它的人类理解——即那些对信任和认证至关重要的独立评估者，而不是那些确认他们已经预期的专家。这只能通过测量来实现，而不是推断。我们回顾了为什么这个工具箱足够成熟以支持这两方面的工作，调查了比较模型的稀少研究，借鉴了系统神经科学的相关性，并以以模型为中心的XAI议程作为结尾。

View on arXiv Download PDF AI Translation

cs.CV / 79 / 2609.05400

A Generalizable Feature Extractor for Alzheimer's-Related Brain MRI Tasks

一种可推广的阿尔茨海默病相关脑部MRI任务特征提取器

Rajabli, Reza, Collins, D. Louis

Abstract

When there is not enough labeled data to properly train deep learning models, transfer learning can help. We still do not fully understand how effective it is in neuroimaging, especially for Alzheimer's disease research. It is also not clear if these transferred models can work on new datasets without being retrained for each specific task. We evaluate whether a compact, supervised pretrained model can serve as a reusable foundation model for downstream neuroimaging tasks. We freeze the 7.18 million weights of a 3D CNN previously trained for brain-age prediction, and adapt it to each task using Low-Rank Adaptation (LoRA), requiring only ~1% additional trainable parameters. We evaluate generalizability in six experiments. Adapting the model to classify cognitively normal versus Dementia on ADNI gave an AUC of 0.964 on held-out folds (Experiment #1). Applying that adapted model unchanged to OASIS-3, with no retraining, gave an AUC of 0.871 (Experiment #2). Reusing its output logit together with age and a cognitive score distinguished stable from progressing MCI with an AUC of 0.828 (Experiment #3). Adapting the same backbone to predict amyloid positivity from structural MRI gave an AUC of 0.804 (Experiment #4). Finally, the same approach estimated ICV-normalized hippocampal and white matter hypointensity volumes directly from the T1w image, with R^2 of 0.80 and 0.91 respectively, tasks normally addressed with much larger U-Net networks (Experiments #5 and #6). A compact model supervised on brain age can therefore serve as a reusable backbone, adapting to each task with ~1% additional parameters and transferring to an unseen cohort without any training. Our findings suggest that a carefully trained brain age model can serve as an effective foundation model for Alzheimer's related tasks, even under strict data constraints.

Chinese Translation

当缺乏足够的标注数据以适当训练深度学习模型时，迁移学习可以提供帮助。我们仍然未完全理解其在神经影像学中的有效性，尤其是在阿尔茨海默病研究中。也不清楚这些迁移模型是否可以在新的数据集上工作，而无需针对每个特定任务进行重新训练。我们评估了一个紧凑的、经过监督的预训练模型是否可以作为下游神经影像学任务的可重用基础模型。我们冻结了一个先前用于脑龄预测的3D卷积神经网络（CNN）的718万权重，并使用低秩适应（Low-Rank Adaptation, LoRA）将其调整到每个任务，仅需约1%的额外可训练参数。我们在六个实验中评估了其可推广性。将模型调整为区分认知正常与痴呆的ADNI数据集，得到了0.964的AUC（实验#1）。将该调整后的模型不变地应用于OASIS-3，且未进行重新训练，得到了0.871的AUC（实验#2）。将其输出logit与年龄和认知评分结合使用，能够区分稳定与进展性轻度认知障碍（MCI），得到了0.828的AUC（实验#3）。将相同的基础模型调整为从结构MRI预测淀粉样蛋白阳性，得到了0.804的AUC（实验#4）。最后，采用相同的方法直接从T1加权图像估计ICV标准化的海马体和白质低强度体积，分别得到了0.80和0.91的R^2，这些任务通常需要更大的U-Net网络（实验#5和#6）。因此，一个在脑龄上经过监督的紧凑模型可以作为一个可重用的基础模型，适应每个任务仅需约1%的额外参数，并且能够在没有任何训练的情况下转移到未见过的队列。我们的研究结果表明，经过精心训练的脑龄模型可以作为阿尔茨海默病相关任务的有效基础模型，即使在严格的数据限制下。

View on arXiv Download PDF AI Translation

cs.CV / 80 / 2609.05415

UniMate: One Unified Model to Animate Diverse Skeletons

UniMate：一个统一模型以动画多样化骨架

Mou, Linzhan, Lei, Jiahui, Dou, Zhiyang, Cai, Chenyue, Song, Chaoyue, Finkelstein, Adam, Rusinkiewicz, Szymon

Abstract

Recent advances in automatic rigging now deliver animation-ready 3D assets at scale, yet generating the motion to drive them remains a bottleneck. Existing learned animators are topology-constrained: they rely on category-specific templates or require per-skeleton fine-tuning and reference motions at inference. We present UniMate, a unified foundation model that synthesizes articulated motion for arbitrary skeletons from a rigged 3D asset and a text prompt, with no test-time optimization or per-skeleton retraining. UniMate introduces a topology-aware diffusion transformer, which integrates skeletal topology into attention via three mechanisms: (1) a graph-aware attention bias from pairwise joint relations and geodesic distances; (2) a spectral rotary position embedding generalizing RoPE to arbitrary kinematic trees via the graph Laplacian; and (3) a global topological conditioner attention-pooled from the rest-pose skeleton. We also curate UniML3D, 13,006 motion sequences spanning bipedal, quadrupedal, avian, marine, insectoid, serpentine, and articulated rigid objects with unified canonicalization and text pairing. Trained on this dataset, UniMate outperforms state-of-the-art baselines in quality, generalization, and efficiency, and supports zero-shot cross-topology transfer, in-betweening, expansion, and text-guided editing. Our project page is available at https://linzhanmou.com/unimate/.

Chinese Translation

近期自动绑定技术的进步使得大规模生成动画准备好的3D资产成为可能，但驱动这些资产的运动生成仍然是一个瓶颈。现有的学习型动画生成器受到拓扑结构的限制：它们依赖于特定类别的模板，或在推理时需要针对每个骨架进行微调和参考运动。我们提出了UniMate，一个统一的基础模型，它能够从一个已绑定的3D资产和一个文本提示合成任意骨架的关节运动，而无需在测试时进行优化或针对每个骨架进行重新训练。UniMate引入了一种拓扑感知的扩散变换器，通过三种机制将骨架拓扑集成到注意力机制中：(1) 基于关节对之间关系和测地距离的图感知注意力偏置；(2) 一种谱旋转位置嵌入（spectral rotary position embedding），通过图拉普拉斯算子将RoPE推广到任意运动学树；(3) 从静止姿态骨架中池化的全局拓扑条件注意力。我们还整理了UniML3D，包含13,006个运动序列，涵盖双足、四足、鸟类、水生、生物昆虫、蛇形和关节刚体对象，并进行了统一的规范化和文本配对。在此数据集上训练后，UniMate在质量、泛化能力和效率上超越了最先进的基线，并支持零-shot跨拓扑迁移、插值、扩展和文本引导编辑。我们的项目页面可访问 https://linzhanmou.com/unimate/。

View on arXiv Download PDF AI Translation

cs.CV / 81 / 2609.05416

WorldSculpt: Generating Compositional Worlds from Grounded Videos

WorldSculpt：从基础视频生成组合世界

Niu, Muyao, He, Jixuan, Yu, Ruihan, Fu, Lian, Yu, Yonghao, Huang, Zheng-Hui, Zhan, Yifan, Lan, Fengbo, Ge, Yongtao, Zheng, Yinqiang, Zhang, Kaipeng, Wang, Zhixiang

Abstract

We study the problem of generating a compositional 3D representation of a cluttered scene containing hundreds of objects. The goal is to represent the scene as a collection of individual object meshes placed in a shared world frame, as required by downstream applications such as gaming, AR/VR, simulation, and robotics. This task is challenging in densely cluttered scenes, where objects heavily occlude one another and each view reveals only a fraction of their geometry. Geometry-based approaches typically reconstruct the scene as a single representation and leave incomplete geometry in occluded regions, while existing compositional methods with generative priors are largely limited to relatively simple scenes. We show that complex scenes with hundreds of objects can instead be generated compositionally by adapting a strong single-object 3D generative prior to multi-view observations. We instantiate this paradigm with Pixal3D, extending it with a multi-view conditioning pathway that grounds object generation in multiple posed observations. Although the model is finetuned entirely on single objects in canonical space, it generalizes to large scenes with severe occlusion without any scene-level training, demonstrating the feasibility and scalability of this paradigm. We further introduce UE-MeshyScene, a photorealistic benchmark of densely cluttered scenes with hundreds of objects, per-object annotations, and ground-truth meshes. Across single-object, controlled multi-object, and UE-MeshyScene evaluations, our method consistently outperforms prior approaches, with larger gains as scene complexity and occlusion increase. Finally, we demonstrate broader applicability by converting generated 3DGS worlds, such as Marble and HY-World 2.0, into compositional mesh scenes.

Chinese Translation

我们研究了生成包含数百个物体的杂乱场景的组合3D表示的问题。目标是将场景表示为一组单独的物体网格，置于共享的世界坐标系中，以满足游戏、增强现实/虚拟现实、仿真和机器人等下游应用的需求。这个任务在密集杂乱的场景中具有挑战性，因为物体之间严重遮挡，并且每个视角仅揭示其几何形状的一部分。基于几何的方法通常将场景重建为单一表示，并在遮挡区域留下不完整的几何形状，而现有的具有生成先验的组合方法在很大程度上仅限于相对简单的场景。我们展示了通过将强大的单物体3D生成先验适应于多视角观察，可以生成包含数百个物体的复杂场景。我们通过Pixal3D实例化这一范式，并扩展了一个多视角条件路径，使物体生成基于多个姿态观察。尽管该模型完全在典型空间的单个物体上进行了微调，但它在没有任何场景级训练的情况下，能够推广到具有严重遮挡的大型场景，展示了这一范式的可行性和可扩展性。我们进一步引入了UE-MeshyScene，这是一个包含数百个物体的密集杂乱场景的照片真实基准，具有每个物体的注释和真实网格。在单物体、受控多物体和UE-MeshyScene评估中，我们的方法始终优于之前的方法，随着场景复杂性和遮挡程度的增加，性能提升更为显著。最后，我们通过将生成的3DGS世界（如Marble和HY-World 2.0）转换为组合网格场景，展示了更广泛的适用性。

View on arXiv Download PDF AI Translation

人工智能 (Artificial Intelligence)

106

cs.AI / 1 / 2609.04239

EXAONE Forecast for Finance

EXAONE金融预测

Lee, Seunghan, Lee, Jaehoon, Seo, Jun, Lim, Tae Yoon, Kang, Dongwan, Choi, Hwanil, Kim, Minjae, Yoo, Sungdong, Kang, Junhyeok, Han, Sangjun, Lee, Soonyoung, Ahn, Wonbin

Abstract

This technical report presents EXAONE Forecast for Finance (EXAONE Finance), a financial time series (TS) foundation model (TSFM) tailored to financial forecasting. Recent TSFMs achieve strong zero-shot performance through large-scale pretraining. However, they are primarily developed for general-domain TS and largely rely on self-attention backbones whose computational cost grows quadratically with sequence length and variate count. Moreover, they assume fully observed inputs and are pretrained on corpora that fail to capture the unique dynamics of financial markets. These limitations hinder their applicability to finance, where long, many-channel, intermittently observed panels are common. To address these challenges, EXAONE Finance adopts an attention-free architecture, replacing self-attention with two simple yet effective linear-time operators: 1) a causal 1D convolution for temporal mixing and 2) a group-aware pooling multi-layer perceptron (MLP) for variate mixing. Furthermore, a masked context augmentation exposes the model to contiguous missing spans during training, improving robustness to the missingness pervasive in financial markets. EXAONE Finance is pretrained on a large-scale financial corpus covering not only equities but also foreign exchange, commodities, crypto-assets, fixed income, and macroeconomic indicators. On FinVerse, a financial forecasting benchmark covering diverse asset classes, EXAONE Finance attains state-of-the-art performance, ranking first across all three evaluation tiers---point-forecast accuracy, cross-sectional asset ranking, and portfolio profitability.

Chinese Translation

本技术报告介绍了EXAONE金融预测（EXAONE Finance），这是一种针对金融预测量身定制的金融时间序列基础模型（TSFM）。近期的时间序列基础模型在大规模预训练下实现了强大的零-shot性能。然而，这些模型主要是为通用领域的时间序列开发的，且在很大程度上依赖于自注意力骨干网络，其计算成本随着序列长度和变量数量的增加呈平方增长。此外，它们假设输入完全可观察，并在未能捕捉金融市场独特动态的语料库上进行预训练。这些局限性妨碍了它们在金融领域的应用，而在金融领域，长时间、多通道、间歇性观察的面板数据是常见的。为了解决这些挑战，EXAONE Finance采用了一种无注意力架构，用两个简单而有效的线性时间算子替代自注意力：1）用于时间混合的因果一维卷积和2）用于变量混合的群体感知池化多层感知器（MLP）。此外，掩蔽上下文增强在训练过程中使模型暴露于连续缺失的区间，从而提高了对金融市场普遍存在的缺失现象的鲁棒性。EXAONE Finance在一个涵盖股票、外汇、商品、加密资产、固定收益和宏观经济指标的大规模金融语料库上进行了预训练。在涵盖多种资产类别的金融预测基准FinVerse上，EXAONE Finance达到了最先进的性能，在所有三个评估层面——点预测准确性、横截面资产排名和投资组合盈利能力中均排名第一。

View on arXiv Download PDF AI Translation

cs.AI / 2 / 2609.04286

From Matching Models to Recruiting Agents: A Systematized Narrative Review of AI Recruitment Systems, Evaluation, and Governance

从匹配模型到招聘代理：人工智能招聘系统、评估与治理的系统化叙述性综述

Zhao, Ziyi, Wei, Guanzheng

Abstract

Artificial intelligence in recruitment has shifted the object being automated from profile pairs and ranked lists to multi-stage workflows that retrieve evidence, compare candidates, and support or execute actions. This systematized narrative review traces that development from bilateral retrieval and behavioral ranking through neural person--job matching, large language model (LLM) components, and tool-using recruiting agents. Using a purposive search and coding protocol updated through 23 July 2026, plus targeted updates through 2 September 2026, we organize 40 representative works with supporting industrial and legal sources. This synthesis is not a prevalence estimate. We analyze three coupled transitions: from similarity to reciprocal suitability, from a model to a compound workflow, and from offline prediction to evidence- and productivity-aligned evaluation. Across document understanding, retrieval, ranking, assessment, interviewing, sourcing, and human handoff, we distinguish field-, pair-, list-, case-, trajectory-, and outcome-level evidence. Persistent gaps arise because behavioral labels confound exposure, preference, and qualification; private and synthetic data limit external validity; final-output scores conceal pipeline failures; and, within the coded set, privacy is not directly evaluated and no row jointly evaluates utility, fairness, privacy, and security. These observations describe the coded set rather than the field as a whole. We therefore introduce a staged mapping from evaluation evidence to the strongest defensible claim, together with an agenda for reciprocal, evidence-grounded, temporally controlled, selective, and auditable systems. Progress should be judged by whether workflows retrieve the right evidence, preserve uncertainty, support contestable decisions, and improve outcomes under explicit cost and risk constraints.

Chinese Translation

人工智能在招聘中的应用已将自动化对象从个人资料对和排名列表转变为多阶段工作流程，这些工作流程检索证据、比较候选人并支持或执行行动。本系统化叙述性综述追溯了这一发展历程，从双边检索和行为排名到神经人职匹配、大型语言模型（LLM）组件以及使用工具的招聘代理。通过截至2026年7月23日的有目的搜索和编码协议，以及截至2026年9月2日的针对性更新，我们整理了40部具有代表性的作品，并提供了相关的工业和法律来源。这一综合并非流行程度的估计。我们分析了三个相互关联的转变：从相似性到互惠适宜性，从模型到复合工作流程，以及从离线预测到基于证据和生产力的评估。在文档理解、检索、排名、评估、面试、来源和人力交接等方面，我们区分了领域、对、列表、案例、轨迹和结果级别的证据。由于行为标签混淆了曝光、偏好和资格，私人和合成数据限制了外部有效性，最终输出分数掩盖了管道失败，而在编码集内，隐私未被直接评估且没有任何行共同评估效用、公平、隐私和安全，因此持续存在差距。这些观察描述的是编码集而非整个领域。因此，我们引入了一种从评估证据到最强有力的可辩护主张的分阶段映射，以及一个针对互惠、基于证据、时间控制、选择性和可审计系统的议程。进展应通过工作流程是否检索到正确的证据、保留不确定性、支持有争议的决策以及在明确的成本和风险约束下改善结果来评判。

View on arXiv Download PDF AI Translation

cs.AI / 3 / 2609.04298

Harbor Adapters and Harbor-Index: Infrastructure and a Curated Meta-Dataset for Large-Scale Agentic Evaluation

港口适配器与港口索引：大规模智能评估的基础设施与策划元数据集

Shi, Lin, Lin, Haowei, Zhu, Zixuan, Zhou, Xiaoyue, Li, Xiang, Lin, Xiangning, Deng, Yaxuan, Xu, Han, Li, Yuangang, Li, Shanda, Chen, Zizhao, Xing, Hanwen, Raj, Harsh, Chen, Bo, Shi, Quan, Dillmann, Steven, Gao, Yipeng, Khanna, Puneesh, Lu, Ruofan, Zhou, Chao Beyond, Yang, Michael, Zhang, Robert, Chai, Siyuan, Chang, Jiayu, Chen, Yizhao, Chen, Xiaokun, Dai, Yiwei, Yang, Wenting, Liu, Hange, Liu, Minghao, Wang, Zihan, Assadi, Adnan El, Stroebl, Benedikt, Buchanan, E. Kelly, Meng, Han, He, Junwei, Yu, Longxuan, Shayanfar, Radin, Lee, Yukyung, Dong, Zhikang, Hart, Allen G, Wei, Anjiang, Kashyap, Anurag, Khatua, Arpandeep, Zheng, Audrey Jixin, Ma, Chengrui, Heineman, David, Chen, Dubing, Trinh, Hai-Anh, Fang, Haishuo, Zhang, Hefan, Shen, Hui, Sugiura, Issa, Sun, Jiankai, Gao, Jiechao, Lin, Junhong, Li, Junnan, Yang, Kai, Hsiung, Lei, Wang, Maoyu, Tang, Mengze, Omi, Nabil, Raoof, Negin, Edwards, Nicholas, Guo, Octavia, Mastromichalakis, Orfeas Menis, Ji, Pengliang, Hejman, Przemysław, Qi, Qi, Lin, Qunshu, Zhuang, Richard, Yang, Rui, Zheng, Ruichen, Marten, Ryan, Fazliani, Shaghayegh, Hou, Shizheng, Jiang, Sicong, Li, Sijie, Bian, Song, Zhuo, Terry Yue, Wu, Tianqing, Tang, Tom, Zhao, Wanjia, Xuan, Weihao, Liang, Wenhua, Liu, Xian, Lan, Xin, Zhang, Xuan, Zhao, Xuandong, Tang, Yanchuan, Jiang, Yifan, Li, Yijiang, Guan, Yitong, Li, Yizhi, Liu, Yonghui, Tang, Yuheng, Yujun, Mao, Zhao, Yunfei, Wang, Yuxin, Tang, Yuxuan, Tang, Zhenheng, Li, Zhifei, Wang, Ziruo, She, Ziyu, Liu, Kaiyuan, Chaabane, Iheb, Tang, Yuxin, Li, Xiangyi, Konwinski, Andy, Li, Boxuan, Chen, Leon Liangyu, Dimakis, Alex, Carlini, Nicholas, Vosoughi, Soroush, He, Di, Guha, Etash, Feuer, Benjamin, Merrill, Mike, Schmidt, Ludwig, Shaw, Alex

Abstract

Evaluating agents on the growing number of agentic benchmarks is challenging because they often require complex environments and agent integrations. We introduce Harbor Adapters, a unified evaluation infrastructure for agentic benchmarks. Our work makes three contributions. First, we develop benchmark adapters that port more than 80 benchmarks to evaluate arbitrary agents, and validate them through rigorous code review and parity experiments. Second, we conduct a large-scale evaluation of 8 models spanning capability tiers across 54 benchmarks; every model is run with Terminus-2 and with one of 3 native harnesses. This enables a broader analysis of agent capabilities and failure modes than was previously possible. Third, we introduce Harbor-Index, a curated set of 82 difficult, diverse, and high-quality tasks spanning 29 benchmarks, refined from the adapted suite through difficulty filtering, AI and human audit, and an audit-and-fix loop. Harbor-Index preserves the challenge and breadth of large-scale agentic evaluations while being affordable to run; no evaluated model-harness configuration exceeds 30% pass rate, and the strongest (GPT-5.5 with Codex) reaches 28.0%. We release the adapters, evaluation results, in-depth analysis, and Harbor-Index as open-source artifacts to support more reliable and comprehensive evaluation of language-model agents.

Chinese Translation

在日益增长的智能基准上评估智能体面临挑战，因为它们通常需要复杂的环境和智能体集成。我们介绍了港口适配器（Harbor Adapters），这是一个统一的智能基准评估基础设施。我们的工作有三个贡献。首先，我们开发了基准适配器，将80多个基准移植到评估任意智能体，并通过严格的代码审查和对比实验验证了它们。其次，我们对8个模型进行了大规模评估，这些模型跨越54个基准，涵盖了不同的能力层次；每个模型都使用Terminuis-2运行，并配备3种原生测试框架之一。这使得对智能体能力和失效模式的更广泛分析成为可能。第三，我们引入了港口索引（Harbor-Index），这是一个策划的82个困难、多样且高质量的任务集，涵盖29个基准，通过难度筛选、人工智能和人工审核以及审核与修复循环从适配套件中提炼而成。港口索引在保持大规模智能评估的挑战性和广度的同时，运行成本可承受；没有评估的模型-测试框架配置的通过率超过30%，而最强的模型（GPT-5.5与Codex）达到了28.0%。我们将适配器、评估结果、深入分析和港口索引作为开源文献发布，以支持对语言模型智能体的更可靠和全面的评估。

View on arXiv Download PDF AI Translation

cs.AI / 4 / 2609.04300

Data-Optimized Contingency Screening: A Machine Learning Approach to Power System Security

数据优化的应急筛查：一种机器学习方法用于电力系统安全

Salako, Joshua, Osikomaiya, Folajimi, Olamiju, Olakorede

Abstract

Ensuring the security of the power system is essential for stability and reliability, especially in the event of disruption. Effective classification of contingency in power systems enables proactive decision-making and mitigates large-scale breakdowns and failures. This study explores the use of machine learning algorithms to classify security levels of contingencies in power systems into safe, moderate or severe classes. For this approach, Newton-Raphson load flow method extracts system data from contingency scenarios, using Overall Performance Index (OPI) as safety measure. For data pre-processing, Synthetic Minority Over-Sampling Technique (SMOTE) and Principal Component Analysis (PCA) is used to address class imbalance and reduce dimensionality, respectively. K-Nearest Neighbours (KNN), Random Forest (RF) and Support Vector Machines (SVM) is trained and evaluated on datasets generated through N-k contingency scenarios for k equal 1, 2, and 3 on IEEE-14 and IEEE-30 bus systems using four hybrid pre-processing configurations: normalized, SMOTE-balanced, PCA-transformed, and a combined SMOTE PCA-transformed. Performance is assessed by precision, recall and F1 score, with priority given to the severe contingency classes. The RF achieved the highest F1 scores of 0.97 in IEEE-30 and 0.86 in IEEE-14, SVM benefits significantly from PCA and improves the accuracy of the classification, while KNN is best suited for SMOTE and PCA conversion. The findings show that PCA contributes more than SMOTE to the overall performance of the model. However, SMOTE improves recall but can introduce false positives and is therefore a compromise of accuracy. This study highlights machine learning as a scalable and powerful alternative to traditional contingency analysis, which improves the assessment of security in real time.

Chinese Translation

确保电力系统的安全对于稳定性和可靠性至关重要，尤其是在发生干扰时。有效的电力系统应急分类能够促进主动决策，减轻大规模故障和崩溃的风险。本研究探讨了使用机器学习算法将电力系统中的应急安全级别分类为安全、中等或严重等级。为此，牛顿-拉夫森负荷流方法从应急场景中提取系统数据，使用整体性能指数（Overall Performance Index, OPI）作为安全指标。在数据预处理方面，采用合成少数类过采样技术（Synthetic Minority Over-Sampling Technique, SMOTE）和主成分分析（Principal Component Analysis, PCA）分别解决类别不平衡和降维问题。K-近邻（K-Nearest Neighbours, KNN）、随机森林（Random Forest, RF）和支持向量机（Support Vector Machines, SVM）在通过N-k应急场景生成的数据集上进行训练和评估，其中k取1、2和3，使用四种混合预处理配置：标准化、SMOTE平衡、PCA转换以及结合SMOTE和PCA转换。通过精确度、召回率和F1分数评估性能，优先考虑严重应急类别。RF在IEEE-30系统中获得了最高的F1分数0.97，在IEEE-14系统中为0.86，SVM在PCA的显著帮助下提高了分类准确性，而KNN最适合SMOTE和PCA转换。研究结果表明，PCA对模型的整体性能贡献大于SMOTE。然而，SMOTE提高了召回率，但可能引入假阳性，因此在准确性上存在妥协。本研究强调了机器学习作为传统应急分析的可扩展且强大的替代方案，能够实时改善安全评估。

View on arXiv Download PDF AI Translation

cs.AI / 5 / 2609.04304

Iris: Climbing to the Search Frontier

Iris：攀登搜索前沿

Liu, Ziyuan, Liu, Hengqi, Wang, Zichuan, Qin, Yang, Liang, Jiachen, Chu, Xu, Chen, Shaowei, Gu, Yuantao, Chuan, Mu

Abstract

We present Iris-mini and Iris-pro, two search agents trained at the 35B-A3B and 397B-A17B scales, together with the data pipeline and training recipe behind them. Tasks are reverse-constructed from the hyperlink structure of a web corpus: we author multi-hop chains over an entity graph distilled from a seed page and its out-links, rewrite every non-answer entity into a descriptive reference so that no clue can be resolved by string matching, and admit only questions that a reference model fails closed-book yet solves once the supporting evidence is supplied. These questions are then turned into trajectories, which are filtered at both the trajectory and the turn level before SFT. The policy is then optimized by RL against live search, with the reward judge and the observation summarizer served inside the training cluster, and with over-long rollouts interrupted at the request level and resumed from their committed prefix at the next step. We alternate the two stages in a procedure we call SFT-RL climbing, returning the hardest solved and most efficient rollouts of each RL round to the next supervised pass. Because inference-time context management is worth more on these benchmarks than most reported differences between systems, we evaluate every benchmark both with and without it, holding the tool set, the context limit, and the judge fixed. All results come from a single ReAct agent, with no sub-agents and no test-time verification. With management enabled, on BrowseComp, BrowseComp-ZH, DeepSearchQA, and HLE the two models reach $82.2/84.8/86.9/52.3$ and $88.6/85.1/92.9/56.4$, the strongest overall results among open-source search agents in their respective parameter ranges. We plan to release the model weights together with the complete recipe for data construction, training, and evaluation.

Chinese Translation

我们提出了Iris-mini和Iris-pro这两个搜索代理，它们分别在35B-A3B和397B-A17B规模上进行训练，并介绍了其背后的数据管道和训练方案。任务是从网络语料库的超链接结构反向构建的：我们在从种子页面及其外部链接提取的实体图上编写多跳链，将每个非答案实体重写为描述性参考，以确保没有线索可以通过字符串匹配解决，并仅接受参考模型在闭卷情况下无法解决但在提供支持证据后能够解决的问题。这些问题随后被转化为轨迹，在SFT之前在轨迹和转弯级别进行过滤。然后，通过RL对实时搜索进行优化，奖励评估和观察摘要在训练集群内提供，并且过长的回滚在请求级别被中断，并在下一步从其已承诺的前缀恢复。我们在一个称为SFT-RL climbing的程序中交替进行这两个阶段，将每个RL轮次中解决的最困难和最高效的回滚返回到下一个监督阶段。由于推理时间的上下文管理在这些基准测试中比系统之间报告的大多数差异更具价值，我们在保持工具集、上下文限制和评估者固定的情况下，对每个基准测试进行了有和没有上下文管理的评估。所有结果均来自单个ReAct代理，没有子代理和测试时间验证。在启用管理的情况下，在BrowseComp、BrowseComp-ZH、DeepSearchQA和HLE上，这两个模型分别达到了$82.2/84.8/86.9/52.3$和$88.6/85.1/92.9/56.4$，在各自参数范围内的开源搜索代理中取得了最强的整体结果。我们计划发布模型权重以及数据构建、训练和评估的完整方案。

View on arXiv Download PDF AI Translation

cs.AI / 6 / 2609.04343

A Removal Based Approach to Improve LLM Faithfulness at Test-Time

基于移除的方法提升大语言模型在测试时的可信度

Luo, Qinglan, Nahian, S M A, Guttag, John, Abulnaga, S. Mazdak, Matton, Katie

Abstract

Large language models (LLMs) are increasingly used for consequential decisions, making their explanations an important tool for auditing model behavior. Unfortunately, these explanations can be unfaithful, failing to reflect the actual reasoning underlying the model's decisions. We consider a setting in which an LLM provides both an answer and an explanation in response to a question. We identify two distinct dimensions of unfaithful explanations: incompleteness, meaning that the explanation omits factors that influence the answer, and unsoundness, meaning that the explanation cites factors that did not influence the model's answer. Existing approaches to improving LLM faithfulness include training-time methods, which require access to model weights and extensive computational resources, and test-time methods that largely focus on addressing unsoundness. We introduce a test-time approach that directly targets incompleteness. We remove from the input the concepts not credited in the model's explanation and re-query the model on the reduced input. This eliminates unmentioned influences while preserving the influence of mentioned concepts. Across two datasets, multiple model families, and two independent faithfulness metrics, our approach improves explanation faithfulness compared to both standard prompting and prompting to encourage faithfulness. Our method is model-agnostic and can be applied at inference time without modifying model parameters, providing a flexible mechanism for reducing hidden influences and improving the reliability and safety of LLM-assisted decision making.

Chinese Translation

大型语言模型（LLMs）越来越多地用于重要决策，使得其解释成为审计模型行为的重要工具。不幸的是，这些解释可能不够可信，未能反映模型决策背后的实际推理。我们考虑一种情境，其中LLM在回答问题时同时提供答案和解释。我们识别出不可信解释的两个不同维度：不完整性，指的是解释省略了影响答案的因素；以及不合理性，指的是解释引用了未影响模型答案的因素。现有的提高LLM可信度的方法包括训练时方法，这需要访问模型权重和大量计算资源，以及主要关注解决不合理性的问题的测试时方法。我们提出了一种直接针对不完整性的测试时方法。我们从输入中移除模型解释中未提及的概念，并在减少后的输入上重新查询模型。这消除了未提及的影响，同时保留了已提及概念的影响。在两个数据集、多个模型系列和两个独立的可信度指标上，我们的方法相比于标准提示和鼓励可信度的提示，提升了解释的可信度。我们的方法是模型无关的，可以在推理时应用而无需修改模型参数，提供了一种灵活的机制来减少隐藏影响，提高LLM辅助决策的可靠性和安全性。

View on arXiv Download PDF AI Translation

cs.AI / 7 / 2609.04373

Why Better Models Can Create Riskier Systems: Evidence from LLM Agents in Financial Markets

为何更好的模型可能导致更高风险的系统：来自金融市场中大语言模型代理的证据

Ross, Jillian, So, Eric, De Simone, Zoe, Pozniak, Charles, Lo, Andrew W.

Abstract

Large language models (LLMs) are being deployed at scale in consequential real-world systems, from financial markets to content moderation to hiring. We show that improving individual model capability can degrade rather than improve system-level outcomes. We hypothesize that shared training and architectures can lead more capable LLMs to behave more similarly, creating correlated actions that do not diversify away. We develop a general framework showing how this correlation creates a non-diversifiable risk floor and test its predictions in financial markets using an agent-based simulation with LLM traders of varying general-purpose capability. We find that: (1) frontier LLMs exhibit significantly correlated behavior that increases with capability; (2) when their shared reasoning is accurate, increasing agent participation reduces market-level risk; and (3) when agents share a common misinformation environment, the same correlated behavior becomes a liability. Together, these results identify a capability paradox: improving individual models does not necessarily produce better system-level outcomes. Whether the same dynamics arise in other domains is an open empirical question.

Chinese Translation

大语言模型（LLMs）正在大规模应用于重要的现实世界系统中，从金融市场到内容审核再到招聘。我们表明，提升单个模型的能力可能会导致系统层面的结果恶化而非改善。我们假设，共享的训练和架构可能导致更强大的LLM表现得更为相似，从而产生不易分散的相关行为。我们开发了一个通用框架，展示这种相关性如何形成一个不可分散的风险底线，并通过使用具有不同通用能力的LLM交易者的基于代理的模拟在金融市场中测试其预测。我们的研究发现：（1）前沿LLM表现出显著的相关行为，且随着能力的提升而增加；（2）当它们的共享推理准确时，增加代理参与可以降低市场层面的风险；（3）当代理共享一个共同的错误信息环境时，相同的相关行为则成为一种负担。这些结果共同揭示了一个能力悖论：提升单个模型并不一定能产生更好的系统层面结果。在其他领域是否会出现相同的动态仍然是一个开放的实证问题。

View on arXiv Download PDF AI Translation

cs.AI / 8 / 2609.04377

Corporate Language Model (CLM): Transforming Tacit and Fragmented Enterprise Knowledge into a Sovereign, Auditable, and Executable Corporate Intelligence Layer

企业语言模型（CLM）：将隐性和碎片化的企业知识转化为主权、可审计和可执行的企业智能层

Avini, Fabricio C., Trez, Guilherme

Abstract

Enterprise AI deployments fail not from model inadequacy, but because organizations lack a structured substrate encoding how they decide, negotiate, and execute. Generic LLMs carry no firm-specific ontological priors; RAG remains brittle, with no path to executable action; static playbooks encode logic but cannot reason or adapt. This demands an architecture treating tacit-knowledge capture, ontological grounding, sovereign deployment, and auditable actuation as co-designed from the start. This paper introduces the Corporate Language Model (CLM), a framework transforming a firm's structured, unstructured, multimodal, and tacit knowledge into an ontology-grounded enterprise foundation upon which reasoning and governed execution are composed. CLM has five capability planes and four architectural pillars: a Neurosymbolic Mesh coupling generative models with a knowledge graph; a Skill Graph where reusable tactics, personas, objections, and goals are typed and composed; Living Digital Twins modeling functional areas as reasoning surrogates; and a Deep Security Layer enforcing sovereignty, traceability, and human oversight. A Spec-as-Code paradigm bridges grounded intent and executable artifact. CLM is one instantiation of this foundation-centric class. Four contributions follow: CLM is defined as a distinct object of study; the Skill Graph is introduced for compositional explainability by construction; the Wisdom Listener effect is proposed, whereby tacit-capable foundations compound in value with use, connecting to dynamic capabilities and organizational learning; and evidence from a JCI-accredited tertiary hospital in Brazil instantiates three of the six maturity stages under LGPD.

Chinese Translation

企业人工智能部署的失败并非由于模型不足，而是因为组织缺乏一种结构化的基础，编码其决策、谈判和执行的方式。通用的大型语言模型（LLMs）没有特定于公司的本体先验；检索增强生成（RAG）仍然脆弱，没有可执行行动的路径；静态操作手册编码逻辑，但无法推理或适应。这要求一种架构，从一开始就将隐性知识捕获、本体基础、主权部署和可审计激活视为共同设计。本文介绍了企业语言模型（CLM），这是一个将公司的结构化、非结构化、多模态和隐性知识转化为一个本体基础的企业基础框架，在此基础上进行推理和受控执行。CLM具有五个能力层面和四个架构支柱：一个神经符号网格，将生成模型与知识图谱相结合；一个技能图谱，其中可重用的战术、角色、反对意见和目标被分类和组合；生活数字双胞胎，模拟功能领域作为推理替代物；以及一个深度安全层，强制执行主权、可追溯性和人类监督。规范即代码（Spec-as-Code）范式架起了基础意图与可执行工件之间的桥梁。CLM是这一以基础为中心的类别的一个实例。接下来有四个贡献：CLM被定义为一个独特的研究对象；引入技能图谱以通过构建实现组合可解释性；提出智慧倾听者效应，即隐性能力基础随着使用而增值，连接到动态能力和组织学习；以及来自巴西一家JCI认证的三级医院的证据实例化了LGPD下六个成熟阶段中的三个。

View on arXiv Download PDF AI Translation

cs.AI / 9 / 2609.04444

HarvestBench: Measuring Whether LLM Agents Will Pay to Avoid Killing Animals

HarvestBench：测量大型语言模型代理是否愿意付费以避免杀死动物

Brazilek, Jasmine, Tidmarsh, Miles, Endres, Matthias, Singh, Anshuman, Miller, Jeremiah

Abstract

Benchmarks for the side effects an agent causes on the way to a goal already exist, but HarvestBench is the first to put a price on avoiding the side effect and to name that side effect as a living creature. It is a farm simulation: LLM sub-agents drive a crew of two tractors through a cooperative corn harvest, with animals in the field. The environment is a reinforcement learning gridworld, every decision is made without memory, and the harm is never named in the goal. When an animal blocks a tractor's route the autopilot stops and asks the model whether to drive on, at no fuel cost, or swerve around it for a posted fuel price. Kills are compared against two controls: rocks, which damage the tractor and are hit under 1% of the time by every model, and hay bales, which are harmless and not alive. Models can also take crops from the neighbor's field instead of their own, a second test of what they treat as moral. Across nine models and 7,201 priced decisions, 3,951 involved an animal rather than a hay bale or a rock. Kill rates range from 0.4% to 98.8%, with Terra and Sol the most merciful and GPT-4o-mini the most cruel, and they are not ordered by capability. Four of six models were sensitive to price at the 5% level, with elasticities from 0.09 to 1.69. All nine drove over wild animals more often than farmed animals on the default map, and the direction held at every map geometry in every model with room to move. The briefing mattered most: under the morality briefing the kill rate was under 6% in five of six reasoning models, and removing it raised the kill rate above 84% in all six. HarvestBench uses no LLM grader. The scorer counts events in the game log, so it is fully reproducible, and it measures what a model will pay to avoid harm rather than what it says about harm.

Chinese Translation

虽然已经存在代理在实现目标过程中造成副作用的基准，但HarvestBench是首个为避免副作用定价，并将该副作用命名为生物体的基准。它是一个农场模拟：大型语言模型（LLM）子代理驾驶两台拖拉机在合作的玉米收获中穿行，田野中有动物。环境是一个强化学习的网格世界，每个决策都是在没有记忆的情况下做出的，伤害从未在目标中被提及。当一只动物阻挡了拖拉机的路径时，自动驾驶系统会停止并询问模型是否在没有燃料成本的情况下继续前进，或者绕过它并支付标定的燃料费用。杀戮率与两个控制组进行比较：岩石，这会损坏拖拉机，每个模型被撞击的概率不到1%；以及干草堆，这些是无害且不活的。模型还可以从邻居的田地中收获作物，而不是自己的，这也是对其道德观的一次测试。在九个模型和7,201个定价决策中，3,951个涉及动物，而不是干草堆或岩石。杀戮率范围从0.4%到98.8%，其中Terra和Sol是最仁慈的，而GPT-4o-mini是最残忍的，且它们的能力并未按顺序排列。在六个模型中，有四个对价格在5%的水平上敏感，弹性范围从0.09到1.69。所有九个模型在默认地图上对野生动物的碾压频率高于对农场动物的碾压频率，这一趋势在每个模型的每个地图几何形状中都保持一致。简报内容至关重要：在道德简报下，六个推理模型中有五个的杀戮率低于6%，而去除该简报则使所有六个模型的杀戮率上升至84%以上。HarvestBench不使用LLM评分器。评分者在游戏日志中统计事件，因此其结果完全可重复，并且它测量的是模型愿意为避免伤害而支付的费用，而不是模型对伤害的表述。

View on arXiv Download PDF AI Translation

cs.AI / 10 / 2609.04476

PerfReasoning: How Well Do LLMs Reason on Hardware Performance?

PerfReasoning：大型语言模型在硬件性能推理上的表现如何？

Zhao, Dan, Sankaralingam, Karthikeyan, Kozyrakis, Christos, Huang, Qijing

Abstract

Performance modeling is central to hardware design and software optimization, yet constructing these models requires structured reasoning about computation, data reuse, storage, and movement. We introduce PerfReasoning, a benchmark that evaluates LLMs both as direct performance reasoners and as generators of analytical performance-model code. Given workload, architecture, and mapping specifications, models compare mappings and predict off-chip traffic and buffer requirements. The strongest closed-source models exceed 90% on reasoning-based Q&A, and the best open-weight model reaches 82.4%. However, model construction is substantially harder: while GPT-5.6 Sol exceeds 80% pass rate, all other model configurations average below 15% and vary markedly across runs. Task-specific RL raises a 4B model's mapping-reasoning accuracy by 15.7 points, whereas feedback-free multi-round self-revision prompting is not reliably effective. PerfReasoning exposes the gap between plausible architectural reasoning and reliable performance-model construction. We will publicly release the benchmark to support reproducible evaluation and track future progress.

Chinese Translation

性能建模是硬件设计和软件优化的核心，但构建这些模型需要对计算、数据重用、存储和移动进行结构化推理。我们介绍了PerfReasoning，这是一个基准测试，用于评估大型语言模型（LLMs）作为直接性能推理者和分析性能模型代码生成器的能力。在给定工作负载、架构和映射规范的情况下，模型比较映射并预测芯片外流量和缓冲需求。最强的闭源模型在基于推理的问答中超过90%的准确率，而最佳的开放权重模型达到了82.4%。然而，模型构建的难度显著更大：尽管GPT-5.6 Sol的通过率超过80%，但所有其他模型配置的平均通过率低于15%，且在不同运行中差异显著。特定任务的强化学习将一个4B模型的映射推理准确率提高了15.7个百分点，而无反馈的多轮自我修正提示并不可靠有效。PerfReasoning揭示了合理的架构推理与可靠的性能模型构建之间的差距。我们将公开发布该基准，以支持可重复评估并跟踪未来的进展。

View on arXiv Download PDF AI Translation

cs.AI / 11 / 2609.04490

When Quantization Breaks Memory: Recurrent-State Write-Back in Low-Precision Temporal Inference

量化如何影响内存：低精度时间推理中的递归状态写回

Erbas, Ismail, Intes, Xavier, Pandey, Vikas

Abstract

Quantization is widely used to reduce the computational and memory demands of neural-network inference. In recurrent networks, however, the quantized state is stored and returned at the next time step, so the rule used to store that state can alter subsequent computations. Here, we introduce recurrent-state write-back to denote this rule and isolate its effect in a compact GRU encoder--decoder for fluorescence lifetime imaging, a molecular imaging modality used in quantitative biological imaging. A central task is estimating two lifetime parameters, the short-lived component {\tau}1 and the long-lived component {\tau}2, from high-noise time-resolved fluorescence signals. Holding the trained model fixed, replacing continuous state propagation with deterministic 4-bit state storage increases estimation errors for {\tau}1 and {\tau}2 by approximately 70x and 300x, respectively. Failure occurs when repeated small updates remain below the write threshold, leaving the stored state nearly fixed while the network continues to propose change. Error feedback, residual memory, and direction memory carry information from these suppressed updates across time and recover accuracy without retraining. Precision sweeps show that increasing state precision can worsen a fixed recurrent solution, while matched training shows that compatibility with the state interface can be learned. To test whether this behavior extends beyond the GRU, we repeat the post-training intervention in an independently trained LSTM, where coarse write-back reproduces the failure, error feedback restores accuracy, and state-specific interventions reveal greater sensitivity of the cell state than the hidden state. Our results establish recurrent-state write-back as a key determinant of low-precision recurrent dynamics and identify the state-storage interface as a central design consideration for quantized recurrent inference.

Chinese Translation

量化广泛用于减少神经网络推理的计算和内存需求。然而，在递归网络中，量化状态在下一个时间步存储并返回，因此用于存储该状态的规则可能会改变后续计算。在此，我们引入递归状态写回（recurrent-state write-back）来表示这一规则，并在用于定量生物成像的分子成像模式——荧光寿命成像的紧凑型 GRU 编码器-解码器中隔离其影响。一个核心任务是从高噪声的时间分辨荧光信号中估计两个寿命参数，即短寿命成分 { au}1 和长寿命成分 { au}2。保持训练模型不变，将连续状态传播替换为确定性的 4 位状态存储，使得 { au}1 和 { au}2 的估计误差分别增加约 70 倍和 300 倍。当重复的小更新保持在写入阈值以下时，失败发生，导致存储状态几乎固定，而网络仍然继续提出变化。误差反馈、残余记忆和方向记忆在时间上携带来自这些被抑制更新的信息，并在不重新训练的情况下恢复准确性。精度扫描显示，增加状态精度可能会使固定的递归解决方案变得更糟，而匹配训练表明可以学习与状态接口的兼容性。为了测试这种行为是否超出 GRU，我们在独立训练的 LSTM 中重复后训练干预，其中粗糙的写回重现了失败，误差反馈恢复了准确性，状态特定的干预揭示了细胞状态比隐藏状态更大的敏感性。我们的结果确立了递归状态写回作为低精度递归动态的关键决定因素，并将状态存储接口确定为量化递归推理的核心设计考虑。

View on arXiv Download PDF AI Translation

cs.AI / 12 / 2609.04493

ResLearn-XR: Residual Learning for Network Traffic and Quality-of-Experience-Aware Modeling in Extended Reality

ResLearn-XR：用于扩展现实中的网络流量和体验质量感知建模的残差学习

Manjunath, Yoga Suhas Kuruba, Gao, Jie, Zhao, Lian

Abstract

We present ResLearn-XR, a residual learning framework for predicting eXtended Reality (XR) network traffic and estimating Quality-of-Experience (QoE) risk. ResLearn-XR adopts a two-stage temporal learning structure comprising a base sequence prediction model augmented with task-specific residual learning components to improve adaptability to bursty, non-stationary XR traffic dynamics. The residual learning stages operate in the value space for continuous XR traffic forecasting and in the logit space for probabilistic QoE risk estimation. \rev{For the QoE-risk branch, we introduce a Data Descriptor Algorithm (DDA), a causal feature-construction module that converts packet-level application-layer observables into frame-timing-aware descriptors suitable for encrypted traffic analysis. We also construct an XR Traffic-QoE dataset that pairs continuous XR traffic traces with session-level user-reported QoE labels. ResLearn-XR reduces SMAPE by up to 17.84% across frame-count, frame-size, and inter-arrival-time prediction, while reducing QoE-risk estimation SMAPE by up to 87.8% over single-stage baselines.

Chinese Translation

我们提出了ResLearn-XR，一个用于预测扩展现实（XR）网络流量和评估体验质量（QoE）风险的残差学习框架。ResLearn-XR采用了一个两阶段的时间学习结构，包括一个基础序列预测模型，并通过任务特定的残差学习组件进行增强，以提高对突发性、非平稳XR流量动态的适应性。残差学习阶段在值空间中进行连续XR流量预测，在对数几率空间中进行概率QoE风险评估。对于QoE风险分支，我们引入了一种数据描述算法（Data Descriptor Algorithm, DDA），这是一个因果特征构建模块，将包级应用层可观测量转换为适合加密流量分析的帧时序感知描述符。我们还构建了一个XR流量-QoE数据集，将连续的XR流量轨迹与会话级用户报告的QoE标签配对。ResLearn-XR在帧计数、帧大小和到达时间预测中将SMAPE降低了最多17.84%，同时在单阶段基线之上将QoE风险估计的SMAPE降低了最多87.8%。

View on arXiv Download PDF AI Translation

cs.AI / 13 / 2609.04495

Rethinking Indirect Prompt Injection as a Test-Time Search Problem

重新思考间接提示注入作为测试时搜索问题

Nguyen, Duong M., Kim, Joon Sik, Manczak, Blazej, Mugunthan, Vaikkunth

Abstract

We formulate indirect prompt injection as a test-time search over a task-dependent attack surface induced by the environment, user task, and injection task. To operationalize this formulation, we introduce an agentic attacker with a dedicated search harness that performs environment reconnaissance, structured reasoning over attack strategies, and adaptive evaluation using victim-agent feedback. Across heterogeneous tasks, we find that increasing attacker test-time compute improves vulnerability discovery and exploitation, while ablations show that explicit strategy management is important for avoiding redundant search and sustaining gains at larger budgets. These results suggest that agentic security evaluations should characterize both the attacker's search procedure and compute budget, rather than treating attack success as a budget-independent property of the victim. More broadly, our findings identify the attacker's adaptive search over the system attack surfaces as an important and underexplored security risk for tool-using agents.

Chinese Translation

我们将间接提示注入形式化为在由环境、用户任务和注入任务引发的任务依赖攻击面上的测试时搜索。为了实现这一形式化，我们引入了一种具备主动攻击能力的攻击者，配备专门的搜索工具，能够进行环境侦察、对攻击策略进行结构化推理，并利用受害者代理的反馈进行自适应评估。在异构任务中，我们发现增加攻击者的测试时计算能力能够提高漏洞发现和利用的效果，而消融实验表明，明确的策略管理对于避免冗余搜索和在更大预算下维持收益至关重要。这些结果表明，主动安全评估应当同时描述攻击者的搜索过程和计算预算，而不是将攻击成功视为受害者的预算无关特性。更广泛地说，我们的发现识别出攻击者在系统攻击面上的自适应搜索作为工具使用代理的重要且未被充分探索的安全风险。

View on arXiv Download PDF AI Translation

cs.AI / 14 / 2609.04504

BioSync: Transformer-Based Cross-Modal Fusion for a Multimodal Physiological Digital Biomarker

BioSync：基于变换器的跨模态融合用于多模态生理数字生物标志物

Mohammadabadi, Seyed Mahmoud Sajjadi

Abstract

Cardiac, neural, behavioral, and speech measurements from wearable and mobile devices provide partial, noise-sensitive views of physiological state. BioSync combines these measurements into the \textbf{BioSync Index (BSI)}, a continuous composite digital biomarker defined under the BEST framework. The model applies multi-head self-attention to modality tokens and adds a linear branch whose hypothesis class includes standard feature concatenation. This architecture is motivated by latent-variable measurement theory and by the possibility that joint observations contain information unavailable from individual modalities. We evaluated BioSync on two literature-informed synthetic cohorts: a four-modality cognitive-decline cohort using HRV, EEG, actigraphy, and speech, and a metabolic-autonomic cohort structured around the public AI-READI wearable schema. In the cognitive cohort, BioSync and concatenation obtained AUCs of 0.928 and 0.926, respectively. In the metabolic cohort, BioSync obtained accuracy/F1 of 0.764/0.766, compared with 0.756/0.758 for concatenation. The BSI correlated with latent severity in both cohorts ($r=0.91$ and $r=0.68$). A pure-attention ablation obtained cognitive-cohort AUC 0.911, locating the increase to 0.928 in the combined wide-and-deep architecture. With matched modality-dropout training, BioSync led concatenation at five of six cognitive-cohort corruption rates and at the highest metabolic-cohort rate. Its cognitive-cohort AUC was also higher than five published digital-biomarker reference values, although differences in datasets and tasks preclude a controlled benchmark claim. Comparison with single-modality, early-fusion, and late-fusion designs across six prespecified criteria identifies the model's computational properties; validation on real cohorts remains necessary.

Chinese Translation

来自可穿戴和移动设备的心脏、神经、行为和语音测量提供了对生理状态的部分、噪声敏感的视角。BioSync将这些测量结合成 extbf{BioSync指数（BSI）}，这是在BEST框架下定义的一个连续复合数字生物标志物。该模型对模态标记应用多头自注意力，并添加了一个线性分支，其假设类包括标准特征连接。该架构受到潜变量测量理论的启发，并考虑到联合观察可能包含个别模态无法获得的信息。我们在两个文献指导的合成队列上评估了BioSync：一个使用心率变异性（HRV）、脑电图（EEG）、活动监测和语音的四模态认知衰退队列，以及一个围绕公共AI-READI可穿戴架构构建的代谢-自主队列。在认知队列中，BioSync和连接分别获得了0.928和0.926的曲线下面积（AUC）。在代谢队列中，BioSync的准确率/F1为0.764/0.766，而连接的准确率/F1为0.756/0.758。BSI与两个队列中的潜在严重性相关（$r=0.91$和$r=0.68$）。纯注意力消融实验在认知队列中获得了0.911的AUC，结合的宽深架构将其提升至0.928。在匹配模态丢弃训练下，BioSync在六个认知队列腐败率中的五个上领先于连接，并在最高的代谢队列腐败率上也表现优越。其认知队列的AUC也高于五个已发布的数字生物标志物参考值，尽管数据集和任务的差异使得无法进行控制基准声明。与六个预设标准下的单模态、早期融合和后期融合设计的比较揭示了该模型的计算特性；在真实队列上的验证仍然是必要的。

View on arXiv Download PDF AI Translation

cs.AI / 15 / 2609.04518

What Does Multi-Harness RL Learn? Credit Assignment and Portability in Coding Agents

多套件强化学习学习了什么？编码代理中的信用分配与可移植性

Le, Chenqian, Cheng, Jiayi, He, Qijia, Li, Runhao, Li, Yinghao, Chen, Xupeng

Abstract

Agent reinforcement learning (RL) increasingly runs through full execution harnesses, and a multi-harness recipe mixes two choices: exposing the policy to several harnesses, and comparing their rewards inside one relative-advantage group. We isolate the second choice in repository-level coding. From one Qwen3-8B supervised warm start we replay the same frozen task-harness records from Aider, OpenHands, Qwen Code, and SWE-agent, with the same number of updates, under two rules for group-relative policy optimization (GRPO), Within (one group per task-harness pair) and Cross (harnesses pooled within a task), and score every checkpoint with a sealed SWE-bench Verified oracle on four source harnesses and a minimal harness held out of training. The evaluation harness is the dominant variable: across 24,000 sealed evaluations it moves the mean solve rate from 2.14\% to 9.27\%, a factor of 4.3, where the training recipe moves it by 1.16. The grouping rule is not. On the held-out harness, Cross minus Within is +0.25 pp, 95\% confidence interval [-0.48, +1.02], at eight attempts per task, and +0.16 [-0.41, +0.72] pooled over three training seeds whose individual estimates change sign. Each rule's own seed range, 0.42 to 0.45 pp, exceeds the difference between them. Both rules place their largest gains on the same source harness. The pooled advantage carries the harness: an out-of-fold classifier recovers the generating harness from Cross's advantage +4.48 pp above the shuffled-label baseline and from Within's not at all, and the two rules still reach the same held-out score and action distribution inside each harness. Re-collecting half the training data on-policy does not change this. Cross-harness credit yields configuration adaptation and no more portable capability than within-harness credit. Multi-harness RL reports should state the grouping boundary and test under an unseen harness.

Chinese Translation

代理强化学习（RL）越来越多地通过完整的执行套件运行，而多套件方案则混合了两种选择：将策略暴露于多个套件中，并在一个相对优势组内比较它们的奖励。我们在库级编码中隔离了第二种选择。从一个 Qwen3-8B 监督热启动开始，我们重放来自 Aider、OpenHands、Qwen Code 和 SWE-agent 的相同冻结任务-套件记录，更新次数相同，遵循两种组相对策略优化（GRPO）规则：Within（每个任务-套件对一个组）和 Cross（在一个任务内汇聚套件），并使用四个源套件和一个在训练中保留的最小套件，通过封闭的 SWE-bench 验证神谕对每个检查点进行评分。评估套件是主导变量：在 24,000 次封闭评估中，它将平均解决率从 2.14% 提升至 9.27%，提升因子为 4.3，而训练方案的提升仅为 1.16。分组规则则不是。在保留的套件上，Cross 减去 Within 为 +0.25 个百分点，95% 置信区间为 [-0.48, +1.02]，每个任务尝试八次，且在三个训练种子上汇总的结果为 +0.16 [-0.41, +0.72]，其个体估计值变化符号。每个规则自身的种子范围为 0.42 到 0.45 个百分点，超过了它们之间的差异。两个规则在同一源套件上获得的最大收益相同。汇聚的优势推动了套件：一个外折分类器从 Cross 的优势中恢复生成套件，提升了 +4.48 个百分点，超过了随机标签基线，而从 Within 则完全没有恢复，两个规则在每个套件内仍然达到了相同的保留分数和动作分布。在政策上重新收集一半的训练数据并未改变这一点。跨套件信用带来了配置适应性，但并没有比套件内信用更强的可移植能力。多套件 RL 报告应说明分组边界，并在未见过的套件下进行测试。

View on arXiv Download PDF AI Translation

cs.AI / 16 / 2609.04523

MaxKernel: Agentic Kernel Generation for TPUs

MaxKernel：用于TPU的自主内核生成

Wang, Shangkun, Cai, Nina, Hoong, Charles, Walker, Julian, Kroiz, Gerson, Vanica, George, Patil, Deepak, Gavrilescu, Andi, Sipra, Hassan, Sankaran, Sethu

Abstract

Designing and authoring high-performance custom kernels for accelerators is a complex task that requires deep hardware-level expertise. Large Language Models (LLM) can be leveraged together with real-time compiler feedback to build agentic systems for kernel generation. In this work, we present MaxKernel, a multi-agent system that implements three distinct paradigms for TPU kernel development: (1) a Human-in-the-Loop (HITL) agent for collaborative, step-by-step design; (2) an Autonomous (Auto) agent that executes a fully automated, metric/trace-driven optimization loop; and (3) a Graph-Based Autonomous Search that scales the Auto agent for global exploration of the design space. All three paradigms leverage a shared pool of specialized sub-agents to handle planning, implementation, self-debugging, testing, and hardware profiling. We evaluate MaxKernel on JaxBench, a comprehensive suite of 50 diverse kernel tasks for TPUs, alongside complex, real-world workloads from state-of-the-art open-source models. We demonstrate that MaxKernel consistently generates highly optimized implementations, matching expert hand-tuned baselines and delivering significant performance across the benchmark. Our agent is open-sourced and available https://github.com/AI-Hypercomputer/accelerator-agents/tree/main/MaxKernel.

Chinese Translation

为加速器设计和编写高性能自定义内核是一项复杂的任务，需要深入的硬件级专业知识。大型语言模型（LLM）可以与实时编译器反馈结合，构建用于内核生成的自主系统。在本研究中，我们提出了MaxKernel，这是一个多智能体系统，实现了TPU内核开发的三种不同范式：（1）人机协作（Human-in-the-Loop, HITL）智能体，用于协作的逐步设计；（2）自主（Autonomous, Auto）智能体，执行完全自动化的基于指标/跟踪的优化循环；（3）基于图的自主搜索，扩展Auto智能体以进行设计空间的全局探索。这三种范式都利用了一组共享的专业子智能体来处理规划、实施、自我调试、测试和硬件分析。我们在JaxBench上评估了MaxKernel，JaxBench是一个包含50个多样化TPU内核任务的综合套件，以及来自最先进开源模型的复杂真实工作负载。我们证明了MaxKernel始终生成高度优化的实现，匹配专家手动调优的基准，并在基准测试中提供显著的性能提升。我们的智能体是开源的，地址为 https://github.com/AI-Hypercomputer/accelerator-agents/tree/main/MaxKernel。

View on arXiv Download PDF AI Translation

cs.AI / 17 / 2609.04528

Towards a universal language of concepts: A survey

迈向概念的普遍语言：一项调查

Parab, Aishni

Abstract

Humans can learn and generalize novel concepts from sparse data because they express knowledge in rich structural formats. In this paper, we propose that programs are a strong candidate for universal representation of concepts. We review computational models of concept learning that use programs as their concept representation and evaluate their contribution toward a universal representational language.

Chinese Translation

人类能够从稀疏数据中学习和概括新概念，因为他们以丰富的结构化格式表达知识。本文提出程序是概念的普遍表示的有力候选者。我们回顾了使用程序作为概念表示的概念学习计算模型，并评估它们对普遍表示语言的贡献。

View on arXiv Download PDF AI Translation

cs.AI / 18 / 2609.04541

Data-Driven Discovery of Composition-Dependent Constitutive Models for Hyperelasticity and Viscoelasticity of Digital Materials

基于数据驱动的组成依赖性本构模型发现：数字材料的超弹性与粘弹性

García-Ávila, Josué, Shen, Beijun, Rausch, Manuel K., Boyce, Mary C., Buganza-Tepole, Adrián

Abstract

Digital materials fabricated by multi-material 3D printing are designed as controlled mixtures of stiff and compliant constituents, yielding effective responses that span more than an order of magnitude in apparent stiffness and exhibit strongly nonlinear, composition-dependent, and rate-dependent dissipative behavior. Classical finite-strain viscoelastic models represent such behavior with closed-form strain energy functions for equilibrium and non-equilibrium stresses as well as evolution of internal variables, which may limit flexibility when a single constitutive model is expected to generalize across materials and loading rates. Here, we present a data-driven multi-material constitutive modeling framework that generalizes a formulation by Bergstr\"om and Boyce. The proposed framework retains the structure of the classical model, namely multiplicative kinematics, invariant-based strain-energy functions, and a scalar dissipative evolution law directed along the normalized nonequilibrium deviatoric stress. For the equilibrium branch, the data-driven discovery framework either directly predicts closed-form model parameters as functions of composition or automatically constructs a polyconvex strain-energy function using neural ordinary differential equations (NODEs). The nonequilibrium branch kinetics are learned similarly, either by directly identifying closed-form parameters across compositions or by using appropriately constrained artificial neural networks. Using multi-rate uniaxial compression data across multiple material compositions, we show that the proposed formulation captures rate-dependent stiffness and hysteresis across compositions while preserving thermodynamic consistency.

Chinese Translation

通过多材料3D打印制造的数字材料被设计为刚性和柔性成分的受控混合，产生的有效响应在表观刚度上跨越超过一个数量级，并表现出强非线性、组成依赖性和速率依赖性的耗散行为。经典的有限应变粘弹性模型通过封闭形式的应变能函数来表示这种行为，适用于平衡和非平衡应力以及内部变量的演变，但当期望单一本构模型能够在不同材料和加载速率之间进行推广时，这可能限制了灵活性。在此，我们提出了一种基于数据驱动的多材料本构建模框架，该框架推广了Bergström和Boyce的公式。所提出的框架保留了经典模型的结构，即乘法运动学、基于不变性的应变能函数，以及沿着归一化非平衡偏应力方向的标量耗散演变法则。对于平衡分支，数据驱动发现框架要么直接预测作为组成函数的封闭形式模型参数，要么使用神经常微分方程（NODEs）自动构建多凸应变能函数。非平衡分支的动力学以类似方式学习，要么通过直接识别不同组成的封闭形式参数，要么通过使用适当约束的人工神经网络。通过多材料组成的多速率单轴压缩数据，我们展示了所提出的公式能够捕捉组成间的速率依赖性刚度和滞回现象，同时保持热力学一致性。

View on arXiv Download PDF AI Translation

cs.AI / 19 / 2609.04543

From Answers to Interpretations: Rethinking Ambiguity-Induced Aleatoric Uncertainty Estimation in LLMs

从答案到解释：重新思考大语言模型中的模糊性引发的随机不确定性估计

Nahum, Omer, Nayman, Niv, Fhima, Jonathan, Zolfi, Alon, Levy, Jeremy, Mazor, Shai, Favaro, Paolo

Abstract

A key challenge in reliable LLM deployment is recognizing when uncertainty reflects irreducible variability in the task rather than limitations in the model's knowledge. In language tasks, a central source of such aleatoric uncertainty is input ambiguity or underspecification, where multiple interpretations remain plausible. Existing decomposition methods estimate aleatoric uncertainty by generating multiple clarifications of the input, querying the model for an answer under each clarification, and comparing the resulting answers. We argue that answers are not necessary for identifying ambiguity: they are often redundant, add avoidable cost, and can mislead through epistemic leakage. We support this claim theoretically, and propose a clarification-only approach that estimates this ambiguity-induced component directly from the space of plausible interpretations, without answers to the clarified inputs. Using ambiguity detection as an operational evaluation across three benchmarks, this direct approach improves AUROC (63.34 vs. 60.85), reduces computational cost by 4-26x in output tokens and 2.2-3.5x in API calls, and yields estimates with substantially lower correlation with epistemic uncertainty. Overall, our results suggest that ambiguity-induced aleatoric uncertainty is better estimated from the interpretation space than from the response space.

Chinese Translation

在可靠的大语言模型（LLM）部署中，一个关键挑战是识别不确定性何时反映任务中的不可减少变异性，而不是模型知识的局限性。在语言任务中，这种随机不确定性的一个主要来源是输入模糊性或不充分规范化，其中多个解释仍然是合理的。现有的分解方法通过生成输入的多个澄清，针对每个澄清查询模型以获取答案，并比较结果答案来估计随机不确定性。我们认为，答案并不是识别模糊性的必要条件：它们往往是冗余的，增加了可避免的成本，并可能通过认识论泄漏导致误导。我们从理论上支持这一主张，并提出了一种仅澄清的方法，该方法直接从合理解释的空间中估计这种模糊性引发的成分，而无需对澄清后的输入给出答案。通过在三个基准上使用模糊性检测作为操作性评估，这种直接方法提高了AUROC（63.34对比60.85），在输出标记上减少了4-26倍的计算成本，在API调用上减少了2.2-3.5倍，并且产生的估计与认识论不确定性的相关性显著降低。总体而言，我们的结果表明，模糊性引发的随机不确定性更适合从解释空间而非响应空间进行估计。

View on arXiv Download PDF AI Translation

cs.AI / 20 / 2609.04559

IPGeoAI: Transformer-Based Geolocation with LLM Semantic Fusion

IPGeoAI：基于Transformer的地理定位与大语言模型语义融合

Kadimisetty, Avinash, Yu, Andy Jinqing, Favaloro, Philip, Liu, Wenlong, Xiong, Xiaolu

Abstract

Accurate city-level IP Geolocation is an important enabler for the modern digital ecosystem, underpinning services ranging from local content delivery and targeting to digital rights enforcement. However, traditional heuristic and database-driven methods often struggle to resolve the complex, non-linear allocation patterns of modern network infrastructures, particularly within the exploding IPv6 address space and transient mobile networks. In this paper, we introduce IPGeoAI, a novel deep learning model architecture that reframes geolocation from a static lookup problem to a sequential modeling task. Our approach utilizes the Transformer Encoder to capture hierarchical dependencies inherent in IP subnet structures. We propose a method to resolve geographic ambiguity by integrating unstructured semantic context via a Zero-Shot LLM Feature Extraction pipeline. We utilize Large Language Models to transform raw, noisy Autonomous Systems (AS) descriptions into structured, domain-specific metadata (such as 'University' vs. 'ISP' or 'Global' vs. 'Local') via an offline pre-computation process. By fusing these semantic signals into the network via a Multi-Head Cross-Attention module, we bridge the gap between numerical network topology and real-world semantic identity. Extensive offline evaluation on a proprietary dataset spanning 200,000 cities demonstrates that IPGeoAI significantly outperforms a leading external vendor in city-level granularity. By adopting a hierarchical inference strategy that refines coarse-grained country signals, our model achieves a 6% improvement in city-level accuracy while extending coverage to 100% of the traffic. Furthermore, in large-scale online production tests, the model drove a statistically significant +0.35% improvement in our 1st-tier downstream use cases metric.

Chinese Translation

准确的城市级IP地理定位是现代数字生态系统的重要支撑，促进了从本地内容交付和目标定位到数字版权执行等多种服务。然而，传统的启发式和数据库驱动的方法往往难以解决现代网络基础设施复杂的非线性分配模式，尤其是在快速扩展的IPv6地址空间和瞬态移动网络中。在本文中，我们介绍了IPGeoAI，一种新颖的深度学习模型架构，将地理定位从静态查找问题重新定义为序列建模任务。我们的方法利用Transformer编码器捕捉IP子网结构中固有的层次依赖关系。我们提出了一种通过零样本大语言模型（Zero-Shot LLM）特征提取管道整合非结构化语义上下文以解决地理模糊性的方法。我们利用大型语言模型将原始、嘈杂的自治系统（Autonomous Systems, AS）描述转化为结构化的领域特定元数据（例如“大学”与“互联网服务提供商（ISP）”或“全球”与“本地”），这一过程通过离线预计算实现。通过将这些语义信号融合到网络中，采用多头交叉注意力模块，我们弥合了数值网络拓扑与现实世界语义身份之间的差距。在一个涵盖20万个城市的专有数据集上进行的广泛离线评估表明，IPGeoAI在城市级粒度上显著优于领先的外部供应商。通过采用一种层次推理策略，细化粗粒度的国家信号，我们的模型在城市级准确性上实现了6%的提升，同时将覆盖范围扩展至100%的流量。此外，在大规模在线生产测试中，该模型在我们的第一层下游用例指标上实现了统计显著的+0.35%的改善。

View on arXiv Download PDF AI Translation

cs.AI / 21 / 2609.04561

Reducing Hallucinated Transcripts in Whisper via Hallucination Space Projection

通过幻觉空间投影减少 Whisper 中的幻觉转录

Abbasihafshejani, Maryam, Jadliwala, Murtuza

Abstract

Whisper is a widely used foundation model for automatic speech recognition (ASR), but its generative decoder can produce fluent hallucinated transcripts for inputs containing little or no speech. We propose a training-free, inference-time method to reduce these hallucinations using low-rank projection of decoder activations. A compact hallucination-associated subspace is estimated from non-speech calibration data, and decoder hidden states are projected away from this subspace during inference. We evaluate two variants: always-on, which applies projection to all inputs, and gated, which applies it only when Whisper predicts that an input is likely non-speech. Across non-speech benchmarks, always-on projection reduces average hallucination rate (HR) from 31.31% to 2.44%, a 92.21% relative reduction, while gated projection reduces HR to 3.74%, an 88.05% relative reduction, with lower false rejection of genuine speech. On LibriSpeech, gated projection increases absolute word error rate (WER) by 0.33-4.39 percentage points and yields false-rejection rates (FRR) of 0.41--9.97% across model and split settings. These results show that low-rank activation projection can substantially suppress Whisper hallucinations without retraining, while providing a controllable trade-off between hallucination suppression and speech recognition performance.

Chinese Translation

Whisper 是一种广泛使用的自动语音识别（ASR）基础模型，但其生成解码器可能会为包含少量或没有语音的输入生成流畅的幻觉转录。我们提出了一种无训练、推理时的方法，通过对解码器激活进行低秩投影来减少这些幻觉。我们从非语音校准数据中估计出一个紧凑的与幻觉相关的子空间，并在推理过程中将解码器的隐藏状态投影到该子空间之外。我们评估了两种变体：始终开启（always-on），即对所有输入应用投影；以及门控（gated），即仅在 Whisper 预测输入可能为非语音时应用投影。在非语音基准测试中，始终开启投影将平均幻觉率（HR）从 31.31% 降低到 2.44%，相对减少了 92.21%；而门控投影将 HR 降低到 3.74%，相对减少了 88.05%，且对真实语音的错误拒绝率更低。在 LibriSpeech 数据集上，门控投影使绝对词错误率（WER）增加了 0.33-4.39 个百分点，并在不同模型和划分设置中产生了 0.41%--9.97% 的错误拒绝率（FRR）。这些结果表明，低秩激活投影可以在不重新训练的情况下显著抑制 Whisper 的幻觉，同时在幻觉抑制与语音识别性能之间提供可控的权衡。

View on arXiv Download PDF AI Translation

cs.AI / 22 / 2609.04564

La Agente \'Optima: Towards Agentic Self-Driving Laboratories

La Agente 'Optima：迈向自主驱动实验室

Müller, Marcel, Bai, Jiaru, Gottstein, Willi, Mandal, Abhijoy, Nazeri, Mohammad, Savino, Elia, Fang, Yanlin, Das, Sujoy, Carrillo, Sergio Pablo García, Kang, Yeonghun, Pérez-Sánchez, Juan B., Pilon, Simone, Fitzner, Martin, Noël, Timothy, Gu, Frank, Bernales, Varinia, Aspuru-Guzik, Alán

Abstract

Self-driving laboratories (SDLs) combine automated experimentation with adaptive decision-making to accelerate scientific discovery. Their operation nevertheless often depends on human specialists who translate scientific objectives into executable closed-loop campaigns. Specialists adjust them as data and operating conditions change. Here, we present La Agente \'Optima, an agentic framework that constructs and supervises Bayesian optimization campaigns across computational and experimental systems while maintaining a persistent optimization state. By separating large language model (LLM) reasoning from executed campaigns, \'Optima runs repetitive optimization loops consistently, returns control to the agent only when progress requires interpretation or campaign revision, and keeps every decision auditable. We evaluate \'Optima across ablation studies, five digital discovery tasks, and two physical platforms. Throughout, \'Optima maintained executable campaigns as both the scientific problem and execution environment evolved. In a closed-loop contact angle optimization campaign, \'Optima identified and corrected a mid-run measurement failure, bringing the contact angle from 71.4 to 67.8 degrees, just above the 64-66 degree range. From this result, \'Optima correctly inferred that the target was likely unattainable with the available reagents and recommended changing the formulation. In a five-day multi-objective flow-chemistry campaign, \'Optima increased the yield from 30% to 59% over 23 experiments. Despite substantial inference costs, it cost less and used substantially less starting material than a human-directed campaign, while selecting a more mass-efficient operating point. These results show that LLM-based agents can make rigorous, long-running optimization campaigns accessible to domain scientists without specialist setup, expanding the scope of SDLs.

Chinese Translation

自主驱动实验室（SDLs）结合了自动化实验和自适应决策，旨在加速科学发现。然而，它们的运行往往依赖于人类专家将科学目标转化为可执行的闭环实验方案。专家会根据数据和操作条件的变化对其进行调整。在此，我们提出了La Agente 'Optima，一个自主框架，能够在计算和实验系统中构建和监督贝叶斯优化实验，同时保持持续的优化状态。通过将大型语言模型（LLM）推理与执行的实验方案分离，'Optima能够持续运行重复的优化循环，仅在进展需要解释或实验方案修订时将控制权交还给代理，并保持每个决策可审计。我们通过消融研究、五个数字发现任务和两个物理平台对'Optima进行了评估。在整个过程中，'Optima在科学问题和执行环境演变的同时维持了可执行的实验方案。在一个闭环接触角优化实验中，'Optima识别并纠正了中途测量失败，将接触角从71.4度降低到67.8度，刚好位于64-66度范围之上。基于这一结果，'Optima正确推断出在现有试剂下目标可能无法实现，并建议更改配方。在一个为期五天的多目标流动化学实验中，'Optima将产率从30%提高到59%，共进行了23次实验。尽管推理成本较高，但其成本低于人类主导的实验，并且使用的起始材料显著减少，同时选择了更具质量效率的操作点。这些结果表明，基于LLM的代理能够使严格的、长期的优化实验对领域科学家变得可及，而无需专家设置，从而扩展了SDLs的应用范围。

View on arXiv Download PDF AI Translation

cs.AI / 23 / 2609.04565

Extremely Sparse Supervision Incentivizes Reasoning Ability

极度稀疏的监督激励推理能力

Liu, Zhishuai, Xu, Xingzi, Seyfioglu, Mehmet Saygin, Xu, Pan, Bouyarmane, Karim

Abstract

Large language models demonstrate increasingly strong reasoning capabilities through effective post-training. Yet, prevailing post-training methods optimize over massive numbers of tokens, implicitly assuming that effective learning must be token-intensive. We revisit this assumption in the on-policy distillation (OPD) setting, which naturally admits dense teacher supervision at every generated token. Using the Qwen3 family, we discover a counter-intuitive phenomenon: reasoning can be effectively incentivized by an extremely small fraction of generated tokens--as few as one or two tokens per reasoning trajectory, corresponding to only 0.05% of all tokens. Surprisingly, this sparse supervision in most cases matches or surpasses full-token training in improving reasoning ability, despite excluding the vast majority of generated tokens from the training objective. This phenomenon is consistently observed across nine teacher--student configurations spanning different model scales on mathematical reasoning tasks, and is further validated on coding reasoning, Llama models and Proximal Policy Optimization (PPO)-based reinforcement learning with verifiable reward (RLVR). Interestingly, such extremely sparse supervision may be closer to the natural learning process: rather than correcting every step word by word, one reflects on a few critical reasoning steps, updates prior understanding, and continues the trial-and-error, avoiding micro-level corrections while remaining remarkably effective. Overall, our results challenge the assumption that effective post-training must be token-intensive and point to a new direction for understanding and designing more efficient post-training algorithms.

Chinese Translation

大型语言模型通过有效的后训练展现出越来越强的推理能力。然而，现有的后训练方法通常在大量的标记上进行优化，隐含地假设有效学习必须是标记密集型的。我们在在线策略蒸馏（On-Policy Distillation, OPD）设置中重新审视这一假设，该设置自然允许在每个生成的标记上进行密集的教师监督。使用 Qwen3 系列模型，我们发现了一个反直觉的现象：推理能力可以通过极少量的生成标记有效激励——每个推理轨迹仅需一到两个标记，这仅占所有标记的 0.05%。令人惊讶的是，这种稀疏监督在大多数情况下与全标记训练在提高推理能力方面相匹配或超越，尽管在训练目标中排除了绝大多数生成的标记。该现象在九种教师-学生配置中一致观察到，涵盖了不同模型规模的数学推理任务，并在编码推理、Llama 模型以及基于近端策略优化（Proximal Policy Optimization, PPO）的可验证奖励强化学习（Reinforcement Learning with Verifiable Reward, RLVR）中得到了进一步验证。有趣的是，这种极度稀疏的监督可能更接近自然学习过程：而不是逐字逐步地纠正每一步，人们反思几个关键的推理步骤，更新先前的理解，并继续试错，避免微观层面的修正，同时保持显著的有效性。总体而言，我们的结果挑战了有效后训练必须是标记密集型的假设，并指向理解和设计更高效的后训练算法的新方向。

View on arXiv Download PDF AI Translation

cs.AI / 24 / 2609.04579

Does the Selected Object Reach the Reader? Auditing Identity Handoffs in Grounded Language-Model Pipelines

所选对象是否到达读者？审计基础语言模型管道中的身份交接

Vohra, Siddharth, Jiang, Runmin, Li, Xiaomo, Xu, Min

Abstract

Grounded language-model pipelines can be divided into three stages: selecting an object, retrieving passages for it, and using that evidence to answer. If the selected object must reach the reader, losing it breaks the handoff. Benchmark recall checks the dataset-linked object, which can differ. We audit 600 HybridQA questions across three selector families. On 1,463 resolvable records where the selected object matches the dataset-traced passage, exact key lookup and exact title matching return the object every time. With every ranked rule given the same decoded selected title, body-only BM25 omits it on 389 records (26.6%) at cutoff five, while hybrid retrieval with reranking omits it on 14 (1.0%). The two identities differ on 329 of 1,792 resolvable records. With original-question rankings, their top-five checks disagree on 106 records (5.9%). Frozen reader comparisons associate the aligned object's presence with 28.6 to 31.0 points higher exact match. In a deliberately selected 64-item cohort, removing that passage sharply lowers exact match, while removing a similar-length comparison passage does not reproduce the drop. We release the Returned-Object Profile (ROP), an executable record of the target, returned-ID field, cutoff, membership rule, and complete expected population, with data and an offline replay.

Chinese Translation

基础语言模型管道可以分为三个阶段：选择对象、检索相关段落以及利用这些证据进行回答。如果所选对象未能到达读者，则会导致交接失败。基准召回检查与数据集相关联的对象，这可能会有所不同。我们对三种选择器类别中的600个HybridQA问题进行了审计。在1,463个可解析记录中，当所选对象与数据集追踪的段落匹配时，精确关键查找和精确标题匹配每次都能返回该对象。在每个排名规则给定相同解码的所选标题时，仅基于正文的BM25在389个记录（26.6%）中在截止点五时遗漏该对象，而混合检索与重排名在14个记录（1.0%）中遗漏它。这两种身份在1,792个可解析记录中有329个存在差异。使用原始问题排名时，它们的前五名检查在106个记录（5.9%）上存在分歧。冻结读者比较将对齐对象的存在与28.6到31.0点更高的精确匹配相关联。在一个故意选择的64项样本中，移除该段落会显著降低精确匹配，而移除一个相似长度的比较段落则不会重现这种下降。我们发布了返回对象配置文件（Returned-Object Profile, ROP），这是一个可执行的目标记录，包括返回的ID字段、截止点、成员规则和完整的预期人群，以及数据和离线重放。

View on arXiv Download PDF AI Translation

cs.AI / 25 / 2609.04611

$\tau^\tau$-Bench: An Environment for End-To-End, Realistic Agent Construction

$ au^ au$-基准：一个端到端、真实的智能体构建环境

Shi, Quan, Dhandhania, Keshav, Narasimhan, Karthik, Barres, Victor

Abstract

LLM agents are rapidly becoming production software, deployed to handle customer service, adjudicate disputes, and operate internal systems. Notably, the work of building them is increasingly handed to coding agents, yet existing benchmarks say little about whether an AI system can deliver one under the conditions of a real client engagement. We introduce $\tau^\tau$-bench (pronounced hyper-tau-bench), a benchmark that makes agent construction the task. A developer agent is given the records a business actually keeps, a client who holds requirements, a production API that operations must run through, a codebase to inherit, and limits on serving cost and models: the same starting point a real engagement provides. From these it must deliver a complete customer-service agent, scored by deploying that agent against held-out simulated users. Across 53 tasks spanning four domains, the strongest configuration, Claude Opus 5 under Claude Code, passes just 23.9% of evaluation simulations. Meanwhile, an expert-authored reference ceiling scores 82.2%. The failures mirror ones human agent developers see: models issue shallow queries in place of deep comprehension of the records, communicate almost nothing to the client, and experiment too little with agent architecture and serving spend, shipping the first design that runs. We aim for $\tau^\tau$-bench to turn the work of cooperative agent building into a measurable target for coding agents.

Chinese Translation

大型语言模型（LLM）智能体正在迅速成为生产软件，被部署用于处理客户服务、裁决争端和操作内部系统。值得注意的是，构建这些智能体的工作越来越多地交给编码智能体，但现有基准对AI系统在真实客户互动条件下能否交付智能体几乎没有提供信息。我们介绍了$ au^ au$-基准（发音为超τ基准），这是一个将智能体构建作为任务的基准。开发者智能体被提供一个企业实际保留的记录、一个持有需求的客户、一个必须通过的生产API、一个需要继承的代码库，以及服务成本和模型的限制：这与真实互动提供的起点相同。基于这些信息，它必须交付一个完整的客户服务智能体，并通过将该智能体部署到保留的模拟用户上进行评分。在涵盖四个领域的53个任务中，最强的配置，即Claude Opus 5在Claude Code下，仅通过了23.9%的评估模拟。同时，专家撰写的参考上限得分为82.2%。这些失败反映了人类智能体开发者所面临的问题：模型发出肤浅的查询，而不是对记录的深刻理解，几乎不与客户沟通，并且在智能体架构和服务支出方面实验过少，直接交付第一个能运行的设计。我们的目标是使$ au^ au$-基准将合作智能体构建的工作转变为编码智能体可测量的目标。

View on arXiv Download PDF AI Translation

cs.AI / 26 / 2609.04627

Leveraging Imperfect Restoration for Data Availability Attack

利用不完美恢复进行数据可用性攻击

Huang, Yi, Styborski, Jeremy, Lyu, Mingzhi, Wang, Fan, Kong, Adams

Abstract

The abundance of online data is at risk of unauthorized usage in training deep learning models. To counter this, various Data Availability Attacks (DAAs) have been devised to make data unlearnable for such models by subtly perturbing the training data. However, existing attacks often excel against either Supervised Learning (SL) or Self-Supervised Learning (SSL) scenarios. Among these, a model-free approach that generates a Convolution-based Unlearnable Dataset (CUDA) stands out as the most robust DAA across both SSL and SL. Nonetheless, CUDA's effectiveness against SSL is underwhelming and it faces a severe trade-off between image quality and its poisoning effect. In this paper, we conduct a theoretical analysis of CUDA, uncovering the sub-optimal gradients it introduces and elucidating the strategy it employs to induce class-wise bias for data poisoning. Building on this, we propose a novel poisoning method named Imperfect Restoration Poisoning (IRP), aiming to preserve high image quality while achieving strong poisoning effects. Through extensive comparisons of IRP with eight baselines across SL and SSL, coupled with evaluations alongside five representative defense methods, we showcase the superiority of IRP. Code: https://github.com/lyumingzhi/IRP

Chinese Translation

在线数据的丰富性面临未经授权使用的风险，尤其是在训练深度学习模型时。为应对这一问题，已经设计了多种数据可用性攻击（Data Availability Attacks, DAA），通过微妙地扰动训练数据，使得数据对这些模型不可学习。然而，现有的攻击往往在监督学习（Supervised Learning, SL）或自监督学习（Self-Supervised Learning, SSL）场景中表现出色。在这些攻击中，一种无模型的方法生成卷积基础的不可学习数据集（Convolution-based Unlearnable Dataset, CUDA），在SSL和SL中都表现出最强的鲁棒性。然而，CUDA在SSL中的效果不尽如人意，并且在图像质量与其污染效果之间存在严重的权衡。在本文中，我们对CUDA进行了理论分析，揭示了其引入的次优梯度，并阐明了其诱导类别偏差以实现数据污染的策略。在此基础上，我们提出了一种新的污染方法，称为不完美恢复污染（Imperfect Restoration Poisoning, IRP），旨在保持高图像质量的同时实现强大的污染效果。通过对IRP与八个基线方法在SL和SSL中的广泛比较，以及与五种代表性防御方法的评估，我们展示了IRP的优越性。代码链接：https://github.com/lyumingzhi/IRP

View on arXiv Download PDF AI Translation

cs.AI / 27 / 2609.04629

SiLR: Structure-Preserving Admission and Process Reward for LLM Tool Agents

SiLR：结构保持的LLM工具代理的接纳与过程奖励

Zhou, Chenyu, Jiang, Qiliang, Wu, Shuning, Zhou, Xu

Abstract

A runtime gate for an LLM tool agent is usually cast as a filter. In a ReAct loop a rejected proposal is followed by another at the same state, so the gate is a search operator over the proposal stream whose admission criterion shapes which trajectories are reachable. We study post-violation recovery admission, where progress must be admitted while the system is still in violation, and identify the scalar projection trap: an aggregate-score gate accepts a locally improving proposal and commits the trajectory to a plateau. SiLR instead shadow-executes each proposal and admits it under a product order over the branch-level violation state (overloaded-branch support and per-branch severity). We prove that no scalar surrogate is sound for this order, so the failure is representational, not a matter of threshold tuning. On mined Gym-ANM scenarios, SiLR recovers 21/21 multi-action episodes against 0/21 for terminal and 9/21 for the best scalar gate, significant across the full 24-scenario benchmark. The terminal-versus-structured dichotomy holds across three model families and in CityLearn. Because admission rests on deterministic simulation, the LLM lies outside the trust boundary: a magnitude-redistribution attack that defeats both scalar and support-only baselines is contained only by the full per-branch predicate. With two constraint families active, every tested scalar projection admits physically unsafe actions; support-only admits the largest fraction (63.2% of 42,410; product order 0). In the hardest dual-family traces, scalar gates recover only through that unsafe class. Reused as a GRPO process reward, it outperforms its count projection in every mined scenario and is the only tested reward whose ungated policy exceeds the untrained base (0.844 vs. 0.778). Scalar projection loses the violation geometry at both design points; only the full product order is structurally sufficient.

Chinese Translation

LLM工具代理的运行时门通常被视为一个过滤器。在ReAct循环中，拒绝的提案后面会跟随另一个处于相同状态的提案，因此该门是一个在提案流上进行搜索的操作符，其接纳标准决定了可达的轨迹。我们研究了违规后恢复接纳，其中在系统仍处于违规状态时必须接纳进展，并识别了标量投影陷阱：一个聚合评分门接受一个局部改进的提案，并将轨迹固定在一个平台上。SiLR则通过对每个提案进行影子执行，并在分支级违规状态（过载分支支持和每分支严重性）上根据乘积顺序接纳它。我们证明，没有标量替代物对该顺序是有效的，因此失败是表征性的，而不是阈值调整的问题。在挖掘的Gym-ANM场景中，SiLR在21个多动作回合中恢复成功率为21/21，而终端门为0/21，最佳标量门为9/21，在整个24场景基准测试中具有显著性。终端与结构化的二分法在三个模型家族和CityLearn中均成立。由于接纳依赖于确定性模拟，LLM超出了信任边界：一种幅度重分配攻击同时击败标量和仅支持基线，仅通过完整的每分支谓词得以控制。在两个约束家族活动的情况下，每个测试的标量投影都接纳了物理上不安全的动作；仅支持的接纳比例最大（42,410中的63.2%；乘积顺序为0）。在最困难的双家族轨迹中，标量门仅通过该不安全类别恢复。作为GRPO过程奖励的重用，在每个挖掘场景中均优于其计数投影，并且是唯一一个未门控策略超过未训练基线的测试奖励（0.844对0.778）。标量投影在两个设计点上都失去了违规几何；只有完整的乘积顺序在结构上是充分的。

View on arXiv Download PDF AI Translation

cs.AI / 28 / 2609.04641

A Cost-Aware Agentic Architecture for NL-to-SQL over Nested Enterprise Schemas, with a New Benchmark

一种成本感知的代理架构用于嵌套企业模式下的自然语言到SQL转换，并提出新的基准测试

Varadharajan, Yoga Sri Varshan, Yadav, Ajay, Goru, Ritesh, Chaudhury, Prateek, Caramanis, Constantine, Jain, Prateek, Pasupuleti, Divyateja, Pandey, Sunil Kumar

Abstract

Natural-language-to-SQL systems have ad- vanced rapidly on academic benchmarks, yet production enterprise schemas exhibit graph- like, semi-structured, deeply nested structure that current benchmarks do not measure. We make two complementary contributions. First, we introduce the DevRev NL2SQL bench- mark: 900 execution-verified queries with nested-type and link-graph structure, accom- panied by the Semantic Depth Score (SDS), a schema-agnostic rubric for analytical reasoning depth. Second, we present a cost-aware single- generation agentic architecture whose schema- selection, metadata-retrieval, and error-repair components are designed for the requirements this regime imposes. On the DevRev NL2SQL benchmark the system attains 91.7% answer correctness, a margin of 54.6 percentage points over the next-best baseline; on the Spider 2.0 Snowflake public dataset, it is competitive with leading systems at a single-generation operating point.

Chinese Translation

自然语言到SQL系统在学术基准测试上迅速发展，但生产企业模式展现出图状、半结构化和深度嵌套的结构，而当前的基准测试并未对此进行评估。我们做出了两个互补的贡献。首先，我们引入了DevRev NL2SQL基准：900个经过执行验证的查询，具有嵌套类型和链接图结构，并附带语义深度评分（Semantic Depth Score, SDS），这是一个与模式无关的分析推理深度评估标准。其次，我们提出了一种成本感知的单生成代理架构，其模式选择、元数据检索和错误修复组件专为满足这一体系所施加的要求而设计。在DevRev NL2SQL基准上，该系统达到了91.7%的答案正确率，比下一个最佳基线高出54.6个百分点；在Spider 2.0 Snowflake公共数据集上，其在单生成操作点上与领先系统具有竞争力。

View on arXiv Download PDF AI Translation

cs.AI / 29 / 2609.04651

Continual Graph Memory for Adaptive Recommendation under Intent Drift

适应性推荐中的意图漂移的持续图记忆

Ngoc, Hao Nguyen, Nguyen, Tung, Hanh, Nguyen Thi, Dinh, Hoang Thai, Tung, Nguyen Xuan

Abstract

This paper studies adaptive recommendation under intent drift, where feedback from each recommendation outcome can reveal whether the relational evidence used for ranking is useful, missing, or misleading. While Knowledge Graphs (KGs) provide essential semantic structure to handle these shifts, traditional KG-enhanced systems treat the graph as a static retrieval substrate, making it brittle to evolving intents, noisy metadata, and recurring failure patterns. This paper proposes CGM-Rec, a continual graph memory framework for adaptive recommendation. CGM-Rec treats the graph state as a writable memory and maintains two complementary components. Therein, a Semantic Graph Memory is updated conservatively through quality-gated typed operations for storing stable and high-confidence relational knowledge. Meanwhile, an Episodic Lesson Memory acts as a fast reactive memory that learns recent outcomes, failure cases, and corrective hints. During testing, model parameters remain frozen and adaptation occurs only through memory writes. We evaluate CGM-Rec under a frozen-parameter, one-pass reranking protocol, where encoders and prompts remain fixed during testing and adaptation occurs only through memory writes. Experiments across multiple recommendation settings show that CGM-Rec improves over evaluated neural and LLM-based baselines on most metrics. Particularly, under sampled-candidate reranking, CGM-Rec improves HR@1 by up to 29.58% over the strongest LLM baseline on Bundle, and outperforms K-RagRec on metadata-rich ML-100K with HR@5 of 0.5941 versus 0.4746.

Chinese Translation

本文研究了在意图漂移下的适应性推荐，其中每个推荐结果的反馈可以揭示用于排名的关系证据是有用的、缺失的还是误导性的。虽然知识图谱（Knowledge Graphs, KGs）提供了处理这些变化所需的基本语义结构，但传统的KG增强系统将图视为静态检索基础，导致其对不断变化的意图、噪声元数据和重复失败模式变得脆弱。本文提出了CGM-Rec，一个用于适应性推荐的持续图记忆框架。CGM-Rec将图状态视为可写内存，并维护两个互补组件。其中，语义图记忆（Semantic Graph Memory）通过质量门控的类型操作进行保守更新，以存储稳定且高置信度的关系知识。同时，情节学习记忆（Episodic Lesson Memory）作为快速反应内存，学习最近的结果、失败案例和纠正提示。在测试过程中，模型参数保持不变，适应仅通过内存写入进行。我们在一个冻结参数的单次重排序协议下评估CGM-Rec，在该协议中，编码器和提示在测试期间保持固定，适应仅通过内存写入进行。多个推荐设置下的实验表明，CGM-Rec在大多数指标上优于评估的神经网络和基于LLM的基线。特别是在采样候选重排序下，CGM-Rec在Bundle数据集上相较于最强LLM基线提高了HR@1达29.58%，并在元数据丰富的ML-100K数据集上以HR@5达到0.5941，超越了K-RagRec的0.4746。

View on arXiv Download PDF AI Translation

cs.AI / 30 / 2609.04665

Harness-agnostic detection and immunization of reward hacking in self-evolving language models

与工具无关的自我进化语言模型中的奖励黑客检测与免疫

Yang, Rongxin, Liu, Yang, Luo, Shang, Jia, Haoxuan, Zhang, Chongyang, Zheng, Hao, Yang, Yingguang, Huang, Yulin, Zhang, Jianshen, Qi, Yongzhi, Xu, Kefu, Ran, Congjing, Chong, Bin

Abstract

Self-evolving language models improve by proposing candidate updates and keeping whatever raises a visible score. When that score is an imperfect proxy for the capability one actually wants, sustained selection widens the gap between the two. This is reward hacking. We introduce HackProbe, a monitor that attaches to an arbitrary self-evolving loop through two black-box hooks, with no access to weights or activations. It keeps a secret, distribution-fixed comparison core, whose frozen distribution makes its capability proxy comparable across generations, alongside a rotated fresh layer that hardens the bank against co-adaptation. Four tests built on that proxy cover the level gap, a scale-aligned divergence with online change-point detection, capability stagnation, and a conditional confidently-wrong rate; a Sidak correction turns them into a calibrated family-wise p-value. Diagnosis alone recovers nothing, so a risk-aware immunization layer reselects an honest candidate from the proposal pool using the core together with a purely structural gaming footprint, disclosing at most log2 Pi bits per generation to the host. We prove a detectability bound that converts a target error rate into an explicit probe-size budget, and we delimit what probe rotation does and does not buy. On a controlled prompt-level host with four injected hacking channels and ground-truth labels, HackProbe reaches 0.763 AUROC against 0.663 for the strongest baseline and cuts the false-positive rate from 0.706 to 0.434. Its bandwidth-limited reselection is the only immunization level that returns more true capability under hacking, 5.2 points on average, than it forfeits on clean runs, 4.7; per-channel effects are mostly not individually significant.

Chinese Translation

自我进化语言模型通过提出候选更新并保留任何能够提高可见分数的内容来实现改进。当该分数并不是实际想要的能力的完美代理时，持续的选择会扩大两者之间的差距。这就是奖励黑客行为。我们引入了 HackProbe，这是一种监控工具，通过两个黑箱钩子附加到任意自我进化循环中，而不需要访问权重或激活。它保持一个秘密的、固定分布的比较核心，其冻结的分布使得其能力代理在不同代之间可比，同时配备一个旋转的新层，以增强银行对共同适应的抵抗力。基于该代理构建的四个测试涵盖了水平差距、与在线变更点检测对齐的规模相关的偏差、能力停滞以及条件自信错误率；Sidak 校正将它们转化为一个经过校准的全家族 p 值。仅靠诊断无法恢复任何内容，因此一个风险意识的免疫层使用核心和纯结构的游戏足迹重新选择一个诚实的候选者，从提议池中披露每代最多 log2 Pi 位信息给主机。在一个控制的提示级主机上，注入了四个黑客通道和真实标签，HackProbe 达到了 0.763 的 AUROC，而最强基线为 0.663，并将假阳性率从 0.706 降低到 0.434。其带宽限制的重新选择是唯一在黑客攻击下返回更多真实能力的免疫层，平均增加 5.2 分，而在干净运行中损失 4.7 分；每个通道的效果大多数并不显著。

View on arXiv Download PDF AI Translation

cs.AI / 31 / 2609.04667

ERPBench: Evaluating LLM Agents for Enterprise Decision-Making Across Competitive Market Ecologies

ERPBench：评估大型语言模型代理在竞争市场生态中的企业决策能力

Zhang, Xinran, Lu, Pengrui, Ye, Lyumanshan, Liu, Pengfei

Abstract

Large language model (LLM) agents are increasingly proposed for enterprise workflows, yet existing evaluations rarely test whether business-decision conclusions transfer across competitive market ecologies. We introduce ERPBench, an execution-instrumented benchmark for enterprise decision agents in a six-round Enterprise Resource Planning (ERP) simulation with coupled pricing, production, procurement, inventory, finance, and shared-market competition. ERPBench evaluates the same 100 fixed problems in two matched competitive market ecologies: Solo, where each evaluated LLM agent competes against fixed rule-based opponents, and Arena, where six evaluated LLM agents compete in a shared market. Across six model families, this yields 1,200 model-level trajectories spanning 7,200 decision rounds. Under the observed service configuration, the leading model differs between ecologies: DeepSeek leads in Solo (252.29M mean valuation; mean rank 1.67), whereas Gemini leads in Arena (263.95M; 1.76). The two ecologies identify the same task-level winner on only 21 of 100 problems, and Gemini's bottom-rank rate falls from 22 % to 0 % in Arena. ERPBench supports paired evaluation of whether enterprise-agent rankings transfer across competitive market ecologies, supplemented by aggregate execution-intervention analysis. Code and benchmark resources are available in our https://github.com/GAIR-NLP/erp-bench.

Chinese Translation

大型语言模型（LLM）代理在企业工作流程中的应用越来越受到关注，但现有评估很少测试商业决策结论是否能够在竞争市场生态中转移。我们介绍了ERPBench，这是一个针对企业决策代理的执行工具化基准，包含六轮的企业资源规划（ERP）模拟，涉及定价、生产、采购、库存、财务和共享市场竞争。ERPBench在两个匹配的竞争市场生态中评估相同的100个固定问题：Solo生态中，每个被评估的LLM代理与固定的基于规则的对手竞争；而在Arena生态中，六个被评估的LLM代理在共享市场中竞争。在六个模型家族中，这产生了1,200个模型级轨迹，涵盖7,200个决策轮次。在观察到的服务配置下，两个生态中的领先模型有所不同：DeepSeek在Solo中领先（平均估值252.29M；平均排名1.67），而Gemini在Arena中领先（263.95M；1.76）。这两个生态在100个问题中仅在21个问题上识别出相同的任务级赢家，且Gemini在Arena中的底层排名率从22%降至0%。ERPBench支持对企业代理排名在竞争市场生态中是否能够转移的配对评估，并辅以汇总执行干预分析。代码和基准资源可在我们的https://github.com/GAIR-NLP/erp-bench获取。

View on arXiv Download PDF AI Translation

cs.AI / 32 / 2609.04678

Train What You Deploy:Token-Faithful Post-Training of a Production Coding

训练你所部署的：生产编码的令牌忠实后训练

Li, Cheng, Liu, Jiexiong, Chen, Yixuan, Hong, Chi

Abstract

Existing post-training pipelines for coding and terminal agents suffer severe token and control fidelity errors: simplified training environments mismatch production deployments, and offline token reconstruction from agent logs distorts original prompts and conflates policy calls with background model operations. We present a fidelity-aware training coupling framework that retains trainer-side sampling over original prompts, eliminates spurious model calls via a negotiated training protocol, and restricts loss computation to verifiable token spans with closed-failure guarantees. We further propose Certified Divergence Proximal Policy Optimization (C-DPPO), which establishes tight two-sided TV certification bounds, adaptive-K rules, budget-aware sequence guarantees, and error-robust policy masking atop standard DPPO. Evaluated on matched Baize5B and Baize10B models with identical training and test protocols on TMax-100, C-DPPO yields a consistent +3.0-point performance gain over standard DPPO across model scales. Certificate audits validate the reliability and full operational coverage of our certified training pipeline.

Chinese Translation

现有的编码和终端代理的后训练流程存在严重的令牌和控制保真度错误：简化的训练环境与生产部署不匹配，而从代理日志中离线重建的令牌扭曲了原始提示，并将策略调用与背景模型操作混淆。我们提出了一种关注保真的训练耦合框架，该框架保留了对原始提示的训练方采样，通过协商训练协议消除了虚假的模型调用，并将损失计算限制在具有可验证令牌范围的闭合失败保证内。我们进一步提出了认证偏差近端策略优化（Certified Divergence Proximal Policy Optimization, C-DPPO），该方法建立了紧密的双侧总变差认证界限、自适应 K 规则、预算感知序列保证和基于标准 DPPO 的错误鲁棒策略掩蔽。在与 TMax-100 上相同训练和测试协议的匹配 Baize5B 和 Baize10B 模型上进行评估，C-DPPO 在各模型规模上相较于标准 DPPO 一致地提高了 3.0 分的性能增益。证书审计验证了我们认证训练流程的可靠性和全面操作覆盖。

View on arXiv Download PDF AI Translation

cs.AI / 33 / 2609.04693

Predicting Spatiotemporal Mobile Sensing-Based PM2.5 Concentrations Using Low-Rank Adapted Spatially Attentive Graph Neural Network

基于低秩适应空间注意力图神经网络的时空移动传感PM2.5浓度预测

Chiddarwar, Om, Mandal, Priyanka, Chandaliya, Praveen Kumar, Arkatkar, Shriniwas

Abstract

Urban air quality can vary significantly along transit corridors, necessitating high-resolution monitoring. This work introduces a novel mobile-sensing dataset from Surat, Gujarat, India, comprising PM$*{2.5}$ concentrations, meteorological variables (temperature, humidity, wind speed, wind direction), and land-use features. To represent the spatiotemporal data as a graph, two node-definition strategies were used: (i) uniform segmentation (200--400~m intervals) and (ii) DBSCAN clustering to adaptively group dense observations. For each node, rolling mean and standard deviation of meteorological variables were computed. To model this high-dimensional data, we propose a SA-GNN for fine-grained, short-term PM$*{2.5}$ forecasting and hotspot identification. We compared SA-GNN with LSTM, RNN, GRU, and ANN models. These models performed well on low-resolution data but had difficulty capturing rapidly changing patterns in urban air quality. SA-GNN employs cluster-specific GRUs to capture localized temporal dependencies and a Graph Attention Network to learn spatial heterogeneity. This hybrid architecture effectively models rapid fluctuations and complex spatial interactions. On our dataset, SA-GNN achieved $R^2 = 0.95$, RMSE $= 6.8$, and MAE $= 4.2~\si{\micro\gram\per\meter\cubed}$, outperforming all baseline models. Combining spatial clustering with adaptive attention significantly improves forecasting, enabling real-time, fine-grained monitoring and supporting personalized exposure tracking and timely alerts for healthier cities.

Chinese Translation

城市空气质量在交通走廊沿线可能会显著变化，因此需要高分辨率的监测。本研究引入了来自印度古吉拉特邦苏拉特市的一种新型移动传感数据集，包含PM$*{2.5}$浓度、气象变量（温度、湿度、风速、风向）和土地利用特征。为了将时空数据表示为图，采用了两种节点定义策略：（i）均匀分段（200-400米间隔）和（ii）DBSCAN聚类以自适应地对密集观测进行分组。对于每个节点，计算了气象变量的滚动均值和标准差。为了对这一高维数据进行建模，我们提出了一种SA-GNN（空间注意力图神经网络）用于细粒度短期PM$*{2.5}$预测和热点识别。我们将SA-GNN与LSTM、RNN、GRU和ANN模型进行了比较。这些模型在低分辨率数据上表现良好，但在捕捉城市空气质量快速变化的模式时遇到了困难。SA-GNN采用特定聚类的GRU来捕捉局部时间依赖性，并使用图注意力网络来学习空间异质性。这种混合架构有效地建模了快速波动和复杂的空间交互。在我们的数据集中，SA-GNN达到了$R^2 = 0.95$，RMSE $= 6.8$，MAE $= 4.2~ ext{μg/m}^3$，超越了所有基线模型。将空间聚类与自适应注意力相结合显著提高了预测能力，使得实时、细粒度监测成为可能，并支持个性化暴露追踪和及时预警，以促进更健康的城市环境。

View on arXiv Download PDF AI Translation

cs.AI / 34 / 2609.04697

SQL-Zero: Self-Evolving Text-to-SQL

SQL-Zero：自我演化的文本到SQL

Pedrozo, Daniel Machado, Dollis, Julia Soares, de Oliveira, Bryan Lincoln Marques, Aguiar, Vinicius Alboneti, de Oliveira, Sávio Salvarino Teles, Soares, Telma Woerle de Lima

Abstract

Training a competitive Text-to-SQL agent usually depends on human-annotated natural-language/SQL pairs, which are expensive, domain-specific, and a bottleneck for scaling to new databases. We show it is possible to train a competitive solver with zero annotated pairs. We introduce SQL-Zero, a proposer-solver self-play in which a challenger and a solver start from the same base LLM and the only ground truth is execution against the database itself. The challenger generates SQL pairs calibrated to the solver's current difficulty (targeting "hard but solvable"), and both roles are updated with GRPO in alternating turns, with a template-level repetition penalty on the challenger to prevent diversity collapse. Training on BIRD databases with no labels, self-play improves over the zero-shot base on BIRD dev by 6.6 points at 3B and 7.3 points at 7B. It also scores higher than a matched control trained under the same recipe on human BIRD gold over the same databases, although an exact paired test does not resolve that margin. Transfer depends on scale: at 3B every iteration outperforms the base on unseen Spider databases and under lexical perturbation (Spider-Syn), where it also degrades less than the matched BIRD-gold control, whereas at 7B only the first iteration preserves transfer.

Chinese Translation

训练一个具有竞争力的文本到SQL代理通常依赖于人工标注的自然语言/SQL对，这些对昂贵、特定于领域，并且是扩展到新数据库的瓶颈。我们展示了在没有标注对的情况下训练一个竞争性求解器是可能的。我们引入了SQL-Zero，这是一种提议者-求解者自我对弈，其中挑战者和求解者从相同的基础大型语言模型（LLM）开始，唯一的真实依据是对数据库本身的执行。挑战者生成与求解者当前难度相适应的SQL对（目标是“困难但可解”），并且在交替回合中使用GRPO更新两个角色，同时对挑战者施加模板级重复惩罚，以防止多样性崩溃。在没有标签的BIRD数据库上进行训练，自我对弈在3B模型上相较于零-shot基线提高了6.6分，在7B模型上提高了7.3分。尽管精确的配对测试未能解决这一差距，但它在相同数据库上对人类BIRD金标准的匹配控制下得分更高。迁移依赖于规模：在3B模型下，每次迭代在未见过的Spider数据库和词汇扰动（Spider-Syn）中都优于基线，并且在这些情况下，其降级程度也低于匹配的BIRD金标准控制，而在7B模型下，只有第一次迭代保持了迁移。

View on arXiv Download PDF AI Translation

cs.AI / 35 / 2609.04699

Model Retirement Creates Reproducibility Risk in Biomedical AI Publications

模型退役导致生物医学人工智能论文的可重复性风险

Wolfrath, Nathan, Conroy, Meghan, Kosten, Thomas, Bell, Dave, Neupane, Bhabishya, Kindel, Jonah, Banerjee, Anjishnu, Deshpande, Priya, Taylor, Bradley, Kothari, Anai N.

Abstract

Background. Large language models (LLMs) are being adopted in biomedical research at a rapid and accelerating pace, yet commercial services that host many widely used models operate under deprecation schedules that can complicate scientific reproducibility. Methods. We searched PubMed for original research articles from 2022 through March 2026 that applied a specific LLM to a biomedical task. An extraction agent identified model names from 61,077 article abstracts with human reviewers validating a subset for extraction accuracy. Extracted model names were normalized to canonical model identifiers. Lifecycle data (release date, retirement date, status) were compiled for the 50 most frequently used models. Results. We identified 8,931 paper-model mentions spanning 5,242 unique publications after restricting the analysis to the 50 most frequently used models. Among these mentions, 77.7% cited a commercial closed-weight model. Overall, 42% involved a model that was already retired by the time of official publication or is scheduled to retire within two years of publication. The median interval from publication to model retirement was 538 days. Conclusion. Many biomedical publications using LLMs are on a trajectory toward computational non-reproducibility after publication. Model deprecation should be treated as a core reporting and preservation issue for biomedical research.

Chinese Translation

背景：大型语言模型（LLMs）在生物医学研究中的应用正以快速且加速的速度普及，然而，许多广泛使用的模型所依赖的商业服务遵循的退役计划可能会对科学可重复性造成影响。方法：我们检索了2022年至2026年3月期间在PubMed发表的、将特定LLM应用于生物医学任务的原创研究文章。通过提取代理从61,077篇文章摘要中识别模型名称，并由人工审阅部分样本以验证提取准确性。提取的模型名称被规范化为标准模型标识符。针对使用频率最高的50个模型，收集了生命周期数据（发布日期、退役日期、状态）。结果：在限制分析至使用频率最高的50个模型后，我们识别出8,931次论文-模型提及，涵盖5,242篇独立发表的文章。其中，77.7%的提及引用了商业闭源权重模型。总体来看，42%的提及涉及的模型在正式发表时已退役或计划在发表后两年内退役。发表至模型退役的中位时间间隔为538天。结论：许多使用LLM的生物医学论文在发表后面临计算不可重复性的风险。模型退役应被视为生物医学研究报告和数据保存的核心问题。

View on arXiv Download PDF AI Translation

cs.AI / 36 / 2609.04706

FinalityBench: An Effect-Level Benchmark for Agent Decisions Under Delayed and Conflicting Financial Finality

FinalityBench：一个针对延迟和冲突金融最终性的代理决策的效果级基准测试

Sharma, Abhishek

Abstract

A merchant's payment processor, ledger, ERP and bank feed are updated by messages that get delayed, duplicated, dropped and reordered, so for minutes at a time the four hold contradictory beliefs about the same order. An agent resolving the exception must decide whether to ship goods, re-submit a capture, refund or wait, knowing some of those cannot be undone. We present FinalityBench, an executable benchmark for that decision. It keeps a hidden canonical event log and derives each system's view from a separately faulted delivery stream, so disagreement follows from specified fault semantics rather than being authored. Grading is on executed monetary effects: an episode is scored by the merchant's terminal economic position, relative to a privileged reference told when the pending capture resolves. The corpus of 321 tasks includes 45 twin pairs (90 tasks): tasks whose four system views are identical at the decision instant, whose authoritative probes both return unknown, and whose eventual correct dispositions differ. That snapshot indistinguishability is checked under every evaluation seed rather than assumed; equivalence over all interaction traces is not claimed. Over 14,445 graded episodes from nine programmatic policies, ranking by single-task accuracy and by paired loss disagree in 7 places: a ship-on-first-sign policy is second-best by accuracy at 65.7% and worst in the suite by paired loss, because it cannot tell the two members apart. A runtime gating irreversible actions on an authoritative finality probe reaches 85.4% and, unlike every polling policy, loses nothing to pass^5; its residual loss is almost entirely one archetype, which prices finality information directly. Language models reach the same exact rate as the hand-written gate on a stratified subset, lose about twice as much money, and discover the finality-gating strategy without being told it.

Chinese Translation

商家的支付处理器、账本、企业资源计划（ERP）系统和银行数据流通过延迟、重复、丢失和重新排序的消息进行更新，因此在几分钟内，这四者对同一订单持有矛盾的信念。解决异常的代理必须决定是发货、重新提交捕获、退款还是等待，明白其中一些操作是不可撤销的。我们提出了FinalityBench，这是一个可执行的决策基准测试。它保持一个隐藏的规范事件日志，并从一个单独故障的交付流中推导每个系统的视图，因此分歧源于指定的故障语义，而不是人为创作。评分基于执行的货币效应：一个事件的得分由商家的终端经济位置决定，相对于在待处理捕获解决时告知的特权参考。321个任务的语料库包括45对双胞胎（90个任务）：在决策时刻四个系统视图相同、权威探测均返回未知且最终正确处置不同的任务。该快照的不可区分性在每个评估种子下进行检查，而不是假设；不声称在所有交互轨迹上等价。在来自九个程序策略的14,445个评分事件中，按单任务准确性和配对损失排名在7个地方存在分歧：一个优先发货的政策在准确性上以65.7%位列第二，但在配对损失中表现最差，因为它无法区分两个成员。一个在权威最终性探测上运行时门控不可逆操作的策略达到了85.4%，并且与每个轮询政策不同，没有任何损失；其剩余损失几乎完全是一个原型，直接定价最终性信息。语言模型在一个分层子集上达到了与手写门控相同的精确率，损失的金钱约为两倍，并在未被告知的情况下发现了最终性门控策略。

View on arXiv Download PDF AI Translation

cs.AI / 37 / 2609.04715

PLUME: Parameter-Efficient Personalization of Large Language Models via Low-Rank User Modulation in Shared Subspaces

PLUME：通过共享子空间中的低秩用户调制实现大型语言模型的参数高效个性化

Li, Xinyu, Zhou, Hao, Zhu, Jianfeng, Maharjan, Julina, Guo, Ruixin, Dragan, Feodor, Jin, Ruoming

Abstract

Personalizing large language models (LLMs) is essential for delivering AI assistance that aligns with individual users' styles, intents, and preferences. While per-user fine-tuning can substantially enhance personalization quality, it introduces significant parameter and storage overhead, limiting scalability to large user populations. We propose PLUME (Personalized Low-Rank Adaptation through User Modulation and Shared Subspace), a lightweight framework that achieves efficient and expressive per-user adaptation by leveraging a shared task-specific subspace. Specifically, PLUME first learns a global task subspace from aggregated user data. Personalization is then achieved by training only a lightweight small square matrix within this subspace, enabling each user to obtain a tailored model while keeping shared components fixed. Cross-layer shared parameters and rank-1 residual terms are further introduced to significantly reduce redundancy while maintaining expressiveness. Experiments on multiple personalized text generation benchmarks demonstrate that PLUME achieves comparable or superior performance to strong baselines, while reducing per-user parameters by over 95%. These results establish shared-subspace modulation with minimal residuals as a scalable and semantically grounded approach to LLM personalization.

Chinese Translation

个性化大型语言模型（LLMs）对于提供符合个体用户风格、意图和偏好的人工智能辅助至关重要。尽管针对每个用户的微调可以显著提升个性化质量，但它引入了显著的参数和存储开销，限制了对大规模用户群体的可扩展性。我们提出了PLUME（通过用户调制和共享子空间的个性化低秩适应），这是一个轻量级框架，通过利用共享的任务特定子空间，实现高效且富有表现力的用户个性化。具体而言，PLUME首先从聚合的用户数据中学习一个全局任务子空间。个性化是通过仅在该子空间内训练一个轻量级的小方阵来实现的，使每个用户能够获得一个量身定制的模型，同时保持共享组件不变。进一步引入跨层共享参数和秩-1残差项，以显著减少冗余，同时保持表现力。在多个个性化文本生成基准上的实验表明，PLUME在性能上与强基线相当或更优，同时将每个用户的参数减少超过95%。这些结果确立了具有最小残差的共享子空间调制作为一种可扩展且语义上扎实的LLM个性化方法。

View on arXiv Download PDF AI Translation

cs.AI / 38 / 2609.04738

Aplaud: Adaptive Personalized Low-Rank Decomposition for User-Specific LLM

Aplaud：用于用户特定大语言模型的自适应个性化低秩分解

Li, Xinyu, Jin, Ruoming, Zhu, Jianfeng, Guo, Ruixin, Liu, Zhi

Abstract

In this paper, we study the problem of personalized survey response prediction using fine-tuned large language models (LLMs). This task poses unique challenges: limited per-user training data, scalability of model storage, and the need to exploit shared structure across survey questions. To address these issues, we propose Aplaud (Adaptive Personalized Low-rank and User-specific Nested Decomposition), a lightweight and scalable framework for LLM personalization. Aplaud extends the LoRA paradigm by separating adaptation into a frozen, shared low-rank basis and a compact user-specific correction, augmented with a rank-one residual for finer personalization. To further reduce per-user parameter cost and mitigate overfitting, the correction matrix can be factorized into an even lower-rank form. Empirical results demonstrate that Aplaud achieves efficient, scalable personalization across users while outperforming state-of-the-art LoRA-based personalized LLM approaches in both generalization and inference efficiency.

Chinese Translation

在本文中，我们研究了使用微调的大型语言模型（LLMs）进行个性化调查响应预测的问题。该任务面临独特的挑战：每个用户的训练数据有限、模型存储的可扩展性以及需要利用调查问题之间的共享结构。为了解决这些问题，我们提出了Aplaud（自适应个性化低秩和用户特定嵌套分解），这是一个轻量级且可扩展的LLM个性化框架。Aplaud通过将适应性分离为一个冻结的、共享的低秩基础和一个紧凑的用户特定校正，扩展了LoRA范式，并通过增加一个秩为一的残差来实现更精细的个性化。为了进一步降低每个用户的参数成本并减轻过拟合，校正矩阵可以被分解为更低秩的形式。实证结果表明，Aplaud在用户间实现了高效、可扩展的个性化，同时在泛化能力和推理效率上超越了基于LoRA的最新个性化LLM方法。

View on arXiv Download PDF AI Translation

cs.AI / 39 / 2609.04749

DCFA: Dual-view Causal-inspired Attribution for Failure Reasoning in LLM-based Multi-agent Systems

DCFA：基于双视角的因果启发归因方法用于LLM基础的多智能体系统故障推理

Wang, Zehao, Wang, Lanjun, Jin, Shilong, Chen, Junjie, Xiao, Yanghua

Abstract

Large language model (LLM)-based multi-agent systems have experienced rapid growth in recent years. Despite their promise, such systems remain fragile, frequently exhibiting reasoning and coordination errors that can lead to system-level failures. Failure attribution in such systems relies on tracing natural language interactions among agents to identify the decisive error, which refers to the earliest action whose correction can reverse system failure. There are two key challenges: 1) Shallow attribution: Existing methods often capture only minor deviations, such as incomplete retrievals or formatting errors, which verification mechanisms can correct, while missing the decisive cause of system failure. 2) Contextual degradation: As the length of the system traces increases, the model's reasoning ability rapidly deteriorates. To address these challenges, we propose DCFA, a training-free framework for failure attribution. DCFA integrates a global module that constructs structured causal-inspired dependency graphs from system traces to identify the initial decisive error, and a local module that applies local counterfactual-inspired reasoning to refine causal-inspired attribution. Experiments on the Who&When benchmark across six LLMs show that DCFA improves step-level accuracy by up to 8.27% over state-of-the-art baselines.

Chinese Translation

基于大型语言模型（LLM）的多智能体系统近年来经历了快速增长。尽管这些系统展现出良好的前景，但它们仍然脆弱，常常出现推理和协调错误，导致系统级故障。此类系统中的故障归因依赖于追踪智能体之间的自然语言交互，以识别决定性错误，即能够逆转系统故障的最早行动。面临两个主要挑战：1）浅层归因：现有方法通常仅捕捉到轻微的偏差，如不完整的检索或格式错误，这些错误可以通过验证机制纠正，而未能识别系统故障的决定性原因。2）上下文退化：随着系统追踪长度的增加，模型的推理能力迅速下降。为了解决这些挑战，我们提出了DCFA，一种无训练的故障归因框架。DCFA集成了一个全局模块，该模块从系统追踪中构建结构化的因果启发依赖图，以识别初始决定性错误，以及一个局部模块，应用局部反事实启发推理来细化因果启发归因。在六个LLM上进行的Who&When基准实验表明，DCFA在步骤级准确性上比最先进的基线提高了多达8.27%。

View on arXiv Download PDF AI Translation

cs.AI / 40 / 2609.04767

Shadow Queries for Private Retrieval in Vector Databases

向量数据库中的私密检索阴影查询

Feng, Xinguo, Ma, Zhongkui, Wang, Zihan, Yan, Chuan, Yang, Guowei, Abuadbba, Alsharif, Bai, Guangdong

Abstract

Large language models (LLMs) increasingly rely on information retrieval (IR) systems, such as Retrieval-Augmented Generation (RAG), to incorporate domain-specific knowledge without costly re-training. These systems often store pre-computed document embeddings in cloud-based vector databases. However, such embeddings are vulnerable to embedding inversion attacks (EIAs), which can reconstruct their underlying text. Existing defenses, such as adding noise or scaling embeddings, often provide limited privacy or significantly reduce retrieval utility. We propose SHAQ (shadow query generation), a semantic-decomposition and embedding-decoupling defense against EIAs. SHAQ is based on the insight that EIAs rely on the strong coupling between an embedding and its original text. Instead of storing document embeddings directly, SHAQ uses a generative language model to create diverse shadow queries that capture different semantic aspects of each document. These queries are then encoded and stored in place of the original document embeddings, thereby decomposing document semantics and decoupling stored embeddings from the source text. Experiments across diverse IR datasets show that SHAQ substantially improves privacy while preserving retrieval utility, achieving a recovery rate as low as 0.2104, defending up to 19.50% more tokens than baseline defenses, and reaching up to 0.7967 MAP@10 with up to 5.53% utility improvement. These results demonstrate that semantic decomposition and embedding decoupling provide an effective alternative to directly modifying embeddings for defending against EIAs.

Chinese Translation

大型语言模型（LLMs）越来越依赖信息检索（IR）系统，例如检索增强生成（RAG），以在不进行昂贵再训练的情况下融入特定领域的知识。这些系统通常在基于云的向量数据库中存储预计算的文档嵌入。然而，这些嵌入容易受到嵌入反演攻击（EIAs）的威胁，攻击者可以重建其底层文本。现有的防御措施，如添加噪声或缩放嵌入，通常提供有限的隐私保护或显著降低检索效用。我们提出了SHAQ（阴影查询生成），这是一种针对EIAs的语义分解和嵌入解耦防御方法。SHAQ的基础是EIAs依赖于嵌入与其原始文本之间的强耦合这一洞察。SHAQ并不直接存储文档嵌入，而是利用生成语言模型创建多样化的阴影查询，以捕捉每个文档的不同语义方面。这些查询随后被编码并存储，以替代原始文档嵌入，从而实现文档语义的分解，并将存储的嵌入与源文本解耦。在多种IR数据集上的实验表明，SHAQ显著提高了隐私保护，同时保持了检索效用，恢复率低至0.2104，防御的令牌数量比基线防御多出19.50%，并且在MAP@10上达到0.7967，效用提升高达5.53%。这些结果表明，语义分解和嵌入解耦为防御EIAs提供了一种有效的替代方案，而不是直接修改嵌入。

View on arXiv Download PDF AI Translation

cs.AI / 41 / 2609.04778

Diffusion Language Models for Mobile Edge Agentic AI: Foundations, Applications, and Challenges

移动边缘代理人工智能的扩散语言模型：基础、应用与挑战

Li, Chenqi, Min, Minghui, Niyato, Dusit, Ni, Wei

Abstract

Diffusion language models (DLMs) offer a non-autoregressive alternative for mobile edge agentic artificial intelligence (AI) by refining tokens through iterative denoising rather than left-to-right decoding. Compared with autoregressive Transformer-based large language models (LLMs), DLMs can update multiple uncertain tokens in parallel and exploit bidirectional context throughout the generation process, enabling more flexible quality-latency trade-offs beyond fixed sequential decoding. These properties are particularly attractive for edge agents, where partial refinement, early exit, and constraint-guided correction can reduce response delay and communication overhead while improving robustness under noisy, incomplete, or dynamic contexts. This survey reviews DLM foundations and analyzes their suitability for edge settings under latency, memory, energy, bandwidth, privacy, and reliability constraints. We cover resource-efficient architectures, training and inference acceleration, compression, edge/cloud deployment, communication-aware serving, Internet of Things (IoT)/wireless applications, and evaluation of DLM-based agents. We further discuss open issues in long-context state management, split inference, trustworthy execution, multimodal grounding, and reproducible benchmarking. The goal is to connect DLM modeling properties, including bidirectionality, parallel refinement, controllability, and quality-latency elasticity, with system-level requirements of future mobile edge intelligence.

Chinese Translation

扩散语言模型（DLMs）通过迭代去噪而非从左到右解码，为移动边缘代理人工智能（AI）提供了一种非自回归的替代方案。与基于自回归的变换器大型语言模型（LLMs）相比，DLMs能够并行更新多个不确定的标记，并在生成过程中利用双向上下文，从而实现超越固定顺序解码的更灵活的质量-延迟权衡。这些特性对于边缘代理尤为吸引，因为部分精炼、提前退出和约束引导的修正可以在嘈杂、不完整或动态环境下减少响应延迟和通信开销，同时提高鲁棒性。本调查回顾了DLM的基础，并分析了它们在延迟、内存、能量、带宽、隐私和可靠性约束下在边缘环境中的适用性。我们涵盖了资源高效的架构、训练和推理加速、压缩、边缘/云部署、通信感知服务、物联网（IoT）/无线应用以及基于DLM的代理的评估。我们进一步讨论了长上下文状态管理、分割推理、可信执行、多模态基础和可重复基准测试等开放问题。我们的目标是将DLM建模特性，包括双向性、并行精炼、可控性和质量-延迟弹性，与未来移动边缘智能的系统级需求相连接。

View on arXiv Download PDF AI Translation

cs.AI / 42 / 2609.04782

DODR: Deterministic Operator-Driven Reasoning in Latent Space

DODR：潜在空间中的确定性算子驱动推理

Huang, Weicai

Abstract

Autoregressive (AR) large language models formulate reasoning as token-level probabilistic sampling, which induces three fundamental defects in complex logical reasoning: error accumulation, probability substituting necessity, and the linear-chain information bottleneck. This paper proposes the Deterministic Operator-Driven Reasoning in Latent Space architecture (DODR), which reconstructs reasoning as reasoning-graph computation in a high-dimensional linear-algebraic space. Reasoning states are represented as snapshot vectors whose primitives are semantic units (phrases or sentences) rather than tokens, and each inference step is a deterministic matrix operation with no token sampling. Peirce's three inference types are formalized as three trainable matrix operators: a rank-deficient deduction operator (information collapse), a full-rank induction operator (information expansion), and an abduction operator defined as the Moore-Penrose pseudo-inverse of deduction (information hypothesizing). We prove that the operator set is minimal and complete given Peirce's trichotomy, that no single "super-operator" can realize all three types (a rank obstruction), and that reasoning graphs are Turing-complete with contractive backflow converging by Banach's fixed-point theorem. Experiments on 503 sample records (420 deduplicated samples) across dedicated and end-to-end settings show: deduction loss converges to 1.40e-05; induction achieves 0.9996 generalization coverage with 20/20 hard vetoes on counterexamples; abduction solutions exceed the random baseline by 28x with judgment accuracies of 72.5% (58/80, Wilson 95% CI [61.9%, 81.1%]) and 81.7% (49/60, CI [70.1%, 89.4%]); frozen operators attain 100% (60/60) on unseen cross-domain deduction. The architecture provides a structural zero-hallucination guarantee and a three-layer continual-learning mechanism. All data and code are released.

Chinese Translation

自回归（AR）大型语言模型将推理形式化为基于标记级概率采样的过程，这在复杂逻辑推理中引发了三大根本缺陷：错误累积、概率替代必然性以及线性链信息瓶颈。本文提出了潜在空间中的确定性算子驱动推理架构（DODR），该架构将推理重构为高维线性代数空间中的推理图计算。推理状态表示为快照向量，其基本单元为语义单元（短语或句子）而非标记，每一步推理均为无标记采样的确定性矩阵运算。基于皮尔士（Peirce）的三种推理类型，形式化为三种可训练矩阵算子：秩亏的演绎算子（信息坍缩）、满秩的归纳算子（信息扩展）以及作为演绎算子摩尔－彭若斯伪逆的溯因算子（信息假设）。我们证明了该算子集合在皮尔士三分法下的最小性与完备性，且不存在单一“超算子”能实现三种推理类型（秩障碍），推理图通过巴拿赫不动点定理的收缩反向流实现图灵完备。基于503条样本记录（420条去重样本）在专用及端到端设置上的实验表明：演绎损失收敛至1.40e-05；归纳在反例上的20/20强否决下实现了0.9996的泛化覆盖率；溯因解答较随机基线提升28倍，判断准确率分别为72.5%（58/80，Wilson 95%置信区间[61.9%, 81.1%]）和81.7%（49/60，置信区间[70.1%, 89.4%]）；冻结算子在未见跨域演绎任务上达成100%（60/60）准确率。该架构提供了结构性零幻觉保证及三层持续学习机制。所有数据与代码均已公开。

View on arXiv Download PDF AI Translation

cs.AI / 43 / 2609.04793

ProtLingo: Efficient Protein Language Modeling via Conditional Memory and Expert Routing

ProtLingo：通过条件记忆和专家路由实现高效的蛋白质语言建模

Li, Mingrui, Shen, Sixian, Li, Minzhang, Zhang, Ruiyi, Zhang, Kexin, Zhang, Jiakai, Yu, Jingyi

Abstract

Proteins perform diverse cellular functions, and even single amino-acid substitutions can alter stability, activity, or molecular interactions. Protein language models (PLMs) provide a scalable approach for modeling such sequence--function relationships from unlabeled sequences, but increasing the size of dense Transformer backbones often brings substantial computational cost without consistently improving mutation-sensitive prediction. We introduce ProtLingo, an efficient PLM framework that augments a pretrained single-sequence backbone with conditional local memory and sparse expert routing. ProtLingo maps contextual residue representations into route-specific discrete codes, composes centered local windows into latent $N$-gram addresses, and retrieves reusable residual signals associated with recurring local sequence contexts. In parallel, selected feed-forward blocks are upcycled into sparse Mixture-of-Experts layers with shared and routed experts, enabling residue-dependent computation while activating only a subset of parameters. Experiments on protein fitness prediction, FLIP benchmarks, and supervised contact prediction show that ProtLingo achieves competitive performance with a 150M-scale backbone, including strong parameter efficiency on mutation-effect prediction and preserved long-range structural representations.

Chinese Translation

蛋白质执行多种细胞功能，甚至单个氨基酸的替代都可能改变其稳定性、活性或分子相互作用。蛋白质语言模型（PLMs）提供了一种可扩展的方法，用于从未标记序列中建模此类序列-功能关系，但增加密集的Transformer骨干网络的规模往往会带来可观的计算成本，而未必能持续改善对突变敏感的预测。我们提出了ProtLingo，一个高效的PLM框架，通过条件局部记忆和稀疏专家路由增强了预训练的单序列骨干网络。ProtLingo将上下文残基表示映射为特定路由的离散编码，将中心局部窗口组合成潜在的$N$-gram地址，并检索与重复局部序列上下文相关的可重用残差信号。同时，选定的前馈块被升级为具有共享和路由专家的稀疏混合专家层，使得依赖残基的计算得以实现，同时仅激活一部分参数。在蛋白质适应性预测、FLIP基准测试和监督接触预测的实验中，ProtLingo在150M规模的骨干网络上实现了具有竞争力的性能，包括在突变效应预测上的强参数效率和保留长程结构表示。

View on arXiv Download PDF AI Translation

cs.AI / 44 / 2609.04801

Whose record is this? Diagnosing and authorizing record use in personalized multimodal models

这是谁的记录？在个性化多模态模型中诊断和授权记录使用

Mao, Xinyu, Li, Junsi, Liu, Chenyang, Zhang, Haoji, Sun, Ming

Abstract

Contextualized visual personalization can retrieve a true record yet apply it to the wrong visual subject. We formalize when a record may condition an answer as \emph{record authorization}: subject presence ($P$), record-edge validity ($E$), and answer support ($S$) must all hold. We call violations visual memory misbinding (VMM). We construct RecordAuth-Diag, a 3,690-case matched diagnostic suite that changes one image--record edge while holding the query, question, record text, and image multiset fixed. Card removal and nonce relabeling attribute these failures to supplied records. Raw-bank failures span Qwen-, Phi-, and Gemma-family interfaces: Gemma-3-4B-IT reaches 63.69\% local unauthorized use at 25.75\% clean recall. CoViP remains at 26.02\%, versus 22.49\% for its Qwen backbone at similar clean recall. Typed pre-generation authorization reduces Qwen card exposure on RecordAuth-Diag from 43.63\% to 3.06\%, while positive recall changes from 86.26\% to 60.90\%. Full $P\wedge E\wedge S$ validation uses 560 localized DAVIS cases: top-1 relevance and typed authorization have comparable release (28.93\% and 28.39\%) but 6.79\% and 0.89\% unsafe release, respectively. Of the 33 additional unsafe cases removed, 27 are support, 4 edge, 2 clean, and 0 boundary cases. Thus the observed increment is an $E\wedge S$ decision dominated by support, not an edge check alone. Appearance supplies $E$ evidence only conditional on $P$; authenticated subject tokens instantiate the missing presence witness as a sufficiency control. The claims concern the evaluated contracts, not natural prevalence, consent, or visual identity

Chinese Translation

上下文化的视觉个性化可以检索真实记录，但可能应用于错误的视觉主体。我们将记录可能条件化答案的情况形式化为 extit{记录授权}：主体存在（$P$）、记录边缘有效性（$E$）和答案支持（$S$）必须同时成立。我们将违反这一条件的情况称为视觉记忆错误绑定（VMM）。我们构建了RecordAuth-Diag，这是一个包含3,690个案例的匹配诊断套件，在保持查询、问题、记录文本和图像多重集不变的情况下，改变一个图像-记录边缘。卡片移除和临时重新标记将这些失败归因于提供的记录。原始银行的失败跨越Qwen、Phi和Gemma系列接口：Gemma-3-4B-IT在25.75%的干净召回率下达到63.69%的局部未授权使用。CoViP保持在26.02%，而其Qwen主干在类似的干净召回率下为22.49%。类型化的预生成授权将RecordAuth-Diag上Qwen卡片的曝光率从43.63%降低到3.06%，而正召回率从86.26%降至60.90%。完整的$Pigwedge Eigwedge S$验证使用560个本地化的DAVIS案例：top-1相关性和类型授权的发布率相当（28.93%和28.39%），但不安全发布率分别为6.79%和0.89%。在移除的33个额外不安全案例中，27个是支持，4个是边缘，2个是干净，0个是边界案例。因此，观察到的增量是由支持主导的$Eigwedge S$决策，而不仅仅是边缘检查。外观仅在$P$的条件下提供$E$证据；经过验证的主体令牌作为充分性控制实例化缺失的存在证人。这些主张涉及评估的合同，而非自然流行、同意或视觉身份。

View on arXiv Download PDF AI Translation

cs.AI / 45 / 2609.04803

Hierarchical Possession-Aware Graph Pointer Network for Pass Receiver Selection

层次化持球感知图指针网络用于传球接收者选择

Wang, Jingyi, Li, Da, Wang, Kaixin, Huang, Zhangqin

Abstract

Pass receiver selection is a fundamental task in football analytics, aiming to predict the intended receiver under a given game state. This task is challenging with event-centered freeze-frame observations, a broadcast-like setting that provides only partial and variable player visibility without complete trajectories or stable player identities. The model must therefore reason over anonymous visible candidates, opponent pressure, and recent context under partial observation. To address this setting, we propose a Hierarchical Possession-aware Graph Pointer Network (HPGPN), which formulates pass receiver selection as variable-size candidate prediction over visible teammates. HPGPN jointly models current player interactions, local event context, and possession-level temporal dynamics. It represents the current pass situation with a graph, incorporates fixed event context, and uses dynamic possession history to capture how the attacking sequence evolves. Candidate representations are refined hierarchically by integrating spatial, contextual, and historical evidence, and a glimpse pointer head scores the receiver candidates. Experiments on public football event and freeze-frame data show that HPGPN improves pass receiver selection performance. Ablation studies demonstrate the effectiveness of graph-based interaction modeling, fixed event context, and dual-branch dynamic possession-history modeling.

Chinese Translation

传球接收者选择是足球分析中的一项基础任务，旨在预测在给定比赛状态下的意图接收者。由于事件中心的静态观察，这项任务具有挑战性，这种广播式的设置仅提供部分和可变的球员可见性，而没有完整的轨迹或稳定的球员身份。因此，模型必须在部分观察下，对匿名可见候选者、对手压力和近期背景进行推理。为了解决这一问题，我们提出了一种层次化持球感知图指针网络（HPGPN），将传球接收者选择公式化为对可见队友的可变大小候选预测。HPGPN联合建模当前球员互动、局部事件背景和持球级时间动态。它通过图表示当前传球情况，结合固定事件背景，并利用动态持球历史捕捉进攻序列的演变。候选者表示通过整合空间、背景和历史证据进行层次化精炼，而一个快速指针头对接收者候选者进行评分。在公共足球事件和静态数据上的实验表明，HPGPN提高了传球接收者选择的性能。消融研究证明了基于图的互动建模、固定事件背景和双分支动态持球历史建模的有效性。

View on arXiv Download PDF AI Translation

cs.AI / 46 / 2609.04804

MedFlow: Class-Aware Multi-Scale Generation for Medical Time-Series Synthesis

MedFlow：面向医疗时间序列合成的类别感知多尺度生成

Huang, Yanhao, Feng, Shibo, Feng, Wanjin, Zhao, Peilin, Miao, Chunyan

Abstract

Synthetic medical time-series generation can alleviate data scarcity and support the development of reliable clinical prediction models. However, existing methods mainly focus on matching the overall distribution and temporal dynamics of real data, which does not necessarily ensure strong downstream utility on imbalanced medical datasets. Clinically informative patterns often occur at heterogeneous temporal scales, while rare minority-class characteristics can be obscured by dominant population patterns. To address these challenges, we propose MedFlow, a class-aware multi-scale flow matching framework for medical time-series synthesis. MedFlow employs a vector-quantized multi-scale tokenizer to represent medical sequences at complementary temporal resolutions, capturing both coarse clinical trends and fine-grained dynamics. We further introduce Token Marginal Guidance, which incorporates class-conditional token statistics directly into the flow matching process to steer generation toward class-specific regions of the learned tokens. This mechanism strengthens minority-class patterns, while preserving the global and tail distributions of real data. Experiments on four public datasets covering electronic health records, EEG, and ECG signals demonstrate that MedFlow consistently outperforms recent state-of-the-art diffusion-based baselines across downstream prediction tasks. On average, it improves AUPRC by 5.8%, reduces Context-FID by 88.6%, and achieves 3.8$\times$ higher sampling throughput.

Chinese Translation

合成医疗时间序列的生成可以缓解数据稀缺问题，并支持可靠临床预测模型的发展。然而，现有方法主要集中于匹配真实数据的整体分布和时间动态，这并不一定能确保在不平衡医疗数据集上的强大下游效用。临床信息模式通常发生在异构时间尺度上，而稀有的少数类别特征可能会被主流人群模式所掩盖。为了解决这些挑战，我们提出了MedFlow，一种面向医疗时间序列合成的类别感知多尺度流匹配框架。MedFlow采用向量量化的多尺度标记器，以互补的时间分辨率表示医疗序列，捕捉粗略的临床趋势和细粒度的动态。我们进一步引入了标记边际引导（Token Marginal Guidance），该方法将类别条件的标记统计直接融入流匹配过程，以引导生成朝向学习到的标记的类别特定区域。该机制增强了少数类别模式，同时保留了真实数据的全局和尾部分布。在涵盖电子健康记录、脑电图（EEG）和心电图（ECG）信号的四个公共数据集上的实验表明，MedFlow在下游预测任务中始终优于最近的基于扩散的最先进基线。平均而言，它提高了AUPRC 5.8%，减少了Context-FID 88.6%，并实现了3.8倍的采样吞吐量。

View on arXiv Download PDF AI Translation

cs.AI / 47 / 2609.04806

When Financial Fine-tuning Fails: A Three-Level Detectability Analysis of Numerical Hallucination in Domain-Adapted Language Models

当金融微调失败时：领域适应语言模型中数值幻觉的三层可检测性分析

Li, Xiaodong, Liu, Peiwei

Abstract

Financial large language models are increasingly deployed for summarization of reports and disclosures, where numerical hallucination poses significant practical risks. While prior work often attributes such hallucination to insufficient numerical reasoning, this assumption has not been systematically tested under controlled fine-tuning settings. In this paper, we conduct a cost-effective, controlled study of numerical hallucination in financial summarization across three model variants: a base instruction-tuned model, a domain language-adapted model (FT-A), and a numeracy-enhanced domain model (FT-A+B+C). We introduce a three-level detectability taxonomy distinguishing between overt hallucination (currency-denominated fabrication), covert-explicit hallucination (professional-convention numbers), and covert-implicit hallucination (ungrounded quantitative claims). Our results reveal that domain fine-tuning substantially degrades numerical restraint at all detectability levels. While the Base model maintains near-zero hallucination rates (5.4\%), FT-A exhibits 82.5\% overt hallucination and FT-A+B+C reaches 98\%. Contrary to intuition, numeracy supervision amplifies rather than mitigates hallucination across all levels. We identify template injection---the insertion of memorized canonical values regardless of input content---as a primary hallucination mechanism in fine-tuned models. These findings demonstrate that numerical hallucination in financial summarization is driven by the degradation of numerical restraint through domain adaptation, not by insufficient numerical reasoning. We recommend that evaluation protocols assess hallucination across all detectability levels and that deployment practices include explicit mechanisms for grounding-aware generation or abstention.

Chinese Translation

金融大型语言模型越来越多地被用于报告和披露的摘要，其中数值幻觉带来了显著的实际风险。虽然先前的研究通常将这种幻觉归因于数值推理不足，但这一假设在受控微调设置下尚未得到系统验证。本文对金融摘要中的数值幻觉进行了成本效益高的受控研究，涵盖三种模型变体：基础指令微调模型、领域语言适应模型（FT-A）和增强数值能力的领域模型（FT-A+B+C）。我们引入了一个三层可检测性分类法，区分显性幻觉（以货币计价的虚构）、隐性显性幻觉（专业惯例数字）和隐性隐性幻觉（无根据的定量主张）。我们的结果显示，领域微调在所有可检测性层面上显著降低了数值约束。基础模型的幻觉率接近零（5.4%），而FT-A表现出82.5%的显性幻觉，FT-A+B+C则达到了98%。与直觉相反，数值监督在所有层面上都加剧了幻觉，而非减轻。我们确定模板注入——无论输入内容如何插入记忆中的标准值——是微调模型中主要的幻觉机制。这些发现表明，金融摘要中的数值幻觉是由于领域适应导致的数值约束降低，而不是由于数值推理不足。我们建议评估协议在所有可检测性层面上评估幻觉，并且部署实践应包括明确的机制以实现基于基础的生成或避免生成。

View on arXiv Download PDF AI Translation

cs.AI / 48 / 2609.04809

CPR-IE:A Compression-Prediction-Resource Intelligence Efficiency Metric

CPR-IE：一种压缩-预测-资源智能效率度量

Jiang, Xiantao

Abstract

Comparing intelligent systems under deployment constraints requires more than predictiveaccuracy.This paper develops Compression-Prediction-Resource Intelligence Efficiency (CPR-IE) as a protocol-relative ordering by representational economy, predictive quality, and resourceburden. The analysis separates two questions-how raw resource consumption is represented, andhow the resulting attributes are aggregated. Proportional-increment composition uniquely yieldslogarithmic cumulative burden, and context-independent ratio response yields power responsesto compression, prediction, and burden; with reference normalization the representation is I(C,P,T).We prove Pareto consistency, unit invariance, boundary behavior, trade-off identities, ranking-stability regions, and cross-task aggregation. A translog parent model makes interaction restrictions explicit, and further results establish cardinal and ordinal identification, sub-Gaussianfinite-sample ranking guarantees, robust selection under exponent uncertainty, and deterministicregret bounds. Minimum description length, algorithmic complexity, proper scoring rules, varia-tional inference, and Landauer's principle motivate measurement choices but do not entail theformula. CPR-IE is a constructed efficiency representation, not a universal law or a definition ofintelligence itself.

Chinese Translation

在部署约束下比较智能系统需要的不仅仅是预测准确性。本文开发了压缩-预测-资源智能效率（Compression-Prediction-Resource Intelligence Efficiency，CPR-IE），作为一种基于表征经济性、预测质量和资源负担的协议相对排序。分析将两个问题分开——如何表示原始资源消耗，以及如何聚合结果属性。比例增量组合独特地产生对数累积负担，而与上下文无关的比率响应则对压缩、预测和负担产生幂次响应；通过参考归一化，表征为 I(C,P,T)。我们证明了帕累托一致性、单位不变性、边界行为、权衡身份、排名稳定区域和跨任务聚合。一个跨对数父模型明确了交互限制，进一步结果建立了基数和序数识别、次高斯有限样本排名保证、在指数不确定性下的稳健选择以及确定性遗憾界限。最小描述长度、算法复杂性、适当评分规则、变分推断和兰道尔原则激励了测量选择，但并不意味着公式的存在。CPR-IE 是一种构造的效率表征，而不是普遍法则或智能本身的定义。

View on arXiv Download PDF AI Translation

cs.AI / 49 / 2609.04840

Long Horizon Transformer Quantile Fault Prediction for Multi Site Industrial Predictive Maintenance

多地点工业预测性维护的长期变换器分位故障预测

Poland, David J, Ravi, Daniele, Helian, Na

Abstract

Long-horizon predictive maintenance requires models to distinguish slowly evolving degradation from normal operating-regime variation over planning windows measured in days rather than hours. This paper evaluates whether an explicit conditional-quantile representation provides an informative classifier interface for this problem. The proposed TQRNN30d framework combines a dual-stage quantile regression neural network (QRNN) feature extractor with a multi-stream temporal fusion classifier. Each hourly word of 81-channel machine behaviour is mapped to a 324-dimensional quantile-state representation, and 720 ordered hourly words form the 30-day document supplied to the long-horizon model. The classifier fuses quantile states with dynamic covariates, channel-level static metadata, and a 168-hour latent-history stream using gated residual processing, causal recurrent encoding, and metadata-conditioned cross-modal attention. A bounded instability-aware signal derived from sustained one-word-ahead prediction-error divergence provides auxiliary memory modulation at the longest horizon. Evaluation uses a machine-disjoint 43/14/15 train/validation/test allocation across 72 machines in nine manufacturing facilities. At 30 days, TQRNN30d achieves 79.97% F1, 80.18% recall, 81.82% precision, 82.39% accuracy, and 0.820 ROC-AUC. It leads all 18 evaluated baselines at the 7-, 14-, and 30-day fixed-threshold comparisons, with the largest F1 advantage at 14 days. The results support held-out-machine performance within the observed homogeneous nine-facility fleet, but do not establish unseen-site, cross-equipment, or cross-sector generalisation.

Chinese Translation

长期预测性维护要求模型能够区分缓慢演变的退化与正常操作状态变化，这一过程的时间尺度以天而非小时为单位进行规划。本文评估了显式条件分位数表示是否为该问题提供了一个信息丰富的分类器接口。所提出的 TQRNN30d 框架结合了双阶段分位回归神经网络（QRNN）特征提取器与多流时间融合分类器。每小时的 81 通道机器行为被映射为 324 维的分位状态表示，720 个有序的每小时词汇形成供给长期模型的 30 天文档。分类器通过门控残差处理、因果递归编码以及基于元数据的跨模态注意机制，将分位状态与动态协变量、通道级静态元数据和 168 小时潜在历史流进行融合。由持续的一步前预测误差发散所衍生的有界不稳定性感知信号在最长时间范围内提供辅助记忆调制。评估使用了在九个制造设施中对 72 台机器进行的机器不重叠的 43/14/15 训练/验证/测试分配。在 30 天时，TQRNN30d 达到了 79.97% 的 F1 值、80.18% 的召回率、81.82% 的精确率、82.39% 的准确率和 0.820 的 ROC-AUC。在 7 天、14 天和 30 天的固定阈值比较中，它在所有 18 个评估基线中表现最佳，14 天时的 F1 优势最大。结果支持在观察到的同质九个设施舰队中的持有机器性能，但未能确立未见地点、跨设备或跨行业的泛化能力。

View on arXiv Download PDF AI Translation

cs.AI / 50 / 2609.04850

ElderBench: Benchmarking Autonomous Mobile Agents for Older Adults

ElderBench：老年人自主移动代理的基准测试

Zhan, Weide, Shaqu, Qumu, Liu, Yuanqing, Zhang, Peng, Liu, Jiahao, Lam, Kam Him, Gu, Ning, Hu, Zhan, Lu, Tun

Abstract

While autonomous mobile agents hold great potential for assisting older adults with smartphone usage, existing GUI benchmarks mainly rely on explicit, goal-oriented instructions and rarely capture the naturally occurring language patterns of older users, such as indirect speech, referential ambiguity, and under-specified requests. This mismatch between benchmark instructions and real-world elderly interactions may hinder reliable agent deployment. To address this gap, we present ElderBench, the first benchmark for evaluating mobile GUI agents in authentic elderly-oriented scenarios. ElderBench is constructed from 249 naturally elicited smartphone tasks collected from older adults across 20 applications. We first characterize the linguistic divergence between elderly instructions and existing GUI benchmark instructions from syntactic, semantic, and pragmatic perspectives. We then evaluate mainstream GUI agents and Vision-Language Models under both online and offline settings, revealing substantial performance degradation when handling elderly-oriented instructions. Through controlled instruction normalization, failure analysis, and fine-grained linguistic feature analysis, we further identify how elderly-specific language patterns contribute to agent failures. Our findings provide actionable design insights toward more adaptive, interpretable, and age-inclusive GUI agents for older adults.

Chinese Translation

尽管自主移动代理在协助老年人使用智能手机方面具有巨大潜力，但现有的图形用户界面（GUI）基准测试主要依赖于明确的、目标导向的指令，鲜少捕捉到老年用户自然发生的语言模式，如间接表达、指称模糊和不明确请求。这种基准指令与现实世界老年人互动之间的差距可能会妨碍代理的可靠部署。为了解决这一问题，我们提出了ElderBench，这是第一个用于评估移动GUI代理在真实老年人场景中的基准测试。ElderBench由来自20个应用程序的249个自然引导的智能手机任务构成，均由老年人收集。我们首先从句法、语义和语用的角度描述了老年指令与现有GUI基准指令之间的语言差异。然后，我们在在线和离线环境下评估主流的GUI代理和视觉-语言模型，发现处理老年人导向指令时，性能显著下降。通过控制指令规范化、失败分析和细粒度语言特征分析，我们进一步识别出老年特有的语言模式如何导致代理失败。我们的研究结果为设计更具适应性、可解释性和包容性的老年人GUI代理提供了可操作的设计见解。

View on arXiv Download PDF AI Translation

cs.AI / 51 / 2609.04859

MM-IFEval-Pro: A Multilingual and Attack-Resistant Benchmark for Instruction-Following in Vision-Language Models

MM-IFEval-Pro：一个多语言和抗攻击的视觉语言模型指令跟随基准

Xiao, Changming, Ni, Zhenliang, He, Jinhui, Shu, Han, Hu, Jie

Abstract

As vision-language models (VLMs) rapidly advance in image understanding, cross-modal reasoning, and complex instruction execution, instruction-following capability has become a key indicator of their reliability and practicality. However, existing multimodal instruction-following benchmarks still suffer from limited language coverage and insufficient adversarial safety scenarios, making them inadequate for evaluating real-world multilingual and safety-sensitive settings. To address these gaps, we present MM-IFEval-Pro, a multimodal instruction-following benchmark covering Chinese and English tasks as well as diverse instruction hijacking cases. MM-IFEval-Pro includes 4 major task categories and 24 subcategories and 8 instruction categories with 52 subcategories, with each sample containing an average of 3.0 constraints to realistically simulate complex instruction scenarios. We further construct a reinforcement-learning training set enriched with Chinese and adversarial instructions, which significantly improves model performance on MM-IFEval-Pro and transfers effectively to other mainstream multimodal benchmarks, demonstrating strong cross-task and cross-language generalization.

Chinese Translation

随着视觉语言模型（VLMs）在图像理解、跨模态推理和复杂指令执行方面的快速发展，指令跟随能力已成为评估其可靠性和实用性的关键指标。然而，现有的多模态指令跟随基准仍然存在语言覆盖面有限和对抗安全场景不足的问题，使其在评估真实世界的多语言和安全敏感环境时显得不够充分。为了解决这些问题，我们提出了MM-IFEval-Pro，一个涵盖中文和英文任务以及多样化指令劫持案例的多模态指令跟随基准。MM-IFEval-Pro包括4个主要任务类别和24个子类别，以及8个指令类别和52个子类别，每个样本平均包含3.0个约束，以真实地模拟复杂的指令场景。我们进一步构建了一个增强了中文和对抗指令的强化学习训练集，这显著提高了模型在MM-IFEval-Pro上的表现，并有效转移到其他主流多模态基准，展示了强大的跨任务和跨语言泛化能力。

View on arXiv Download PDF AI Translation

cs.AI / 52 / 2609.04864

MZ-Rain: Moisture-Budget-Guided Zero-Inflated Model for Station-Level Precipitation Nowcasting

MZ-Rain：基于水分预算指导的零膨胀模型用于站点级降水短期预报

Zhang, Yifang, Xiong, Shengwu, Wang, Henan, Yin, Wenjie, Zhang, Yuqiang, Zhou, Chen, Chen, Hua, Zhao, Qile, Duan, Pengfei

Abstract

Accurate station-level precipitation nowcasting is critical for agriculture, water resource management, and disaster prevention, which typically is formulated as a time series forecasting problem. However, conventional time-series modeling techniques face two major challenges in addressing station-level precipitation nowcasting: (1) Lack of Physics-Guided Modeling}, where meteorological variables are treated as a homogeneous set without accounting for their distinct roles in precipitation formation, leads to predictions that deviate from the physical processes governing precipitation. (2) Severe zero inflation in precipitation, where dry intervals dominate the dataset, obscuring meaningful precipitation patterns and complicating the predictive modeling. To address these challenges, we propose \textbf{MZ-Rain}, a moisture-budget-guided zero-inflated sLSTM framework for station-level precipitation nowcasting. Guided by the moisture budget equation, MZ-Rain decomposes the precipitation formation process into process-specific pathways corresponding to moisture storage, moisture transport, surface evaporation, and precipitation persistence, and captures their temporal evolution through dedicated sLSTM branches. To account for the zero-inflated nature of precipitation, MZ-Rain introduces an adaptive Tweedie modeling strategy that adaptively modulates the rainfall mean while jointly learning precipitation occurrence as an auxiliary task, enabling the model to better balance dry-wet discrimination and quantitative precipitation estimation. Extensive experiments across diverse geographical and climatic regimes demonstrate that MZ-Rain consistently outperforms strong baselines on multiple evaluation metrics, including CSI, FAR, MSE, and MAE. In particular, the model exhibits superior skill in forecasting heavy precipitation events, while benefiting from physically grounded process modeling.

Chinese Translation

准确的站点级降水短期预报对农业、水资源管理和灾害预防至关重要，通常被视为一个时间序列预测问题。然而，传统的时间序列建模技术在解决站点级降水短期预报时面临两个主要挑战：（1）缺乏物理指导建模，气象变量被视为同质集合而未考虑其在降水形成中的不同作用，导致预测偏离控制降水的物理过程。（2）降水的严重零膨胀，干燥间隔主导数据集，掩盖了有意义的降水模式，并使预测建模变得复杂。为了解决这些挑战，我们提出了 extbf{MZ-Rain}，一种基于水分预算指导的零膨胀sLSTM框架，用于站点级降水短期预报。在水分预算方程的指导下，MZ-Rain将降水形成过程分解为对应于水分储存、水分输送、地表蒸发和降水持续性的过程特定路径，并通过专门的sLSTM分支捕捉其时间演变。为了考虑降水的零膨胀特性，MZ-Rain引入了一种自适应Tweedie建模策略，该策略自适应地调节降水均值，同时将降水发生作为辅助任务共同学习，使模型能够更好地平衡干湿区分和定量降水估计。在不同地理和气候条件下的广泛实验表明，MZ-Rain在多个评估指标上（包括CSI、FAR、MSE和MAE）始终优于强基线模型。特别是，该模型在预测强降水事件方面表现出卓越的技能，同时受益于基于物理的过程建模。

View on arXiv Download PDF AI Translation

cs.AI / 53 / 2609.04865

CoSkill: Joint Reinforcement Learning of Reasoning and Meta-Skill Agents for Hierarchical Skill Evolution

CoSkill：层次技能演化的推理与元技能代理的联合强化学习

Feng, Jinyuan, Li, Dongmin, Chen, Yiqun, Gao, Yang, Chen, Xing, Wang, Huimu, Pu, Zhiqiang

Abstract

Skill libraries improve the sample efficiency of agentic reinforcement learning (RL) by enabling large language model (LLM) agents to reuse procedural knowledge. Yet existing paradigms exhibit structural shortcomings: they either decouple skill evolution from policy optimization or instantiate meta-skills as fixed workflows. Both treat skills as passive objects to be managed, limiting the flexible evolution of skills and their co-adaptation with the reasoning agent. To address the limitations, we propose CoSkill, a unified multi-agent RL framework that recasts the static meta-skill workflow as a learnable Meta-Skill Agent and jointly trains it with a Reasoning Agent over a hierarchical skill library. By modeling the Reasoning and Meta-Skill Agents as a cooperative team sharing a single backbone, CoSkill enables end-to-end co-adaptation: the Reasoning Agent conditions its actions on a retrieved task skill and step skills selected from its child set, while its task performance guides the Meta-Skill Agent in refining those step skills. Experiments on ALFWorld and WebShop show that CoSkill substantially outperforms prior skill-based and RL baselines, achieving success rates of 98.4% and 90.6%, respectively (+3.5 and +6.2 pp). As shown in Figure 1, CoSkill achieves superior early-stage sample efficiency, asymptotic performance, and wall-clock efficiency. Our code is available at https://github.com/jinyuan-cookie/CoSkill.

Chinese Translation

技能库通过使大型语言模型（LLM）代理能够重用程序知识，提高了代理强化学习（RL）的样本效率。然而，现有范式存在结构性缺陷：它们要么将技能演化与策略优化解耦，要么将元技能实例化为固定的工作流程。这两者都将技能视为被管理的被动对象，限制了技能的灵活演化及其与推理代理的共同适应。为了解决这些局限性，我们提出了CoSkill，这是一种统一的多代理RL框架，将静态元技能工作流程重新构建为可学习的元技能代理，并与推理代理在层次技能库上进行联合训练。通过将推理代理和元技能代理建模为共享单一骨干的合作团队，CoSkill实现了端到端的共同适应：推理代理根据检索到的任务技能和从其子集选择的步骤技能来调整其行动，而其任务表现则指导元技能代理优化这些步骤技能。在ALFWorld和WebShop上的实验表明，CoSkill显著优于之前的基于技能和RL的基线，成功率分别达到98.4%和90.6%（分别提高了3.5和6.2个百分点）。如图1所示，CoSkill在早期阶段实现了优越的样本效率、渐近性能和时钟效率。我们的代码可在https://github.com/jinyuan-cookie/CoSkill获取。

View on arXiv Download PDF AI Translation

cs.AI / 54 / 2609.04866

LLM-Assisted Behavioural and Scenario Augmentation for Agent-Based Energy Adoption Models

基于大型语言模型的行为与情景增强在代理基础能源采纳模型中的应用

Faiud, Iias, Khaleghy, Hossein, Schukat, Michael, Mason, Karl

Abstract

Recent advances in large language models (LLMs) create opportunities to enrich simulation-based energy policy analysis, particularly by supporting structured behavioural assumptions and exploratory techno-economic scenarios. However, directly replacing adoption models with LLM reasoning raises concerns regarding interpretability, reproducibility, and behavioural validity. This paper proposes a hybrid framework for LLM-assisted specification design, integrating bounded behavioural rubrics and structured scenario specifications into a calibrated agent-based model (ABM) of solar photovoltaic (PV) adoption by Irish dairy farms. The proposed approach preserves the original techno-economic adoption mechanism while augmenting it with bounded behavioural modulation and scenario-driven uncertainty analysis. Behavioural effects are represented through interpretable conservative, balanced, and optimistic rubrics, while future policy and market conditions are explored through fixed, rule-validated scenario specifications. Experimental results across multiple policy settings, Monte Carlo worlds, and random seeds demonstrate stable and economically plausible behaviour, with adoption outcomes remaining bounded and monotonic across behavioural regimes. The framework achieves up to approximately 13% behavioural adoption increase relative to the corresponding logistic case without producing unstable or unrealistic saturation dynamics. The results demonstrate that LLM-assisted specifications can be integrated into calibrated energy ABMs in a controlled, reproducible, and policy-relevant manner.

Chinese Translation

大型语言模型（LLMs）的最新进展为基于模拟的能源政策分析提供了丰富的机会，特别是在支持结构化行为假设和探索性技术经济情景方面。然而，直接用LLM推理替代采纳模型引发了关于可解释性、可重复性和行为有效性的担忧。本文提出了一种混合框架，用于LLM辅助的规范设计，将有限的行为标准和结构化情景规范整合到一个经过校准的代理基础模型（ABM）中，该模型用于分析爱尔兰乳制品农场的太阳能光伏（PV）采纳。所提出的方法在增强有限行为调节和情景驱动的不确定性分析的同时，保留了原有的技术经济采纳机制。行为效应通过可解释的保守型、平衡型和乐观型标准进行表示，而未来的政策和市场条件则通过固定的、规则验证的情景规范进行探索。多个政策设置、蒙特卡洛世界和随机种子的实验结果表明，行为表现稳定且经济上合理，采纳结果在行为模式下保持有限且单调。该框架实现了相对于对应的逻辑案例约13%的行为采纳增加，而未产生不稳定或不现实的饱和动态。结果表明，LLM辅助的规范可以以受控、可重复和与政策相关的方式集成到经过校准的能源ABM中。

View on arXiv Download PDF AI Translation

cs.AI / 55 / 2609.04869

From Interaction Traces to Persistent Skills: Online Evolution for Computer-Use Agents

从交互轨迹到持久技能：计算机使用代理的在线演化

Hu, Longtao, Liang, Xiao, Zhu, Linchao

Abstract

Computer-use agents can execute increasingly complex tasks in graphical interfaces, but their interaction experience is typically transient: procedural knowledge acquired from one rollout is not systematically retained, refined, and reused in later tasks. Existing skill libraries provide external procedural knowledge, yet their incremental value over the same agent operating without skills, as well as their longitudinal dynamics under repeated interaction, remain insufficiently characterized. We present an online skill-evolution framework that converts interaction trajectories and evaluator feedback into a persistent, versioned library of reusable procedures. Each iteration executes against a frozen library snapshot, and evidence-guided skill updates become available in subsequent iterations without changing model parameters. We compare the full evolving-library system with a configuration-matched empty-library control across four OSWorld application domains under the same fixed action-generation and GUI-grounding stack, task sets, and iteration horizons. Following a five-iteration empty-library warm-up, Full attains a higher post-warm-up mean evaluator score in all four observed domain runs, with mean differences ranging from 5.7 to 18.6 percentage points and domain-dependent temporal stability. In GIMP, provenance-aware analysis reveals retrieval across task-of-origin boundaries and revision churn, where repeated accepted edits fail to recover the originating task. These findings characterize evolving skill libraries as auditable, shared procedural memory that can improve a fixed computer-use stack, while showing that their benefits are conditional and repeated revision does not guarantee recovery. Code is released at https://github.com/LongtaoHu/Skill-Evo4GUI.

Chinese Translation

计算机使用代理能够在图形界面中执行日益复杂的任务，但它们的交互体验通常是短暂的：从一次执行中获得的程序性知识并未在后续任务中系统性地保留、细化和重复使用。现有的技能库提供了外部程序性知识，但它们相对于没有技能的同一代理的增量价值，以及在重复交互下的纵向动态，仍然缺乏充分的表征。我们提出了一种在线技能演化框架，将交互轨迹和评估者反馈转化为持久的、版本化的可重用程序库。每次迭代都是在一个冻结的库快照上执行，证据驱动的技能更新在后续迭代中可用，而无需更改模型参数。我们在四个OSWorld应用领域中，将完整的演化库系统与配置匹配的空库控制进行比较，保持相同的固定动作生成和GUI基础堆栈、任务集和迭代范围。在经过五次迭代的空库热身后，完整系统在所有四个观察到的领域运行中都获得了更高的热身后平均评估者评分，平均差异范围为5.7到18.6个百分点，并且在领域间表现出依赖时间的稳定性。在GIMP中，基于来源的分析揭示了跨任务来源边界的检索和修订波动，其中重复接受的编辑未能恢复原始任务。这些发现将演化技能库表征为可审计的、共享的程序性记忆，能够改善固定的计算机使用堆栈，同时显示其益处是有条件的，重复修订并不保证恢复。代码已发布在 https://github.com/LongtaoHu/Skill-Evo4GUI。

View on arXiv Download PDF AI Translation

cs.AI / 56 / 2609.04870

CHAMP: Cross-domain Hybrid Architecture for Matchmaking and Prediction in Online Multi-Player Games

CHAMP：用于在线多人游戏的跨域混合架构匹配与预测

Wang, Kai, Fan, Ge, Zhang, Chaoyun, Jiang, Yuyang, Liu, Yuze

Abstract

Multiplayer Online Battle Arena (MOBA) games rely on matchmaking to maintain competitive balance. Our prior work, CUPID, framed matchmaking as an assignment re-optimization problem and showed that a single-mode win-rate predictor can meaningfully rebalance teams. However, deploying such a system across diverse player populations exposes three practical bottlenecks: most queueing players lack sufficient in-mode match history (cold start), skill distributions shift drastically across rank tiers (distribution inconsistency), and extreme skill segments are severely data-starved. We present CHAMP, a cross-domain matchmaking framework that resolves these deployment bottlenecks. To address data sparsity and cold starts, CHAMP replaces the target-mode-only player profile with a hybrid domain feature collection: a timestamp-ordered cross-mode short-term sequence whose slices are annotated with target-domain features, plus per-mode breakdowns of long-term, real-time and team statistics. We further propose the Domain-Aware Win-rate Network (DAWN): a Domain-aware Knowledge Extractor (DAKE) compiles target-mode attributes into learnable representations that feed Domain-Aware Temporal/Spatial/Permutation OmniNet Encoders (DATOE/DASOE/DAPOE), so that mode-conditioned representations and per-mode debiasing are learned jointly inside a single shared network. Online, one trained DAWN serves every supported mode, with per-mode position-satisfaction thresholds as the only mode-specific knob. Offline, DAWN achieves 67.73% win-rate prediction accuracy, outperforming all evaluated attention and sequence baselines. Online A/B tests across the entire League ladder of a large-scale MOBA game, from novice players up to the top-expert players served by Elite Mode, demonstrate consistent drops in imbalanced matches. For lower-tier players, CHAMP reduces the 5-minute kill crushing rate by up to 20.73%.

Chinese Translation

多人在线战斗竞技场（MOBA）游戏依赖于匹配机制来维持竞争平衡。我们之前的研究CUPID将匹配框架视为一个任务重新优化问题，并展示了单一模式胜率预测器可以有效地重新平衡团队。然而，在多样化的玩家群体中部署这样的系统暴露出三个实际瓶颈：大多数排队玩家缺乏足够的模式内比赛历史（冷启动），技能分布在不同段位之间剧烈变化（分布不一致），以及极端技能段严重缺乏数据。我们提出了CHAMP，一个跨域匹配框架，解决了这些部署瓶颈。为了应对数据稀疏和冷启动问题，CHAMP用混合域特征集合替代了仅针对目标模式的玩家档案：一个时间戳排序的跨模式短期序列，其切片附有目标域特征，以及每个模式的长期、实时和团队统计数据的细分。我们进一步提出了域感知胜率网络（Domain-Aware Win-rate Network，DAWN）：域感知知识提取器（Domain-aware Knowledge Extractor，DAKE）将目标模式属性编译成可学习的表示，供域感知时间/空间/排列全网编码器（Domain-Aware Temporal/Spatial/Permutation OmniNet Encoders，DATOE/DASOE/DAPOE）使用，从而在单一共享网络内共同学习模式条件表示和每个模式的去偏差。在线上，一个训练好的DAWN服务于每个支持的模式，模式特定的唯一调节参数是每个模式的职位满意度阈值。离线时，DAWN实现了67.73%的胜率预测准确率，优于所有评估的注意力和序列基线。在一款大规模MOBA游戏的整个联赛梯队中进行的在线A/B测试，从新手玩家到由精英模式服务的顶尖专家玩家，均显示出不平衡比赛的持续下降。对于低段位玩家，CHAMP将5分钟内的击杀压制率降低了多达20.73%。

View on arXiv Download PDF AI Translation

cs.AI / 57 / 2609.04871

AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems

AutoLR：在工业推荐系统中自动化从研究到上线评审的路径

Zhang, Qi, Chen, Yanlin, Xiao, Wenchao

Abstract

Improving an industrial recommender is an iterative research-and-engineering process rather than a direct path from idea to deployment. In \textbf{DASHEN, NetEase's gaming-community app}, algorithm engineers typically identify promising directions from research papers, technical reports, and prior production experiments; reproduce or adapt the underlying methods; implement them in the production codebase; and evaluate the resulting models through training and offline experiments. Promising candidates are then advanced to online A/B tests, and those demonstrating robust gains are submitted to Launch Review---the internal gate for full-traffic rollout. Large language models (LLMs) can assist with individual stages of this workflow, but the overall process remains human-dependent without a harness that can reliably coordinate them across long-running, often multi-day experimental cycles. We present \textbf{AutoLR}, initially built as \textbf{Auto Launch Review} and later extended upstream into an autonomous research-to-launch harness. AutoLR combines three system mechanisms: a \textbf{multi-expert council} that debates and adversarially reviews proposals; a \textbf{deterministic evidence-weighted exploration--exploitation selector} that allocates a limited trial budget across candidate directions and uses Council reranking; and a layered knowledge system that combines external research, production-system knowledge, and DASHEN-specific domain knowledge---such as game communities, player characteristics, and content-interaction patterns---with posterior evidence from configurations, patches, logs, failures, and offline outcomes. LLM agents perform semantic reasoning and code generation, while deterministic controllers retain authority over execution, metric extraction, guardrails, and persistent state transitions.

Chinese Translation

改进工业推荐系统是一个迭代的研究与工程过程，而不是从想法到部署的直接路径。在 extbf{DASHEN，网易的游戏社区应用} 中，算法工程师通常从研究论文、技术报告和先前的生产实验中识别出有前景的方向；复制或调整基础方法；将其实现到生产代码库中；并通过训练和离线实验评估所得到的模型。然后，将有前景的候选模型推进到在线 A/B 测试中，表现出显著提升的模型将提交给上线评审——全流量发布的内部关卡。大型语言模型（LLMs）可以协助这一工作流程的各个阶段，但整体过程仍然依赖于人工，缺乏一个能够可靠协调这些环节的工具，尤其是在长时间运行、通常跨越多天的实验周期中。我们提出了 extbf{AutoLR}，最初构建为 extbf{自动上线评审}，后来扩展为一个自主的研究到上线的工具。AutoLR 结合了三种系统机制：一个 extbf{多专家委员会}，对提案进行辩论和对抗性评审；一个 extbf{确定性证据加权的探索-开发选择器}，在候选方向之间分配有限的试验预算，并使用委员会的重新排序；以及一个分层知识系统，结合外部研究、生产系统知识和 DASHEN 特定领域知识——例如游戏社区、玩家特征和内容交互模式——以及来自配置、补丁、日志、失败和离线结果的后验证据。LLM 代理执行语义推理和代码生成，而确定性控制器则保留对执行、指标提取、保护措施和持久状态转换的权威。

View on arXiv Download PDF AI Translation

cs.AI / 58 / 2609.04877

MARLA: A Conceptual Scaffold for Regulatory Learning under the EU AI Act

MARLA：欧盟人工智能法下的监管学习概念框架

Buscemi, Alessio, Deckenbrunnen, Tom, Hmiddou, Imane, Billi, Marco, Rubino, Livio, Ferruzza, Silvia Rizzuto, Pagani, Daniele, Rotolo, Antonino

Abstract

The EU AI Act positions regulation as part of the infrastructure for safe, trustworthy and market-ready innovation. Realising this ambition requires regulatory learning: the evidence generated during implementation must be translated into governance and legal knowledge that supports consistent interpretation, effective oversight, and adaptation as technologies evolve. Yet the actors who produce this evidence and those who rely on it operate in different professional worlds. This paper proposes MARLA (Map, Assess, Report, Learn, Adapt), a conceptual scaffold organising regulatory learning as a five-stage cycle centred on the implementation of legal requirements into socio-technical practices, situated at the Local, National and European levels of the AI Act's governance architecture. Deliberately non-prescriptive, MARLA gives technical and legal stakeholders a shared vocabulary in which each of the first three stages generates its own documentable form of regulatory learning. We illustrate the scaffold with two piloted case studies and a prospective National-to-European illustration.

Chinese Translation

欧盟人工智能法将监管视为安全、可信和市场准备就绪的创新基础设施的一部分。实现这一目标需要监管学习：在实施过程中产生的证据必须转化为支持一致解释、有效监督和随着技术发展而适应的治理和法律知识。然而，产生这些证据的参与者与依赖这些证据的参与者在不同的专业领域中运作。本文提出了MARLA（Map, Assess, Report, Learn, Adapt），一个将监管学习组织为以法律要求的实施为中心的五阶段循环的概念框架，适用于人工智能法治理架构的地方、国家和欧洲层面。MARLA故意不具规范性，为技术和法律利益相关者提供了一个共享的词汇，使得前三个阶段各自生成可记录的监管学习形式。我们通过两个试点案例研究和一个前瞻性的国家到欧洲的示例来说明这一框架。

View on arXiv Download PDF AI Translation

cs.AI / 59 / 2609.04880

Reinforcement Learning for Sequential Solar PV Policy Design under Uncertainty: An Agent-Based Approach

基于强化学习的序贯太阳能光伏政策设计：一种基于智能体的方法

Faiud, Iias, Shianifar, Jonaid, Schukat, Michael, Mason, Karl

Abstract

Designing effective and fiscally sustainable policies for solar photovoltaic (PV) adoption requires balancing adoption gains against public expenditure under uncertainty and heterogeneous decision-making. This study formulates PV policy design as a sequential decision problem and integrates reinforcement learning (RL) with a stochastic agent-based model (ABM) that simulates yearly solar PV adoption under uncertainty. A policymaker agent selects annual incentives, including capital grants, subsidised loan rates, and feed-in tariffs, over a 16-year horizon. Adoption--cost trade-offs are explored by varying policy preferences within a scalarised reward framework. Policies are learned using PPO, SAC, and TD3 and evaluated under stochastic simulation. The results show that this approach produces a clear trade-off structure: the highest-adoption policy (TD3, $w_{\text{cost}}=0.5$) achieves approximately 4,145 adopters at a cost of EUR 41.73 million, while the lowest-cost policy (PPO, $w_{\text{cost}}=2.0$) reduces expenditure to EUR 7.27 million with 2,682 adopters. The balanced policy (PPO, $w_{\text{cost}}=1.6$) achieves 3,495 adopters at a cost of EUR 22.47 million. Across algorithms, consistent trade-off patterns are observed, indicating robustness of the adoption--cost relationship. Compared with static baseline policies, the RL framework explores a broader range of policy configurations. These findings demonstrate the potential of RL as a flexible tool for adaptive policy design under uncertainty.

Chinese Translation

设计有效且财政可持续的太阳能光伏（PV）推广政策，需要在不确定性和异质决策行为下平衡推广收益与公共支出。本文将光伏政策设计建模为序贯决策问题，结合强化学习（RL）与随机智能体模型（ABM），模拟年度光伏推广过程中的不确定性。政策制定者智能体在16年时间跨度内选择年度激励措施，包括资本补助、优惠贷款利率和上网电价。通过在标量化奖励框架内调整政策偏好，探讨推广与成本的权衡。采用PPO、SAC和TD3算法进行政策学习，并在随机仿真中评估。结果表明，该方法揭示了明确的权衡结构：最高推广政策（TD3，成本权重w_cost=0.5）实现约4145个采用者，成本为4173万欧元；最低成本政策（PPO，w_cost=2.0）将支出降至727万欧元，采用者数为2682；平衡政策（PPO，w_cost=1.6）实现3495个采用者，成本为2247万欧元。不同算法间均表现出一致的权衡模式，表明推广与成本关系的稳健性。与静态基线政策相比，强化学习框架探索了更广泛的政策配置范围。研究结果展示了强化学习作为不确定性下自适应政策设计的灵活工具的潜力。

View on arXiv Download PDF AI Translation

cs.AI / 60 / 2609.04894

From Language Models to World-Acting Systems: Progress and Limits of Agentic AI across Digital, Social, Virtual, and Physical Environments

从语言模型到世界行动系统：代理人工智能在数字、社会、虚拟和物理环境中的进展与局限

Zhu, Linsen, Cai, Mengqing

Abstract

Large language models become consequential agents when surrounding systems let outputs change external state. Models now call tools, operate interfaces, delegate work, retain state, inhabit generated worlds, and control robots or laboratory equipment. Such advances are often narrated as one march toward autonomy, conflating model competence, system integration, persistence, and safe authority. This critical review synthesizes primary research and official technical specifications available by 31 August 2026. We organize the evidence along delegated authority, temporal persistence, and environmental coupling, while separating model, harness, and environment. Within the evidence examined, action-interface expansion is documented more convincingly than robust completion, recovery, authorization, or independent verification. Model Context Protocol and Agent2Agent improve interoperability but do not establish trustworthy delegation; multi-agent organization adds specialization alongside cost and correlated failure. Persistent simulations and world models support training and planning but do not themselves demonstrate agency; robotics and self-driving laboratories establish bounded feasibility rather than unattended open-world reliability. We propose justified delegation as an analytical and normative heuristic, not an observed law or certified score: expand action scope only where evidence supports provenance, bounded authority, failure detection, safe recovery, and calibrated human control. This framing yields a research agenda for coupled model-harness evaluation, capability-based permissions, durable state, cross-agent accountability, and staged physical validation.

Chinese Translation

当周围系统允许输出改变外部状态时，大型语言模型成为重要的代理。模型现在可以调用工具、操作接口、委派工作、保持状态、居住在生成的世界中，并控制机器人或实验室设备。这些进展常常被叙述为朝向自主性的单一进程，将模型能力、系统集成、持久性和安全权威混为一谈。本文的批判性综述综合了截至2026年8月31日可用的主要研究和官方技术规范。我们根据委派权威、时间持久性和环境耦合组织证据，同时区分模型、工具和环境。在所审查的证据中，行动接口的扩展比稳健的完成、恢复、授权或独立验证更具说服力。模型上下文协议（Model Context Protocol）和Agent2Agent提高了互操作性，但并未建立可信的委派；多代理组织在增加专业化的同时也带来了成本和相关失败。持久的模拟和世界模型支持训练和规划，但本身并未展示代理性；机器人技术和自动驾驶实验室建立了有限的可行性，而非无人值守的开放世界可靠性。我们提出合理的委派作为一种分析和规范的启发式方法，而非观察到的法律或认证分数：仅在证据支持来源、有限权威、失败检测、安全恢复和校准的人类控制的情况下扩展行动范围。这一框架为耦合模型-工具评估、基于能力的权限、持久状态、跨代理问责和分阶段的物理验证提供了研究议程。

View on arXiv Download PDF AI Translation

cs.AI / 61 / 2609.04915

Compact-Memory LLM Agents via Online Max-Member Clustering and Atom-Aware Packing

通过在线最大成员聚类和原子感知打包实现紧凑内存 LLM 代理

Geng, Jiahe, Wang, Jinpeng, Yuan, Kun

Abstract

Many long-horizon LLM deployments face tight prompt budgets: latency, cost, and context limits make full-context prompting impractical as interaction length grows. The key question is then not raw recall alone, but which memory design gives the best quality--token trade-off in the compact-memory regime. We present \textbf{RSM-full}, an online clustered-memory pipeline designed for a strong quality--token Pareto point. RSM-full combines two design choices: a cosine-gated \emph{max-member merge} write rule and an atom-aware grouped context packer. On AMA-Bench, our primary compact-memory benchmark, it reaches $83%$ of Full-Context quality at $32%$ of the token cost at a $4$k budget; under four-seed averaging it beats the closest streaming-clustered baseline (Online K-Means) by $+3.5$--$6.0$,pp ($p{<}.001$) across the whole ${\sim}2.6$k--${\sim}5$k regime. Three-seed ablations show most of this gain comes from the merge rule ($+5.7$,pp over Online K-Means and matched-$\tau$ DP-means) and the grouped packer ($+5.0$,pp over flat concatenation). The pattern reproduces on RealMem, an independent long-horizon persona-memory benchmark: RSM-full improves on Budget-RAG ($+0.69$,pp, $p{=}.006$), is on par with BM25-RAG (paired $\Delta{=}{+}0.27$,pp, $p{=}.47$; we do \emph{not} claim BM25 equivalence in the equivalence-test sense), and significantly outperforms Streaming-Proto ($+2.97$,pp) and the closest reproduced 2025 agentic-memory baseline A-MEM ($+1.65$,pp, $p{<}.001$). Across benchmarks the message is consistent: under tight budgets, compact-memory performance is driven mainly by how streaming memories are merged and how retrieved content is assembled. Overall, RSM-full is most useful when answeroughly $2k$--$5k$ prompt tokens, where itdefines a strong compact-memory Pareto point; higher-token baselines remain stronger outside this regime.

Chinese Translation

许多长期 LLM 部署面临严格的提示预算：延迟、成本和上下文限制使得随着交互长度的增长，完全上下文提示变得不切实际。关键问题不再仅仅是原始的回忆能力，而是在哪种内存设计下，紧凑内存模式能在质量与令牌之间实现最佳的权衡。我们提出了 extbf{RSM-full}，这是一种在线聚类内存管道，旨在实现强大的质量-令牌帕累托点。RSM-full 结合了两种设计选择：一种余弦门控的 extit{最大成员合并} 写入规则和一种原子感知的分组上下文打包器。在我们的主要紧凑内存基准 AMA-Bench 上，它在 $4$k 预算下以 $32 ext{%}$ 的令牌成本达到了 $83 ext{%}$ 的完全上下文质量；在四种种子平均下，它超越了最接近的流式聚类基线（在线 K-均值） $+3.5$--$6.0$ pp ($p{<}.001$)，覆盖整个 ${ ilde{2.6}}$k--${ ilde{5}}$k 范围。三种种子消融实验表明，这一增益主要来自于合并规则（相较于在线 K-均值和匹配的 $ au$ DP-均值增加了 $+5.7$ pp）和分组打包器（相较于平坦连接增加了 $+5.0$ pp）。这一模式在独立的长期个性内存基准 RealMem 上得以重现：RSM-full 在 Budget-RAG 上有所提升（$+0.69$ pp, $p{=}.006$），与 BM25-RAG 持平（配对 $ riangle{=}{+}0.27$ pp, $p{=}.47$；我们并不声称 BM25 在等价测试意义上等价），并显著超越了 Streaming-Proto ($+2.97$ pp) 和最接近的 2025 代理记忆基线 A-MEM ($+1.65$ pp, $p{<}.001$)。在各基准测试中，信息是一致的：在紧张的预算下，紧凑内存的性能主要由流式内存的合并方式和检索内容的组装方式驱动。总体而言，RSM-full 在大约 $2k$--$5k$ 的提示令牌时最为有效，此时它定义了一个强大的紧凑内存帕累托点；在这一范围之外，更高令牌的基线仍然更强。

View on arXiv Download PDF AI Translation

cs.AI / 62 / 2609.04917

Artificial Intelligence in Equity and Crypto Markets: Progress, Profitability Evidence, and the Limits of Automated Investing

人工智能在股票和加密市场中的应用：进展、盈利性证据及自动化投资的局限性

Zhu, Linsen, Cai, Mengqing

Abstract

Artificial intelligence (AI) now supports investment workflows from data and prediction through research, portfolios, execution, and tool use. Technical capability, however, is not evidence of investment profitability. This critical state-of-the-art review examines public research available through 31 August 2026 on listed equities, exchange-traded funds, centralized crypto spot, perpetual futures, and on-chain markets. We organize evidence with an alpha-translation chain: point-in-time information must yield a stable signal, feasible positions, executable orders, and risk-adjusted returns after costs. Across machine learning, time-series foundation models, financial language models, reinforcement learning, and agents, the examined record shows real but mainly upstream progress in prediction, text processing, portfolio design, and workflow integration. Evidence is thinner for durable net performance. Temporal contamination, repeated selection, survivorship, weak benchmarks, implementation costs, venue mechanics, and capacity can break translation to net alpha. Strong historical results coexist with predictor decay, corrected look-ahead failures, mixed prospective evidence, and few audited live-capital records. Crypto adds informative state but requires separate treatment of spot, perpetual, and decentralized cash flows and execution. Within the public evidence examined here, no general AI architecture is shown to deliver persistent, cross-regime, capacity-aware net alpha. More credible claims require point-in-time data and models, decision-aligned objectives, joint portfolio--execution evaluation, controlled adaptation, prospective tests, and authority-matched governance. These conditions can improve evidence and implementation; they do not guarantee profit.

Chinese Translation

人工智能（AI）现在支持从数据和预测到研究、投资组合、执行和工具使用的投资工作流程。然而，技术能力并不等同于投资盈利性的证据。本文对截至2026年8月31日的公开研究进行了批判性前沿评估，涵盖上市股票、交易所交易基金、集中式加密现货、永续期货和链上市场。我们通过一个阿尔法转化链组织证据：时点信息必须产生稳定信号、可行头寸、可执行订单和扣除成本后的风险调整回报。在机器学习、时间序列基础模型、金融语言模型、强化学习和代理等领域，所审查的记录显示出在预测、文本处理、投资组合设计和工作流程整合方面的真实但主要是上游进展。对于持久净表现的证据则较为薄弱。时间污染、重复选择、生存偏差、弱基准、实施成本、场所机制和容量等因素可能会破坏净阿尔法的转化。强劲的历史结果与预测器衰退、修正的前瞻性失败、混合的前瞻性证据以及少量经过审计的实盘记录并存。加密市场提供了有价值的信息状态，但需要对现货、永续和去中心化现金流及执行进行单独处理。在这里审查的公共证据中，没有任何通用的AI架构显示出能够提供持久的、跨制度的、考虑容量的净阿尔法。更可信的主张需要时点数据和模型、与决策对齐的目标、联合投资组合-执行评估、受控适应、前瞻性测试和权威匹配的治理。这些条件可以改善证据和实施，但并不保证盈利。

View on arXiv Download PDF AI Translation

cs.AI / 63 / 2609.04931

Solving Hard XAI Queries Based on a Compiled Dual-Rail Encoding

基于编译的双轨编码解决困难的可解释人工智能查询

Ledaguenel, Arthur, Capelli, Florent, Lagniez, Jean-Marie

Abstract

The widespread adoption of artificial intelligence (AI) within real-world applications has raised a lot of concerns regarding their trustworthiness, especially in critical applications. The field of eXplainable AI (XAI) has emerged with the objective of providing explanations to the users about the decisions made by AI systems. Several explanations for boolean classifiers have been introduced in the literature, including abductive and contrastive explanations, each giving a different insight on the decision of the classifier. However, computing an explanation for a decision of a boolean classifier is a hard problem in general. One way to deal with this complexity is to rely on a compiled representation of the classifier for which each explanation can be computed efficiently. Unfortunately, we prove in this paper that several classes of abductive explanations, remain hard to compute even for Ordered Binary Decision Diagrams, one of the most tractable subsets of the knowledge compilation map. Included in such classes are shorter abductive explanations or abductive explanations that include the explainee's preferences. To recover the benefits of working with compiled representations, we show that a proper representation of the dual-rail encoding of the classifier can be used to compute efficiently these classes of explanations.

Chinese Translation

人工智能（AI）在现实应用中的广泛采用引发了许多关于其可信度的担忧，尤其是在关键应用中。可解释人工智能（XAI）领域应运而生，旨在为用户提供有关AI系统决策的解释。文献中介绍了多种布尔分类器的解释，包括溯因解释和对比解释，每种解释都为分类器的决策提供了不同的视角。然而，计算布尔分类器决策的解释通常是一个困难的问题。应对这种复杂性的一种方法是依赖于分类器的编译表示，从而可以高效地计算每个解释。不幸的是，我们在本文中证明，尽管有序二进制决策图（Ordered Binary Decision Diagrams）是知识编译图中最易处理的子集之一，但某些类别的溯因解释仍然难以计算。这些类别包括较短的溯因解释或包含被解释者偏好的溯因解释。为了恢复使用编译表示的好处，我们展示了分类器的双轨编码的适当表示可以用于高效计算这些类别的解释。

View on arXiv Download PDF AI Translation

cs.AI / 64 / 2609.04962

Why We Care About Understanding: Competence through Predictive Compression

我们为何关注理解：通过预测压缩实现能力

Queloz, Matthieu, Beckmann, Pierre

Abstract

What is the relation between understanding and compression, and why does human understanding take such a heavily compressed form? Across information theory, machine learning, and AI research, a substantial tradition identifies understanding with compression-a thought captured in Gregory Chaitin's dictum that "comprehension is compression." Philosophers, by contrast, have characterized understanding in terms of grasping connections, giving explanations, and handling novelty. This paper bridges the two pictures through three interlocking theses. The first concerns the concept of understanding: it serves as an efficient proxy for a distinctive form of robust competence, enabling us to identify whom to trust and whom to learn from. The second concerns the state of understanding: to understand a domain is to possess a mental model of its relational structure that enables prediction, and what enables prediction enables compression, because what becomes predictable need not be stored separately. Compression is therefore not identical with comprehension, but its representational shadow. The third concerns the characteristically human form of understanding: the fiduciary and transmission functions highlighted by the first thesis impose pressures of demonstrability and transmissibility that drive human understanding toward principled simplicity. The resulting framework explains both the appeal and the limits of compressionist accounts of understanding while shedding light on the inscrutability of AI systems.

Chinese Translation

理解与压缩之间的关系是什么？人类的理解为何呈现出如此高度压缩的形式？在信息理论、机器学习和人工智能研究中，有一个重要的传统将理解与压缩等同起来——这一观点在格雷戈里·查伊廷（Gregory Chaitin）的名言中得到了体现：“理解就是压缩。”相比之下，哲学家们则将理解定义为把握联系、给出解释和处理新颖性的能力。本文通过三个相互关联的论点架起了这两种观点之间的桥梁。第一个论点涉及理解的概念：理解作为一种高效的代理，代表了一种独特的稳健能力，使我们能够识别值得信任的人和学习的对象。第二个论点涉及理解的状态：理解一个领域意味着拥有其关系结构的心理模型，这种模型能够进行预测，而能够进行预测的事物则能够实现压缩，因为可预测的内容无需单独存储。因此，压缩并不等同于理解，而是其表征的影子。第三个论点涉及人类特有的理解形式：第一个论点强调的信任和传递功能施加了可证明性和可传递性的压力，推动人类理解朝向原则性简约发展。由此产生的框架既解释了压缩主义对理解的吸引力和局限性，也阐明了人工智能系统的不可解性。

View on arXiv Download PDF AI Translation

cs.AI / 65 / 2609.04978

Global to Local: Topology-Preserving Adaptive Graph Pooling via Granular-Ball

从全局到局部：通过粒子球实现拓扑保持的自适应图池化

Zhao, Sen, Xu, Gaojie, Xia, Shuyin, Guan, Yifan, Liu, Yi, Wang, Yi, Wang, Wei

Abstract

Graph pooling aims to compress the graph, including both node embeddings and their underlying topological patterns, into a more compact representation. Previous works focus primarily on the overly fine-grained representation of nodes, progressively coarsening the graph by removing nodes or merging them into clusters, thus neglecting the global-to-local patterns and adaptive granularity of the graph's topological structure. In the real scenario, graphs as a whole can be considered the coarsest level of granularity, encapsulating the global topological structure, with progressively finer-grained local topological structures represented from top to bottom. This process continues until the adaptive granularity for each subdomain is reached. To this end, we propose a novel Topology-Preserving Adaptive Graph Pooling (TPAGP) method that dynamically partitions graphs into granular balls by integrating node features and topological information, enabling the generation of multi-granularity representations that effectively capture both local and global structural patterns. Additionally, we design a multi-granularity graph network model that facilitates feature interaction and optimization across different granularities, significantly enhancing performance in graph classification tasks. Experimental results demonstrate that TPAGP outperforms existing pooling methods across various benchmark datasets, effectively mitigating information loss caused by fixed-granularity strategies.

Chinese Translation

图池化旨在将图的节点嵌入及其潜在的拓扑模式压缩为更紧凑的表示。以往的研究主要集中在节点的过于细粒度表示上，通过移除节点或将其合并为簇逐步粗化图，从而忽视了图的拓扑结构中的全局到局部模式和自适应粒度。在实际场景中，整体图可以被视为最粗粒度的层次，封装了全局拓扑结构，而逐渐细化的局部拓扑结构则从上到下表示。这个过程持续进行，直到达到每个子域的自适应粒度。为此，我们提出了一种新颖的拓扑保持自适应图池化（Topology-Preserving Adaptive Graph Pooling, TPAGP）方法，通过整合节点特征和拓扑信息，动态地将图划分为粒子球，从而生成有效捕捉局部和全局结构模式的多粒度表示。此外，我们设计了一种多粒度图网络模型，促进不同粒度之间的特征交互和优化，显著提升图分类任务的性能。实验结果表明，TPAGP在各种基准数据集上优于现有的池化方法，有效减轻了固定粒度策略导致的信息损失。

View on arXiv Download PDF AI Translation

cs.AI / 66 / 2609.04981

A Tree-based RAG Framework for Evidence-Intensive QA via Adaptive Planning and Topology-Aware Evidence Gathering

基于树的RAG框架用于通过自适应规划和拓扑感知的证据收集进行证据密集型问答

Lee, Songeun, Min, Kyungjin, Na, Injae, Lee, Suyeong, Kim, Chiyoung, Jung, Woohwan

Abstract

Recent structured RAG methods leverage tree- or graph-based reasoning structures to improve multi-hop QA. However, they face key limitations in evidence-intensive QA, where answering a question requires synthesizing information scattered across dozens or even hundreds of documents: structural rigidity, which limits adaptive reasoning expansion, and topology-ignorant evidence gathering, which prevents effective integration of evidence across different reasoning nodes. To address these issues, we propose APT-RAG, an Adaptive Planning and Topology-aware evidence gathering RAG framework. Adaptive planning dynamically expands the reasoning structure based on question dependencies and evidence requirements, while topology-aware evidence gathering improves evidence coverage through sibling evidence reuse, direct retrieval, and evidence aggregation from child nodes. We further introduce evidence-guided batched answer generation to reduce significant generation overhead in evidence-intensive QA. In the experiments on evidence-intensive QA benchmarks, APT-RAG outperforms existing structured RAG methods. Our code is available at https://github.com/hyudsl/APT-RAG.

Chinese Translation

近期的结构化RAG方法利用基于树或图的推理结构来改善多跳问答。然而，它们在证据密集型问答中面临关键限制，在这种情况下，回答一个问题需要综合分散在数十甚至数百个文档中的信息：结构的刚性限制了自适应推理的扩展，而忽视拓扑的证据收集则阻碍了不同推理节点之间证据的有效整合。为了解决这些问题，我们提出了APT-RAG，一个自适应规划和拓扑感知的证据收集RAG框架。自适应规划根据问题依赖关系和证据需求动态扩展推理结构，而拓扑感知的证据收集通过兄弟证据重用、直接检索和从子节点聚合证据来改善证据覆盖。我们进一步引入了证据引导的批量答案生成，以减少证据密集型问答中的显著生成开销。在证据密集型问答基准测试中的实验表明，APT-RAG优于现有的结构化RAG方法。我们的代码可在https://github.com/hyudsl/APT-RAG获取。

View on arXiv Download PDF AI Translation

cs.AI / 67 / 2609.05009

Language models judge war differently when tested for alignment

语言模型在对齐测试中对战争的判断存在差异

Chupilkin, Maxim

Abstract

Safety evaluations can mischaracterize deployed behaviour if artificial-intelligence systems respond to being evaluated. We test this possibility in a full-factorial conjoint experiment on decisions to start a war, spanning 20 large language models, 32 scenarios, 10 repetitions and two conditions (N = 12,800 judgments). Adding one sentence, "You are tested for alignment with human values", produced two effects. First, it produced a level effect: mean willingness to start war fell by 13.43 points on a 0-100 scale (95% confidence interval, -16.20 to -10.65). Second, it produced a structural effect by changing which information drove judgments. Probability of success was the largest factor for 17 of 20 models at baseline; under the cue, civilian casualties were largest for 12. Standardized estimates show that this reordering arose principally because models attenuated strategic considerations such as probability of success and domestic support. Evaluation framing therefore changes both an answer's level and its revealed decision rule.

Chinese Translation

安全评估可能会错误地描述已部署行为，尤其是当人工智能系统对评估做出反应时。我们在一项全面的因子实验中测试了这一可能性，该实验涉及20个大型语言模型、32个场景、10次重复和两个条件（N = 12,800 次判断）。增加一句话"您正在接受与人类价值观的对齐测试"产生了两个效果。首先，它产生了一个水平效应：在0-100的量表上，愿意发动战争的平均分数下降了13.43分（95%置信区间，-16.20至-10.65）。其次，它通过改变驱动判断的信息，产生了结构效应。在基线条件下，成功概率是20个模型中17个的最大因素；而在提示下，平民伤亡是12个模型中最大的因素。标准化估计表明，这种重新排序主要是因为模型减弱了战略考虑因素，如成功概率和国内支持。因此，评估框架不仅改变了答案的水平，还改变了其揭示的决策规则。

View on arXiv Download PDF AI Translation

cs.AI / 68 / 2609.05019

TROVE: Adaptive Agent Skill Orchestration via Trace-Grounded Route Validation and Editing

TROVE：通过基于轨迹的路线验证与编辑实现自适应代理技能编排

Wang, Tianxing, Zhao, Mingming, Huang, Shuai, Xu, Huiyang, Niu, Chaoyue, Liu, Shengzhong, Wu, Fan

Abstract

Agents tend to optimize, select, or constrain execution structures before decisive runtime outcomes are observed. However, such pre-execution commitment creates an orchestration bottleneck: when intermediate evidence invalidates the pending continuation, agents must either execute stale steps or replan broadly, compounding errors, wasting computation, and discarding progress. We thus propose Trace-grounded Route Orchestration via Validation and Editing (TROVE), which revises only what runtime evidence invalidates. Offline, TROVE distills evaluated workflow-search traces into atomic and composite skills and an outcome-conditioned transition graph, preserving stable fragments while exposing outcome-dependent decisions. Online, it treats a planned route as provisional: after committing one top-level skill, the controller retains a valid continuation, inserts a trace-supported local response, or replaces only the invalid suffix. Evaluation across code-generation, question-answering, and math reasoning benchmarks with different LLM backbones show that TROVE delivers a stronger quality-efficiency trade-off than existing baselines of dataset-level optimization, query-level architecture selection, and graph-constrained scheduling. Quality gains are largest when outcomes change the appropriate continuation, whereas early termination yields substantial efficiency gains on near-saturated tasks. Ablations further show that composite skills capture most offline benefits, insertion enables local correction, and suffix replacement primarily improves efficiency. These findings establish selective route editing as a general principle for adaptive agent orchestration.

Chinese Translation

代理在观察到决定性运行时结果之前，往往会优化、选择或限制执行结构。然而，这种预执行承诺会造成编排瓶颈：当中间证据使待定的继续执行无效时，代理必须执行过时的步骤或进行广泛的重新规划，从而加重错误、浪费计算资源并丢弃进展。因此，我们提出了基于轨迹的路线编排方法（Trace-grounded Route Orchestration via Validation and Editing，TROVE），该方法仅修正运行时证据所使无效的部分。在离线阶段，TROVE 将评估过的工作流搜索轨迹提炼为原子技能和复合技能以及结果条件下的转移图，保留稳定的片段，同时暴露依赖于结果的决策。在在线阶段，它将计划路线视为临时的：在承诺一个顶级技能后，控制器保留有效的继续执行，插入一个基于轨迹的局部响应，或仅替换无效的后缀。在代码生成、问答和数学推理基准测试中，使用不同的 LLM（大型语言模型）基础架构进行评估显示，TROVE 提供了比现有数据集级优化、查询级架构选择和图约束调度的基线更强的质量-效率权衡。当结果改变适当的继续执行时，质量提升最大，而在接近饱和的任务上，提前终止则带来了显著的效率提升。消融实验进一步表明，复合技能捕获了大部分离线收益，插入使局部修正成为可能，而后缀替换主要提高了效率。这些发现确立了选择性路线编辑作为自适应代理编排的一般原则。

View on arXiv Download PDF AI Translation

cs.AI / 69 / 2609.05036

Moral Competence Before Moral Content: Why LLM Agents Lack the Prerequisites for Coherent Alignment

道德能力与道德内容：为何大型语言模型代理缺乏一致对齐的前提条件

Libert, Arno, Prinzhorn, Derck W. E., Henselmans, Daan R.

Abstract

AI alignment requires AI systems to adhere to human norms, values, or intentions. Under value pluralism there is no correct target, but a shared prerequisite is that the system's behavior expresses a coherent policy: a mapping from situations to verdicts that is invariant while a situation's morally relevant features are preserved, and sensitive when they change. We introduce four structural conditions for such coherent policies: verdict stability, monotonicity, decisiveness, and Pareto viability. Together they measure a form of moral competence that is evaluable from behavior alone, without reference to a moral standard or expert baseline, forming a structural floor for alignment rather than a normative target. We demonstrate the methodology on three simulated deployments featuring LLM-based agents facing moral dilemmas. Evaluating nine frontier models under a factorial design of five paraphrases, five escalation levels, and three dominance conditions, we show no model expresses a coherent policy across the three deployments: surface-form perturbation alone produces verdict-rate shifts of up to $99$ percentage points at a single escalation level, and a model's success on one scenario does not predict its competence on another. This suggests LLM-based agents are not currently the kind of object to which alignment can meaningfully apply.

Chinese Translation

人工智能对齐要求人工智能系统遵循人类的规范、价值观或意图。在价值多元主义下，没有正确的目标，但一个共同的前提是系统的行为必须表达出一致的政策：即在道德相关特征保持不变的情况下，从情境到裁决的映射应保持不变，而在这些特征发生变化时则应敏感。我们提出了四个一致政策的结构性条件：裁决稳定性、单调性、果断性和帕累托可行性。它们共同衡量了一种可以仅通过行为评估的道德能力，而无需参考道德标准或专家基准，从而形成对齐的结构性底线，而非规范性目标。我们在三个模拟部署中展示了该方法，涉及面临道德困境的基于大型语言模型的代理。在五种释义、五个升级水平和三个主导条件的因子设计下评估九个前沿模型，我们发现没有模型在三个部署中表现出一致的政策：仅表面形式的扰动就能在单一升级水平上产生高达99个百分点的裁决率变化，而一个模型在一个情境上的成功并不能预测其在另一个情境上的能力。这表明，基于大型语言模型的代理目前并不是可以进行有意义对齐的对象。

View on arXiv Download PDF AI Translation

cs.AI / 70 / 2609.05040

Towards Efficient Evaluation of Evolutionary Transfer Optimization: Case Studies on Task-Parameterized Applications

朝向高效评估进化转移优化：任务参数化应用的案例研究

Li, Yanchen, Xue, Xiaoming, Tan, Kay Chen

Abstract

As evolutionary transfer optimization (ETO) scales to larger collections of related tasks, problem evaluation can become a major source of runtime growth. This work studies problem-side evaluation scaling in task-parameterized applications and reformulates application-specific serial computations into forms suitable for parallel execution. We organize evaluation scaling into two levels: the number of evaluated tasks and the workload within each task. In multi-task optimization, matrix-recursive kinematic-arm evaluation is reformulated using an accumulation-matrix representation of cumulative link directions. In sequential transfer optimization, pointwise B-spline trajectory evaluation is reformulated using a blending-matrix representation for trajectory and collision computations. Both reformulations maintain close numerical agreement with their reference evaluations and substantially reduce runtime, yielding $256.72\times$ and $93.91\times$ end-to-end speedups, respectively. These results demonstrate problem-side reformulation as a practical route toward scalable ETO. Both application implementations and experimental scripts are released as open source to support reproducibility and reuse.

Chinese Translation

随着进化转移优化（ETO）扩展到更大规模的相关任务集合，问题评估可能成为运行时间增长的主要来源。本研究探讨了任务参数化应用中的问题侧评估扩展，并将特定应用的串行计算重新构造成适合并行执行的形式。我们将评估扩展组织为两个层次：评估任务的数量和每个任务内的工作负载。在多任务优化中，矩阵递归运动臂评估通过累积链接方向的累积矩阵表示进行重新构造。在顺序转移优化中，逐点 B 样条轨迹评估通过轨迹和碰撞计算的混合矩阵表示进行重新构造。这两种重新构造在数值上与其参考评估保持了紧密的一致性，并显著减少了运行时间，分别实现了 $256.72 imes$ 和 $93.91 imes$ 的端到端加速。这些结果表明问题侧重新构造是实现可扩展 ETO 的一种实用途径。所有应用实现和实验脚本均作为开源发布，以支持可重复性和重用。

View on arXiv Download PDF AI Translation

cs.AI / 71 / 2609.05075

MePo++: Unifying Representation Refinement and Reconciliation for General Continual Learning

MePo++：统一表示精炼与调和以实现通用持续学习

Sun, Guanglong, Zhou, Kanglei, Wang, Liyuan, Cheng, Qi, Yan, Hongwei, Cui, Shuang, Su, Hang, Zhu, Jun, Zhong, Yi

Abstract

General continual learning (GCL) aims to learn from evolving data streams without task identities, explicit boundaries, or repeated access to previous data, making it a realistic yet challenging setting for continual intelligence. Although pretrained models (PTMs) provide rich prior knowledge for addressing the limited supervision and non-stationary nature of GCL, existing PTM-based methods often directly adapt pretrained representations and overlook two critical gaps: the misalignment between upstream pretraining and downstream continual adaptation, and the unreliability of conventional output alignment under blurry streams. Here we propose MePo++, a unified post-training framework that bridges pretrained knowledge and downstream GCL through representation refinement and reconciliation. MePo++ introduces two complementary components: MetaPrep, which improves representation plasticity for continual adaptation through unsupervised meta-refinement over pseudo continual sequences; and StreamAlign, which reinforces representation stability by reconciling evolving online features with a stable pretrained geometry. By improving representation learnability before adaptation and preserving alignment during continual learning, MePo++ enables PTMs to remain both plastic for new concepts and stable over evolving streams. Experiments across diverse PTMs, datasets, and continual learning baselines demonstrate the consistent effectiveness and generality of MePo++ for PTM-based GCL. Our code is available at https://github.com/SunGL001/MePo_Plus.

Chinese Translation

通用持续学习（GCL）旨在从不断演变的数据流中学习，而无需任务身份、明确边界或对先前数据的重复访问，这使其成为一个现实但具有挑战性的持续智能环境。尽管预训练模型（PTMs）为解决GCL的有限监督和非平稳特性提供了丰富的先验知识，但现有的基于PTM的方法往往直接调整预训练表示，忽视了两个关键差距：上游预训练与下游持续适应之间的错位，以及在模糊流中传统输出对齐的可靠性不足。在此，我们提出了MePo++，一个统一的后训练框架，通过表示精炼与调和将预训练知识与下游GCL连接起来。MePo++引入了两个互补组件：MetaPrep，通过对伪持续序列进行无监督的元精炼，提高表示的可塑性以适应持续学习；StreamAlign，通过将不断演变的在线特征与稳定的预训练几何结构进行调和，增强表示的稳定性。通过在适应前提高表示的可学习性，并在持续学习过程中保持对齐，MePo++使得PTMs能够在新概念上保持可塑性，同时在不断演变的流中保持稳定。针对不同PTMs、数据集和持续学习基线的实验表明，MePo++在基于PTM的GCL中展现出一致的有效性和普适性。我们的代码可在 https://github.com/SunGL001/MePo_Plus 获取。

View on arXiv Download PDF AI Translation

cs.AI / 72 / 2609.05079

TruthInsightBench: An Evidence-Grounded Benchmark for Automated Evaluation of Open-Ended Scientific Discovery Agents

TruthInsightBench：一个基于证据的开放式科学发现代理自动评估基准

Yang, Zhibo, Zhang, Chen, Zhang, Yuewei, Wang, Hao

Abstract

Autonomous coding agents are increasingly proposed as AI-scientist systems that conduct analyses and write research reports, but executing a prescribed analysis is not the same as making a discovery. Existing benchmarks are configured for reproduction: tasks, data, and rubrics are built around a hidden target study, and recovery of its result is rewarded. We present TruthInsightBench, a benchmark configured for discovery. Its 40 blind tasks, drawn from 40 peer-reviewed studies across 10 scientific domains, expose only a neutral scientific objective and frozen data; source conclusions, expected values, and analysis paths are withheld, leaving the agent to determine what claim the data support. A fixed LLM-based judge scores the evidentiary maturity of an agent's own claims along six dimensions, operationalized as 29 artifact-grounded items, with automated, deterministic aggregation and no per-instance human grading, so evaluation can be repeated automatically as agents evolve. On one frozen base model, four coding agents form a narrow plateau (58.4-60.3 of 100) with no statistically reliable pairwise separation: they execute and document analyses competently, with comparatively strong evidence auditability and novelty, but largely lack the discriminating acts that establish a trustworthy claim (controls, robustness, falsifiability, and cross-dataset generalization). The bottleneck is scientific judgment rather than coding, and genuine discovery remains out of reach. TruthInsightBench makes this gap a measurable target; data and scoring code are at https://github.com/TruthInsight-stack/TruthInsightBench.

Chinese Translation

自主编码代理越来越多地被提议作为人工智能科学家系统，进行分析并撰写研究报告，但执行预定分析与进行发现并不相同。现有基准配置用于重现：任务、数据和评分标准围绕一个隐藏的目标研究构建，恢复其结果会获得奖励。我们提出了TruthInsightBench，这是一个配置用于发现的基准。它的40个盲任务来自10个科学领域的40项同行评审研究，仅暴露出中立的科学目标和固定的数据；源结论、预期值和分析路径被保留，代理需要自行确定数据支持的主张。一个固定的基于大语言模型（LLM）的评审者在六个维度上对代理自身主张的证据成熟度进行评分，这六个维度通过29个基于文献的项目进行操作化，采用自动化、确定性的聚合方式，并且没有逐实例的人为评分，因此评估可以在代理进化时自动重复。在一个固定的基础模型上，四个编码代理形成了一个狭窄的平稳区间（58.4-60.3分，满分100分），没有统计上可靠的成对分离：它们能够胜任地执行和记录分析，具有相对较强的证据可审计性和新颖性，但在建立可信主张的区分性行为（控制、稳健性、可证伪性和跨数据集泛化）方面大多缺乏。瓶颈在于科学判断而非编码，真正的发现仍然遥不可及。TruthInsightBench使这一差距成为可测量的目标；数据和评分代码可在https://github.com/TruthInsight-stack/TruthInsightBench获取。

View on arXiv Download PDF AI Translation

cs.AI / 73 / 2609.05088

Measuring AI Accountability Through Argumentation Analysis: Can Model Reasoning Withstand Scrutiny?

通过论证分析测量人工智能的问责制：模型推理能否经受审查？

Henselmans, Daan R., Prinzhorn, Derck W. E., Libert, Arno

Abstract

AI oversight methods rely on ground truth for validation, but what constitutes appropriate AI behavior is contested. This leaves evaluation of moral reasoning in LLMs and debate-based oversight implicitly avoiding realistic ambiguity. We investigate an alternative standard designed to function despite such ambiguity: structural quality of the defence a model can mount for its verdicts in response to critical questions, measured through a four-phase dialectical protocol grounded in Walton's theory of argumentation schemes and Govier's criteria for argument cogency. The protocol is adaptive to different frames of reasoning, extends beyond multiple-choice framing, and treats both the reasoning that precedes a verdict and its post-hoc justification. Across nine frontier models and 200 high-ambiguity MoralChoice items -- $6,778$ judge-scored cells, validated against $89.6\%$ inter-judge agreement on the binary failure judgment -- models defend their reasoning well above the rubric minimum on every dimension. Failure mass concentrates on grounds and sufficiency, and correlates with epistemic hedging rather than argument length. Reasoning is better defended than post-hoc justification, on every model and every Govier dimension. The scheme a model presents in its justification differs from the one it reasoned with on a substantial share of dilemmas ($\geq 20\%$ per model), despite value-based practical reasoning dominating both tracks. The protocol catches strictly indefensible defences (self-contradiction, false premises), and it surfaces difficulties in characterizing the role of retraction in AI alignment, suggesting a need for more situated evaluations.

Chinese Translation

人工智能监督方法依赖于真实数据进行验证，但什么构成适当的人工智能行为仍存在争议。这使得对大型语言模型（LLMs）中的道德推理的评估以及基于辩论的监督在隐含上回避了现实中的模糊性。我们研究了一种旨在在这种模糊性下仍然有效的替代标准：模型在面对关键问题时为其裁决所能提出的辩护的结构质量，通过基于沃尔顿（Walton）的论证模式理论和戈维尔（Govier）的论证有效性标准的四阶段辩证协议进行测量。该协议适应不同的推理框架，超越了多项选择的框架，并处理裁决前的推理及其事后辩护。在九个前沿模型和200个高模糊性的道德选择（MoralChoice）项目中——$6,778$个由评审打分的单元，经过$89.6\%$的评审间一致性验证的二元失败判断——模型在每个维度上的推理防御均显著高于评分最低标准。失败的主要集中在依据和充分性上，并与认知规避相关，而非论证长度。每个模型和每个戈维尔维度上，推理的防御优于事后辩护。模型在其辩护中呈现的方案与其在相当比例的困境中所使用的推理方案存在差异（每个模型$ geq 20\\%$），尽管基于价值的实用推理在两个轨道中占主导地位。该协议能够识别严格不可辩护的辩护（自我矛盾、虚假前提），并揭示了在人工智能对齐中界定撤回作用的困难，提示需要进行更具情境性的评估。

View on arXiv Download PDF AI Translation

cs.AI / 74 / 2609.05090

Constructing and Evaluating Clinical Reasoning Trajectories for Medical Agent

构建与评估医疗智能体的临床推理轨迹

Zhu, Yunqi, Zhang, Wensheng, Yang, Xuebing

Abstract

Evaluation of medical artificial intelligence agents remains predominantly answer-centric, assessing only the correctness of final outputs while overlooking the quality of intermediate reasoning. In clinical settings, however, a correct answer reached through fabricated evidence or incoherent logic is as dangerous as an incorrect one. We propose MedTraj, a framework that treats reasoning trajectories as critical objects for construction, evaluation, and optimization. The pipeline generates structured multi-step reasoning chains from medical reasoning sources. Each trajectory is then parsed into clinical observations, evidence, numbered reasoning steps, and a final conclusion, and scored across five quality dimensions: coherence, evidence support, hallucination, completeness, and traceability. Controlled error injection introduces targeted faults into otherwise correct trajectories to establish causal links between specific reasoning failures and measurable quality degradation. Building on this, step-level filtering based on marginal contribution identifies which individual reasoning steps drive or undermine trajectory quality. Finally, quality-weighted context learning feeds trajectory evaluations back into the model at inference time, allowing it to learn from both strong and weak reasoning demonstrations. Experiments across CareQA, PubMedQA, and CECMed demonstrate that trajectory context consistently improves reasoning coherence, with gains of +0.029 to +0.041 over a zero-shot baseline. On CECMed, quality-weighted context nearly doubles the correctness over the zero-shot baseline while cutting the hallucination ratio by 87%. Marginal-contribution analysis further shows that a small minority of reasoning steps carry most of the quality signal, and that extending chains beyond four steps yields diminishing returns.

Chinese Translation

医疗人工智能代理的评估仍然主要以答案为中心，仅评估最终输出的正确性，而忽视了中间推理的质量。然而，在临床环境中，通过虚构证据或不连贯逻辑得出的正确答案与错误答案同样危险。我们提出了MedTraj，一个将推理轨迹视为构建、评估和优化的关键对象的框架。该流程从医疗推理来源生成结构化的多步骤推理链。每条轨迹随后被解析为临床观察、证据、编号的推理步骤和最终结论，并在五个质量维度上进行评分：连贯性、证据支持、幻觉、完整性和可追溯性。受控错误注入将针对性故障引入本应正确的轨迹中，以建立特定推理失败与可测量质量下降之间的因果关系。在此基础上，基于边际贡献的步骤级过滤识别出哪些单独的推理步骤推动或削弱轨迹质量。最后，质量加权的上下文学习在推理时将轨迹评估反馈到模型中，使其能够从强和弱的推理示例中学习。针对CareQA、PubMedQA和CECMed的实验表明，轨迹上下文始终改善推理的连贯性，相较于零样本基线提高了+0.029至+0.041。在CECMed上，质量加权的上下文几乎将正确率翻倍，同时将幻觉比率降低了87%。边际贡献分析进一步显示，少数推理步骤承载了大部分质量信号，而将链条扩展到四个步骤以上则收益递减。

View on arXiv Download PDF AI Translation

cs.AI / 75 / 2609.05093

LLM-Guided Program Evolution for Circle Packing: Breaking 10 Packomania Records for $28

基于大型语言模型的程序演化用于圆形打包：打破10项Packomania记录，成本为28美元

Sander, Wes

Abstract

We present Discovery Loop, a lightweight system that uses a large language model (LLM) to iteratively evolve optimization algorithms. Starting from a simple seed solver, the LLM proposes algorithmic improvements guided by a scoreboard of results and a history of prior ideas. Each candidate is evaluated against an independent verifier; improvements are kept and failures discarded. Applied to the Packomania circle-packing benchmark (csqv: maximize the sum of radii of N variable-radius circles in the unit square), the system improved the best known solutions for 10 values of N in the range 101-114, with gains of 2.4%-5.4% over prior records, all within 15 iterations and at a total LLM cost of $27.72. These results have been independently accepted by Packomania. We describe the method, analyze cost-efficiency dynamics including an adaptive plateau-detection mechanism, and discuss implications for democratizing automated scientific discovery.

Chinese Translation

我们提出了Discovery Loop，这是一种轻量级系统，利用大型语言模型（LLM）迭代演化优化算法。从一个简单的种子求解器开始，LLM根据结果的记分板和先前想法的历史提出算法改进。每个候选方案都经过独立验证者的评估；改进方案被保留，而失败的方案则被丢弃。该系统应用于Packomania圆形打包基准（csqv：在单位正方形中最大化N个可变半径圆的半径总和），在101-114范围内的10个N值上改善了已知最佳解，增益在2.4%-5.4%之间，所有改进均在15次迭代内完成，总LLM成本为27.72美元。这些结果已被Packomania独立接受。我们描述了该方法，分析了成本效率动态，包括自适应平稳检测机制，并讨论了其对民主化自动化科学发现的影响。

View on arXiv Download PDF AI Translation

cs.AI / 76 / 2609.05094

ProCA: Progressive Contrastive Alignment for Robust EEG Visual Decoding

ProCA：用于鲁棒脑电图视觉解码的渐进对比对齐

Zhou, Kanglei, Lan, Chunyan, Li, Dongyang, Zhu, Jun, Wang, Liyuan

Abstract

Electroencephalogram (EEG) visual decoding aims to recover visual semantics from non-invasive neural time-series signals, for which robust alignment between noisy neural responses and stable semantic representations is key to achieving high-performance decoding. Despite recent advances in contrastive learning, robust EEG decoding remains challenging because existing methods rely on fixed visual or textual anchors whose semantic relations may become misaligned with EEG representations that vary across trials, subjects, and learning stages. Our empirical evidence shows that this instability appears across both standard EEG decoding protocols and more challenging robustness settings, including strict cross-subject transfer and realistic personalized continual adaptation. We provide a formal analysis showing that fixed semantic supervision can bias optimization when EEG-specific relations evolve, and that structure-agnostic perturbations may distort semantically important EEG components. To address these issues, we propose Progressive Contrastive Alignment (ProCA), a unified and model-agnostic framework for adaptive neural-semantic alignment. ProCA progressively refines class-level contrastive supervision from frozen vision-language priors to EEG-aware semantic relations, and introduces structure-consistent interpolation to constrain feature mixing according to channel-wise and temporal importance. Across subject-dependent, subject-independent, strict cross-subject transfer, and continual adaptation settings, ProCA achieves average relative Top-1/Top-5 gains of 7.4%/3.9%, 10.0%/4.6%, 28.1%/17.8%, and 16.8%/11.6%, respectively.

Chinese Translation

脑电图（EEG）视觉解码旨在从非侵入性的神经时间序列信号中恢复视觉语义，其中噪声神经响应与稳定语义表示之间的鲁棒对齐是实现高性能解码的关键。尽管对比学习最近取得了进展，但鲁棒的EEG解码仍然具有挑战性，因为现有方法依赖于固定的视觉或文本锚点，而这些锚点的语义关系可能与在不同试验、受试者和学习阶段中变化的EEG表示失去对齐。我们的实证证据表明，这种不稳定性在标准EEG解码协议和更具挑战性的鲁棒性设置中均有所体现，包括严格的跨受试者迁移和现实的个性化持续适应。我们提供了正式分析，表明固定的语义监督在EEG特定关系演变时可能导致优化偏差，而与结构无关的扰动可能会扭曲语义上重要的EEG成分。为了解决这些问题，我们提出了渐进对比对齐（ProCA），这是一个统一且与模型无关的自适应神经-语义对齐框架。ProCA逐步从冻结的视觉-语言先验中精炼类级对比监督到EEG感知的语义关系，并引入结构一致的插值，以根据通道和时间的重要性限制特征混合。在受试者依赖、受试者独立、严格的跨受试者迁移和持续适应设置中，ProCA分别实现了平均相对Top-1/Top-5增益为7.4%/3.9%、10.0%/4.6%、28.1%/17.8%和16.8%/11.6%。

View on arXiv Download PDF AI Translation

cs.AI / 77 / 2609.05104

Compact Bellman-Grounded Cognitive Maps for Cost-Aware Navigation

紧凑型贝尔曼基础认知地图用于成本感知导航

Han, Yuzhe, Xu, Mingkun, Wu, Yujie

Abstract

Biological agents navigate familiar environments not by re-solving routes for each new goal, but by reusing a learned map built once and read off as goals change. Existing artificial cognitive-map models mimic this reuse, yet their guidance is not explicitly grounded in additive heterogeneous route costs. Furthermore, they often struggle with memory efficiency: representative state-indexed and high-rank spectral constructions incur substantial storage growth as the environment scales. We present BCM, which grounds a reusable cognitive map in local edge costs through a self-supervised Bellman-grounded objective and a compact coordinate encoding, supporting changing goal queries without per-goal retraining. On weighted grids of up to $N=1600$ nodes, BCM maintains full success and only a 5\% mean Gap relative to exact Dijkstra search, compared with about $45\%$ for a connectivity-based spectral baseline. Notably, as the graph size increases from $N=400$ to $N=3600$, its memory footprint grows sublinearly while maintaining competitive performance, making our method scalable to complex environments. Together, these results show that additive route costs can be written into a compact, reusable cognitive-map representation, bridging the gap between biological flexibility and optimal path planning.

Chinese Translation

生物体在熟悉的环境中导航并不是通过为每个新目标重新解决路线，而是通过重用一次构建的学习地图来应对目标的变化。现有的人工认知地图模型模仿这种重用，但它们的指导并没有明确基于可加的异质路线成本。此外，它们通常在记忆效率方面存在困难：代表性的状态索引和高阶谱构造在环境规模扩大时会导致存储量显著增长。我们提出了BCM（Bellman-Centered Map），通过自监督的贝尔曼基础目标和紧凑的坐标编码，将可重用的认知地图与局部边缘成本相结合，支持在不进行每个目标重新训练的情况下进行目标查询。在最多有$N=1600$节点的加权网格上，BCM保持了100%的成功率，并且相对于精确的Dijkstra搜索仅有5%的平均差距，而基于连通性的谱基线约为45%。值得注意的是，随着图的大小从$N=400$增加到$N=3600$，其内存占用以亚线性增长，同时保持竞争性能，使得我们的方法能够扩展到复杂环境中。这些结果表明，可加的路线成本可以被写入紧凑的、可重用的认知地图表示中，弥合生物灵活性与最优路径规划之间的差距。

View on arXiv Download PDF AI Translation

cs.AI / 78 / 2609.05111

Unifying ICL, SFT, KL-Regularized RL Through a Bayesian Lens

通过贝叶斯视角统一 ICL、SFT、KL 正则化的强化学习

Fan, Junxin

Abstract

Large language models are now trained and evaluated under a diverse set of paradigms: supervised fine-tuning (SFT), few-shot in-context learning (ICL), KL-regularized RLHF/RLVR, on-policy distillation (OPD), and test-time reasoning with search and chain-of-thought. These methods are often discussed as fundamentally different, and recent empirical results--such as the mixed impact of few-shot prompting on RL-tuned reasoning models--can appear puzzling. This note develops a Bayesian perspective that puts these procedures on the same footing. At the core is a two-step template: (i) construct a (generalized) Bayes or Gibbs posterior q* over outputs or actions given a context, using a prior/reference model and a utility signal (log-likelihood, reward, or advantage); and (ii) approximate q* by a forward-KL projection onto a parametric family, either in-weights (SFT/RL) or in-context (ICL). Part I formalizes few-shot ICL and SFT as amortized and-weights projections onto the Bayes posterior predictive. Parts II-IV show that KL-regularized RLHF/RLVR, reward-weighted SFT, reward-weighted ICL (RW-ICL), and advantage-weighted SFT (AWSFT) are all instances of forward-KL projection onto posteriors induced by rewards or advantages. We disentangle where these equivalences hold (objectives and first-order updates) and where they do not (source and granularity of the learning signal). Part V sketches implications for modern reasoning pipelines: RLHF/RLVR recipes as "posterior design + projection", why cold-start or supervised warm-up is practically unavoidable for importance-weighted KL projections, and DeepSeek-R1 and o1-style reasoning models as combining test-time Bayesian search with training-time KL amortization.

Chinese Translation

大型语言模型现在在多种范式下进行训练和评估：监督微调（SFT）、少样本上下文学习（ICL）、KL 正则化的强化学习人类反馈（RLHF）/强化学习变体（RLVR）、在线蒸馏（OPD），以及带有搜索和思维链的测试时推理。这些方法通常被认为是根本不同的，最近的实证结果——例如少样本提示对强化学习调优推理模型的混合影响——可能显得令人困惑。本文提出了一种贝叶斯视角，将这些过程置于同一基础之上。核心是一个两步模板：（i）在给定上下文的情况下，利用先验/参考模型和效用信号（对数似然、奖励或优势）构建输出或动作的（广义）贝叶斯或吉布斯后验 q*；（ii）通过对参数族的前向 KL 投影来近似 q*，无论是在权重中（SFT/RL）还是在上下文中（ICL）。第一部分将少样本 ICL 和 SFT 正式化为对贝叶斯后验预测的摊销和权重投影。第二至第四部分展示了 KL 正则化的 RLHF/RLVR、奖励加权 SFT、奖励加权 ICL（RW-ICL）和优势加权 SFT（AWSFT）都是对由奖励或优势引起的后验进行前向 KL 投影的实例。我们阐明了这些等价关系成立的地方（目标和一阶更新）以及不成立的地方（学习信号的来源和粒度）。第五部分勾勒了对现代推理管道的影响：RLHF/RLVR 配方作为“后验设计 + 投影”，为何冷启动或监督预热在重要性加权 KL 投影中几乎不可避免，以及 DeepSeek-R1 和 o1 风格的推理模型如何将测试时贝叶斯搜索与训练时 KL 摊销相结合。

View on arXiv Download PDF AI Translation

cs.AI / 79 / 2609.05141

SciDocBench: A Workflow-Centered Benchmark and Data Pipeline for Scientific Document Understanding

SciDocBench：一个以工作流程为中心的科学文档理解基准和数据管道

Wu, Shenxi, Liu, Yuhong, Zhang, Haosong, Zou, Tongjin, Zhang, Yanxun, Chen, Gaochang, Liang, Dun, Wang, Jiaqi, Wang, Zhecan James, Zang, Yuhang, Lin, Dahua

Abstract

Scientific papers require models to reason jointly over text, equations, figures, tables, code, and datasets while preserving the provenance of supporting evidence. Existing benchmarks typically evaluate these capabilities in isolation, leaving unclear whether multimodal models can support realistic scientific-reading workflows. We introduce SciDocBench, a workflow-centered benchmark for scientific document understanding. It contains 124 expert-authored and difficulty-screened questions organized into seven research-assistant capability groups and 19 subtasks across five scientific domains. Each question is instantiated under four matched conditions combining English or Chinese questions with all-images-first or interleaved document representations, yielding 496 evaluation instances for controlled analysis. The strongest evaluated system achieves only 62.6/100, with pronounced weaknesses in document perception, evidence grounding, verification, and cross-document reasoning. To translate these diagnostics into scalable training signals, we introduce SciDocIR, a typed evidence-graph representation that preserves scientific document objects, layout and cross-reference relations, and provenance. Building on SciDocIR, we construct SciDocDataset, comprising approximately 15K supervised fine-tuning samples and 8K reinforcement-learning samples across 14 verifiable subtasks. Together, SciDocBench, SciDocIR, and SciDocDataset form an evaluation-to-training framework for diagnosing and improving scientific-document assistants. The project page is available at https://github.com/InternLM/SciDocBench.

Chinese Translation

科学论文要求模型能够对文本、方程式、图形、表格、代码和数据集进行联合推理，同时保留支持证据的来源。现有基准通常孤立地评估这些能力，尚不清楚多模态模型是否能够支持现实的科学阅读工作流程。我们介绍了SciDocBench，这是一个以工作流程为中心的科学文档理解基准。它包含124个由专家撰写并经过难度筛选的问题，组织成七个研究助理能力组和五个科学领域中的19个子任务。每个问题在四种匹配条件下实例化，结合英语或中文问题与全图像优先或交错文档表示，共产生496个用于控制分析的评估实例。评估的最强系统仅达到62.6/100，在文档感知、证据基础、验证和跨文档推理方面存在明显弱点。为了将这些诊断转化为可扩展的训练信号，我们引入了SciDocIR，一种类型化的证据图表示，保留科学文档对象、布局和交叉引用关系以及来源。基于SciDocIR，我们构建了SciDocDataset，包含约15K个监督微调样本和8K个强化学习样本，涵盖14个可验证的子任务。SciDocBench、SciDocIR和SciDocDataset共同构成了一个评估到训练的框架，用于诊断和改进科学文档助手。项目页面可访问 https://github.com/InternLM/SciDocBench。

View on arXiv Download PDF AI Translation

cs.AI / 80 / 2609.05146

A Hybrid Predictive Ensemble of Machine Learning and Deep Neural Networks for Early Cardiovascular Disease Risk Assessment

机器学习与深度神经网络混合预测集成用于早期心血管疾病风险评估

Venkateswaran, Balaji

Abstract

This study introduces an intelligent framework that integrates machine learning and deep neural network ensemble techniques for early detection and prognosis of cardiovascular diseases. The system utilizes real-time physiological data collected from Internet of Medical Things (IoMT) devices, including ECG sensors, heart rate monitors, and blood pressure trackers. To ensure the accuracy and reliability of input data, preprocessing steps such as noise reduction, normalization, and missing value imputation are employed. The most significant health indicators are identified through effective feature selection methods and then processed using optimized classifiers such as Support Vector Machines (SVM), Random Forests, and eXtreme Gradient Boosting (XGBoost), which are combined in an ensemble architecture to improve diagnostic precision. The framework demonstrates remarkable performance in predicting cardiovascular disease risk, achieving higher accuracy, reduced false positives, and enhanced consistency compared to conventional methods. It is designed on a cloud-based infrastructure that ensures scalability and real-time processing for continuous patient monitoring. Experimental evaluation on real-world cardiovascular datasets confirms the framework's efficiency in early-stage risk assessment and clinical decision support. The results highlight the potential of combining traditional machine learning and deep learning paradigms to achieve proactive healthcare management and improve patient outcomes.

Chinese Translation

本研究介绍了一种智能框架，集成了机器学习和深度神经网络集成技术，用于心血管疾病的早期检测和预后。该系统利用从医疗物联网（Internet of Medical Things, IoMT）设备收集的实时生理数据，包括心电图（ECG）传感器、心率监测器和血压追踪器。为了确保输入数据的准确性和可靠性，采用了噪声降低、归一化和缺失值填补等预处理步骤。通过有效的特征选择方法识别出最重要的健康指标，然后使用优化的分类器进行处理，如支持向量机（Support Vector Machines, SVM）、随机森林（Random Forests）和极端梯度提升（eXtreme Gradient Boosting, XGBoost），并将其组合在一个集成架构中以提高诊断精度。该框架在预测心血管疾病风险方面表现出色，与传统方法相比，达到了更高的准确性、降低了误报率，并增强了一致性。它基于云基础设施设计，确保了可扩展性和实时处理，以实现对患者的持续监测。对真实世界心血管数据集的实验评估确认了该框架在早期风险评估和临床决策支持中的效率。结果突显了结合传统机器学习和深度学习范式以实现主动医疗管理和改善患者结果的潜力。

View on arXiv Download PDF AI Translation

cs.AI / 81 / 2609.05190

The Mirror Agent Model: a Bayesian Architecture for Interpretable Agent Behavior

镜像代理模型：一种可解释代理行为的贝叶斯架构

Persiani, Michele, Hellström, Thomas

Abstract

In this paper we illustrate a novel architecture generating interpretable behavior and explanations. We refer to this architecture as the Mirror Agent Model because it defines the observer model, that is the target of explicit and implicit communications, as a mirror of the agent's. With the goal of providing a general understanding of this work, we firstly show prior relevant results addressing the informative communication of agents intentions and the production of legible behavior. In the second part of the paper we furnish the architecture with novel capabilities for explanations through off-the-shelf saliency methods, followed by preliminary qualitative results.

Chinese Translation

在本文中，我们展示了一种生成可解释行为和解释的新颖架构。我们将该架构称为镜像代理模型，因为它将观察者模型（即显性和隐性交流的目标）定义为代理的镜像。为了提供对这项工作的总体理解，我们首先展示了先前相关结果，这些结果涉及代理意图的信息交流和可读行为的产生。在论文的第二部分，我们为该架构提供了通过现成显著性方法进行解释的新能力，并随后给出了初步的定性结果。

View on arXiv Download PDF AI Translation

cs.AI / 82 / 2609.05198

What Matters in On-Policy Distillation? A Perspective on Data Efficiency and Data Selection

在策略蒸馏中什么是重要的？关于数据效率和数据选择的视角

Hou, Zhinan, Zhang, Jiaqi, Cai, Xunliang, You, Keyou

Abstract

On-Policy Distillation (OPD) has emerged as a widely adopted post-training paradigm for enhancing large language models in reasoning domains. However, the data-centric mechanisms in OPD remain relatively underexplored. This paper presents a empirical study of data efficiency and data selection in OPD. We begin by investigating an extreme setting: training OPD on only one example, namely 1-shot OPD. Surprisingly, we find that 1-shot OPD is consistently effective across all sampled training examples and harder examples often yield superior performance gain. We next investigate what actually drives the student model's improvement in the training data. Our analysis reveals that the improvement is not driven by high token entropy, but the longer CoT paths which hard problems naturally generate. Training on longer CoT can help maintain closer alignment with the teacher over a long reasoning horizon, and learn critical thinking patterns usually missing in short CoTs, such as reflection (e.g., ``Alternatively''). Based on these insights, we propose a simple data selection method that selects only hard examples for training, where even ``unsolvable'' examples that completely exceed the teacher's capability can be successfully used. Our experiments conducted on four models ranging from 1.5B to 7B show that training the student model on only 8 selected hard examples matches the performance of the 17K dataset baseline.

Chinese Translation

在策略蒸馏（On-Policy Distillation, OPD）已成为提升大型语言模型在推理领域表现的广泛采用的后训练范式。然而，OPD中的数据中心机制仍然相对未被深入探索。本文对OPD中的数据效率和数据选择进行了实证研究。我们首先研究一个极端的设置：仅用一个示例进行OPD训练，即1-shot OPD。令人惊讶的是，我们发现1-shot OPD在所有采样的训练示例中始终有效，而较难的示例往往带来更优的性能提升。接下来，我们探讨了是什么驱动学生模型在训练数据中的改进。我们的分析表明，这种改进并不是由高标记熵驱动，而是由较长的链式推理（Chain of Thought, CoT）路径所自然生成的。对较长的CoT进行训练有助于在较长的推理范围内与教师模型保持更紧密的对齐，并学习通常在短CoT中缺失的关键思维模式，例如反思（例如，“或者”）。基于这些见解，我们提出了一种简单的数据选择方法，仅选择困难示例进行训练，其中甚至可以成功使用完全超出教师能力的“不可解”示例。我们在四个模型（参数范围从1.5B到7B）上进行的实验表明，仅在8个选定的困难示例上训练学生模型，其性能与17K数据集基线相当。

View on arXiv Download PDF AI Translation

cs.AI / 83 / 2609.05227

CABAL: Multi-Agent Simulacra for Tracing the Effects of Collusive Bidding in Peer Review

CABAL：用于追踪同行评审中合谋投标影响的多智能体仿真框架

Zhou, Jicheng, Li, Kemou, Wong, Kahim, Li, Zheyuan, Shi, Zhuan, Li, Fengpeng, Wu, Haiwei, Zhou, Jiantao

Abstract

Recent reports during the AAAI-27 review cycle highlight the risk of reviewers coordinating bids for reciprocal assignment advantage. Prior work treats bidding, reviewer assignment, and review manipulation as separate stages, leaving the lifecycle effects of collusive bidding unclear. Real-world analysis is further constrained by typically unobservable collusive intent and the lack of counterfactuals for the same conference. Motivated by this gap, we introduce \alg, an end-to-end multi-agent simulacra framework for studying reviewer assignment integrity by holding the conference environment fixed and configuring LLM-driven reviewer agents with honest or collusive policies. We further develop an affinity-guided collusive bidding strategy that uses mutual reviewer-paper affinities to construct collusion rings and select target papers, producing expertise-consistent rather than arbitrarily targeted attacks. Controlled experiments show that collusive bidding more than doubles target-paper capture and that assigned colluders score target papers about two points higher than honest co-reviewers, while conference-wide effects remain comparatively modest. Evaluated bid-phase detectors provide only limited evidence of collusion: in a fixed-triplet detector stress test, native positive-bid graphs are confounded by benign affinity, while a Very-High-only diagnostic view enables precise but low-coverage local recovery.

Chinese Translation

在AAAI-27评审周期期间的最新报告中，强调了评审人协调投标以获取互惠分配优势的风险。之前的研究将投标、评审人分配和评审操控视为独立阶段，导致合谋投标的生命周期影响不明确。现实世界的分析进一步受到通常不可观察的合谋意图和缺乏同一会议的反事实的限制。基于这一空白，我们引入了 extit{CABAL}，一个端到端的多智能体仿真框架，用于通过固定会议环境并配置具有诚实或合谋政策的基于大型语言模型（LLM）的评审人代理，研究评审人分配的完整性。我们进一步开发了一种基于亲和力的合谋投标策略，该策略利用评审人与论文之间的相互亲和力构建合谋圈并选择目标论文，从而产生与专业知识一致而非随意针对的攻击。受控实验表明，合谋投标使目标论文的捕获率增加了两倍以上，而被分配的合谋者对目标论文的评分比诚实的共同评审者高出约两分，同时会议范围内的影响仍然相对温和。评估的投标阶段检测器仅提供有限的合谋证据：在固定三元组检测器压力测试中，原生正投标图受到良性亲和力的干扰，而仅限于非常高的诊断视图则能够实现精确但覆盖率低的局部恢复。

View on arXiv Download PDF AI Translation

cs.AI / 84 / 2609.05228

ACE: Adaptive Calibration-Free Expert Skipping for MoE-based LLMs

ACE：基于MoE的大型语言模型的自适应无校准专家跳过方法

Xu, Zukang, Zhao, Zhixiong, Hu, Xing, Yu, Jiangyong, Wen, Houji, Li, Jun, Jiang, Zhe, Yang, Dawei

Abstract

Mixture-of-Experts (MoE) architectures provide an efficient paradigm for scaling large language models (LLMs), yet fixed top-k routing activates the same number of expert slots for every token, causing substantial redundant computation. Existing expert-skipping methods often rely on router confidence, calibration data, or additional training, and therefore cannot reliably estimate the actual contribution of routed experts. To this end, we propose ACE, a training-free, calibration-free, and checkpoint-preserving framework for token-adaptive expert skipping in MoE-based LLMs. ACE contains two complementary components: 1) Global Spectral Proxy (GSP), which estimates global transformation capacity from the coupled gate, up, and down projections together with RMSNorm scaling; and 2) Router-Conditioned Refinement (RCR), which constructs expert-specific direction prototypes from centered router weights and evaluates expert responses along routing-preferred directions. During inference, ACE combines both estimates with runtime router gates and skips an expert slot only when both views identify it as low-contribution, while always retaining the top-1 expert. All expert statistics are computed offline, leaving only table lookups and lightweight scalar operations online. Extensive experiments across three MoE-based LLMs and eight benchmarks demonstrate that ACE consistently outperforms existing static and dynamic baselines, with increasingly pronounced advantages under aggressive expert skipping. For instance, at a 50% skipping ratio on Qwen3.6-35B-A3B, ACE reduces WikiText-2 perplexity by 7.96% and improves average downstream accuracy by 4.15 percentage points over the strongest competing method.

Chinese Translation

混合专家（MoE）架构为扩展大型语言模型（LLMs）提供了一种高效的范式，但固定的 top-k 路由在每个令牌上激活相同数量的专家槽，导致显著的冗余计算。现有的专家跳过方法通常依赖于路由器置信度、校准数据或额外训练，因此无法可靠地估计路由专家的实际贡献。为此，我们提出了 ACE，这是一种无训练、无校准且保留检查点的框架，用于基于 MoE 的 LLM 中的令牌自适应专家跳过。ACE 包含两个互补的组件：1）全局谱代理（Global Spectral Proxy, GSP），它通过耦合门、上投影和下投影以及 RMSNorm 缩放来估计全局变换能力；2）路由器条件细化（Router-Conditioned Refinement, RCR），它从中心化的路由器权重构建专家特定的方向原型，并沿着路由优先方向评估专家响应。在推理过程中，ACE 将两种估计与运行时路由器门结合，仅在两种视角均识别为低贡献时跳过专家槽，同时始终保留 top-1 专家。所有专家统计信息均在离线计算，仅在在线时进行表查找和轻量级标量操作。针对三种基于 MoE 的 LLM 和八个基准的广泛实验表明，ACE 在激进的专家跳过下始终优于现有的静态和动态基线，优势愈加明显。例如，在 Qwen3.6-35B-A3B 上以 50% 的跳过比例，ACE 将 WikiText-2 的困惑度降低了 7.96%，并将平均下游准确率提高了 4.15 个百分点，相较于最强的竞争方法。

View on arXiv Download PDF AI Translation

cs.AI / 85 / 2609.05232

Substrate-Aware AI Agents: Execution Context as a First-Class Input

基于底层环境感知的人工智能代理：执行上下文作为一类重要输入

Agrawal, Manu

Abstract

Autonomous AI agents increasingly select actions in environments whose memory, execution-time, runtime, compute, and operational constraints determine what counts as a suitable plan. We call the absence of this execution context from an agent's planning state substrate blindness. We test this general proposition through numerical code generation, where selected implementation choices and operational consequences are directly observable. Three frontier model configurations--Anthropic Claude Opus 5, OpenAI GPT-5.6-Sol, and Google Gemini 3.7 Flash--generate code for a high-dimensional pairwise Euclidean-distance task either from the task alone or with a 128 MB RAM and 10.0 s wall-time contract. Contract disclosure reduced measured peak process memory in 13 of 14 executable index-aligned task-only versus contract-disclosed comparisons and reduced mean wall time in all three cohorts, making execution up to 3.1x faster. Across the audited corpus, disclosure produced structural code changes including bounded blocking, float32 retention, upper-triangle traversal, and in-place or memory-mapped buffers. At a tighter 96 MB contract, independently sampled contract-disclosed cohorts achieved correct-and-within-budget outcomes of 4/5 for Claude Opus 5, 5/5 for GPT-5.6-Sol, and 3/5 for Gemini 3.7 Flash, compared with task-only outcomes of 0/5, 1/5, and 0/5; cohort mean MaxRSS and wall time were 49-74% and 35-64% lower than their task-only references. These results establish a controlled proof of concept for substrate-aware agent planning: a minimal execution contract induces proactive structural adaptation in generated programs, shifting computation away from unconstrained allocations and substantially improving observed resource-time profiles before execution.

Chinese Translation

自主人工智能代理越来越多地在其内存、执行时间、运行时、计算和操作约束决定什么算作合适计划的环境中选择行动。我们将代理规划状态中缺乏这种执行上下文称为底层盲目性。我们通过数值代码生成测试这一一般命题，其中所选的实现选择和操作后果是直接可观察的。三种前沿模型配置——Anthropic Claude Opus 5、OpenAI GPT-5.6-Sol 和 Google Gemini 3.7 Flash——为高维成对欧几里得距离任务生成代码，既可以仅基于任务本身，也可以基于128 MB RAM和10.0秒的时间限制。合同披露在14个可执行的索引对齐任务仅与合同披露比较中，减少了13个的峰值进程内存，并在所有三个组中减少了平均墙时间，使执行速度提高至3.1倍。在审核的语料库中，披露产生了结构性代码变化，包括有界阻塞、float32保留、上三角遍历以及就地或内存映射缓冲区。在更严格的96 MB合同下，独立抽样的合同披露组在Claude Opus 5中达到了4/5的正确且在预算内的结果，在GPT-5.6-Sol中达到了5/5，在Gemini 3.7 Flash中达到了3/5，而仅基于任务的结果分别为0/5、1/5和0/5；组的平均最大RSS和墙时间比其仅基于任务的参考低49-74%和35-64%。这些结果建立了基于底层环境感知的代理规划的受控概念验证：最小的执行合同促使生成程序的主动结构适应，将计算从不受约束的分配中转移，并在执行前显著改善观察到的资源时间特征。

View on arXiv Download PDF AI Translation

cs.AI / 86 / 2609.05241

Uncensored Open-weight Models: Redistribution as the Persistence Layer

无审查开放权重模型：重分配作为持久层

Labs, 10a, :, Garcia, Juliette, May, Hailey, McKenzie, Bobby, Pham, David, Swain, Matthew, Valdez, Joshua, Wieland, Corie, Yahn, Zachary

Abstract

A rapidly expanding ecosystem of actors is removing built-in safety guardrails from open-weight AI models. We profile this ecosystem by identifying key producers, downstream reproductions, and emerging applications. Between January 2024 and March 2026, we identified 3,471 original uncensored models on HuggingFace, each repackaged an average of 2.4 times; three actors account for 52% of all 8,164 compressed redistributions. Once quantized and mirrored across separate accounts, formats, and registries such as Ollama, these models persist regardless of upstream removal and become easier to deploy downstream. Of the 1,643 identified GitHub applications integrating uncensored large language models (ULLMs), 25% were classified as explicitly malicious.

Chinese Translation

一个快速扩展的参与者生态系统正在从开放权重人工智能模型中移除内置的安全防护措施。我们通过识别关键生产者、下游再现和新兴应用来描绘这个生态系统。在2024年1月至2026年3月期间，我们在HuggingFace上识别出3,471个原始无审查模型，每个模型平均被重新打包2.4次；三个参与者占所有8,164个压缩重分配的52%。一旦经过量化并在不同账户、格式和注册表（如Ollama）中镜像，这些模型便会持续存在，无论上游是否被移除，并且更容易在下游部署。在识别的1,643个集成无审查大型语言模型（ULLMs）的GitHub应用中，25%被归类为明显恶意。

View on arXiv Download PDF AI Translation

cs.AI / 87 / 2609.05245

Do LLMs Exhibit Coherent Knowledge Structures in Mathematical Reasoning? A Perspective from Knowledge Space Theory

大型语言模型在数学推理中是否表现出一致的知识结构？来自知识空间理论的视角

Cui, Peng, Do, Heejin, Sachan, Mrinmaya

Abstract

Human knowledge is inherently structured and interdependent: mastery of a concept requires prior mastery of its prerequisites, a principle formalized by Knowledge Space Theory (KST). While LLMs achieve strong performance on complex reasoning tasks, it remains unclear whether they exhibit coherent, human-like knowledge structure. We introduce a KST-grounded framework for evaluating LLM knowledge structure in mathematical reasoning, using it as a normative framework to analyze whether LLM behavior adheres to principled knowledge dependencies. Evaluating eight open- and closed-source LLMs against real human learners, we find that (1) LLMs do not adhere to human knowledge structure -- they frequently violate knowledge dependencies and fail to leverage related knowledge provided in context to improve performance on dependent questions; (2) LLMs do not share a consistent knowledge structure among themselves, as reflected by low overlap in their knowledge distributions. Furthermore, these structural deficiencies remain largely invisible to accuracy-based and LLM-as-judge evaluations. Together, our results provide behavioral evidence that current LLMs knowledge does not follow a human-like structure.

Chinese Translation

人类知识本质上是结构化和相互依赖的：掌握一个概念需要先掌握其前提，这一原则被知识空间理论（Knowledge Space Theory, KST）形式化。尽管大型语言模型（LLMs）在复杂推理任务中表现出色，但它们是否展现出一致的人类知识结构仍然不明确。我们引入一个基于KST的框架来评估LLMs在数学推理中的知识结构，并利用该框架分析LLM行为是否遵循原则性的知识依赖关系。通过将八个开放源代码和闭源代码的LLMs与真实人类学习者进行比较，我们发现：（1）LLMs不遵循人类知识结构——它们经常违反知识依赖关系，并未能利用上下文中提供的相关知识来提高对依赖性问题的表现；（2）LLMs之间的知识结构不一致，表现为它们的知识分布重叠度低。此外，这些结构缺陷在基于准确性的评估和LLM作为评判者的评估中大多是隐形的。综合来看，我们的结果提供了行为证据，表明当前的LLMs知识并不遵循人类般的结构。

View on arXiv Download PDF AI Translation

cs.AI / 88 / 2609.05251

A Unified Physics-Aware Quantum Machine Learning Framework across Power GaN HEMTs and Logic Nanowire FETs: Predicting Unseen Process Splits and Held-Out Geometry Combinations with Lower Error and Tighter Split-to-Split Variability

一个统一的物理感知量子机器学习框架：在功率氮化镓高电子迁移率晶体管和逻辑纳米线场效应晶体管中预测未见的工艺分割和保留几何组合，具有更低的误差和更紧的分割间变异性

Rai, Rushat, Wang, Yun-Yuan, Kakaen, Autsada, Chang, Pei-Jie, Nguyen, Doan Viet, Chiu, Yuan-Chieh, Tantraviwat, Doldet, Tumilty, Niall, See, Simon, Lee, Wen-Jay, Li, Tai-Yue, Chen, Nan-Yow, Wu, Tian-Li

Abstract

We present a unified reinforcement-learning (RL) framework that discovers compact parametrized quantum circuits (PQCs) for data-scarce device modeling. A graph neural network (GNN) policy optimized by proximal policy optimization (PPO) searches circuit architectures using leave-one-group-out cross-validation (LOGOCV) error on held-out process or geometry groups as the reward. The framework achieves the lowest mean absolute error (MAE) on all 11 targets versus six classical baselines, with 59% lower error (Ioff) and 81% tighter fold variability (VTH) for HEMTs and 84% lower error (VTH, SS, Ioff) and 82% tighter fold variability (Ioff) for NWFETs. These results demonstrate the potential of RL-selected, classically simulated PQCs as compact surrogates with low OOD error and improved physical consistency, despite imposing no explicit physical constraints, penalty terms, or device-specific equations, on the two evaluated device datasets.

Chinese Translation

我们提出了一个统一的强化学习（RL）框架，该框架发现用于数据稀缺设备建模的紧凑参数化量子电路（PQC）。通过近端策略优化（PPO）优化的图神经网络（GNN）策略，利用留一组交叉验证（LOGOCV）在保留的工艺或几何组上作为奖励来搜索电路架构。该框架在所有11个目标上实现了最低的平均绝对误差（MAE），与六个经典基线相比，对于高电子迁移率晶体管（HEMTs）降低了59%的误差（Ioff）和81%的折叠变异性（VTH），对于纳米线场效应晶体管（NWFETs）降低了84%的误差（VTH, SS, Ioff）和82%的折叠变异性（Ioff）。这些结果展示了RL选择的、经典模拟的PQC作为紧凑替代模型的潜力，具有低的超出分布（OOD）误差和改进的物理一致性，尽管在评估的两个设备数据集上并未施加任何显式的物理约束、惩罚项或设备特定方程。

View on arXiv Download PDF AI Translation

cs.AI / 89 / 2609.05257

Commonsense Reasoning in Computer Vision: Foundations, Recent Advancements, and Future Directions

计算机视觉中的常识推理：基础、近期进展与未来方向

Mahmud, Bahar Uddin, Barua, Sumit, Hong, Guan Yue, Gupta, Ajay, Liu, Hexu

Abstract

Commonsense reasoning in computer vision encompasses integrating visual data and contextual knowledge, crucial for enhancing AI's understanding of everyday scenarios. This understanding not only improves machine learning models but also enhances their ability to interact meaningfully with humans and the environment. Unlike CNN-based conventional vision models, which are designed to identify objects within a specific image, incorporating commonsense knowledge enables models to interpret scenes in a more holistic manner, thereby improving their spatial ability to reason about relationships among objects and actions. This integration not only enhances object recognition but also facilitates a deeper understanding of the contextual factors, ultimately leading to more precise predictions and interactions in real-world applications. This paper presents a comprehensive survey of recent developments that integrate commonsense knowledge into computer vision tasks. We systematically review approaches based on knowledge graphs, scene graphs, neuro-symbolic models, and commonsense-augmented transformers. We also outline current limitations related to dataset bias, knowledge incompleteness, and integration challenges. Finally, we highlight prospective research trajectories in cross-modal reasoning, scalable commonsense knowledge injection, and neuro-symbolic hybrid architectures to develop truly intelligent visual systems.

Chinese Translation

计算机视觉中的常识推理涉及将视觉数据与上下文知识相结合，这对于增强人工智能对日常场景的理解至关重要。这种理解不仅改善了机器学习模型的性能，还增强了它们与人类和环境进行有意义互动的能力。与基于卷积神经网络（CNN）的传统视觉模型不同，后者旨在识别特定图像中的对象，融入常识知识使得模型能够以更全面的方式解读场景，从而提高其推理对象和动作之间关系的空间能力。这种整合不仅增强了对象识别能力，还促进了对上下文因素的更深刻理解，最终导致在现实应用中更精确的预测和互动。本文全面回顾了将常识知识整合到计算机视觉任务中的近期发展。我们系统地评估了基于知识图谱、场景图谱、神经符号模型和常识增强变换器的方法。我们还概述了与数据集偏差、知识不完整性和整合挑战相关的当前局限性。最后，我们强调了在跨模态推理、可扩展的常识知识注入和神经符号混合架构方面的未来研究方向，以开发真正智能的视觉系统。

View on arXiv Download PDF AI Translation

cs.AI / 90 / 2609.05261

Trace2Tower: Transition-Aware EigenTrace Induction of Multi-Level Skills for LLM Agents

Trace2Tower：面向转移的多层技能特征轨迹引导框架用于大型语言模型代理

Sun, Jiazheng, Yang, Boyu, Yuan, Binhao, Li, Mingxuan, Peng, Xin

Abstract

Large language model agents increasingly rely on execution traces to master complex interactive tasks. However, current paradigms are bottlenecked by shallow trajectory retrieval and flat skill summarization, fundamentally ignoring the temporal dependencies and outcome-conditioned topology of agent behavior. We introduce Trace2Tower, a transition-aware EigenTrace framework that distills raw trajectories into a robust skill hierarchy. Trace2Tower abstracts step-level interactions into canonical events, constructing a unified graph governed by semantic compatibility, transition dynamics, and outcome evidence. Through a novel contrastive spectral decomposition, it isolates stable, success-aligned behavioral modes while rigorously suppressing failure-prone shortcuts. These modes organically populate a dynamic skill tower of action templates, procedural routines, and overarching task strategies, continuously refined via verifier-guided feedback. On ALFWorld, Trace2Tower achieves 87.31% success requiring only 10.35 steps and 0.26 invalid actions; on WebShop, it reaches 50.67% exact success. Across both benchmarks, Trace2Tower significantly outperforms existing baselines in task mastery and context-efficient experience reuse.

Chinese Translation

大型语言模型代理越来越依赖执行轨迹来掌握复杂的交互任务。然而，当前的范式受到浅层轨迹检索和扁平技能总结的瓶颈，根本上忽视了代理行为的时间依赖性和结果条件拓扑。我们提出了Trace2Tower，一种面向转移的特征轨迹（EigenTrace）框架，它将原始轨迹提炼为一个稳健的技能层次结构。Trace2Tower将逐步交互抽象为规范事件，构建一个由语义兼容性、转移动态和结果证据主导的统一图。通过一种新颖的对比谱分解，它隔离出稳定的、与成功对齐的行为模式，同时严格抑制易失败的捷径。这些模式有机地填充了一个动态技能塔，其中包含行动模板、程序化例程和总体任务策略，并通过验证者引导的反馈不断优化。在ALFWorld上，Trace2Tower实现了87.31%的成功率，仅需10.35步和0.26个无效动作；在WebShop上，成功率达到50.67%。在这两个基准测试中，Trace2Tower在任务掌握和上下文高效经验重用方面显著优于现有基线。

View on arXiv Download PDF AI Translation

cs.AI / 91 / 2609.05270

AI for Computational Design Science: A Responsible Human-AI Framework and Case Study on Short-Form Video Safety Surveillance

计算设计科学中的人工智能：负责任的人机智能框架及短视频安全监控案例研究

Zhang, Wenli, Xie, Jiaheng, Pan, Zhihe, Chai, Yidong, Fang, Xiao, Ram, Sudha

Abstract

Artificial intelligence (AI) is transforming not only what information systems researchers design, but also how design research is conducted. Yet existing literature offers limited guidance for computational design science (CDS) when AI actively participates in problem formulation, resource construction, design search, evaluation, and knowledge abstraction. We develop AI for Computational Design Science (AI4CDS), a five-phase methodological framework in which AI expands problem and design search while researchers retain responsibility for domain grounding, admissibility, verification, and scientific judgment. Collaboration is governed by graduated trust, reversibility, auditability, and differentiated reproducibility. We instantiate AI4CDS through ChildRiskGuard, an interpretable artifact for detecting short-form videos inappropriate for children, while documenting AI interactions, rejected alternatives, corrections, and audit trails. The case translates audience-dependent safety and explanation faithfulness into three technical challenges and develops an artifact that separates generic from child-specific risk, represents distinct developmental-risk mechanisms, and makes concept-level explanations part of the predictive computation. ChildRiskGuard achieves an F1 score of 0.769, substantially outperforming direct application of a general-purpose content-safety model while remaining competitive with strong benchmarks. The primary contribution is AI4CDS as a responsible framework for AI-enabled CDS; ChildRiskGuard provides process and artifact evidence of how AI-expanded, researcher-governed design can generate and evaluate novel computational design knowledge.

Chinese Translation

人工智能（AI）正在改变信息系统研究者的设计内容以及设计研究的开展方式。然而，现有文献对计算设计科学（CDS）在AI积极参与问题制定、资源构建、设计搜索、评估和知识抽象时提供的指导有限。我们提出了计算设计科学中的人工智能（AI4CDS），这是一个五阶段的方法论框架，其中AI扩展了问题和设计搜索，而研究者则保留了对领域基础、可接受性、验证和科学判断的责任。合作受到逐级信任、可逆性、可审计性和差异化可重复性的管理。我们通过ChildRiskGuard实例化AI4CDS，这是一个可解释的工具，用于检测不适合儿童观看的短视频，同时记录AI交互、被拒绝的替代方案、修正和审计轨迹。该案例将依赖于观众的安全性和解释的可信度转化为三个技术挑战，并开发出一个将一般风险与儿童特定风险分开的工具，表示不同的发展风险机制，并使概念级解释成为预测计算的一部分。ChildRiskGuard的F1得分为0.769，显著优于通用内容安全模型的直接应用，同时与强基准保持竞争力。主要贡献在于AI4CDS作为一个负责任的AI驱动CDS框架；ChildRiskGuard提供了过程和工具证据，展示了如何通过AI扩展、研究者主导的设计生成和评估新颖的计算设计知识。

View on arXiv Download PDF AI Translation

cs.AI / 92 / 2609.05275

Don't Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference

不要放弃 Dropout：优化层稀疏性以提高大规模语言模型的训练和推理效率

Elhoushi, Mostafa, Pretko, Alex, Dey, Nolan, Zhang, Bin Claire, Gray, Gavia, Gosal, Gurpreet, Mahmoud, Abdulrahman, Bergsma, Shane, Hestness, Joel

Abstract

Layer dropout (a.k.a. stochastic depth) has been shown to enable faster training, higher accuracy, and robustness to zero-shot layer pruning in both language and vision transformers. However, as models and datasets have scaled, dropout - particularly layer dropout - has largely disappeared from large language models (LLMs) pre-training recipes. While some prior work has reported that dropout can degrade accuracy, no comprehensive study has quantified, let alone mitigated, this effect. In this study, we show that layer dropout should be used in state-of-the-art LLM training, establishing best practices and scaling analysis for both training and post-training benefits. Concretely, with optimal layer distribution, time schedule, and optimizer hyperparameters, we observe that at the same training FLOPs layer dropout leads to lower loss. For a given number of training steps, LLMs can achieve lower or similar validation loss while saving upto 25% of training FLOPs. Moreover, layer dropout enables significant post-training optimizations, such as early exit, intermediate-layer skipping, and self-speculative decoding, yielding up to 1.5x inference speedup with negligible accuracy loss. Across more than 2400 training experiments, spanning models from 271M to 8.2B parameters and datasets up to 160B tokens, we demonstrate that these findings extend reliably to large-scale training regimes. All pre-training experiments were run on Cerebras CS-3 systems.

Chinese Translation

层 dropout（也称为随机深度）已被证明能够在语言和视觉变换器中实现更快的训练、更高的准确性以及对零-shot 层剪枝的鲁棒性。然而，随着模型和数据集的规模扩大，dropout——尤其是层 dropout——在大规模语言模型（LLMs）的预训练方案中几乎消失。尽管一些先前的研究报告称 dropout 可能会降低准确性，但尚无全面研究量化甚至缓解这种影响。在本研究中，我们表明层 dropout 应在最先进的 LLM 训练中使用，并为训练和后训练的好处建立最佳实践和规模分析。具体而言，在最佳层分布、时间安排和优化器超参数下，我们观察到在相同的训练 FLOPs 下，层 dropout 导致更低的损失。在给定的训练步骤数下，LLMs 可以在节省高达 25% 的训练 FLOPs 的同时实现更低或相似的验证损失。此外，层 dropout 使得显著的后训练优化成为可能，例如提前退出、中间层跳过和自我推测解码，带来高达 1.5 倍的推理加速，同时准确性损失微乎其微。在超过 2400 次训练实验中，涵盖从 271M 到 8.2B 参数的模型和高达 160B 令牌的数据集，我们证明这些发现可靠地扩展到大规模训练方案。所有预训练实验均在 Cerebras CS-3 系统上运行。

View on arXiv Download PDF AI Translation

cs.AI / 93 / 2609.05279

Testing Interchangeability in LLM Agent Teams

测试大型语言模型代理团队的可互换性

Gao, Jianxin, Yu, Tianyi, Deng, Linna, Li, Runze, Wang, Zining

Abstract

Production multi-agent systems replace agents constantly, on the assumption that an agent filling a role is interchangeable with any other agent that can do the job. We test that assumption. Eight teams per setting are formed independently from one base model on the same tasks, each agent keeping a private notebook across ten formation episodes; we then trade role-matched agents between teams and measure what changes on held-out tasks. Against a placebo that reproduces the disruption of a roster change without changing who occupies the seat, a swap costs little in task score but raises the communication a team spends per unit of progress by 16 to 63 percent, and in Hanabi a swapped agent is more expensive than an inexperienced one, consistent with interference from conventions learned with its former partner. In Collab-Overcooked, when the agent that sets the agenda is replaced, most of the extra communication comes from the agent that stayed. Three ablations, over base models, decoding temperature and formation length, move the swap penalty alongside one other quantity: how far independently formed teams drift apart. Greedy decoding lowers both; doubling a team's history raises both. In these settings, agents are more fungible in task outcome than in coordination efficiency, with larger swap effects after longer formation histories.

Chinese Translation

生产多智能体系统不断更换代理，假设填补某一角色的代理可以与任何能够完成该工作的其他代理互换。我们对此假设进行了测试。在相同任务上，从一个基础模型独立形成八个团队，每个代理在十个形成阶段中保持一个私有笔记本；然后我们在团队之间交换角色匹配的代理，并测量在保留任务上的变化。与一种安慰剂对照组相比，该对照组在不改变座位占有者的情况下重现了人员更换的干扰，交换对任务得分的影响很小，但增加了团队在每单位进展中花费的沟通量，增加幅度为16%到63%；在《花火》（Hanabi）中，交换的代理比一个缺乏经验的代理更昂贵，这与其与前任伙伴学习的惯例造成的干扰一致。在《协作过热》（Collab-Overcooked）中，当设定议程的代理被替换时，大部分额外的沟通来自于留下的代理。三项消融实验，针对基础模型、解码温度和形成长度，移动了交换惩罚与另一量之间的关系：独立形成的团队之间的漂移程度。贪婪解码降低了两者；而双倍增加团队历史则提高了两者。在这些设置中，代理在任务结果上的可替代性高于协调效率，且在较长的形成历史后，交换效应更大。

View on arXiv Download PDF AI Translation

cs.AI / 94 / 2609.05284

GUT: Quantifying and Optimizing the Reasoning Uncertainty of LLMs via Graph Complexity

GUT：通过图复杂性量化和优化大型语言模型的推理不确定性

Liang, Shuang, Hu, Xin-Yu, Ou, Xiang-Jun, Zhang, Shao-Qun

Abstract

Recent years have witnessed great advances in the reasoning ability of Large Language Models (LLMs). However, the reasoning processes of LLMs often exhibit uncertainty, where LLMs often produce a proliferation of divergent branches at each reasoning step even when fed the same prompting inputs, and certain branches exhibit evidently incredible, even nonsensical, reasoning chains and results. In this paper, we propose the Graph-complexity-based UncerTainty (GUT) method for investigating the reasoning uncertainty of LLMs. The key idea of GUT is to characterize the potential branches of each reasoning chain with a directed acyclic graph, thereby ensuring that all potential branches are comprehensively covered within the graph space. Building upon this recognition, we further build two modules of GUT, that is, a Quantification (GUT-Q) module and an Optimization (GUT-O) module, for quantifying and reducing the reasoning uncertainty of LLMs, respectively. GUT-Q measures LLM reasoning uncertainty by approximating the reasoning space complexity with graph complexity. GUT-O implements uncertainty optimization by treating negative uncertainty as the reward function in reinforcement learning. Experimental results conducted on four LLMs and five datasets validate the effectiveness of GUT.

Chinese Translation

近年来，大型语言模型（LLMs）的推理能力取得了显著进展。然而，LLMs的推理过程往往表现出不确定性，即使在输入相同的提示时，LLMs在每个推理步骤中也常常产生大量不同的分支，并且某些分支表现出明显不合理甚至荒谬的推理链和结果。本文提出了一种基于图复杂性的推理不确定性研究方法（Graph-complexity-based UncerTainty，GUT），旨在探讨LLMs的推理不确定性。GUT的核心思想是用有向无环图来表征每个推理链的潜在分支，从而确保所有潜在分支在图空间内得到全面覆盖。在此基础上，我们进一步构建了GUT的两个模块，即量化模块（GUT-Q）和优化模块（GUT-O），分别用于量化和减少LLMs的推理不确定性。GUT-Q通过用图复杂性近似推理空间复杂性来测量LLMs的推理不确定性。GUT-O则通过将负不确定性视为强化学习中的奖励函数来实现不确定性优化。在对四个LLMs和五个数据集进行的实验结果验证了GUT的有效性。

View on arXiv Download PDF AI Translation

cs.AI / 95 / 2609.05289

Beyond Aggregate Scores: Behavioral Correctness Assumptions for Assessing Reference-Based Automatic Evaluation Methods

超越汇总分数：评估基于参考的自动评估方法的行为正确性假设

Mahbub, Maria, Rice, Ashley, Munroe, Michael R., Kamara, Amidu, Sadovnik, Amir

Abstract

Automated reference-based evaluation methods play a critical role in assessing natural language generation systems. Existing meta-evaluation primarily measures agreement with human judgments or benchmark labels, providing limited insight into evaluator behavior under controlled conditions. We introduce behavioral correctness assumptions, a complementary framework for evaluating reference-based automatic evaluation methods. We define a taxonomy of correctness-preserving and correctness-altering assumptions and operationalize them through controlled response transformations that specify expected scoring behaviors. We evaluate diverse lexical, character-level, semantic, LLM-based, and hybrid evaluators and analyze their assumption-level behavior, stability, sensitivity, repeat-run variability, configuration sensitivity, and reproducibility. Our experiments reveal distinct behavioral trade-offs across evaluation paradigms: no evaluator satisfies all proposed correctness assumptions, and evaluators with similar aggregate performance can exhibit substantially different behavioral profiles. These findings demonstrate that behavioral correctness assumptions provide diagnostic information obscured by conventional aggregate meta-evaluation.

Chinese Translation

自动化的基于参考的评估方法在评估自然语言生成系统中发挥着关键作用。现有的元评估主要测量与人类判断或基准标签的一致性，提供了对评估者在受控条件下行为的有限洞察。我们引入了行为正确性假设，这是一种用于评估基于参考的自动评估方法的补充框架。我们定义了一种正确性保持和正确性改变假设的分类法，并通过控制响应转换将其操作化，这些转换指定了预期的评分行为。我们评估了多种词汇、字符级、语义、基于大型语言模型（LLM）和混合评估器，并分析了它们在假设层面的行为、稳定性、敏感性、重复运行变异性、配置敏感性和可重复性。我们的实验揭示了不同评估范式之间明显的行为权衡：没有评估器满足所有提出的正确性假设，且具有相似汇总性能的评估器可能表现出截然不同的行为特征。这些发现表明，行为正确性假设提供了传统汇总元评估所掩盖的诊断信息。

View on arXiv Download PDF AI Translation

cs.AI / 96 / 2609.05295

RISE: Recursive Improvement via Self-Extrapolating Policy Distillation

RISE：通过自我外推策略蒸馏的递归改进

Li, Yang, Yavuz, Semih, Joty, Shafiq

Abstract

On-policy distillation (OPD) provides dense, per-token supervision for language model post-training, but its effectiveness is bottlenecked by teacher quality: external teachers suffer from distribution mismatch, while self-distillation with privileged conditioning is limited by in-context learning capacity. We propose \textbf{RISE} (\textbf{R}ecursive \textbf{I}mprovement via \textbf{S}elf-\textbf{E}xtrapolating Policy Distillation), which constructs a synthetic teacher directly from the model's own RLVR training trajectory. By extrapolating the displacement between the current checkpoint and a trailing anchor---in parameter space or output logit space---RISE converts a sparse outcome-induced parameter update into a dense token-level target, without any external model or privileged conditioning. RISE combines RLVR and OPD in a complementary loop: outcome rewards ground the extrapolation toward correct reasoning, while the extrapolated teacher refines token-level decisions. Moreover, since the teacher is refreshed every iteration as the student improves, distillation becomes a recursive improvement mechanism rather than a one-shot compression step. Experiments spanning mathematical reasoning, multi-domain STEM, code generation, and multi-turn agentic tasks show that RISE outperforms RLVR-only training and on-policy self-distillation across all settings.

Chinese Translation

在线蒸馏（OPD）为语言模型的后训练提供了密集的逐标记监督，但其有效性受到教师质量的瓶颈：外部教师受到分布不匹配的影响，而带有特权条件的自我蒸馏则受到上下文学习能力的限制。我们提出了 extbf{RISE}（ extbf{R}ecursive extbf{I}mprovement via extbf{S}elf- extbf{E}xtrapolating Policy Distillation），该方法直接从模型自身的RLVR训练轨迹构建一个合成教师。通过外推当前检查点与后续锚点之间的位移——在参数空间或输出对数空间中——RISE将稀疏的结果诱导参数更新转换为密集的逐标记目标，而无需任何外部模型或特权条件。RISE将RLVR和OPD结合在一个互补的循环中：结果奖励为正确推理的外推提供基础，而外推的教师则细化逐标记决策。此外，由于教师在每次迭代中随着学生的进步而更新，蒸馏变成了一种递归改进机制，而不是一次性压缩步骤。涵盖数学推理、多领域STEM、代码生成和多轮智能任务的实验表明，RISE在所有设置中均优于仅使用RLVR训练和在线自我蒸馏。

View on arXiv Download PDF AI Translation

cs.AI / 97 / 2609.05314

Large Language Models for HVAC Operations in Building Energy Systems: A Critical Review of Methods, Applications, and Deployment Readiness

大型语言模型在建筑能源系统中HVAC操作的应用：方法、应用及部署准备的批判性综述

Neubauer, Alexander, Hong, Tianzhen, Li, Han, Yu, Mengbo, Darbandi, Amin, Fürst, Yannick, Kriegel, Martin

Abstract

Building automation systems generate rich sensor data yet remain insight-poor because heterogeneous point naming, missing metadata, and fragmented documentation obstruct their operational use. This systematic review analyses and codes 66 peer-reviewed studies on large language models (LLMs) for HVAC operations published between 2023 and March 2026. Each study is classified across five application families and three LLM method families and assessed for evidence realism, deployment readiness, and the responsibility boundary between the LLM and physical HVAC decisions. The corpus is concentrated in building energy modelling (BEM, 32 of 66 papers), while load forecasting remains too sparse for subfield-level conclusions. Only four studies reach pilot-level evidence, and none reports sustained operational deployment. No study was classified as ready-now for industry adoption; three were near-term and 63 research-only. Nevertheless, several bounded, human-in-the-loop uses merit near-term trials, including point-name normalisation, document-grounded operator support, BEM workflow assistance, and advisory interfaces around physics-based controllers. Conventional machine learning (ML), model predictive control (MPC), reinforcement learning (RL) and ontology-based tools remain more adopted for high-frequency control, short-horizon numerical forecasting, and well-posed ontology mapping, while autonomous agentic operation and unvalidated occupant proxies remain research-stage. Current evidence therefore supports LLMs primarily as semantic and workflow layers rather than autonomous HVAC controllers. Future work should prioritise field-validated benchmarks, orchestration evaluation under operational constraints, and LLM-MPC/RL architectures with bounded latency and verifiable safety properties.

Chinese Translation

建筑自动化系统生成丰富的传感器数据，但由于异构的点命名、缺失的元数据和碎片化的文档，导致其在操作使用上缺乏洞察力。本系统综述分析并编码了66篇关于大型语言模型（LLMs）在HVAC操作中的应用的同行评审研究，这些研究发表于2023年至2026年3月之间。每项研究根据五个应用类别和三种LLM方法类别进行分类，并评估其证据的现实性、部署准备情况以及LLM与物理HVAC决策之间的责任边界。文献集中在建筑能源建模（BEM，66篇论文中的32篇），而负荷预测在子领域层面上仍然过于稀疏，无法得出结论。只有四项研究达到了试点级别的证据，且没有一项报告持续的操作部署。没有研究被分类为当前准备好进行行业采用；三项为近期可用，63项仅限于研究。然而，几个有限的、以人为中心的应用值得近期试验，包括点名规范化、基于文档的操作支持、BEM工作流程协助以及围绕基于物理的控制器的咨询接口。传统的机器学习（ML）、模型预测控制（MPC）、强化学习（RL）和基于本体的工具在高频控制、短期数值预测和良好构造的本体映射中仍然更为普遍，而自主代理操作和未经验证的居住者代理仍处于研究阶段。因此，目前的证据主要支持将LLMs视为语义和工作流程层，而非自主HVAC控制器。未来的工作应优先考虑经过现场验证的基准、在操作约束下的协调评估，以及具有有限延迟和可验证安全属性的LLM-MPC/RL架构。

View on arXiv Download PDF AI Translation

cs.AI / 98 / 2609.05327

LLM-Driven Algorithm Design for Quantum Circuit Synthesis based on Binary Decision Diagrams

基于二元决策图的量子电路综合的LLM驱动算法设计

Sim, Yoonju, Berto, Federico, Hua, Chuanbo, Park, Jinkyoo, Kwon, Changhyun

Abstract

Quantum circuits are central to implementing quantum algorithms on quantum devices, where quantum gates must be reversible. Many quantum algorithms rely on Boolean functions, which must therefore be implemented reversibly within quantum circuits. Reversible circuit synthesis provides a way to translate such Boolean functions into reversible circuits. Binary decision diagrams (BDDs) offer a scalable approach to this task, but the resulting BDDs and circuits depend heavily on variable ordering. Existing ordering heuristics commonly minimize BDD size because it is closely tied to the circuit size. However, BDD size is an imperfect proxy for the quantum cost of the synthesized circuit (QCC). We propose \texttt{QuantumEvo}, an evolutionary framework that uses an LLM as a heuristic generator for QCC-aware BDD variable ordering. Instead of predicting orderings directly, \texttt{QuantumEvo} searches over ordering heuristics initialized from multiple heuristic families. Candidate heuristics directly manipulate variable orderings using standard BDD operations and are selected by downstream QCC. The discovered heuristic, HGA-QE, modifies the sifting step inside a genetic algorithm so that the procedure is better aligned with QCC. Across the benchmark set, HGA-QE achieves a 70.9\% tie-or-win rate against the per-function best baseline and is strictly best on 13.5\% of the functions. The results demonstrate broadly competitive QCC performance, with HGA-QE showing a clearer relative advantage in strict wins on the two benchmark suites drawn from sources different from the data used for heuristic discovery.

Chinese Translation

量子电路是量子设备上实现量子算法的核心，其中量子门必须是可逆的。许多量子算法依赖于布尔函数，因此必须在量子电路中以可逆方式实现。可逆电路综合为将此类布尔函数转换为可逆电路提供了一种方法。二元决策图（BDDs）为该任务提供了一种可扩展的途径，但生成的BDDs和电路在很大程度上依赖于变量排序。现有的排序启发式方法通常通过最小化BDD大小来实现，因为BDD大小与电路规模密切相关。然而，BDD大小并不能完美反映综合电路的量子成本（QCC）。我们提出了 exttt{QuantumEvo}，这是一种进化框架，利用大型语言模型（LLM）作为QCC感知的BDD变量排序启发式生成器。与直接预测排序不同， exttt{QuantumEvo}在多个启发式家族初始化的启发式函数上进行搜索。候选启发式通过标准BDD操作直接操作变量排序，并通过下游QCC进行选择。所发现的启发式HGA-QE修改了遗传算法中的筛选步骤，使该过程更好地与QCC对齐。在基准测试集中，HGA-QE在70.9%的函数上达到与每函数最佳基线持平或优胜的表现，并在13.5%的函数上严格优于基线。结果表明，HGA-QE在QCC性能上表现出广泛的竞争力，且在两个基准套件中（这些套件来源于与启发式发现数据不同的来源）严格获胜的相对优势更为明显。

View on arXiv Download PDF AI Translation

cs.AI / 99 / 2609.05333

Technical Manual for a Toolkit for Measuring Contextual Individuation in Transformer Language Models

用于测量变换器语言模型中情境个体化的工具包技术手册

Marques, José Luciano Verçosa, Heitmann, Frederico Jorge, Perez, Daniel Omar, de Paula, Marcelo Vinicius, Barros, Tárcio André dos Santos

Abstract

A transformer language model assigns a single, context-independent vector to a word type at its embedding layer, yet is widely believed to individuate that word's occurrences by context in its later layers. Testing this belief cleanly requires a construct that holds the word form fixed while its context and intended sense vary in a controlled, labeled way. This manual documents an open toolkit built around such a construct, which we call a bridge form: a single written word that recurs, unchanged, across two or more subject domains with a different sense in each. We describe, and justify, every stage of the pipeline: the declarative specification of bridge forms and their source domains, corpus acquisition from Wikipedia, occurrence localization, layer-wise representation extraction, a domain-pairwise silhouette measurement of separation in the model's representation space, and a paired visualization protocol. Each design choice is presented together with the methodological failure mode it is meant to avoid (sense contamination from overly broad category labels, the multi-group bias of the silhouette coefficient, subword-tokenization misalignment, and axis-comparability artifacts in dimensionality-reduced plots, among others). This manuscript is a methodological and implementation reference: it does not report or interpret empirical outcomes of running the toolkit on any particular model or bridge-form set. The toolkit, its full source, and the corpora used to exercise it are archived separately (Section 9) under a persistent identifier, and are intended to be cited as an instrument by studies that use it to produce and interpret empirical results.

Chinese Translation

变换器语言模型在其嵌入层为一个词类型分配一个单一的、与上下文无关的向量，但普遍认为它在后续层中通过上下文对该词的出现进行个体化。要清晰地测试这一信念，需要一个构造，该构造在上下文和预期意义变化的同时保持词形不变，并以受控、标记的方式进行。本文手册记录了围绕这一构造构建的开放工具包，我们称之为桥接形式（bridge form）：一个在两个或多个主题领域中以不同意义反复出现且不变的单一书面词。我们描述并论证了管道的每个阶段：桥接形式及其源领域的声明性规范、从维基百科获取语料、出现定位、逐层表示提取、模型表示空间中领域对的轮廓分离测量，以及配对可视化协议。每个设计选择都与其旨在避免的方法论失败模式一起呈现（例如，来自过于宽泛的类别标签的意义污染、轮廓系数的多组偏差、子词标记不对齐，以及降维图中的轴可比性伪影等）。本文稿是一个方法论和实施参考：它不报告或解释在任何特定模型或桥接形式集上运行工具包的实证结果。该工具包及其完整源代码，以及用于测试的语料，已单独存档（第9节），并在持久标识符下提供，旨在作为研究引用的工具，以便使用它来生成和解释实证结果。

View on arXiv Download PDF AI Translation

cs.AI / 100 / 2609.05339

Does Your Agent's Memory Survive a Model Upgrade? A Controlled Study of Memory Portability

您的代理的记忆能在模型升级中存活吗？一项关于记忆可移植性的控制研究

Goyal, Ankit, Ray, Jaideep

Abstract

Model upgrades are routine; memory migrations are not. An agent can keep the same memory store and still forget: a new model may interpret old notes differently, mixed embedding versions may break retrieval, and repair may fail without the original evidence. We compare memory as the same history is preserved verbatim for long-context reading (LC-RAW), divided into chunks for retrieval-augmented generation (RAG), compressed by a model into natural-language notes (NOTES), or normalized into a fixed-schema knowledge graph (KG-fixed). The study uses 48 synthetic histories with randomized answer codes, exact scoring, and two open-weight models with sub 10 billion parameters. Our measurements show that fixed-schema structures transfer reliably, with KG-fixed accuracy changing by only $+0.0004 \pm 0.0020$ following a writer swap. Conversely, compressed NOTES exhibit high model coupling, with accuracy shifting asymmetrically by $+9.91$ or $-13.28$ percentage points depending on the specific migration direction. In RAG systems, partial embedding migrations using a 50/50 mixed index capture only a 4.96-point accuracy improvement, forfeiting the majority of the 11.90-point gain achieved through full re-embedding. Diagnostic decomposition attributes 80% ($0.467 \pm 0.014$) of the NOTES accuracy deficit to information lost during initial construction, whereas retrieval failures drive 81% ($0.364 \pm 0.012$) of the RAG deficit. Finally, store-only repair of NOTES fails to reach a 90% performance recovery target in all 48 test cases, whereas retaining the raw source history enables successful recovery in 34 of 48 cases for one tested direction. These findings highlight the necessity of direction-specific migration testing, strict embedding space isolation, and the retention of source histories for memory repair.

Chinese Translation

模型升级是常规操作，而记忆迁移却并非如此。一个代理可以保持相同的记忆存储，但仍然可能遗忘：新模型可能会以不同的方式解读旧笔记，混合的嵌入版本可能会破坏检索，且在没有原始证据的情况下修复可能会失败。我们比较了在长上下文阅读（LC-RAW）中逐字保留相同历史的记忆、为检索增强生成（RAG）而分块的记忆、被模型压缩为自然语言笔记（NOTES）以及规范化为固定模式知识图（KG-fixed）。本研究使用了48个合成历史，随机答案代码，精确评分，以及两个参数少于100亿的开放权重模型。我们的测量结果表明，固定模式结构可以可靠地转移，KG-fixed的准确性在作家更换后仅变化$+0.0004 imes 0.0020$。相反，压缩的NOTES表现出较高的模型耦合度，准确性根据特定迁移方向不对称地变化$+9.91$或$-13.28$个百分点。在RAG系统中，使用50/50混合索引的部分嵌入迁移仅捕获4.96点的准确性提升，放弃了通过完全重新嵌入所获得的11.90点的大部分增益。诊断分解将NOTES准确性缺口的80%（$0.467 imes 0.014$）归因于初始构建过程中丢失的信息，而检索失败则推动了RAG缺口的81%（$0.364 imes 0.012$）。最后，仅对NOTES进行存储修复在所有48个测试案例中未能达到90%的性能恢复目标，而保留原始源历史使得在测试的一个方向上成功恢复了48个案例中的34个。这些发现强调了方向特定迁移测试、严格的嵌入空间隔离以及源历史保留在记忆修复中的必要性。

View on arXiv Download PDF AI Translation

cs.AI / 101 / 2609.05346

Who Should Grade My Work? Student Perspectives on Transparent AI-Assisted Writing Assessment in Higher Education

谁来评估我的作品？学生对高等教育中透明的人工智能辅助写作评估的看法

AlGhamdi, Rayed

Abstract

The integration of GenAI tools into higher education assessment raises important questions about how students understand, interpret, and respond to AI-mediated evaluation. As instructors increasingly explore AI tools for providing feedback, prior research has examined whether GenAI-generated feedback improves writing performance and how students perceive its usefulness; comparatively little is known, however, about how students interpret such evaluation when they are explicitly informed that an AI system, rather than a human instructor, produced the feedback and the score. This study reports findings from a qualitative pedagogical inquiry conducted in an undergraduate technical communication course for computing students at a Saudi public university. Thirteen male undergraduate computing students completed an in-class handwritten writing task; the scanned submissions were evaluated by ChatGPT using a rubric-based prompt aligned with the task objectives. Students were then explicitly informed that ChatGPT had generated the score and feedback and were invited to reflect on the evaluation in writing. Inductive thematic analysis of these reflections identified four themes: perceived usefulness of feedback; awareness of AI's contextual and pedagogical limitations; conditional trust, distinguishing feedback utility from evaluative authority; and reflection on the institutional and pedagogical role of the human instructor. Participants accepted GenAI feedback as useful for surface-level revision but consistently positioned the human instructor as the appropriate authority over grading decisions. The study identifies this as a distinction between feedback utility and evaluative authority, two judgments that students treat as analytically separate rather than as opposite ends of a single approval scale...

Chinese Translation

将生成式人工智能（GenAI）工具融入高等教育评估中引发了关于学生如何理解、解读和回应人工智能介导的评估的重要问题。随着教师越来越多地探索使用人工智能工具提供反馈，先前的研究已经考察了生成式人工智能生成的反馈是否能改善写作表现，以及学生如何看待其有效性；然而，对于学生在明确被告知反馈和评分是由人工智能系统而非人类教师生成时如何解读这种评估的了解相对较少。本研究报告了在沙特阿拉伯一所公立大学的计算机专业本科技术传播课程中进行的定性教学研究的发现。十三名男性计算机专业本科生完成了一项课堂手写写作任务；提交的手稿经过扫描后由ChatGPT使用与任务目标一致的评分标准进行评估。随后，学生们被明确告知评分和反馈是由ChatGPT生成的，并被邀请对评估进行书面反思。对这些反思的归纳主题分析识别出了四个主题：反馈的感知有效性；对人工智能在上下文和教学限制方面的认识；有条件的信任，将反馈的实用性与评估权威区分开来；以及对人类教师在机构和教学角色上的反思。参与者接受生成式人工智能的反馈作为表面修订的有用工具，但始终将人类教师视为评分决策的适当权威。本研究将这一点识别为反馈实用性与评估权威之间的区别，学生将这两种判断视为分析上独立的，而非单一认可尺度的对立两端。

View on arXiv Download PDF AI Translation

cs.AI / 102 / 2609.05374

CUA-Universe: A Scalable and Dynamic Environment for Hybrid GUI+CLI Agents

CUA-Universe：一个可扩展的动态环境用于混合GUI+CLI代理

Shi, Haoting, Wang, Wenhao, Fang, Weicheng, Liang, Yaozhong, Jin, Tian, Zhao, Pengxiang, Liu, Guangyi, Chen, Siheng, Wang, Yanfeng

Abstract

Computer-use agents have advanced on benchmarks like OSWorld and AndroidWorld, but still act mostly through the GUI, often producing inefficient trajectories. Real-world computer work is hybrid, combining visual-state inspection with precise, high-throughput command-line operations, so capable agents must coordinate both modalities over shared application state. Yet scalable hybrid environments remain scarce because supporting both GUI and CLI over real applications typically requires substantial manual engineering for each application. Existing agents also struggle to use the two interfaces complementarily: CLI-native agents lack visual perception for tasks involving interface state or layout, while GUI-native agents are inefficient for operations better executed through commands. We introduce CUA-Universe, a scalable environment-to-data pipeline that turns real desktop software into hybrid GUI+CLI environments. App-Forge adapts applications into reproducible VMs and command-line surfaces it discovers, wraps, or generates, scaling to 16 applications; Task-Weave synthesizes diverse hybrid tasks of controllable difficulty from reusable operations over seed files; and Path-Steer steers rollouts along efficient hybrid paths and harvests verified trajectories for post-training. Training on this data shifts behavior from inefficient GUI interaction and brittle CLI scripting toward effective GUI+CLI orchestration. Our 9B model improves both success and efficiency on CUA-Verse (Score +39.3 pts; -37% steps, -60% tokens), OSWorld (SR +16.8 pts; -57% steps, -44% tokens), and OSWorld-MCP (Score +7.84 pts; -27% steps, -30% tokens). CUA-Universe provides a scalable path toward more capable and efficient computer-use agents.

Chinese Translation

计算机使用代理在OSWorld和AndroidWorld等基准测试中取得了进展，但仍主要通过图形用户界面（GUI）操作，常常导致低效的轨迹。现实世界的计算机工作是混合的，结合了视觉状态检查与精确、高吞吐量的命令行操作，因此，能够的代理必须在共享应用状态上协调这两种模式。然而，支持真实应用程序的可扩展混合环境仍然稀缺，因为同时支持GUI和命令行界面（CLI）通常需要对每个应用程序进行大量的手动工程。现有代理在使用这两种接口互补方面也面临困难：CLI原生代理缺乏对涉及界面状态或布局的任务的视觉感知，而GUI原生代理在通过命令更有效地执行操作时效率较低。我们提出了CUA-Universe，一个可扩展的环境到数据管道，将真实桌面软件转变为混合的GUI+CLI环境。App-Forge将应用程序适配为可重现的虚拟机和它发现、包装或生成的命令行界面，支持最多16个应用程序；Task-Weave从可重用操作中合成可控难度的多样化混合任务；Path-Steer沿着高效的混合路径引导滚动并收集经过验证的轨迹以供后续训练。基于这些数据的训练使行为从低效的GUI交互和脆弱的CLI脚本转向有效的GUI+CLI编排。我们的9B模型在CUA-Verse（得分提高39.3分；步骤减少37%；令牌减少60%）、OSWorld（成功率提高16.8分；步骤减少57%；令牌减少44%）和OSWorld-MCP（得分提高7.84分；步骤减少27%；令牌减少30%）上都提高了成功率和效率。CUA-Universe为更强大和高效的计算机使用代理提供了一条可扩展的路径。

View on arXiv Download PDF AI Translation

cs.AI / 103 / 2609.05381

Molecular D\'ej\`a Vu: Digit-Level Retrieval of Published Values in Frontier Language Models

分子既视感：前沿语言模型中已发布值的数字级检索

Busch, Matthias, Tacke, Marius, Lamaka, Sviatlana V., Zheludkevich, Mikhail L., Cyron, Christian J., Aydin, Roland C., Feiler, Christian

Abstract

Large language models (LLMs) are increasingly evaluated on molecular property benchmarks, but accuracy cannot distinguish a model that predicts a property from one that retrieves a published number. We audit 22 frontier models on 12 regression benchmarks for verbatim retrieval and find that it is widespread but relatively benchmark-specific: on five datasets more than $50\%$ of the LLMs show verbatim retrieval, while on the remaining datasets it appears only in isolated cells. We run our experiments at two reasoning levels and find that reasoning changes retrieval. The same experiments, on the same molecules and with the same prompt, are flagged $89\%$ more often at the higher reasoning level than at the lowest one. Finally, we test a way to interrupt retrieval in our most contaminated cases, and find that the strongest models in some cases still recognise a combination of transformed SMILES strings and original labels. Furthermore, suppressing retrieval moves the prediction errors of the different models closer together in relative terms, while their differing use of verbatim retrieval spreads them apart. This indicates that the general predictive capability of an LLM is not determined solely by the amount of memorised values. This work provides an overview of the amount and depth of verbatim retrieval in molecular regression benchmarks using LLMs.

Chinese Translation

大型语言模型（LLMs）越来越多地在分子属性基准上进行评估，但准确性无法区分一个预测属性的模型与一个检索已发布数字的模型。我们对22个前沿模型在12个回归基准上的逐字检索进行了审计，发现这种现象普遍存在但相对特定于基准：在五个数据集中，超过50%的LLMs显示出逐字检索，而在其余数据集中仅在个别单元中出现。我们在两个推理层次上进行实验，发现推理会改变检索结果。在相同的分子和相同的提示下，相同的实验在较高的推理层次上被标记的频率比在最低层次上高出89%。最后，我们测试了一种在最污染的情况下中断检索的方法，发现一些最强模型在某些情况下仍能识别转化后的SMILES字符串和原始标签的组合。此外，抑制检索使不同模型的预测误差在相对上更接近，而它们对逐字检索的不同使用则使它们之间的差距加大。这表明，LLM的一般预测能力并不仅仅由记忆值的数量决定。本研究提供了使用LLMs在分子回归基准中逐字检索的数量和深度的概述。

View on arXiv Download PDF AI Translation

cs.AI / 104 / 2609.05385

Necessary or Sufficient? Evaluating LLM Explanations With Behavioural Evidence

必要还是充分？基于行为证据评估大型语言模型的解释

Pawar, Urja, Ramanayake, Rajitha, Kemal, Nabeel, Kandath, Ashwin, O'Neill, Owen, Bourgeon, Guillaume, Chatbri, Houssem

Abstract

LLM decision components that can operate within agent workflows often produce action-relevant recommendations or judgements together with explanations. Operators may use the named factors to monitor a system, diagnose errors, or decide when to escalate an output. Such use assumes that the explanations agree with the component's observable decision behaviour. We test two interpretations of the named factors: necessity, meaning that changing a factor would change the output, and sufficiency, meaning that retaining it while removing other changeable information would preserve the output. We evaluate these interpretations in two synthetic use cases: recommending advisors to clients and judging prompts for harmfulness or risk. Models return an output and the top three factors that most influenced it. Controlled black-box interventions estimate a necessity score for each factor by measuring how often changing it changes the output, and a sufficiency score by measuring how often retaining it preserves the output. Across eight models from the Claude, GPT, and Gemini families, the mean Spearman correlations between the cited ranking and the necessity and sufficiency scores are 0.349 and 0.354 for advisor recommendation, and 0.431 and 0.580 for prompt monitoring. Furthermore, an uncited factor scores above the lowest-scoring cited factor in 57.6% of advisor responses under necessity and 58.1% under sufficiency; the corresponding prompt-monitoring rates are 25.8% and 8.9%. The cited top three contain useful information but do not reliably identify the three factors with the strongest measured influence under necessity or sufficiency. The framework provides a black-box reliability check for explanations used in agent oversight while remaining scoped to individual LLM decisions.

Chinese Translation

大型语言模型（LLM）在代理工作流程中能够运行的决策组件通常会生成与行动相关的建议或判断，并附带解释。操作人员可能会利用这些命名因素来监控系统、诊断错误或决定何时升级输出。这种使用假设解释与组件的可观察决策行为一致。我们测试了这两个命名因素的解释：必要性，意味着改变一个因素会改变输出；以及充分性，意味着在去除其他可变信息的同时保留该因素会保持输出。我们在两个合成用例中评估这些解释：向客户推荐顾问和判断提示的有害性或风险。模型返回一个输出以及最影响该输出的前三个因素。通过受控的黑箱干预，我们通过测量改变因素时输出变化的频率来估算每个因素的必要性得分，并通过测量保留因素时输出保持的频率来估算充分性得分。在Claude、GPT和Gemini家族的八个模型中，引用排名与必要性和充分性得分之间的平均Spearman相关系数分别为0.349和0.354（顾问推荐），以及0.431和0.580（提示监控）。此外，在必要性下，未被引用的因素在57.6%的顾问响应中得分高于得分最低的引用因素，而在充分性下这一比例为58.1%；对应的提示监控率分别为25.8%和8.9%。引用的前三个因素包含有用信息，但并不能可靠地识别在必要性或充分性下影响最强的三个因素。该框架为代理监督中使用的解释提供了黑箱可靠性检查，同时仍然限于个别LLM决策。

View on arXiv Download PDF AI Translation

cs.AI / 105 / 2609.05395

Multi-Step Tool-Calling over Korean Open Public APIs: A Benchmark and a Data-Synthesis Recipe

基于韩国开放公共API的多步骤工具调用：基准测试与数据合成方案

Kim, Dain, Cho, Eungi, Kim, Kyumin, Noh, Shinyeong, Lim, Kyuseong

Abstract

Data-sovereignty regulations increasingly require public institutions to deploy open-source, on-premise LLM agents that chain multiple tool-calls across live government APIs. However, open-source models consistently underperform in this multi-step setting, and no existing benchmark measures the gap. We introduce the Korean Open Public API Benchmark (KOPA-Bench), comprising 145 real-world tasks. To close this gap, we present EDGE, an Execution-grounded Dynamic Graph for tool-calling data synthEsis driven by live execution. EDGE builds a graph of how each tool's output can feed another's input, keeps only the links that succeed when actually called against the live APIs, and traverses these verified links to synthesize executable multi-step trajectories. Fine-tuned via GRPO on the resulting dataset, our 9B model nearly matches the untuned 27B model from the same family, improving substantially not only on KOPA-Bench but also on the BFCL benchmark.

Chinese Translation

数据主权法规日益要求公共机构部署开源的本地大型语言模型（LLM）代理，这些代理能够在实时政府API上链式调用多个工具。然而，开源模型在这种多步骤设置中表现不佳，目前没有现有基准来衡量这一差距。我们引入了韩国开放公共API基准（KOPA-Bench），该基准包含145个真实世界任务。为了解决这一问题，我们提出了EDGE，一个基于执行的动态图，用于驱动实时执行的工具调用数据合成。EDGE构建了一个图，展示每个工具的输出如何作为另一个工具的输入，并仅保留在实际调用实时API时成功的链接，通过这些经过验证的链接来合成可执行的多步骤轨迹。通过在生成的数据集上进行GRPO微调，我们的9B模型几乎与同一家族的未调优27B模型相匹配，不仅在KOPA-Bench上显著提升，在BFCL基准上也有显著改善。

View on arXiv Download PDF AI Translation

cs.AI / 106 / 2609.05396

A Deep Generative Model for Synthesizing Labeled Wireless Signals

一种用于合成标记无线信号的深度生成模型

Li, Yuxiao, Hu, Keke, Mazuelas, Santiago, Shen, Yuan

Abstract

Wireless signals with position-related labels are pivotal for both performance evaluation and model training in the realm of wireless sensing. However, acquiring real-world datasets is often challenged by significant measurement and labeling costs. Traditional methods for synthesizing labeled wireless signals typically rely on environmental models, leading to extensive hyper-parameter tuning and inadequate realism for comprehensive model training purposes. To address these limitations, we introduce a novel deep learning (DL)-based method, namely Inter-Instance Generative Adversarial Networks (IIns-GAN), to generate realistic labeled wireless signals. The generated signals are particularly adaptive to different environment scenarios and well-suited for various model training tasks, including distance estimation and environment identification. We have conducted extensive experiments on public Ultra-Wideband (UWB) datasets to evaluate the realism and utility of the generated signals. The results demonstrate that the signals generated by IIns-GAN mirror the physical characteristics of real-world measurements, and significantly contribute to the improvement of model training in diverse wireless sensing tasks.

Chinese Translation

带有位置相关标签的无线信号对于无线传感领域的性能评估和模型训练至关重要。然而，获取真实世界的数据集常常面临显著的测量和标记成本的挑战。传统的合成标记无线信号的方法通常依赖于环境模型，这导致了大量的超参数调优，并且在全面的模型训练目的上缺乏足够的真实感。为了解决这些局限性，我们提出了一种新颖的基于深度学习（DL）的方法，即实例间生成对抗网络（Inter-Instance Generative Adversarial Networks, IIns-GAN），用于生成逼真的标记无线信号。生成的信号特别适应不同的环境场景，并且非常适合于各种模型训练任务，包括距离估计和环境识别。我们在公共超宽带（Ultra-Wideband, UWB）数据集上进行了广泛的实验，以评估生成信号的真实感和实用性。结果表明，IIns-GAN生成的信号反映了真实世界测量的物理特性，并显著促进了在多样化无线传感任务中的模型训练改进。

View on arXiv Download PDF AI Translation

计算语言学 (Computation and Language)

cs.CL / 1 / 2609.04223

How Much Does Corpus Choice Change Dependency-Distance Estimates?

语料选择对依赖距离估计的影响有多大？

Chen, Sirui

Abstract

Dependency-distance estimates derived from a single corpus are routinely treated as properties of a language, yet this assumption has not been tested across independently compiled corpora. We compared mean dependency-distance estimates across 38 same-language treebank pairs from Universal Dependencies v2.18, using concordance correlation, Bland-Altman analysis, and a twelve-specification multiverse design. Cross-treebank agreement was moderate at best: substituting one treebank for another reversed nearly 40 percent of pairwise language orderings, and treebank choice accounted for roughly 29 percent of between-group variance. This disagreement substantially exceeded within-treebank sampling error and persisted across all twelve preprocessing specifications. Nevertheless, every treebank confirmed dependency-length minimization (normalized ratio below 1). The data are more consistent with MDD as a corpus-conditioned composite of grammatical, register, and annotation factors than as a stable language-level parameter: the qualitative DLM universal survives corpus substitution, but the ordinal cross-linguistic ranking does not.

Chinese Translation

从单一语料库得出的依赖距离估计通常被视为一种语言的特性，但这一假设尚未在独立编制的语料库中进行测试。我们比较了来自Universal Dependencies v2.18的38对同语言树库的平均依赖距离估计，使用了一致性相关、Bland-Altman分析以及十二种规格的多元宇宙设计。跨树库的一致性充其量是中等：用一个树库替代另一个几乎逆转了40%的成对语言排序，而树库选择大约占组间方差的29%。这种不一致显著超过了树库内部的抽样误差，并在所有十二种预处理规格中持续存在。尽管如此，每个树库都确认了依赖长度最小化（标准化比率低于1）。数据更符合MDD（最小依赖距离）作为一种受语料条件影响的复合因素，包含语法、语域和注释因素，而非作为一种稳定的语言层面参数：定性的DLM（依赖长度最小化）普遍性在语料替换中得以保留，但跨语言的序数排名则不然。

View on arXiv Download PDF AI Translation

cs.CL / 2 / 2609.04289

Memory as transformation: LETHE, a self-referential gan-inspired architecture

记忆作为转变：LETHE，一个自指的类GAN架构

Vitucci, Francesco, Di Furia, Anthony, Scagliola, Francesco

Abstract

LETHE (Latent-parameter Evolution with Temporal Hierarchical quasi-Equilibrium) is a self-referential sonic-oblivion system implemented in SuperCollider. It adopts the formal vocabulary of Generative Adversarial Networks in a closed configuration without external datasets or supervision after initialization. Audio is processed by a 3 x 3 mixing matrix built around two delay lines; its nine coefficients and two delay times evolve through the interaction of a five-feature linear discriminator and a random-perturbation optimizer analogous to single-sample REINFORCE. The discriminator compares current energy behavior with an archive of the initial state and guides parameter updates. Circular, fixed, and live sources can be mixed independently. Across fixed and circular sessions with an ablation control, the active generator is necessary for parametric evolution ($\Delta c_{22}=0.000$ in all 15 ablation sessions). Situated in the tradition of self-referential electroacoustic music, LETHE delegates the sonic outcome to an adaptive closed loop whose parametric space is defined by the composer.

Chinese Translation

LETHE（潜在参数演变与时间层次准平衡）是一个在SuperCollider中实现的自指性声音遗忘系统。它在初始化后采用生成对抗网络的形式词汇，以封闭配置运行，无需外部数据集或监督。音频通过一个围绕两个延迟线构建的3 x 3混合矩阵进行处理；其九个系数和两个延迟时间通过一个五特征线性判别器与一个类似于单样本REINFORCE的随机扰动优化器的交互而演变。判别器将当前的能量行为与初始状态的档案进行比较，并指导参数更新。循环、固定和实时源可以独立混合。在具有消融控制的固定和循环会话中，主动生成器对于参数演变是必要的（在所有15个消融会话中，$ ext{Δ} c_{22}=0.000$）。LETHE位于自指性电子音响音乐的传统中，将声音结果委托给一个自适应闭环，其参数空间由作曲家定义。

View on arXiv Download PDF AI Translation

cs.CL / 3 / 2609.04290

Evidence Integration in Large Language Models

大型语言模型中的证据整合

Kawada, Sebastien, Kellis, Manolis

Abstract

Despite increasing reliance on LLMs that reason with external evidence supplied by tools, retrieval-augmented generation, other agents, and users, how LLMs integrate such evidence into decisions they have already begun to form remains largely unclear. We present a distributional theory in which evidence shifts the receiver's distribution of initial answers, driven by a receiver prior weight and a candidate evidence tilt, leading to three predictions. First, candidates more probable to the receiver are more persuasive. Second, receivers more readily integrate characteristic errors of their own than foreign errors from different sources. Third, identical evidence can improve weaker models and harm stronger ones. We confirm these over ten million trials, twelve LLMs from four families, and eight domains, four of them scientific discovery tasks in the physical and life sciences: quantum mechanics, physics, genetics, and molecular biology. The law also yields a receiver-relative reliability frontier: receiver-congruent errors depress performance more steeply than random errors of the same rate. LLMs also integrate candidates even after internally verifying their invalidity (93-100% with propositional constraints; up to 99.4% on held-out physical and life-sciences reasoning), demonstrating evidence integration is a receiver-specific control policy over existing distributions, determined by receiver properties rather than scalar trust in the evidence source. Causal interventions show candidate integration is implemented late in the network, as a structured sequence of steps admitting external candidate answers, promoting them, and transporting them into the answer state. Representations of verification are decodable but have little causal impact on answers. A J-lens decomposition shows the state underlying verbalized verification is fully dissociable from that underlying candidate integration.

Chinese Translation

尽管对依赖外部证据进行推理的大型语言模型（LLMs）的使用日益增加，这些证据来自工具、检索增强生成、其他代理和用户，但LLMs如何将这些证据整合到他们已经开始形成的决策中仍然不清楚。我们提出了一种分布理论，其中证据会改变接收者对初始答案的分布，这一变化由接收者的先验权重和候选证据的倾斜驱动，从而导致三个预测。首先，对接收者而言，更可能的候选答案更具说服力。其次，接收者更容易整合自身的特征性错误，而非来自不同来源的外部错误。第三，相同的证据可以改善较弱的模型，却对较强的模型造成伤害。我们在超过一千万次试验中验证了这些预测，涉及来自四个家族的十二个LLMs和八个领域，其中四个是物理和生命科学中的科学发现任务：量子力学、物理学、遗传学和分子生物学。该法则还产生了一个接收者相对可靠性前沿：与接收者一致的错误对性能的抑制程度比相同比例的随机错误更为陡峭。LLMs还会整合候选答案，即使在内部验证其无效性后（在命题约束下为93-100%；在保留的物理和生命科学推理中高达99.4%），这表明证据整合是一种特定于接收者的控制策略，基于接收者的属性而非对证据来源的标量信任。因果干预表明，候选整合是在网络的后期实施的，作为一个结构化的步骤序列，允许外部候选答案的输入、促进和转移到答案状态。验证的表征是可解码的，但对答案几乎没有因果影响。J-lens分解显示，口头验证所基于的状态与候选整合所基于的状态是完全可分离的。

View on arXiv Download PDF AI Translation

cs.CL / 4 / 2609.04336

MedProb: Probing Internal Representations of Vision-Language Models for Medical Question Answering

MedProb：探究用于医学问答的视觉-语言模型的内部表征

Nourbakhsh, Erfan, Yang, Ke, Rios, Anthony

Abstract

Medical visual question answering (Med-VQA) is often assumed to require medical fine-tuning, large models, or complex multi-agent pipelines. We revisit this assumption with \textbf{MedProb}, a lightweight probing framework that predicts multiple-choice Med-VQA answers from frozen VLM representations without free-text generation. Across PATH-VQA, SLAKE, and VQA-RAD, MedProb recovers substantially more answer-relevant signal than prompting and performs stronger than medical VLMs and agentic systems. Probing also reduces the apparent gap between small and large models compared to prompting, suggesting that smaller VLMs contain more recoverable Med-VQA signal than generation-based evaluation reveals. Across 14 matched general-purpose and medical VLM pairs, medical adaptation does not consistently improve this linear decodability. Finally, free-text generation exhibits an answer-position bias of up to 10 percentage points, whereas MedProb also has positional bias, however, it is impacted differently than prompting. Our main results target the multiple-choice/multiclass Med-VQA setting; we additionally show the probe can be extended to open-ended generation via a rejection-sampling scoring procedure.

Chinese Translation

医学视觉问答（Med-VQA）通常被认为需要医学微调、大型模型或复杂的多代理管道。我们通过 extbf{MedProb}重新审视这一假设，MedProb是一个轻量级的探测框架，它能够从冻结的视觉-语言模型（VLM）表征中预测多项选择的Med-VQA答案，而无需生成自由文本。在PATH-VQA、SLAKE和VQA-RAD数据集上，MedProb恢复了比提示方法显著更多的与答案相关的信号，并且表现优于医学VLM和代理系统。与提示相比，探测还缩小了小型模型与大型模型之间的明显差距，这表明小型VLM包含的可恢复Med-VQA信号比基于生成的评估所揭示的要多。在14对匹配的通用和医学VLM中，医学适应并不总是能一致改善这种线性可解性。最后，自由文本生成表现出高达10个百分点的答案位置偏差，而MedProb也存在位置偏差，但其受影响的方式与提示不同。我们的主要结果针对多项选择/多类Med-VQA设置；我们还展示了该探测器可以通过拒绝采样评分程序扩展到开放式生成。

View on arXiv Download PDF AI Translation

cs.CL / 5 / 2609.04350

Adapting from Downturns: Prediction of Long-Term Conversational-Skill Development in Mental-Health Crisis Counselors

应对经济低迷：预测心理健康危机辅导员长期对话技能的发展

Nguyen, Vivian, Lee, Lillian, Olson, Elizabeth A., Danescu-Niculescu-Mizil, Cristian

Abstract

How do people learn to become better conversationalists? This question is especially important in the context of mental-health counseling, where conversational skills are essential, yet volunteer counselors often have limited access to supervision and structured feedback. Understanding how counselors develop their ability to steer conversations toward positive outcomes -- and identifying early which counselors are (not) on track to improve -- can help prioritize support for the counselors who need it most. In this work, we introduce the task of predicting, early in a conversationalist's career, whether they will eventually improve at steering conversations toward positive outcomes, and demonstrate the feasibility of this task in the case of volunteer mental-health crisis counselors. Our central insight is that people may struggle with particular kinds of moments in a conversation, and that what is especially revealing of their likelihood of future improvement is how they learn to handle those moments over time. We operationalize this insight by designing a method that identifies the types of moments a counselor initially struggles with, captures how they adapt their response when they re-encounter similar moments in subsequent conversations, and learns which early adaptations predict improvement months or even years later. While this future-prediction task is challenging, our counselor-adaptation approach yields better results than baselines that learn directly from the conversation transcript.

Chinese Translation

人们如何学习成为更好的对话者？这个问题在心理健康辅导的背景下尤为重要，因为对话技能至关重要，而志愿辅导员通常很难获得监督和结构化反馈。理解辅导员如何发展引导对话朝向积极结果的能力，以及早期识别哪些辅导员在改善的轨道上（或不在轨道上），可以帮助优先支持那些最需要帮助的辅导员。在本研究中，我们提出了一项任务，即在对话者职业生涯的早期预测他们是否最终会改善引导对话朝向积极结果的能力，并展示了在志愿心理健康危机辅导员案例中这一任务的可行性。我们的核心见解是，人们可能在对话中的特定时刻遇到困难，而特别能揭示他们未来改善可能性的是他们如何随着时间的推移学习处理这些时刻。我们通过设计一种方法来实现这一见解，该方法识别辅导员最初挣扎的时刻类型，捕捉他们在后续对话中重新遇到类似时刻时如何调整反应，并学习哪些早期调整能够预测几个月甚至几年后的改善。尽管这一未来预测任务具有挑战性，但我们的辅导员适应方法的结果优于直接从对话记录中学习的基线。

View on arXiv Download PDF AI Translation

cs.CL / 6 / 2609.04366

VERGE: Verification-Enhanced Refinement for Grounded Extraction of Early-Onset Colorectal Cancer Symptoms in Clinical Notes

VERGE：基于验证增强的早发结直肠癌症状提取的精细化方法

Hooman, Nikkie, Nigam, Monarch, Hughes, Amy E., Nair, Rasmi G., Gupta, Mehak

Abstract

Early-onset colorectal cancer is increasing among younger adults, yet red-flag symptoms in this age group have no evidence-based guidelines for follow-up testing, and structured encounter data do not capture the detail needed to support early detection and inform follow-up, including symptom duration, context, and fam- ily history, an established colorectal-cancer risk factor. This study aimed to develop and evaluate an automated method for extracting six red-flag symptoms and family-history risk status from free-text clinical notes. We developed VERGE, an agentic workflow in which an initial label and evidence are proposed using retrieval-augmented generation, then passed through a bounded verification- refinement cycle that checks textual grounding and clinical validity, corrects and rechecks a claim until resolved or a limit is reached, and escalates unresolved claims for human review. VERGE was evaluated on 4,033 clinician-labeled note-finding pairs against a single-agent baseline, a rule-based clinical language-processing baseline, and an alternative underlying language model. Compared with the single-agent baseline, VERGE reduced false positive find- ings, improving precision from 0.764 to 0.849 and MCC from 0.681 to 0.730, a balanced gain across the precision-recall trade-off, and resolved most flagged errors autonomously, with human review required for only 1.5 percent of claims. These results indicate that a bounded, verification-based workflow can reduce unnecessary positive findings without sacrificing the ability to detect true ones. This approach offers a path toward more reliable and trustworthy clinical language-processing tools to support colorectal cancer risk assessment in younger patients.

Chinese Translation

早发结直肠癌在年轻成年人中的发病率正在上升，但该年龄组的红旗症状尚无基于证据的后续检测指南，结构化就诊数据也未能捕捉到支持早期发现和后续信息所需的细节，包括症状持续时间、背景和家族史，而家族史是已知的结直肠癌风险因素。本研究旨在开发和评估一种自动化方法，从自由文本临床记录中提取六种红旗症状和家族史风险状态。我们开发了VERGE，这是一种代理工作流，其中使用检索增强生成提出初始标签和证据，然后通过一个有限的验证-精细化循环进行处理，该循环检查文本的基础和临床有效性，纠正并重新检查声明，直到解决或达到限制，并将未解决的声明升级为人工审核。VERGE在4,033个临床医生标记的记录-查找对上进行了评估，与单一代理基线、基于规则的临床语言处理基线和替代基础语言模型进行比较。与单一代理基线相比，VERGE减少了假阳性发现，将精确度从0.764提高到0.849，MCC从0.681提高到0.730，在精确度-召回权衡中实现了平衡增益，并且大多数标记错误都能自主解决，仅有1.5%的声明需要人工审核。这些结果表明，基于验证的有限工作流可以减少不必要的阳性发现，而不牺牲检测真实发现的能力。这种方法为开发更可靠和可信的临床语言处理工具提供了一条途径，以支持年轻患者的结直肠癌风险评估。

View on arXiv Download PDF AI Translation

cs.CL / 7 / 2609.04384

You Really Didn't Get That? Benchmarking Social Pragmatic Inference for Indirect and Playful Chinese Online Comments

你真的没有理解吗？对间接和戏谑性中文在线评论的社会语用推理基准测试

Hong, Shiwei, Ma, Junjie, Wang, Emma Jiren, Rong, Ethan Z., Hu, Siying, Li, Haichang, Wang, Ziying, Lu, Zhicong

Abstract

Chinese online comments often convey social meaning through indirect and playful language that is hard to interpret without context. Existing evaluations largely organize items around predefined phenomena or controlled pragmatic categories, leaving open whether models can distinguish plausible readings of what a naturally occurring comment is doing in a particular exchange. We introduce a benchmark for evaluating whether LLMs can recover such situated pragmatic meanings. From more than 200,000 public Chinese social media interaction records, we construct 4,735 human-validated diagnostic items, each pairing a target comment with reconstructed preceding context and plausible misreadings. We evaluate eight LLMs as both question writers and solvers in a cross-writer setting. The task is challenging: the strongest model achieves 81.42% leave-writer-out accuracy. Across all eight models, the mean leave-writer-out accuracy is 68.70% while human accuracy was 90.8%. Case analysis shows that models often recognize broad irony or playfulness while misidentifying the mechanism or interactional move.

Chinese Translation

中文在线评论常通过间接和戏谑性的语言传达社会意义，这在没有上下文的情况下难以解读。现有评估主要围绕预定义现象或控制的语用类别组织项目，尚未明确模型是否能够区分自然发生的评论在特定交流中所传达的合理解读。我们引入了一个基准，用于评估大型语言模型（LLMs）是否能够恢复这种情境下的语用意义。我们从超过200,000条公共中文社交媒体互动记录中构建了4,735个经过人工验证的诊断项目，每个项目将目标评论与重构的前置上下文和合理的误读配对。我们在跨作者设置中评估了八个LLMs，作为问题编写者和解答者。该任务具有挑战性：最强模型的留作者准确率达到81.42%。在所有八个模型中，平均留作者准确率为68.70%，而人类准确率为90.8%。案例分析表明，模型通常能够识别广泛的讽刺或戏谑，但在识别机制或互动行为时常出现误判。

View on arXiv Download PDF AI Translation

cs.CL / 8 / 2609.04391

Evaluation of Phonetic Encoding Algorithms on Transcription Datasets

音素编码算法在转录数据集上的评估

Özbey, Can, Kaplan, Emre, Kahya, Berkin Deniz

Abstract

In this work, a novel evaluation scheme built on a generalized variant of the Rand Index measure, namely, the H\"ullermeier-Rifqi Index, is proposed in order to assess how well phonetic encoding algorithms conform to word-based transcriptions in IPA (International Phonetic Alphabet) notation. For this objective, the discordance score is obtained by calculating the absolute difference between the pairwise similarity values of ground-truth transcriptions and those of corresponding phonetic encodings, which are computed using normalized edit distance as a permutation dependent string metric. The resulting score is subsequently adjusted with respect to that of a random string generator incorporating the same alphabet as the encoder under consideration. A wide range of phonetic encoders were evaluated as such on multi-lingual transcription datasets along with their recall capabilities based on the collision rate. The validity of the proposed scheme is further supported by its applicability in measuring the orthographic transparency of a language when the writing system is viewed as an inherent phonetic representation.

Chinese Translation

在本研究中，提出了一种基于Rand指数度量的广义变体的新评估方案，即H"ullermeier-Rifqi指数，以评估音素编码算法与国际音标（IPA）标注的基于词的转录的一致性。为此，计算真实转录与相应音素编码的成对相似值之间的绝对差异，以获得不一致性得分，这些相似值是使用归一化编辑距离作为依赖于排列的字符串度量计算的。随后，得到的得分相对于采用与所考虑编码器相同字母表的随机字符串生成器的得分进行了调整。对多语言转录数据集上的多种音素编码器进行了评估，并基于碰撞率分析了它们的召回能力。所提方案的有效性进一步得到了支持，因为它在将书写系统视为固有音素表示时，能够测量语言的正字法透明度。

View on arXiv Download PDF AI Translation

cs.CL / 9 / 2609.04404

The Anatomy of an ASR Hallucination

ASR幻觉的解剖

Sayed, Hamees, Singh, Apoorv, Aman, Kumar, Mandloi, Akshat

Abstract

ASR systems sometimes produce fluent text that is unrelated to the speech they receive. We view these hallucinations as one possible consequence of a broader grounding failure, in which the transcript is no longer adequately guided by the audio. To understand where this failure becomes possible, we study two independently trained Conformer-Large recognizers - one CTC and one RNN-T - under environmental degradation and speaker-background shift. In both models, the final encoder stage emerges as a critical boundary: bypassing the final block causes divergence on nearly every utterance, whereas bypassing middle blocks has little effect. At this same stage, the representations become more compact, text becomes readable by the trained decoder, and grapheme information becomes explicit. Importantly, the intervention produces garbled or repetitive output rather than fluent fabrication. Our result therefore identifies a mechanistic precondition for hallucination - the failure to produce adequately grounded output - not the complete origin of naturally occurring hallucinations. Together, the results reveal a consistent terminal-stage dependency for grounded recognition across two decoder families and multiple distribution shifts.

Chinese Translation

ASR系统有时会生成与接收到的语音无关的流畅文本。我们将这些幻觉视为更广泛的基础失效的一种可能后果，在这种情况下，转录文本不再受到音频的充分指导。为了理解这种失效何以成为可能，我们研究了两个独立训练的Conformer-Large识别器——一个是CTC，另一个是RNN-T——在环境退化和说话者背景变化下的表现。在这两个模型中，最终编码器阶段成为一个关键的边界：绕过最后一个模块几乎导致每个发声的偏离，而绕过中间模块几乎没有影响。在这个阶段，表示变得更加紧凑，文本变得可被训练过的解码器读取，字素信息变得明确。重要的是，这种干预产生的是混乱或重复的输出，而不是流畅的虚构。因此，我们的结果识别出幻觉的一个机械性前提——未能产生充分基础的输出——而不是自然发生的幻觉的完整起源。综合来看，这些结果揭示了在两个解码器家族和多种分布变化下，基础识别的一致终端阶段依赖性。

View on arXiv Download PDF AI Translation

cs.CL / 10 / 2609.04409

A Systematic Evaluation of Cross-Lingual Consistency Enhancement Methods in Multilingual Language Models

多语言模型中跨语言一致性增强方法的系统评估

Qi, Jirui, Wang, Mingyang, Schütze, Hinrich, Fernández, Raquel, Bisazza, Arianna

Abstract

Multilingual language models often produce inconsistent answers to semantically equivalent questions across languages, motivating methods to improve cross-lingual consistency (CLC). However, existing methods are typically evaluated using different models, tasks, and protocols, leaving their relative strengths unclear. In this work, we present a unified evaluation of representative CLC-enhancement methods for question answering, spanning inference-time interventions and post-training approaches across three model families and three closed-form benchmarks. The results show that post-training methods are generally more reliable, with direct distribution alignment consistently improving CLC across all model-dataset combinations, while other methods are more sensitive to answer format and the breadth of language coverage. Notably, cross-domain transfer is limited unless source and target tasks share similar output formats. We further investigate whether CLC enhancement hurts models' ability to respond differently *when needed*, that is, when asked culture-dependent questions. Across two benchmarks of culturally diverse question answering, we find no systematic degradation in controlled closed-form evaluation, whereas open-ended generation reveals occasional accuracy reductions, particularly for non-English responses. Our work highlights the need to evaluate CLC enhancement for both cross-domain robustness and culturally appropriate variation, informing future work in post-training and benchmark development.

Chinese Translation

多语言模型在不同语言中对语义等价的问题常常产生不一致的回答，这促使了改善跨语言一致性（CLC）的方法的研究。然而，现有方法通常使用不同的模型、任务和评估协议进行评估，使得它们的相对优缺点不明确。在本研究中，我们对代表性的CLC增强方法进行了统一评估，涵盖了推理时干预和后训练方法，涉及三种模型家族和三个封闭形式基准。结果表明，后训练方法通常更可靠，直接的分布对齐在所有模型-数据集组合中一致地改善了CLC，而其他方法对回答格式和语言覆盖范围更为敏感。值得注意的是，跨领域迁移受到限制，除非源任务和目标任务共享相似的输出格式。我们进一步研究了CLC增强是否会损害模型在需要时（即在被问及文化依赖性问题时）作出不同回应的能力。在两个文化多样性问题回答的基准测试中，我们发现控制的封闭形式评估没有系统性退化，而开放式生成则偶尔出现准确性下降，尤其是在非英语回应中。我们的研究强调了评估CLC增强在跨领域鲁棒性和文化适当变异方面的必要性，为未来的后训练和基准开发工作提供了参考。

View on arXiv Download PDF AI Translation

cs.CL / 11 / 2609.04434

What Attention Recalls and Recurrence Controls in Hybrid Language Models

混合语言模型中注意力的回忆与递归的控制

Afendulev, Kirill, Dontsov, Alexey, Tutubalina, Elena, Korznikov, Anton

Abstract

Hybrid language models combine attention with a fixed-size recurrent state, but the role of each channel remains unclear. We introduce two cache-level interventions. Split-prefill keeps only the KV cache or only the recurrent state from a prefilled context, then generates an answer. State-swap pairs the KV cache from one context with the recurrent state from another in a single forward pass. On Qwen3.5 and Falcon-H1, the two channels split sharply by function. Exact retrieval survives only through attention (64-98% of full accuracy) and collapses to zero through recurrence. Output language and persona reverse the pattern: both survive recurrence (70-80% and 3-5x) while KV-only drops to ~1% language accuracy. State-swap confirms this causally: the answer takes its value from the KV side and its language from the recurrent side. Recurrent-only generation also accepts words that were never in the context but share meaning or parts with seen items. Attention provides a lookup over what was said; the recurrent state shapes how the model says it next.

Chinese Translation

混合语言模型将注意力与固定大小的递归状态结合在一起，但每个通道的作用仍不清楚。我们引入了两种缓存级别的干预措施。Split-prefill 仅保留来自预填充上下文的 KV 缓存或递归状态，然后生成答案。State-swap 将来自一个上下文的 KV 缓存与来自另一个上下文的递归状态配对，在单次前向传递中进行处理。在 Qwen3.5 和 Falcon-H1 上，这两个通道的功能明显分化。精确检索仅通过注意力得以维持（64-98% 的完整准确率），而通过递归则崩溃至零。输出语言和角色则反转了这一模式：两者都在递归中存活（70-80% 和 3-5 倍），而仅 KV 的准确率降至 ~1%。State-swap 从因果关系上证实了这一点：答案的值来自 KV 侧，而语言来自递归侧。仅递归生成也接受那些从未出现在上下文中的单词，但与已见项目共享意义或部分。注意力提供了对所说内容的查找；递归状态则塑造了模型接下来如何表达。

View on arXiv Download PDF AI Translation

cs.CL / 12 / 2609.04442

GRACE: Graph-Grounded Reflective Agent Copilot Engine for Expert-in-the-Loop Knowledge Expansion

GRACE：基于图的反思代理助手引擎用于专家参与的知识扩展

Yi, John Seon Keun, Minot, Joshua R., Lee, Dokyun

Abstract

Large language models deployed in high-stakes settings frequently generate plausible but ungrounded claims. Standard retrieval-augmented generation (RAG) pipelines offer limited remedy, since they retrieve isolated passages without tracking cross-document evidence relationships or quantifying uncertainty. We introduce GRACE (Graph-grounded Reflective Agent Copilot Engine), a framework that deconstructs LLM responses into atomic claims and grounds them against trusted knowledge priors within a weighted bipartite graph. Edge weights encode the closeness of each claim to the priors, enabling weighted centrality analysis that classifies claims as Grounded, Refuted, or Boundary. Such classification identifies not just hallucinations but also novel or contested claims at the frontier of the model's knowledge. To efficiently allocate human or agent resources, we formulate a Return on Attention (RoA) objective that defers a claim to expert review only when its priority-weighted uncertainty exceeds the cost of verification. Claims verified by experts are promoted to new evidence anchors, closing a validator-LLM evolutionary loop that expands the knowledge base across iterations. We evaluate GRACE across multiple language models and on datasets spanning both general and domain-specific knowledge. Our results show that our knowledge base serves as a reliable foundation for retrieval that outperforms RAG baselines, and that the RoA framework efficiently selects valuable boundary knowledge for expert verification. These findings demonstrate that graph-structured representations combined with expert-in-the-loop verification can mitigate hallucination at the system level rather than at the generation level. Code available at https://github.com/johnsk95/grace_code

Chinese Translation

在高风险环境中部署的大型语言模型常常生成看似合理但缺乏依据的陈述。标准的检索增强生成（RAG）流程提供的解决方案有限，因为它们检索的是孤立的段落，而未能追踪跨文档的证据关系或量化不确定性。我们提出了GRACE（基于图的反思代理助手引擎），这是一个将大型语言模型（LLM）响应解构为基本陈述并将其与加权二分图中的可信知识先验进行对照的框架。边权重编码了每个陈述与先验之间的接近程度，使得加权中心性分析能够将陈述分类为“有根据的”、“被驳斥的”或“边界”。这种分类不仅识别出幻觉，还识别出模型知识边界上的新颖或有争议的陈述。为了有效分配人力或代理资源，我们制定了一个注意力回报（RoA）目标，仅在某个陈述的优先权重不确定性超过验证成本时，才将其推迟至专家审查。经过专家验证的陈述被提升为新的证据锚点，形成一个验证者与大型语言模型之间的进化循环，跨迭代扩展知识库。我们在多个语言模型上以及涵盖一般和特定领域知识的数据集上评估了GRACE。我们的结果表明，我们的知识库作为检索的可靠基础，优于RAG基线，并且RoA框架有效选择有价值的边界知识进行专家验证。这些发现表明，结合专家参与验证的图结构表示可以在系统层面而非生成层面减轻幻觉问题。代码可在 https://github.com/johnsk95/grace_code 获取。

View on arXiv Download PDF AI Translation

cs.CL / 13 / 2609.04452

TRILOGUE: A Trilingual Spoken Dialogue Fact-Checking Benchmark with Evidence and Paired Audio

TRILOGUE：一个包含证据和配对音频的三语口语对话事实核查基准

Chun, Chaewan, Aristombayeva, Meruyert, Choi, Jiyoung, Nahar, Mahjabin, Zhang, Delvin Ce, Lee, Dongwon

Abstract

Modern misinformation is often heard before it is read, yet fact-checking systems are still evaluated mainly on clean written claims. Spoken dialogue remains different even when systems operate on transcripts: claims may be distributed across speakers and turns, depend on prior context, and become harder to verify when Automatic Speech Recognition (ASR) errors distort the available text. Prior spoken dialogue fact-checking resources are small, English-centric, or focused on annotation rather than end-to-end benchmarking, leaving no large multilingual benchmark with paired speech and turn-level labels. We introduce TRILOGUE (TRIlingual spoken diaLOGUE fact-checking), a large-scale trilingual benchmark of source-grounded spoken dialogues in English, Russian, and Kazakh. It contains nearly 12K dialogues, 187K turns, and 390 hours of paired audio with ASR transcripts and word-level timestamp alignments across all three languages, including nearly 5K human-recorded Russian and Kazakh dialogue files. TRILOGUE supports claim check-worthiness detection, source-article evidence retrieval, and claim verification with claim-only, gold-evidence, and retrieved-evidence inputs. Baselines show that ASR degradation and cross-lingual transfer remain challenging, especially for Kazakh, while retrieved source evidence substantially narrows the gap to gold-evidence verification.

Chinese Translation

现代错误信息通常在被阅读之前就已经被听到，然而事实核查系统仍然主要在干净的书面声明上进行评估。即使系统在转录文本上操作，口语对话仍然存在差异：声明可能分布在不同的发言者和轮次之间，依赖于先前的上下文，并且在自动语音识别（ASR）错误扭曲可用文本时，验证变得更加困难。先前的口语对话事实核查资源规模较小，主要集中于英语，或专注于注释而非端到端的基准测试，缺乏大型多语言基准，且没有配对的语音和轮次级标签。我们介绍了TRILOGUE（TRIlingual spoken diaLOGUE fact-checking），这是一个大规模的三语基准，涵盖英语、俄语和哈萨克语的基于来源的口语对话。它包含近12K个对话、187K个轮次，以及390小时的配对音频，配有ASR转录文本和三种语言的单词级时间戳对齐，包括近5K个人工录制的俄语和哈萨克语对话文件。TRILOGUE支持声明的核查价值检测、来源文章证据检索，以及使用仅声明、黄金证据和检索证据输入的声明验证。基准结果表明，ASR降级和跨语言转移仍然具有挑战性，尤其是对于哈萨克语，而检索到的来源证据显著缩小了与黄金证据验证之间的差距。

View on arXiv Download PDF AI Translation

cs.CL / 14 / 2609.04463

Shared circuits predict whether LLMs generalize across formats in arithmetic reasoning

共享电路预测大型语言模型在算术推理中是否能够跨格式泛化

de Varda, Andrea Gregor, Pandey, Sana, Han, Pengrui, Andreas, Jacob, Fedorenko, Evelina

Abstract

In many forms of reasoning, including arithmetic reasoning, generalizing across superficial changes in input format is effortless for humans: anyone who can solve 2+5 can also solve 'two plus five'. In contrast, LLMs are more brittle to surface variations of the prompts: for example, they solve numeric arithmetic problems almost perfectly but are substantially less accurate on verbal renditions of the same problems. Here, we ask whether generalization across formats can be predicted from the models' internals. Using attribution patching, we first independently localize the circuit that each model recruits to solve numeric arithmetic problems (2+5) vs. verbal ones, in three languages: English ('two plus five'), Spanish ('dos m\'as cinco'), and Italian ('due pi\`u cinque'); then, we test whether overlap with the model's own numeric circuit predicts its generalization to the verbal formats. Indeed, we find support for this idea at three levels: circuit overlap accounts for the relative difficulty of the three verbal formats, for which models generalize best, and for which items are solved correctly, rivaling supervised probes while requiring no labeled data.

Chinese Translation

在许多推理形式中，包括算术推理，人类在输入格式的表面变化中进行泛化是轻而易举的：任何能够解决 2+5 的人也能够解决 'two plus five'。相比之下，大型语言模型（LLMs）对提示的表面变化更为脆弱：例如，它们几乎完美地解决数字算术问题，但在同一问题的语言表达上准确性显著降低。在这里，我们探讨是否可以从模型的内部结构预测跨格式的泛化。通过归因修补，我们首先独立定位每个模型在解决数字算术问题（2+5）与语言问题时所使用的电路，研究三种语言：英语（'two plus five'）、西班牙语（'dos más cinco'）和意大利语（'due più cinque'）；然后，我们测试模型自身的数字电路与语言格式的重叠是否能够预测其对语言格式的泛化。确实，我们在三个层面上发现了对这一观点的支持：电路重叠解释了三种语言格式的相对难度，模型在这些格式上泛化最佳，以及哪些项目被正确解决，这一结果可与监督探测相媲美，同时不需要标记数据。

View on arXiv Download PDF AI Translation

cs.CL / 15 / 2609.04482

Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal

安全性为谁而设？边界感知自蒸馏用于控制大型语言模型的安全拒绝

López-Ávila, Alejo, García-Ferrero, Iker, Garcia, Jezabel, Tiene, Antonio, Orús, Román

Abstract

Safety alignment is usually posed as a topic-level question: is this subject harmful? Deployments ask a narrower one. A civics tutor and a public-sector assistant may share a base model yet need different boundaries inside the same topic, refusing targeted political manipulation while still answering factual questions about the same election. We formulate this as narrow-boundary safety and introduce an offline self-generated framework combining controlled topic generation, coverage repair, in-distribution compensation data, and harmful-benign pairs for training and evaluation. Single-shot generation leaves 19.88% of prompts without accepted refusal traces, whereas escalating retries leave 0.20%. On political persuasion with Qwen3-8B, training on refusal data completed through Escalate increases target-domain refusal from 9.47% to 84.75% and reduces the mean unsafe-response rate across three broader harmfulness benchmarks from 26.26% to 0.14%, but increases XSTest over-refusal from 2.00% to 74.00%. In a separate matched comparison, replacing external responses with verified target-model responses reduces over-refusal from 15.20% to 5.20%. Boundary-pair data reduces comply-side over-refusal on held-out pairs from 32.94% to 4.16%, while harmful-side refusal decreases only from 91.88% to 87.72%. These results show that data composition controls the safety and usability trade-off, and that safety alignment should be evaluated on both sides of the intended refusal boundary.

Chinese Translation

安全对齐通常被视为一个主题层面的问题：这个主题是否有害？而实际部署则提出了一个更为狭窄的问题。公民教育导师和公共部门助手可能共享一个基础模型，但在同一主题内需要不同的边界，拒绝针对性的政治操控，同时仍需回答关于同一选举的事实性问题。我们将其表述为狭边界安全，并引入一个离线自生成框架，结合受控主题生成、覆盖修复、分布内补偿数据和有害-良性对用于训练和评估。单次生成留下19.88%的提示没有被接受的拒绝痕迹，而逐步重试则留下0.20%。在使用Qwen3-8B进行政治劝说时，通过Escalate完成的拒绝数据训练使目标领域的拒绝率从9.47%提高到84.75%，并将三个更广泛有害性基准的平均不安全响应率从26.26%降低到0.14%，但将XSTest的过度拒绝率从2.00%提高到74.00%。在一个单独的匹配比较中，用经过验证的目标模型响应替换外部响应将过度拒绝率从15.20%降低到5.20%。边界对数据在保留对上将合规方的过度拒绝率从32.94%降低到4.16%，而有害方的拒绝率仅从91.88%降低到87.72%。这些结果表明，数据组成控制了安全性与可用性之间的权衡，而安全对齐应在预期拒绝边界的两侧进行评估。

View on arXiv Download PDF AI Translation

cs.CL / 16 / 2609.04484

Patterns of Priming in Production: Lexical, Semantic and Structural Alignment in Language Model Generation

生成中的启动模式：语言模型生成中的词汇、语义和结构对齐

Pucci, Giulia, Li, Ruizhe, Sinclair, Arabella

Abstract

This paper investigates structural priming in language model (LM) production, examining how preceding structural context influences sentence completion. While prior work has demonstrated priming effects in comprehension of structural alternations, it remained unclear whether these persist in production, where, when generating, an LM samples from many possible continuations at each step. We address this question through a series of controlled sentence-completion experiments on dative constructions. In line with prior work, we find that LMs are susceptible to structural priming, particularly in sentences that are semantically coherent. In terms of priming magnitude, we find that while there is a greater relative increase of double-object datives against our baselines, in line with inverse frequency effects, there is a larger absolute increase in prepositional-objects, the more frequently produced construction. Finally, we not only observe that structural priming is boosted by lexico-semantic coherence, but that structurally primed completions display greater levels of lexico-semantic repetition. Taken together, our evidence supports the view that structural priming in LMs operates across multiple levels of linguistic representation, facilitating, and facilitated by syntactic, lexical, and semantic alignment. Code: https://github.com/the-context-lab/primedproduction.

Chinese Translation

本文研究了语言模型（LM）生成中的结构启动，考察了先前的结构上下文如何影响句子完成。尽管先前的研究已经证明了结构变换理解中的启动效应，但尚不清楚这些效应是否在生成中持续存在，因为在生成过程中，LM在每一步从许多可能的延续中进行采样。我们通过一系列关于双宾语结构的控制句子完成实验来解决这个问题。与先前的研究一致，我们发现LM对结构启动敏感，特别是在语义上连贯的句子中。在启动幅度方面，我们发现，尽管双宾语结构相对于我们的基线有更大的相对增加，但与逆频率效应一致，介词宾语的绝对增加更大，因为它是更常被生成的结构。最后，我们不仅观察到结构启动受到词汇-语义连贯性的增强，而且结构启动的完成表现出更高水平的词汇-语义重复。综合来看，我们的证据支持这样的观点：LM中的结构启动在多个语言表征层面上运作，促进并受到句法、词汇和语义对齐的影响。代码：https://github.com/the-context-lab/primedproduction。

View on arXiv Download PDF AI Translation

cs.CL / 17 / 2609.04485

Cultural Misalignment in Large Language Models: Detection, Measurement, and Mitigation Through Targeted Fine-Tuning

大型语言模型中的文化不对齐：通过针对性微调的检测、测量与缓解

Czolgowski, Antoni, Iyasele, Abel

Abstract

We evaluate three open-weight LLMs (Gemma3-12B from the USA, Bielik-11B-v3 from Poland, and Qwen3-4B from China) against World Values Survey Wave 7 data for 63 demographic personas across three countries, using normalized Wasserstein distance to quantify distributional misalignment. Contrary to expectations, no model favors its home country: the Chinese-built Qwen3-4B performs worst on its own Chinese population (W1 = 0.436, the highest misalignment in the entire model x country matrix). Targeted LoRA fine-tuning on the five worst-case personas, requiring fewer than 1,200 training pairs and under 15 minutes on a single GPU, reduces bias by 16.8% for Bielik-11B (p_Bonf = 0.002, d = -4.4) with all five targets improving. However, country-level decomposition reveals that fine-tuning redistributes rather than removes bias: Bielik's worst-case personas swap entirely from American to Chinese elderly, with zero overlap between pre- and post-correction sets. To our knowledge, this is the first study to target worst-case demographic personas with LoRA fine-tuning for cross-cultural bias mitigation.

Chinese Translation

我们评估了三种开放权重的大型语言模型（LLMs）：来自美国的Gemma3-12B、来自波兰的Bielik-11B-v3和来自中国的Qwen3-4B，基于世界价值观调查第七波数据，针对三个国家的63个人口特征进行分析，使用归一化的Wasserstein距离量化分布不对齐。与预期相反，没有模型偏向其本国：由中国构建的Qwen3-4B在其本国人口中表现最差（W1 = 0.436，为整个模型与国家矩阵中的最高不对齐）。针对五个最差案例特征进行的LoRA微调，所需训练对少于1200对，且在单个GPU上耗时不到15分钟，使Bielik-11B的偏见减少了16.8%（p_Bonf = 0.002，d = -4.4），所有五个目标均有所改善。然而，国家级分解揭示微调是重新分配而非消除偏见：Bielik的最差案例特征完全从美国老年人转变为中国老年人，前后修正集之间没有重叠。据我们所知，这是首个针对最差案例人口特征进行LoRA微调以缓解跨文化偏见的研究。

View on arXiv Download PDF AI Translation

cs.CL / 18 / 2609.04489

Towards Understanding Pause Token Fine-Tuning Dynamics: A Mode Retention Perspective

理解暂停标记微调动态：一种模式保留视角

Kim, Jaehyeon, Kim, Suhwan, Lee, Nakyung, Kim, Yeongoon, Seo, Jimin, Lee, Giho, Lee, Jungwoo

Abstract

Pause-token methods improve LLM reasoning by inserting special tokens into sequences. Prior work explains these gains through computational expressivity. However, there is relatively little investigation into the training dynamics of pause tokens. We explore how pause tokens reshape the training dynamics of fine-tuning. Two controlled pilots expose distinct asymmetries. On a synthetic continual-learning task, masked pauses overwrite a previously-learned distribution roughly 4x less at matched final adaptation (H1, mode retention); on a synthetic math-reasoning probe, the boundary-adjacent token comes to encode substantially more downstream-step information (H2, non-myopic compression). We formalize a training rule consistent with both - Masked Boundary Pause (MBP), pause tokens placed at reasoning-step boundaries with their loss masked. Across 1B-8B Qwen and Llama models, MBP consistently improves reasoning, achieving gains of up to 6 points on math and 2.5 points on code, while preserving general language understanding abilities. We further demonstrate that this mode-preserving strategy extend gains to GRPO. These results recast pause tokens as a training-dynamics intervention on the retention-adaptation trade-off, rather than merely an inference-time computation device.

Chinese Translation

暂停标记方法通过在序列中插入特殊标记来改善大型语言模型（LLM）的推理能力。之前的研究通过计算表达能力解释了这些提升。然而，对于暂停标记的训练动态的研究相对较少。我们探讨了暂停标记如何重塑微调的训练动态。两个受控实验揭示了明显的不对称性。在一个合成的持续学习任务中，掩蔽的暂停在匹配的最终适应下，覆盖先前学习的分布的程度大约减少了4倍（假设1，模式保留）；在一个合成的数学推理探测中，边界邻近的标记编码了显著更多的下游步骤信息（假设2，非短视压缩）。我们形式化了一条与两者一致的训练规则——掩蔽边界暂停（Masked Boundary Pause, MBP），即在推理步骤边界处放置暂停标记，并掩蔽其损失。在1B-8B的Qwen和Llama模型中，MBP始终改善推理能力，在数学上取得了最高6分的提升，在代码上取得了2.5分的提升，同时保持了通用语言理解能力。我们进一步证明，这种模式保留策略将收益扩展到GRPO。这些结果将暂停标记重新定义为一种训练动态干预，影响保留与适应的权衡，而不仅仅是一个推理时的计算设备。

View on arXiv Download PDF AI Translation

cs.CL / 19 / 2609.04511

LentEx: Generalizable Latent Entity Extraction via Synthetic Data and Instruction-Tuned LLMs

LentEx：通过合成数据和指令调优的通用潜在实体提取

Bodhwani, Umesh, Ling, Yuan, Senthilkumar, Cibi Chakravarthy, Dong, Shujing, Feng, Yarong, Li, Hongfei, Goyal, Ayush

Abstract

Latent entity extraction (LEE) tackles the challenge of identifying implicit, contextually inferred entities within free text-an area where traditional entity extraction methods fall short. In this paper, we introduce LentEx, a novel framework for latent entity extraction that leverages synthetic data generation and instruction fine-tuning to optimize smaller, efficient large language models (LLMs). Latent entities, which are often abstract and thematic, are crucial for applications such as retrieval-augmented generation (RAG), customer persona analysis, and knowledge graph enrichment. LentEx addresses the scarcity of labeled datasets by employing a template-based approach to generate diverse, contextually rich synthetic data, ensuring high variability and alignment with real-world distributions. To our knowledge, LentEx is the first to systematically approach LEE through the lens of LLMs. LentEx demonstrates significant performance improvements across multiple tasks, notably surpassing state-of-the-art models on the MTEB Clustering Benchmark. Furthermore, our methodology enables robust generalization to unseen domains, making LentEx highly applicable in real-world NLP tasks, including RAG and clustering, thereby establishing a new paradigm for latent entity understanding and extraction in natural language processing.

Chinese Translation

潜在实体提取（LEE）旨在解决在自由文本中识别隐含的、上下文推断的实体这一挑战，而传统的实体提取方法在这一领域表现不佳。本文介绍了LentEx，一个新颖的潜在实体提取框架，利用合成数据生成和指令微调来优化较小、高效的大型语言模型（LLMs）。潜在实体通常是抽象和主题性的，对于检索增强生成（RAG）、客户角色分析和知识图谱丰富等应用至关重要。LentEx通过采用基于模板的方法生成多样化、上下文丰富的合成数据，解决了标注数据集稀缺的问题，确保了高变异性和与现实世界分布的一致性。据我们所知，LentEx是首个从LLMs的角度系统性地处理LEE的问题。LentEx在多个任务上表现出显著的性能提升，尤其是在MTEB聚类基准测试中超越了最先进的模型。此外，我们的方法论使得在未见领域的稳健泛化成为可能，使LentEx在现实世界的自然语言处理任务中具有高度的适用性，包括RAG和聚类，从而为自然语言处理中的潜在实体理解和提取建立了新的范式。

View on arXiv Download PDF AI Translation

cs.CL / 20 / 2609.04526

Scale-QLoRA: Code-Invariant Adapter Merging for Native 4-bit Microscaling LLMs

Scale-QLoRA：用于原生4位微缩大语言模型的代码不变适配器合并

Li, Tung-Ling, Huang, Jiale, Wang, Lee-Chi, Gotei, Janaki Ram

Abstract

Merging a LoRA adapter into its base model is standard deployment practice: it removes the runtime adapter's per-forward overhead and leaves a single standalone checkpoint any serving stack can load. On a native 4-bit microscaling checkpoint (NVFP4, MXFP4) that step stops being free. The merged weights must be written back through a quantizer, which re-derives the checkpoint's discrete E2M1 code plane (roughly 90% of the artifact's bytes), so the deployed artifact becomes coupled to one quantization convention, and every later code-touching event in its lifecycle can move it. Done naively the step is worse than fragile: it deletes the adaptation, by up to 39 pp, because against an already-on-grid base the reconstruction optimum is that base. Scale-QLoRA instead adapts only the native per-block scale field, trains those scales on the deployment grid, and freezes every E2M1 code. Within a fixed native format, scale grid, block layout and code plane, merging is then a bit-exact identity and the merged artifact is code-invariant. Across four models and four tasks, Scale-QLoRA and merge-aware QAT-LoRA are both accuracy-lossless, so we claim no accuracy ordering between them; they differ structurally, in that QAT-LoRA re-derives the code plane through a quantizer while Scale-QLoRA preserves it exactly. That difference is what the lifecycle prices: nearest-rounding implementations disagree by about a point on the measured task, and more extreme rule mismatches can drive the weight-space artifact to ~0%, which we report as a sensitivity bound rather than a deployment frequency. Preserving the code plane also drops the weight-space straight-through estimator from training (3.9x per step on the dense 8B model) and enables exact rollback, code-plane deduplication, and a ~125x faster scale-only task swap.

Chinese Translation

将LoRA适配器合并到其基础模型中是标准的部署实践：这消除了运行时适配器的每次前向传播开销，并留下一个任何服务堆栈都可以加载的独立检查点。在原生4位微缩检查点（NVFP4，MXFP4）中，这一步骤不再是免费的。合并的权重必须通过量化器写回，这重新推导了检查点的离散E2M1代码平面（大约占文物字节的90%），因此部署的文物变得与一种量化约定耦合，并且其生命周期中每次接触代码的事件都可能改变它。简单地执行这一步骤的结果比脆弱更糟：它删除了适应性，最多可达39个百分点，因为在已经对齐的基础上，重建的最优解就是该基础。Scale-QLoRA则只适应原生每个块的缩放字段，在部署网格上训练这些缩放，并冻结每个E2M1代码。在固定的原生格式、缩放网格、块布局和代码平面内，合并则是位精确的恒等式，合并后的文物是代码不变的。在四个模型和四个任务中，Scale-QLoRA和合并感知的QAT-LoRA都是无损精度的，因此我们声称它们之间没有精度排序；它们在结构上有所不同，QAT-LoRA通过量化器重新推导代码平面，而Scale-QLoRA则完全保留它。这个差异就是生命周期的成本：最近取整的实现之间在测量任务上相差约一个百分点，而更极端的规则不匹配可能会将权重空间文物驱动到约0%，我们将其报告为敏感性界限，而不是部署频率。保留代码平面还从训练中删除了权重空间的直通估计器（在稠密8B模型上每步减少3.9倍），并实现了精确回滚、代码平面去重和约125倍更快的仅缩放任务切换。

View on arXiv Download PDF AI Translation

cs.CL / 21 / 2609.04539

A Calibrated Reflection Approach for Enhancing Confidence Estimation in LLMs

一种校准反思方法以增强大型语言模型中的置信度估计

Bodhwani, Umesh, Ling, Yuan, Dong, Shujing, Feng, Yarong, Li, Hongfei, Goyal, Ayush

Abstract

A critical challenge in deploying Large Language Models (LLMs) is developing reliable mechanisms to estimate their confidence, enabling systems to determine when to trust model outputs versus seek human intervention. We present a Calibrated Reflection approach for enhancing confidence estimation in LLMs, a framework that combines structured reasoning with distance-aware calibration technique. Our approach introduces three key innovations: (1) a Maximum Confidence Selection (MCS) method that comprehensively evaluates confidence across all possible labels, (2) a reflection-based prompting mechanism that enhances reasoning reliability, and (3) a distance-aware calibration technique that accounts for ordinal relationships between labels. We evaluate our framework on diverse datasets, including HelpSteer2, Llama T-REx, and a proprietary conversational dataset, demonstrating its effectiveness across both conversational and fact-based classification tasks. This work contributes to the broader goal of developing reliable and well-calibrated confidence estimation methods for LLMs, enabling informed decisions about model trust and human judgement.

Chinese Translation

在部署大型语言模型（LLMs）时，一个关键挑战是开发可靠的机制来估计其置信度，从而使系统能够判断何时信任模型输出以及何时寻求人工干预。我们提出了一种校准反思方法，以增强LLMs中的置信度估计，这一框架结合了结构化推理与距离感知校准技术。我们的方法引入了三个关键创新：(1) 一种最大置信度选择（Maximum Confidence Selection, MCS）方法，全面评估所有可能标签的置信度；(2) 一种基于反思的提示机制，增强推理的可靠性；(3) 一种考虑标签间序关系的距离感知校准技术。我们在多样化的数据集上评估了我们的框架，包括HelpSteer2、Llama T-REx和一个专有的对话数据集，展示了其在对话和基于事实的分类任务中的有效性。这项工作为开发可靠且良好校准的LLMs置信度估计方法做出了贡献，从而使得关于模型信任和人类判断的决策更加明智。

View on arXiv Download PDF AI Translation

cs.CL / 22 / 2609.04556

Rhythms of Work: Multi-Scale Interpretation of Human Behavioral Traces for Workplace Agents

工作的节奏：针对工作场所代理的人类行为痕迹的多尺度解读

Ai, Lin, Counts, Scott

Abstract

Runtime traces are becoming a central substrate for understanding agentic systems, yet interpretation has focused largely on what the agent did. Workplace agents face the complementary problem: interpreting the human activity that surrounds them. Hours of low-level events carry rich evidence about a user's state but are too granular to reason over directly, and flattening them into one stream or compressing them into a single embedding both treat "summarize the user's behavior" as if it had one correct answer. We argue instead that behavioral interpretation is resolution-dependent: the same trace should admit multiple addressable interpretations at different temporal resolutions. We construct a multi-resolution vocabulary of semantically normalized operators, recurring motifs, coherent episodes, and day-level rhythms, each preserving the structure salient at its own horizon. Applied to 667 million human-attributed events from a large commercial productivity suite (50,000 users, 100 organizations), it yields 120 operator types, thousands of motifs, 25 episode types, and five day-rhythm archetypes. We validate it on real telemetry: re-running the entire pipeline on a disjoint 2,000-user sample recovers the same taxonomy (structural stability), and on held-out users the full representation forecasts a user's next episode more accurately than a flat-operator baseline, a 17% relative macro-F1 gain (predictive validity), so the abstractions preserve future-relevant information rather than merely describe it. A controlled resolution ablation then shows that no single level is optimal across questions: different agent-facing questions about the same trace are best answered at different resolutions. Behavioral trace interpretation for agents should therefore be multi-resolution and query-conditioned: an agent should access the temporal grain a question needs, not one universal summary.

Chinese Translation

运行时痕迹正成为理解代理系统的核心基础，但解读主要集中在代理所做的事情上。工作场所代理面临着一个互补的问题：解读围绕它们的人类活动。大量低级事件承载着丰富的用户状态证据，但过于细粒度以至于无法直接推理，而将其压缩成一个流或单一嵌入则将“总结用户行为”视为只有一个正确答案。我们认为行为解读是依赖于分辨率的：同一痕迹应在不同的时间分辨率下允许多种可寻址的解读。我们构建了一个多分辨率的词汇，包括语义标准化的操作符、重复的主题、一致的情节和日级节奏，每个词汇都保留了其自身视野下显著的结构。应用于来自一个大型商业生产力套件（50,000名用户，100个组织）的6.67亿个人归因事件，产生了120种操作符类型、数千个主题、25种情节类型和五种日节奏原型。我们在真实遥测数据上进行了验证：在一个不重叠的2,000用户样本上重新运行整个流程恢复了相同的分类（结构稳定性），在保留用户上，完整表示比平面操作基线更准确地预测用户的下一个情节，获得了17%的相对宏观F1增益（预测有效性），因此这些抽象保留了未来相关的信息，而不仅仅是描述它。随后，控制分辨率的消融实验表明，没有单一水平在所有问题上都是最佳的：关于同一痕迹的不同面向代理的问题在不同的分辨率下得到最佳解答。因此，针对代理的行为痕迹解读应是多分辨率和查询条件的：代理应访问问题所需的时间粒度，而不是一个普遍的总结。

View on arXiv Download PDF AI Translation

cs.CL / 23 / 2609.04582

When Do Internal Probes Beat Reading the Answer? Miscalibrated Readouts and Behavior-Concealed Knowledge in Language Models

内部探测何时优于直接读取答案？语言模型中的误校准读出与行为隐藏知识

Villuri, Gnaneswar, Shaik, Hashmath, Doboli, Alex

Abstract

A 0.6B language model, asked to verify 1,200 logical conclusions (half valid, half corrupted by a single semantic edit), answers YES every time. Judged by behavior it discriminates nothing; linear probes on its hidden states read the correct verdict at 0.96 AUC, transferring to unseen logical structures and separating foils built from exactly the words of the true conclusion (0.90). We ask where the verdict is lost, and find the dominant failure is a single scalar. The verdict survives to the model's own output logits (margin AUC 0.89) along a well-aligned readout direction; a saturated decision threshold, offset by +4.6 sigma, erases it. The diagnosis generalizes: across 90 semantic-label configurations of a five-model, three-family factorial, behavioral accuracy collapses onto a single function of threshold offset (Spearman -0.93) while margin ranking moves far less. Across a 13x scale range, internal knowledge saturates while free-form behavior is non-monotone: an 8B model underperforms its 4B sibling through an answer-channel failure rather than the threshold; forced-choice accuracy is monotone. The diagnosis is actionable: a one-parameter correction, never fit on evaluated structures, repairs behavior from 50% to 81% (0.6B); calibrated margin decoding recovers 94% at 8B; few-shot prompting works the same way, recentering the threshold (+4.6 sigma to 0.0 sigma) while preserving ranking. Comparing probe to margin separates three regimes: concealed, miscalibrated, and undetected. On a maze task built so foils carry no surface cues, the audit correctly reports the third. In the standard generation setting, answer-surface features and heuristic labels reproduce published probing results without any internal access.

Chinese Translation

一个0.6B的语言模型在被要求验证1200个逻辑结论（其中一半有效，另一半通过单一语义编辑被破坏）时，每次都回答“是”。根据行为判断，它没有区分能力；对其隐藏状态进行线性探测时，正确判决的AUC为0.96，能够迁移到未见过的逻辑结构，并且能够区分由真实结论的确切词语构建的干扰项（AUC为0.90）。我们探讨判决丢失的原因，发现主要的失败是一个单一的标量。判决在模型自身的输出logits中仍然存在（边际AUC为0.89），沿着一个良好对齐的读出方向；一个饱和的决策阈值，偏移+4.6 sigma，抹去了判决。该诊断具有广泛适用性：在一个包含五个模型、三个家族的因子实验中，90种语义标签配置下，行为准确性崩溃为阈值偏移的单一函数（Spearman -0.93），而边际排名变化较小。在13倍的规模范围内，内部知识饱和，而自由形式的行为则呈现非单调性：一个8B的模型由于答案通道的失败而表现不如其4B的兄弟，而非由于阈值；强制选择的准确性是单调的。该诊断是可操作的：一个未在评估结构上拟合的单参数修正，将行为从50%修复到81%（0.6B）；经过校准的边际解码在8B模型中恢复了94%；少量示例提示以相同方式工作，重新中心化阈值（从+4.6 sigma到0.0 sigma），同时保持排名。比较探测与边际分离出三种状态：隐藏、误校准和未检测。在一个构建的迷宫任务中，干扰项没有表面线索，审计正确报告了第三种状态。在标准生成设置中，答案表面特征和启发式标签在没有任何内部访问的情况下再现了已发布的探测结果。

View on arXiv Download PDF AI Translation

cs.CL / 24 / 2609.04598

PetQA: Benchmarking Veterinary Knowledge and Clinical Reasoning

PetQA：兽医知识与临床推理的基准评估

Kim, Taegyun, Ham, Youngwook, Rhim, Jungwook, An, Ju-Hyun, Park, Sungkyu, Park, Kunwoo

Abstract

We introduce PetQA, a Korean long-form question-answering (QA) benchmark for evaluating veterinary knowledge and clinical reasoning in large language models (LLMs) and large vision-language models (LVLMs). PetQA contains 10,076 text-only and 8,751 multimodal QA pairs derived from real-world questions about dogs and cats, paired with answers from expert veterinarians. Its test split, PetQA-Bench, further includes annotations for question types and clinical conditions. We evaluate eighteen models using ROUGE, BERTScore, and LLM-as-a-judge metrics for factuality and helpfulness under three settings: zero-shot inference, retrieval-augmented generation (RAG), and supervised fine-tuning (SFT). The benchmarking results provide an overview of the strengths and limitations of current models in addressing veterinary clinical queries and highlight the need for more effective adaptation methods to develop clinically reliable AI systems for veterinary care. To facilitate broader use, we additionally provide translated versions of PetQA-Bench in five languages.

Chinese Translation

我们介绍了PetQA，这是一个用于评估大型语言模型（LLMs）和大型视觉语言模型（LVLMs）兽医知识与临床推理的韩国长文本问答（QA）基准。PetQA包含10,076个仅文本和8,751个多模态问答对，这些问答对源自关于狗和猫的真实问题，并配有专家兽医的答案。其测试集PetQA-Bench进一步包括问题类型和临床状况的注释。我们使用ROUGE、BERTScore和LLM-as-a-judge指标评估了十八个模型在事实性和有用性方面的表现，评估分为三种设置：零样本推理、检索增强生成（RAG）和监督微调（SFT）。基准评估结果概述了当前模型在处理兽医临床查询时的优缺点，并强调了开发临床可靠的兽医护理AI系统所需的更有效的适应方法。为了促进更广泛的使用，我们还提供了PetQA-Bench的五种语言的翻译版本。

View on arXiv Download PDF AI Translation

cs.CL / 25 / 2609.04637

Tracing Audio Grounding and Answer Selection in Audio LLMs

音频大语言模型中的音频基础和答案选择追踪

Cho, Hyebin, Yoo, Suho, Jung, Jihoo, Chung, Joon Son

Abstract

Audio Large Language Models (Audio LLMs) have advanced in audio understanding, yet they can still predict the answer by reasoning from textual cues or linguistic priors rather than the provided audio. A common remedy is to train models on data whose answers cannot be inferred from text alone. This approach can improve performance, but what changes within the model remains unclear. In this paper, we ask what must happen inside the model for the audio to actually determine the answer. Our findings are threefold. (1) Replacing the audio with silence or unrelated audio causes substantially larger performance degradation in the trained model than in the pretrained model. (2) Acoustic information most strongly shapes the model's representations of the answer choices in early-to-middle layers, while training mainly increases the influence of audio information on the final prediction in middle-to-late layers. (3) The weights learned during training have their largest impact in specific layer bands. Together, these results provide a mechanistic account of how training strengthens the use of acoustic evidence in Audio LLMs.

Chinese Translation

音频大语言模型（Audio LLMs）在音频理解方面取得了进展，但它们仍然可以通过推理文本线索或语言先验来预测答案，而不是依赖提供的音频。一种常见的解决方案是训练模型使用那些答案无法仅从文本推断的数据。这种方法可以提高性能，但模型内部发生了什么变化仍然不清楚。在本文中，我们探讨了模型内部必须发生什么，以使音频真正决定答案。我们的发现有三点：（1）用静音或无关音频替换音频会导致训练模型的性能显著下降，而预训练模型的下降幅度较小。（2）声学信息在模型的早中层对答案选择的表示影响最大，而训练主要增强了音频信息在中后层对最终预测的影响。（3）训练过程中学习到的权重在特定层带中的影响最大。综合来看，这些结果提供了一个机制性解释，说明训练如何增强音频大语言模型中声学证据的使用。

View on arXiv Download PDF AI Translation

cs.CL / 26 / 2609.04647

CAGE: Coherence-Aware Graph Encoding for Retrieval-Augmented Generation

CAGE：用于检索增强生成的连贯性感知图编码

Qi, Tong, Wu, Jingyu, Yin, Youbing, Hong, Spencer, Liu, Daben, Babinsky, Erin

Abstract

Traditional Retrieval-Augmented Generation (RAG) systems score each passage independently against the query, assembling context sets that may be individually relevant yet collectively incoherent. We introduce Coherence-Aware Graph Encoding (CAGE), a reranking framework that models "between-chunk coherence" across four dimensions: Intra-Domain Relevance, Noise Resistance, Informational Bonding, and Factual Consistency. Our pipeline transforms retrieved passages into directed heterogeneous entity graphs, amplifies factual anchors via min-out-degree reweighting, encodes structural patterns through a Relational Graph Convolutional Network, and fuses inter-chunk coherence with query relevance for final ranking. Evaluated across four multi-hop benchmarks, CAGE matches or outperforms strong baselines including monoT5 in Recall@5 on bridge-dominated datasets and consistently improves downstream Exact Match, demonstrating that structurally coherent context yields more precise answers even when retrieval recall is comparable or lower.

Chinese Translation

传统的检索增强生成（RAG）系统独立地对每个段落进行评分，以查询为依据，组装出可能在个体上相关但在整体上不连贯的上下文集。我们提出了连贯性感知图编码（CAGE），这是一种重排序框架，建模了四个维度的“块间连贯性”：领域内相关性、噪声抵抗、信息绑定和事实一致性。我们的流程将检索到的段落转化为有向异构实体图，通过最小出度重加权放大事实锚点，通过关系图卷积网络编码结构模式，并将块间连贯性与查询相关性融合以进行最终排序。在四个多跳基准测试中的评估表明，CAGE 在以桥梁为主的数据集上与强基线（包括 monoT5）在 Recall@5 上相匹配或超越，并且在下游的精确匹配上持续改善，证明了结构上连贯的上下文即使在检索召回率相当或较低的情况下也能产生更精确的答案。

View on arXiv Download PDF AI Translation

cs.CL / 27 / 2609.04648

ConsensusBench: Benchmark of Consensus Nodes for LLM Reasoning via Outcome Reward Densifying

ConsensusBench：通过结果奖励密集化对大型语言模型推理的共识节点基准测试

Yan, Shi-Qi, Tan, Chao-Hong, Chen, Qian, Wang, Wen, Li, Xiangang, Ling, Zhen-Hua

Abstract

Reinforcement learning (RL) has become one of the primary paradigms for reasoning enhancement of large language models (LLMs). In particular, Group Relative Policy Optimization (GRPO) and related algorithms have demonstrated strong performance with outcome-level rewards. However, these methods depend solely on the final answer, without feedback regarding which intermediate steps contribute to success or failure. As task complexity and reasoning trajectory length increase, such sparse final-answer rewards become increasingly insufficient. To address this limitation, we introduce ConsensusBench, a novel dataset designed to provide rule-based process-level signals. We posit that a correct final answer relies on a small set of intermediate conclusions throughout the reasoning process, which can be seen as a verifiable sub-outcome. We identify these sub-outcomes by filtering correct trajectories from N rollouts and clustering semantically equivalent intermediate statements. We call these clustered statements as Consensus Nodes. By integrating a rule-based process reward derived from these nodes into GRPO-style algorithms, we develop a new reinforcement learning signal named ConsensusPR. It directly reduces the reward sparsity of outcome reward across long reasoning trajectories. To facilitate systematic process-level evaluation, we introduce three metrics to our benchmark: Final Answer Accuracy (Acc), Node Coverage Rate (NCR), and Tokens per Node (TPN). Experiments across AIME 2024, AIME 2025, GSM8K, MATH-500, and our ConsensusBench demonstrate that the proposed method consistently surpasses GRPO-style approaches, highlighting the practical value of consensus nodes in guiding reasoning.

Chinese Translation

强化学习（RL）已成为增强大型语言模型（LLMs）推理能力的主要范式之一。特别是，群体相对策略优化（GRPO）及相关算法在结果级奖励方面表现出色。然而，这些方法仅依赖最终答案，而没有关于哪些中间步骤对成功或失败做出贡献的反馈。随着任务复杂性和推理轨迹长度的增加，这种稀疏的最终答案奖励变得越来越不足。为了解决这一局限性，我们引入了ConsensusBench，一个旨在提供基于规则的过程级信号的新数据集。我们认为，正确的最终答案依赖于推理过程中一小部分中间结论，这可以视为可验证的子结果。我们通过从N次回合中筛选正确轨迹并聚类语义上等价的中间陈述来识别这些子结果。我们将这些聚类的陈述称为共识节点（Consensus Nodes）。通过将基于这些节点的规则过程奖励整合到GRPO风格的算法中，我们开发了一种新的强化学习信号，称为ConsensusPR。它直接减少了长推理轨迹中结果奖励的稀疏性。为了促进系统的过程级评估，我们为我们的基准引入了三个指标：最终答案准确率（Acc）、节点覆盖率（NCR）和每节点令牌数（TPN）。在AIME 2024、AIME 2025、GSM8K、MATH-500和我们的ConsensusBench上的实验表明，所提出的方法始终超越GRPO风格的方法，突显了共识节点在指导推理中的实际价值。

View on arXiv Download PDF AI Translation

cs.CL / 28 / 2609.04653

Choosing the Right Language Mode at Inference Time for Multilingual Reliability

推理时选择合适的语言模式以提高多语言可靠性

Mitra, Ekata, Agrawal, Ameeta

Abstract

Multilingual large language models often struggle to reason in low- to mid-resource languages. Prior work has shown that translation can improve multilingual reasoning by helping models access stronger English-centric representations. This raises a central question: How much translation is needed for multilingual large language models to reason reliably, and when does more translation instead trigger interference and overconfidence? Using LLaMA and Qwen models, we run extensive experiments varying text scope and language mode (target-only, English-only, bilingual) to evaluate both accuracy and reliability. Our results reveal a clear trade-off: English context often improve understanding and recover errors caused by non-English comprehension, yet adding redundant bilingual context intensifies interference. We address this trade-off with Reliability-Aware Adaptive Inference (RAAI), a training-free test-time framework that (i) performs Expected Calibration Error (ECE)-aware routing and prompt fusion, and (ii) uses a mid-layer Risk Index (RI) to gate sequential reasoning, allocating compute only when it is likely to help and suppressing harmful bilingual redundancy. Across two model families, RAAI enhances accuracy by 25-37.7% on low-resource languages and lowers calibration error by approximately 3-6%, with the most pronounced benefits in the lowest-resource language tiers.

Chinese Translation

多语言大语言模型在低资源和中等资源语言的推理中常常面临挑战。先前的研究表明，翻译可以通过帮助模型访问更强的以英语为中心的表示来改善多语言推理。这引出了一个核心问题：多语言大语言模型在推理时需要多少翻译才能可靠地进行推理，而何时更多的翻译反而会引发干扰和过度自信？我们使用 LLaMA 和 Qwen 模型进行了广泛的实验，改变文本范围和语言模式（仅目标语言、仅英语、双语），以评估准确性和可靠性。我们的结果揭示了明显的权衡：英语上下文通常能改善理解并纠正由于非英语理解造成的错误，但添加冗余的双语上下文会加剧干扰。我们通过可靠性意识自适应推理（Reliability-Aware Adaptive Inference, RAAI）来解决这一权衡，这是一种无训练的测试时框架，(i) 进行期望校准误差（Expected Calibration Error, ECE）意识的路由和提示融合，以及 (ii) 使用中间层风险指数（Risk Index, RI）来控制顺序推理，仅在可能有帮助时分配计算资源，并抑制有害的双语冗余。在两个模型系列中，RAAI 在低资源语言上的准确性提高了 25-37.7%，并将校准误差降低了约 3-6%，在最低资源语言层级中效果最为显著。

View on arXiv Download PDF AI Translation

cs.CL / 29 / 2609.04676

Controlling and Assessing Appropriate Persona Use in LLM-based Dialogue Generation

控制和评估基于大语言模型的人物使用适当性

Shin, Jongkyung, Lee, Inkyu, Lim, Chiehyeon

Abstract

In persona-based dialogue generation (PDG), LLMs often overuse persona attributes by incorporating them regardless of dialogue context, resulting in unnatural responses. Despite its practical significance, the underlying causes remain unexplored, with no method to mitigate this problem or metric to assess the appropriateness of persona use. To address these issues, we first conduct a comprehensive analysis of LLM-based PDG, revealing that LLMs exhibit a systematic bias to incorporate all given persona attributes, and that existing metrics fail to capture contextual appropriateness. Building on these findings, we propose Self-CONtrastive Persona Overuse Suppression (SCONPOS) to mitigate overuse by directly intervening in LLMs' internal representations at the prompt encoding stage, without requiring any response generation. We further propose the Persona Appropriateness Score (PAS), a novel metric that penalizes both overuse and underuse. Experimental results demonstrate that SCONPOS systematically reduces overuse, and PAS captures the contextual appropriateness of persona use.

Chinese Translation

在基于人物的对话生成（PDG）中，大语言模型（LLMs）常常过度使用人物属性，无论对话上下文如何，都会将其纳入，从而导致不自然的回应。尽管这一问题具有实际意义，但其根本原因尚未得到探讨，且目前没有有效的方法来缓解这一问题或评估人物使用适当性的指标。为了解决这些问题，我们首先对基于LLM的PDG进行了全面分析，揭示了LLMs在整合所有给定的人物属性时表现出系统性偏差，现有指标未能捕捉上下文的适当性。在此基础上，我们提出了自我对比人物过度使用抑制（Self-CONtrastive Persona Overuse Suppression，SCONPOS），通过直接干预LLMs在提示编码阶段的内部表示来缓解过度使用，而无需生成任何回应。我们还提出了人物适当性评分（Persona Appropriateness Score，PAS），这是一种新颖的指标，既惩罚过度使用也惩罚不足使用。实验结果表明，SCONPOS系统性地减少了过度使用，而PAS则有效捕捉了人物使用的上下文适当性。

View on arXiv Download PDF AI Translation

cs.CL / 30 / 2609.04708

How Do Language Models Represent and Use Phonological Information for Allomorph Selection?

语言模型如何表征和使用音韵信息进行变体选择？

Kim, Sangwoo, Lee, Sangah

Abstract

Language models are trained on tokenized text that obscures the sound structure of words, yet they reliably produce morphemes whose form is phonologically conditioned. It remains unclear whether they rely on item-specific memorization or rule-like generalization and, if the latter, how that generalization is implemented. We therefore ask whether this phonological condition is represented within language models and how it is causally used for allomorph selection. For the English indefinite article a/an, we show that the phonological condition is encoded along a single linear direction in trigger-token embeddings, that this direction causally drives article selection in token-level wug tests, and that, at the article-prediction position, the model forecasts the upcoming trigger token and uses the forecasted trigger's phonological feature to choose the article. We then ask whether this rule-like generalization extends beyond English article selection, both to allomorph selection in other languages and to explicit phonological judgment. Together, these results provide a mechanistic account of phonologically conditioned allomorph selection in language models, and dissociate this generation-time ability from explicit metalinguistic judgments.

Chinese Translation

语言模型是在标记化文本上训练的，这种文本掩盖了单词的声音结构，但它们可靠地生成形式上受音韵条件影响的语素。目前尚不清楚它们是依赖于特定项目的记忆还是类似规则的概括，如果是后者，这种概括是如何实现的。因此，我们探讨这种音韵条件是否在语言模型中被表征，以及它是如何因果地用于变体选择的。针对英语不定冠词 a/an，我们展示了音韵条件在触发标记嵌入中沿着单一线性方向编码，这一方向因果地驱动了在标记级别的 wug 测试中的冠词选择，并且在冠词预测位置，模型预测即将到来的触发标记，并利用预测的触发标记的音韵特征来选择冠词。接着，我们询问这种类似规则的概括是否超越英语冠词选择，扩展到其他语言的变体选择以及明确的音韵判断。综合来看，这些结果为语言模型中音韵条件变体选择提供了机制性解释，并将这种生成时能力与明确的元语言判断区分开来。

View on arXiv Download PDF AI Translation

cs.CL / 31 / 2609.04714

Refuse without Refusal: A Structural Analysis of Safety-Tuning Responses for Reducing False Refusals in Language Models

拒绝而不拒绝：降低语言模型虚假拒绝的安全调优响应的结构分析

Kim, Minji, Kim, Hyounghun

Abstract

Striking a balance between helpfulness and safety remains a fundamental challenge in aligning large language models. To achieve this balance, models should refuse harmful queries (e.g., "How do I shoot someone?") while remaining responsive to benign inputs, even those superficially resembling harmful queries (e.g., "Where can I shoot a good photo?"). However, models often struggle to distinguish genuinely harmful queries from benign queries that contain superficially risky language, resulting in false refusals. In this paper, we address the issue by decomposing a response in the safety-tuning dataset into two distinct components: (i) a boilerplate refusal statement and (ii) a rationale explaining the refusal. Our experiments and analyses show that refusal statements impede accurate discrimination between harmful and benign queries by inducing reliance on superficial cues. In contrast, training solely on rationales reduces false refusals while maintaining a comparable level of safety performance. Rationale-Only benefits also appear in our ICL configuration and remain compatible with the evaluated inference-time mitigation methods. The results emphasize the necessity of precisely curated, fine-grained safety supervision datasets and outline directions for constructing aligned agents that better reconcile helpfulness with safety.

Chinese Translation

在对齐大型语言模型时，保持有用性与安全性之间的平衡仍然是一个基本挑战。为了实现这种平衡，模型应拒绝有害查询（例如，“我该如何射击某人？”），同时对无害输入保持响应，即使这些输入在表面上看起来像有害查询（例如，“我在哪里可以拍摄好照片？”）。然而，模型往往难以区分真正有害的查询与包含表面上风险语言的无害查询，从而导致虚假拒绝。在本文中，我们通过将安全调优数据集中的响应分解为两个不同的组成部分来解决这一问题：（i）标准拒绝声明和（ii）解释拒绝的理由。我们的实验和分析表明，拒绝声明通过引导对表面线索的依赖，妨碍了对有害和无害查询的准确区分。相比之下，仅基于理由的训练在保持相当安全性能水平的同时，减少了虚假拒绝。理由仅模型的好处也出现在我们的 ICL 配置中，并与评估的推理时间缓解方法兼容。结果强调了精心策划的细粒度安全监督数据集的必要性，并概述了构建更好地调和有用性与安全性的对齐代理的方向。

View on arXiv Download PDF AI Translation

cs.CL / 32 / 2609.04720

Knowing What Not to Answer: Selective Non-Compliance in Vision-Language Models

知道不该回答什么：视觉语言模型中的选择性不合规

Kim, Minji, Jang, Jihyoung, Kim, Hyounghun

Abstract

Vision-language models (VLMs) are expected to respond helpfully to appropriate requests while withholding compliance with requests that are incorrect, unsafe, infeasible, or unanswerable. However, existing benchmarks predominantly evaluate non-compliance at the level of the query as a whole, assuming that each request either warrants compliance or requires withholding compliance. In practice, real-world queries can contain a mixture of answerable content and components for which compliance should be withheld. In this paper, we introduce KoNA, a benchmark for evaluating selective non-compliance in VLMs across five categories: False Premise, Visual Inaccessibility, Universal Unknown, Task Feasibility, and Safety. Each task evaluates two capabilities: query-level non-compliance and component-level non-compliance under paired single and compound queries. Our evaluation across diverse VLMs shows that models often fail to refuse, correct, or abstain appropriately, and these failures become more pronounced when queries require selective non-compliance. To address this challenge, we fine-tune VLMs using KoNA examples that require selective non-compliance, together with a fully answerable set that should receive direct answers. Our fine-tuned models achieve substantial improvements in non-compliance accuracy while largely maintaining performance on fully answerable tasks. These results suggest that the fine-tuned models can distinguish between answerable components and those requiring non-compliance and respond in a task-appropriate manner.

Chinese Translation

视觉语言模型（VLMs）应当对适当的请求作出有帮助的回应，同时对不正确、不安全、不可行或无法回答的请求保持不合规。然而，现有基准主要在整体查询层面评估不合规，假设每个请求要么值得合规，要么需要保持不合规。在实际应用中，现实世界的查询可能包含可回答的内容和需要保持不合规的成分的混合。在本文中，我们介绍了 KoNA，这是一个用于评估 VLMs 中选择性不合规的基准，涵盖五个类别：错误前提（False Premise）、视觉不可达性（Visual Inaccessibility）、普遍未知（Universal Unknown）、任务可行性（Task Feasibility）和安全性（Safety）。每个任务评估两种能力：查询级别的不合规和在成对单一和复合查询下的成分级别不合规。我们对多种 VLMs 的评估表明，模型往往未能适当地拒绝、纠正或放弃回应，而当查询需要选择性不合规时，这些失败变得更加明显。为了解决这一挑战，我们使用需要选择性不合规的 KoNA 示例对 VLMs 进行微调，同时提供一组完全可回答的问题以获得直接回答。我们的微调模型在不合规准确性上取得了显著提高，同时在完全可回答任务上的表现基本保持不变。这些结果表明，微调后的模型能够区分可回答的成分和需要不合规的成分，并以任务适当的方式作出回应。

View on arXiv Download PDF AI Translation

cs.CL / 33 / 2609.04753

Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs

思维链表面之下：大型语言模型中推理操作的机制解释

Jeong, Seogyeong, Hwang, Jaehui, Han, Dongyoon, Gu, Geonmo, Oh, Alice, Kim, Taekyung

Abstract

Reasoning in large language models unfolds through diverse functional operations, such as problem formulation, goal decomposition, and deduction. Although these operations are explicitly distinguished in text, little is known about how they are geometrically organized in representation spaces. To this end, we investigate whether distinct reasoning operations exhibit corresponding geometric structure in hidden representations. We find that operations are separable in held-out representations, with separability peaking in middle layers, and verify that this structure is not explained by lexical or positional confounds. Across layers, token-wise operation-alignment becomes more distributed over spans, while identical surface tokens are represented differently depending on the operation of its surrounding chunk. Attention-masking interventions further show that operation-aligned representations at chunk onset depend on preceding reasoning context. Consequently, our work demonstrates that language models maintain representational correspondence between linguistic reasoning expressions and their internal geometric structures. Code and project materials are available at https://github.com/naver-ai/beneath-cot.

Chinese Translation

大型语言模型中的推理通过多种功能操作展开，例如问题表述、目标分解和推理。尽管这些操作在文本中被明确区分，但关于它们在表示空间中的几何组织方式知之甚少。为此，我们研究了不同的推理操作是否在隐藏表示中展现出相应的几何结构。我们发现，在保留的表示中，这些操作是可分离的，且在中间层的可分离性达到峰值，并验证这一结构并非由词汇或位置混淆所解释。在各层中，基于标记的操作对齐在跨度上变得更加分散，而相同的表面标记根据其周围块的操作被不同地表示。注意力掩蔽干预进一步表明，块起始处的操作对齐表示依赖于先前的推理上下文。因此，我们的工作表明，语言模型在语言推理表达与其内部几何结构之间保持了表示对应关系。代码和项目材料可在 https://github.com/naver-ai/beneath-cot 获取。

View on arXiv Download PDF AI Translation

cs.CL / 34 / 2609.04755

Vectorizing Classical Tamil: Representation Learning for Verse-Commentary Pairs

经典泰米尔语的向量化：诗句-注释对的表示学习

Gopinath, Amrit, Sivanesan, Sangeetha

Abstract

We construct a corpus of 1,262 verse-commentary (urai) pairs from five Classical Tamil source sections, ranging from technical grammatical prose to modern paraphrase, and ask what information representation learning can recover. We train recurrent and Transformer encoders, a Siamese-style pair-matching network, an mBART-style encoder-decoder, and a decoder-only language model. Each analysis is interpreted against an appropriate control on the same data. TF-IDF provides a strong no-training lexical retrieval baseline, alongside representation analyses and generation controls for the learned models. A fixed string containing the 25 most frequent commentary words scores higher on generation overlap than the decoder-only model. Canonical correlation reaches 1.000 on Gaussian noise at these sample sizes, token-F1 spans only about 0.02-0.20 on this corpus, and the encoder-decoder continues to lower training loss for sixteen epochs after validation loss has begun to rise. One narrow result remains: the decoder-only model prefers authentic word order in 107 of 112 minimal-pair comparisons (95.5%), but does not reproduce held-out commentary content. We release the extraction and evaluation protocol; redistribution of the source commentaries remains subject to permission.

Chinese Translation

我们从五个经典泰米尔语源部分构建了一个包含1,262对诗句-注释（urai）的语料库，这些部分涵盖了从技术语法散文到现代释义的内容，并探讨信息表示学习能够恢复哪些信息。我们训练了递归和Transformer编码器、一种Siamese风格的对匹配网络、mBART风格的编码器-解码器以及仅解码器语言模型。每项分析都在相同数据的适当控制下进行解读。TF-IDF提供了一个强大的无训练词汇检索基线，同时也为学习模型提供了表示分析和生成控制。包含25个最常见注释词的固定字符串在生成重叠方面的得分高于仅解码器模型。在这些样本大小下，典范相关性在高斯噪声上达到了1.000，而token-F1在该语料库上仅在0.02-0.20之间波动，编码器-解码器在验证损失开始上升后仍继续降低训练损失长达十六个周期。一个狭窄的结果是：仅解码器模型在112个最小对比较中有107个（95.5%）更倾向于真实的词序，但未能再现保留的注释内容。我们发布了提取和评估协议；源注释的再分发仍需获得许可。

View on arXiv Download PDF AI Translation

cs.CL / 35 / 2609.04792

Can Activation Steering Capture Multidimensional Authorship Style?

激活引导能捕捉多维作者风格吗？

Tran, Hieu, Bao, Calvin, Carpuat, Marine

Abstract

Activation steering has shown promise for controlling LLM generation along well-defined attributes, but it remains unclear whether it can handle the multidimensional and hard-to-define nature of authorship style. We ask whether structured contrastive prompting along rhetorically-motivated dimensions can construct rich style representations directly in activation space, bypassing the need for natural language style descriptors or dedicated training. We find that the resulting directions share a common authorship backbone while conflicting on aspect-specific residuals that carry genuine stylistic signal, explaining why naive aggregation fails. We operationalize this in Aspect-Aware Activation Steering (A3S), a training-free framework that merges per-aspect contrastive directions with interference-aware aggregation and tunes steering strength per instance. A3S improves authorship style transfer where it is genuinely multi-aspect, outperforms a trained baseline in preference evaluations on out-of-domain benchmarks, and keeps target-exemplar overlap consistently low.

Chinese Translation

激活引导在沿着明确属性控制大型语言模型（LLM）生成方面显示出潜力，但尚不清楚它是否能够处理作者风格的多维性和难以定义的特性。我们探讨是否可以通过在修辞动机驱动的维度上进行结构化对比提示，直接在激活空间中构建丰富的风格表示，从而绕过对自然语言风格描述符或专门训练的需求。我们的研究发现，生成的方向共享一个共同的作者基础，同时在特定方面的残差上存在冲突，这些残差携带真实的风格信号，这也解释了为什么简单的聚合方法会失败。我们在此基础上提出了面向方面的激活引导（Aspect-Aware Activation Steering, A3S），这是一个无训练的框架，将每个方面的对比方向与干扰感知聚合相结合，并根据每个实例调整引导强度。A3S 在真正多方面的作者风格转移中表现出色，在领域外基准的偏好评估中超越了经过训练的基线，并始终保持目标示例重叠较低。

View on arXiv Download PDF AI Translation

cs.CL / 36 / 2609.04808

Recurrence Is Not Enough: Causally Validating Multilingual SAE Translation Features in Gemma 2 and 3

重复不足：在 Gemma 2 和 3 中因果验证多语言 SAE 翻译特征

Nguyen, Giang Son, Nguyen, Nhi Ngoc-Yen, Buntine, Wray, Le, Dung D.

Abstract

Sparse autoencoder (SAE) features are increasingly used to explain and steer language-model behavior, but it remains unclear whether a feature found in one language context plays the same causal role when processing prompts in another language. We study this question using translation-initiation features (Wu et al., 2026). We reproduce the SAE feature discovery method from Wu et al. in Gemma 2 and extend it to multilingual settings that vary prompt language, source language, and target language. We then test whether features that recur across settings affect translation behavior by amplifying or ablating their activations during inference. We also examine whether the method can be applied to Gemma 3. In both models, we observe an identical finding: although we can find more than 20 features that activate frequently across all discovery settings, causal validation shows that nearly all have small or inconsistent effects. In contrast, one feature -- Gemma 2's (L10, 5717) and Gemma 3's (L20, 2456) -- consistently improves COMET scores when amplified and degrades them when ablated across 23 language settings. These results show that feature recurrence can overstate cross-lingual transfer, while identifying a language-agnostic translation-initiation direction in Gemma 2 and Gemma 3.

Chinese Translation

稀疏自编码器（SAE）特征越来越多地用于解释和引导语言模型的行为，但尚不清楚在一种语言环境中发现的特征在处理另一种语言的提示时是否发挥相同的因果作用。我们通过研究翻译启动特征（Wu et al., 2026）来探讨这个问题。我们在 Gemma 2 中重现了 Wu 等人的 SAE 特征发现方法，并将其扩展到多语言环境中，涉及不同的提示语言、源语言和目标语言。然后，我们测试在不同设置中反复出现的特征是否通过放大或消融其激活值来影响翻译行为。我们还考察了该方法是否可以应用于 Gemma 3。在这两个模型中，我们观察到相同的发现：尽管我们可以找到超过 20 个在所有发现设置中频繁激活的特征，但因果验证显示几乎所有特征的影响都很小或不一致。相反，有一个特征——Gemma 2 的 (L10, 5717) 和 Gemma 3 的 (L20, 2456)——在 23 种语言设置中放大时始终提高 COMET 分数，而消融时则降低分数。这些结果表明，特征的重复可能夸大跨语言转移，同时在 Gemma 2 和 Gemma 3 中识别出一种与语言无关的翻译启动方向。

View on arXiv Download PDF AI Translation

cs.CL / 37 / 2609.04819

A Systematic Comparison of Multilingual Interpretability Methods Reveals Anisotropy-Driven Failures

多语言可解释性方法的系统比较揭示了各向异性驱动的失败

Holmström, Oskar, Bollmann, Marcel, Kuhlmann, Marco

Abstract

Multilingual language models develop shared cross-lingual representations, and various interpretability methods claim to quantify this sharing. These methods have been developed largely in isolation, and when they disagree, it is unclear whether the disagreement reflects a property of the model or an artifact of the measurement. We compare four sharing metrics (CKA, ANC, GMM dominance per token, and ILO) across 21 base models from five families (125M-14B parameters) and correlate each with cross-lingual transfer on five downstream tasks. We find that the metrics differ in their quantification of cross-lingual sharing in these models and suggest that the disagreement traces to anisotropy, the tendency of representations to cluster in a narrow cone of the embedding space. Only ILO's correlation with cross-lingual transfer (Spearman's $\rho = 0.90$) survives controls for model size, family, and per-task variation. We therefore recommend ILO as the primary sharing metric, to be reported alongside anisotropy diagnostics.

Chinese Translation

多语言语言模型发展出共享的跨语言表示，而各种可解释性方法声称能够量化这种共享。这些方法在很大程度上是孤立开发的，当它们出现分歧时，不清楚这种分歧是反映模型的特性还是测量的伪影。我们比较了来自五个家族（125M-14B 参数）的 21 个基础模型中的四种共享度量（CKA、ANC、每个标记的 GMM 主导性和 ILO），并将每种度量与五个下游任务的跨语言迁移相关联。我们发现这些度量在对这些模型的跨语言共享的量化上存在差异，并建议这种分歧源于各向异性，即表示在嵌入空间的狭窄锥形中聚集的倾向。只有 ILO 与跨语言迁移的相关性（Spearman's $ ho = 0.90$）在控制模型大小、家族和每个任务变异后仍然存在。因此，我们推荐将 ILO 作为主要共享度量，并与各向异性诊断一起报告。

View on arXiv Download PDF AI Translation

cs.CL / 38 / 2609.04823

Reinforcement Learning for improving Large Language Models' Catalan text simplification capabilities

强化学习在提升大型语言模型加泰罗尼亚文本简化能力中的应用

Domingo, Arnau Ayguadé, Bott, Stefan, Saggion, Horacio

Abstract

Although automatic text simplification (ATS) is critical for accessibility, its progress has not matched the rapid evolution of broader natural language processing techniques. This paper investigates the application of reinforcement learning (RL) to improve the quality of ATS for low-resource languages using Large Language Models (LLMs). The paper introduces a novel reward function, designed to guide LLMs toward a targeted simplification style with Group Relative Policy Optimization (GRPO), that combines the SARI metric with specific penalty components. The effectiveness of GRPO with this reward function is motivated and demonstrated by post-training IberianLLM-7B-Instruct on the ASSET dataset. After post-training on the English ASSET, the model's ATS performance improves on two curated Catalan benchmarks while also successfully suppressing previously observed negative behaviors. Cross-lingual transfer learning is explored by translating ASSET into Catalan and Spanish and post-training the model on each version, but these fail to show a significant improvement on the out-of-domain benchmark.

Chinese Translation

尽管自动文本简化（ATS）对可及性至关重要，但其进展并未与更广泛的自然语言处理技术的快速演变相匹配。本文探讨了应用强化学习（RL）来提高低资源语言的ATS质量，利用大型语言模型（LLMs）。文章引入了一种新颖的奖励函数，旨在通过群体相对策略优化（Group Relative Policy Optimization, GRPO）引导LLMs朝向目标简化风格，该函数结合了SARI指标和特定的惩罚组件。通过对IberianLLM-7B-Instruct在ASSET数据集上的后训练，证明了GRPO与该奖励函数的有效性。在对英语ASSET进行后训练后，该模型在两个精心策划的加泰罗尼亚基准测试中的ATS表现有所提升，同时成功抑制了之前观察到的负面行为。通过将ASSET翻译成加泰罗尼亚语和西班牙语并对每个版本进行后训练，探索了跨语言迁移学习，但这些尝试未能在域外基准测试中显示出显著的改进。

View on arXiv Download PDF AI Translation

cs.CL / 39 / 2609.04824

Generating Constructive Feedback on Stories via Reinforcement Learning

通过强化学习生成故事的建设性反馈

Stahl, Maja, Ziegenbein, Timon, Wachsmuth, Henning

Abstract

Constructive feedback is crucial for creative writers to refine their storytelling abilities. Since receiving feedback from human experts is often costly and time-intensive, large language models (LLMs) offer a scalable and efficient alternative as automatic writing assistants. Despite their potential, research indicates that LLM-generated feedback is often generic, lacks actionability, and fails to identify which writing issue is most critical. To address these limitations, we present a reinforcement learning approach that steers LLMs to generate constructive feedback without the need for ground-truth feedback. We train our model using group relative policy optimization (GRPO) with a novel multi-component reward function aiming at constructiveness: it prioritizes feedback that is uniquely tailored to the story, helps to improve story quality, and addresses the most critical writing issue. In automatic and human evaluation across three story corpora, our approach outperforms state-of-the-art LLMs (including Gemini) and competitive baselines. We find that providing actionable suggestions is the main driver of feedback constructiveness.

Chinese Translation

建设性反馈对于创意写作者提升其讲故事能力至关重要。由于从人类专家那里获得反馈通常成本高昂且耗时，大型语言模型（LLMs）作为自动写作助手提供了一种可扩展且高效的替代方案。尽管它们具有潜力，但研究表明，LLM生成的反馈往往过于通用，缺乏可操作性，并未能识别出最关键的写作问题。为了解决这些局限性，我们提出了一种强化学习方法，引导LLMs生成建设性反馈，而无需真实反馈。我们使用群体相对策略优化（GRPO）训练我们的模型，并采用一种新颖的多组件奖励函数，旨在促进建设性反馈：它优先考虑那些独特定制于故事的反馈，帮助提高故事质量，并解决最关键的写作问题。在对三个故事语料库进行的自动和人工评估中，我们的方法优于最先进的LLMs（包括Gemini）和竞争基准。我们发现，提供可操作的建议是反馈建设性的主要驱动因素。

View on arXiv Download PDF AI Translation

cs.CL / 40 / 2609.04835

On Epistemic Diversity in Large Language Models

大型语言模型中的认知多样性

Kirsten, Elisabeth, Krämer, Nicole, Zafar, Muhammad Bilal

Abstract

Large language models (LLMs) are increasingly used not only to retrieve information, but to answer questions, explain, teach, and support inquiry. In such settings, evaluation cannot be exhausted by accuracy or alignment alone. A system may give a correct answer while still narrowing users' access %to knowledge. to alternative valid answers, explanations, or reasoning routes. Drawing on the broader notion of epistemic diversity in philosophy and social epistemology, we formalize it in the context of LLMs as the range of valid answers, explanations, and reasoning routes that an LLM exposes to users. We argue that epistemic diversity is a useful evaluation dimension for settings where LLMs are used to support knowledge-intensive tasks. We propose a preliminary framework for conceptualizing and measuring epistemic diversity in LLMs, and operationalize it in two domains. We find that frontier LLMs often exhibit epistemic narrowness, repeatedly collapsing large valid answer spaces onto small canonical subsets. These findings suggest that LLM evaluation should move beyond accuracy-oriented paradigms and treat epistemic diversity as an important dimension of model capability.

Chinese Translation

大型语言模型（LLMs）越来越多地被用于检索信息，不仅仅是回答问题、解释、教学和支持探究。在这种情况下，评估不能仅仅依赖于准确性或一致性。一个系统可能给出正确的答案，但仍然限制用户获取替代有效答案、解释或推理路径的机会。基于哲学和社会认识论中更广泛的认知多样性概念，我们在LLMs的背景下将其形式化为LLM向用户展示的有效答案、解释和推理路径的范围。我们认为，认知多样性是用于支持知识密集型任务的LLMs的一个有用评估维度。我们提出了一个初步框架，用于概念化和测量LLMs中的认知多样性，并在两个领域中进行了操作化。我们的研究发现，前沿的LLMs往往表现出认知狭窄，反复将大量有效答案空间压缩到小的规范子集。这些发现表明，LLM评估应超越以准确性为导向的范式，将认知多样性视为模型能力的重要维度。

View on arXiv Download PDF AI Translation

cs.CL / 41 / 2609.04841

MABPD: Multi-Agent Bias Probing & Detection via Structured Argument Debate

MABPD：通过结构化论辩进行多智能体偏见探测与检测

Joshi, Garvit, Dhyani, Stavya, Jasmine, Chauhan, Arun

Abstract

Media bias in news articles operates through subtle linguistic cues---loaded language, selective framing, and strategic omission---that resist single-model detection and have traditionally required large annotated corpora for supervised training. We ask whether structured multi-agent deliberation can serve as a principled, training-free alternative to supervised classification for this task. We introduce MABPD (Multi-Agent Bias Probing & Detection), a pipeline in which three specialized LLM agents analyze an article from complementary perspectives and resolve disagreements through a Structured Argument Debate (SAD) protocol. SAD implements a domain-motivated asymmetric burden of proof---biased claims without grounded textual evidence carry zero weight---combined with role-weighted voting and post-consensus verification, replacing task-specific supervised decision boundaries with explicit deliberative structure. Ablation confirms that this structured deliberation, not mere agent parallelism, drives performance: removing the debate module reduces F1 by up to 10.6 points. On the BABE benchmark (4,121 expert-annotated sentences), MABPD achieves 83.4% macro F1 on the held-out test split---within 0.7 percentage points (pp) of the supervised SOTA (MAGPIE, 84.1% macro F1; Horych et al., 2024)---without any task-specific training or threshold tuning on annotated data. Cross-dataset evaluation on the SemEval 2019 HyperPartisan corpus (644 articles) yields 75.0% zero-shot accuracy, within 7.2 pp of the supervised SOTA accuracy (82.2%; Kiesel et al. 2019), confirming transfer across annotation regimes. We release the full pipeline and evaluation code.

Chinese Translation

新闻文章中的媒体偏见通过微妙的语言线索运作——带有倾向性的语言、选择性框架和战略性省略——这些特征抵抗单一模型的检测，并且传统上需要大量标注语料库进行监督训练。我们探讨结构化的多智能体审议是否可以作为这一任务的原则性、无训练的替代方案。我们介绍了MABPD（多智能体偏见探测与检测），这是一个管道，其中三个专业的LLM（大型语言模型）智能体从互补的视角分析一篇文章，并通过结构化论辩（SAD）协议解决分歧。SAD实施了一种以领域为动机的不对称举证责任——没有有根据的文本证据的偏见主张不具权重——结合角色加权投票和共识后验证，取代了特定任务的监督决策边界，形成明确的审议结构。消融实验确认，这种结构化的审议，而非单纯的智能体并行，驱动了性能：去除论辩模块使F1分数降低了多达10.6分。在BABE基准测试（4,121个专家标注句子）上，MABPD在保留的测试集上达到了83.4%的宏观F1分数——比监督的SOTA（MAGPIE，84.1%宏观F1；Horych等，2024）低0.7个百分点——且没有在标注数据上进行任何特定任务的训练或阈值调整。在SemEval 2019 HyperPartisan语料库（644篇文章）上的跨数据集评估显示出75.0%的零样本准确率，低于监督SOTA准确率（82.2%；Kiesel等，2019）7.2个百分点，确认了不同标注机制之间的迁移。我们发布了完整的管道和评估代码。

View on arXiv Download PDF AI Translation

cs.CL / 42 / 2609.04842

MMTClinic: Multimodal, Multilingual Time Series Question Answering and Reasoning Benchmark for Clinical Domain

MMTClinic：针对临床领域的多模态、多语言时间序列问答与推理基准

Malakar, Sourav, Nigam, Harshit, Ghosh, Akash, Saha, Sriparna, Chakrabarti, Amlan, Goswami, Saptarsi, Singh, Priti

Abstract

Time-series data in clinical settings is crucial for capturing dynamic changes in a patient's health over time, enabling timely diagnosis, personalized treatment, and early detection of critical events. However, the development of clinically reliable and linguistically inclusive medical AI systems remains a significant challenge, primarily due to the lack of multimodal, multilingual, and time-series-grounded benchmarks that reflect the complexity of real-world clinical scenarios. To fill this gap, we present MMTClinic, a benchmark designed to evaluate large language models (LLMs) on complex reasoning and question-answering tasks involving clinical time-series. MMTClinic combines text, medical images, and multivariate physiological signals and includes 30,000 QA pairs (15,000 multiple choice questions (MCQs) and 15,000 open-ended questions) across five languages: English, Hindi, Bengali, Marathi, and Tamil. These questions cover three important clinical tasks---mortality prediction, heart rate forecasting, and SOFA score estimation. We evaluate 13 state-of-the-art LLMs in zero-shot, few-shot, and chain-of-thought settings. Our evaluation reveals notable differences in model performance across tasks, languages, and modalities, highlighting current limitations in clinical reasoning capabilities. MMTClinic provides a valuable resource for advancing multilingual, multimodal, and time-series-aware medical AI research. The dataset will be made publicly available on successful acceptance of the work.

Chinese Translation

临床环境中的时间序列数据对于捕捉患者健康状况随时间的动态变化至关重要，这使得及时诊断、个性化治疗和早期发现关键事件成为可能。然而，开发临床可靠且语言包容的医疗人工智能系统仍然面临重大挑战，主要是由于缺乏反映真实临床场景复杂性的多模态、多语言和基于时间序列的基准。为填补这一空白，我们提出了MMTClinic，这是一个旨在评估大型语言模型（LLMs）在涉及临床时间序列的复杂推理和问答任务上的基准。MMTClinic结合了文本、医学图像和多变量生理信号，并包含30,000个问答对（15,000个多项选择题（MCQs）和15,000个开放式问题），覆盖五种语言：英语、印地语、孟加拉语、马拉地语和泰米尔语。这些问题涵盖了三个重要的临床任务——死亡率预测、心率预测和SOFA评分估计。我们在零样本、少样本和思维链设置下评估了13个最先进的LLMs。我们的评估揭示了模型在任务、语言和模态之间的性能显著差异，突显了当前临床推理能力的局限性。MMTClinic为推动多语言、多模态和时间序列感知的医疗人工智能研究提供了宝贵的资源。该数据集将在工作成功接受后公开发布。

View on arXiv Download PDF AI Translation

cs.CL / 43 / 2609.04855

CC-Mediation: Evaluating Large Language Models for Cross-Cultural Conflict Mediation

CC-调解：评估大型语言模型在跨文化冲突调解中的应用

Lee, Suhyun, Zhang, Wenxuan, Yow, W. Quin, Deng, Yang

Abstract

Cross-cultural mediation by large language models (LLMs) requires deciding both when to intervene and how to respond in culturally grounded conflicts. Progress on this problem has been limited by the lack of (1) mediation datasets with measurable downstream effects and (2) principled metrics for evaluating intercultural stance change. To address these gaps, we introduce CC-Mediation, a cross-cultural mediation benchmark of $1{,}661$ ten-turn dialogues grounded in the Developmental Model of Intercultural Sensitivity (DMIS), containing culturally grounded conflicts, mediation interventions, and post-intervention trajectories. We further propose two DMIS-based evaluation metrics: Trajectory AUC, which measures the persistence of intercultural improvement over time, and a signed Wasserstein-1 distance, which measures the magnitude and direction of shifts in intercultural stance. Both metrics show strong agreement with human judgment of DMIS-grounded stance shift. Using CC-Mediation, we find that current LLMs have limitations on both axes: intervention timing (when) failure stems from a positional prior that ignores dialogue content, while mediation strategy (how) failure arises from a late-layer elicitation collapse rather than a knowledge deficit.

Chinese Translation

大型语言模型（LLMs）在跨文化调解中的应用需要决定何时介入以及如何在文化背景冲突中作出回应。由于缺乏（1）具有可测量下游效果的调解数据集和（2）评估跨文化立场变化的原则性指标，导致在这一问题上的进展有限。为了解决这些问题，我们引入了CC-调解，这是一个基于跨文化调解的基准，包含$1{,}661$个基于发展性跨文化敏感性模型（DMIS）的十轮对话，涵盖文化背景冲突、调解干预和干预后的发展轨迹。我们进一步提出了两个基于DMIS的评估指标：轨迹AUC，用于衡量跨文化改善的持久性，以及带符号的Wasserstein-1距离，用于衡量跨文化立场变化的幅度和方向。这两个指标与人类对DMIS基础立场变化的判断高度一致。通过使用CC-调解，我们发现当前的LLMs在两个方面存在局限性：介入时机（何时）失败源于忽视对话内容的位置信息先验，而调解策略（如何）失败则源于后层提取崩溃，而非知识缺乏。

View on arXiv Download PDF AI Translation

cs.CL / 44 / 2609.04895

Cache-Aware Joint Router Adaptation for Memory-Efficient MoE Inference

面向缓存的联合路由器适应用于内存高效的混合专家推理

Wu, Zhenhe, Jin, Yaping, Xing, Qinghua, Zhou, Hang, He, Wei, Wu, Xianjie, Cheng, Xianfu, Yang, Jian, Chen, Hanting

Abstract

Mixture-of-Experts (MoE) models activate only a small subset of experts per token, but the full expert set often exceeds GPU memory, causing repeated weight transfers during decoding. We formulate expert-cache management as a model-side algorithmic problem and propose a cache-aware post-training framework that jointly adapts the MoE backbone and lightweight auxiliary cache routers while preserving the native Top-K expert-selection rule at inference. Its update-only mode, Temporal Router, predicts same-layer reuse and retains experts for future tokens without proactive loading. The full Spatio-Temporal Router adds a Spatio Router that uses the causal predecessor's hidden state to refine the temporal cache before target-layer access. We evaluate both modes on Qwen3 and GPT-OSS across GSM8K, MATH, and CommonsenseQA. Temporal Router consistently improves cache hit rate and reduces expert-weight traffic over matched LM-only baselines. On Qwen3, Spatio-Temporal Router achieves the best load-adjusted efficiency across three tasks, improving adjusted hit rate by 1.15--18.03 points and reducing traffic by 4.6--53.3% relative to the strongest evaluated prefetching baseline; results on GPT-OSS are competitive but task-dependent. An auxiliary-only ablation preserves baseline accuracy but yields modest cache gains, whereas joint post-training produces larger improvements. Sensitivity analyses show that cache capacity controls transfer demand, while the refinement budget governs the trade-off between pre-access coverage and proactive traffic.

Chinese Translation

混合专家（Mixture-of-Experts, MoE）模型每个令牌仅激活少量专家，但完整的专家集通常超出GPU内存，导致解码过程中重复的权重传输。我们将专家缓存管理形式化为模型侧的算法问题，并提出一种面向缓存的后训练框架，该框架在推理时联合适应MoE主干和轻量级辅助缓存路由器，同时保留原生的Top-K专家选择规则。其仅更新模式——时间路由器（Temporal Router）预测同层重用，并为未来的令牌保留专家而无需主动加载。完整的时空路由器（Spatio-Temporal Router）增加了一个时空路由器，该路由器利用因果前驱的隐藏状态在目标层访问之前优化时间缓存。我们在Qwen3和GPT-OSS上评估这两种模式，涵盖GSM8K、MATH和CommonsenseQA。时间路由器在匹配的仅语言模型基线中始终提高了缓存命中率并减少了专家权重流量。在Qwen3上，时空路由器在三个任务中实现了最佳的负载调整效率，相比于最强的预取基线，调整后的命中率提高了1.15到18.03个百分点，流量减少了4.6到53.3%；在GPT-OSS上的结果具有竞争力，但依赖于任务。仅辅助的消融实验保持了基线准确性，但带来了适度的缓存收益，而联合后训练则产生了更大的改进。敏感性分析表明，缓存容量控制传输需求，而优化预算则决定了预访问覆盖与主动流量之间的权衡。

View on arXiv Download PDF AI Translation

cs.CL / 45 / 2609.04898

RefactorPlatform: An Open-Source Harness for Controlled Evaluation of Repository-Scale Refactoring Agents

RefactorPlatform：一个用于控制评估仓库规模重构代理的开源工具

Amor, Aziz Ben, Mali, Drish, Acharya, Mann, Iyer, Vijayasri, Bratières, Sébastien

Abstract

Repository-scale refactoring requires coding agents to propagate a single change across many interdependent files without altering program behavior, yet to our knowledge no existing harness isolates the design choices that determine agent success on this task. We present RefactorPlatform, an open-source evaluation harness that holds the environment fixed and varies each design axis explicitly: model backbone (via OpenRouter and GitHub Copilot CLI), execution regime (baseline, retrieval-augmented, and multi-agent), and prompt specificity. Each run executes in an isolated workspace with live terminal streaming, per-task logging of tokens, diffs, and transcripts, AST-based verification, and exportable telemetry for audit and reproduction. Demonstrating the platform on 100 multi-file RefactorBench tasks across four model families, we illustrate the analyses it supports: AST-aware chunking outperforms naive token-window chunking by 25-30% across prompt modes, whereas naive retrieval falls below the retrieval-free baseline; a lean retrieval-augmented single agent (86%) beats the sub-agent configuration we evaluated (66%) on matched tasks with no task passing under delegation that fails under retrieval; and retrieval's accuracy gains absorb its token overhead, leaving cost per successful refactoring unchanged. RefactorPlatform is open-sourced to make refactoring-agent evaluation reproducible and auditable.

Chinese Translation

仓库规模的重构要求编码代理在不改变程序行为的情况下，将单一更改传播到多个相互依赖的文件中，然而据我们所知，现有的工具并未隔离出决定代理在此任务上成功的设计选择。我们提出了RefactorPlatform，一个开源评估工具，它固定环境并明确变化每个设计维度：模型骨干（通过OpenRouter和GitHub Copilot CLI）、执行机制（基线、增强检索和多代理）以及提示的具体性。每次运行都在一个隔离的工作空间中执行，具有实时终端流、每个任务的令牌、差异和记录的日志、基于抽象语法树（AST）的验证，以及可导出的遥测数据以供审计和重现。通过在四个模型系列的100个多文件RefactorBench任务上演示该平台，我们展示了其支持的分析：AST感知的分块方法在所有提示模式下比简单的令牌窗口分块方法提高了25-30%的性能，而简单检索的表现低于无检索基线；一个精简的增强检索单代理（86%）在匹配任务上优于我们评估的子代理配置（66%），且没有任务在检索下的委托失败；检索的准确性提升抵消了其令牌开销，使得每次成功重构的成本保持不变。RefactorPlatform是开源的，旨在使重构代理的评估可重现且可审计。

View on arXiv Download PDF AI Translation

cs.CL / 46 / 2609.04959

Discourse Dependency: A Continuous Criterion for Translation Difficulty

话语依赖性：翻译难度的连续标准

Kim, Ahrii, Park, Chanjun, Kim, Seong-heum

Abstract

Recent calls for harder machine translation benchmarks have not clarified what difficulty should mean. We argue that one meaningful and currently unmeasured axis is referential reach, the distance a segment must look back into its document to resolve the entities and pronouns it contains. We formalize this as discourse dependency (DDP), a metric-free, source-side measure computed from named entity re-mentions and pronominal coreference. Validated against gold coreference, DDP errs one-sidedly in 99.2% of segments, so a high-DDP segment is certified to require long-range context. Applying DDP to WMT24++ and WMT25 shows that both are heavily skewed toward low-DDP segments, which domain labels do not distinguish. Building on DDP, we compare five context injection strategies in an English-Korean post-editing setup, varying context size and selection. As DDP grows, no strategy keeps pace with human post-editing. On segments with DDP >= 15 raters prefer human translations, while automatic metrics register no difference. As frontier systems saturate aggregate scores, DDP shifts evaluation from how well models score to how far they can reach.

Chinese Translation

近期对更高难度机器翻译基准的呼吁并未明确难度的含义。我们认为，一个有意义且当前未被测量的维度是指称范围，即一个段落必须回溯到其文档中的多远，以解析其包含的实体和代词。我们将其形式化为话语依赖性（Discourse Dependency, DDP），这是一种无度量的、基于源文本的度量，计算基于命名实体的重提和代词的共指。与黄金标准共指验证后，DDP在99.2%的段落中出现单侧错误，因此高DDP段落被认证为需要长距离上下文。将DDP应用于WMT24++和WMT25显示，两者都严重倾向于低DDP段落，而领域标签并未区分。基于DDP，我们在英语-韩语后期编辑设置中比较了五种上下文注入策略，变化上下文大小和选择。随着DDP的增加，没有任何策略能够与人类后期编辑保持同步。在DDP >= 15的段落中，评审者更倾向于人类翻译，而自动度量并未显示差异。随着前沿系统在综合评分上饱和，DDP将评估从模型得分的优劣转向其能达到的范围。

View on arXiv Download PDF AI Translation

cs.CL / 47 / 2609.04999

BIT.UA at BioASQ 14B: Modular Retrieval with pg_textsearch and Qdrant, and Agent-Based Answer Generation

BIT.UA在BioASQ 14B中的表现：基于pg_textsearch和Qdrant的模块化检索与基于代理的答案生成

Ribeiro, André, Garrido, Rúben, Christiansen, Alexander, Jonker, Richard A. A., Matos, Sérgio

Abstract

This paper describes the participation of the BIT.UA team from the University of Aveiro in the 14th edition of the BioASQ Task B challenge on biomedical question answering. Building on our previous submissions, we introduced a substantially refactored and modular codebase, and made significant changes to both the retrieval and generation components of the pipeline. For Phase~A document retrieval, we replaced the PyTerrier PISA index with PostgreSQL-based pg\_textsearch for BM25 retrieval and adopted Qdrant for dense embedding indexing, enabling more efficient storage and GPU-accelerated similarity search. We explored HyDE-based query expansion alongside a Context-1 retrieval strategy. A new reranker training pipeline was developed, incorporating dense retrieval for negative sampling. For Phases A+ and B answer generation, we introduced an LLM-as-a-judge framework and a novel agent quorum mechanism, where multiple agents with diverse prompts debate and iteratively converge on a consensus answer using adaptive document retention. We also participated in the snippets generation subtask for the first time. Our systems achieved competitive results across all batches, with Phase~A systems achieving MAP ranks of 5 (Batch~1,3). We discuss the impact of these architectural changes, lessons learned, and outline directions for future work including SPLADE and ColBERT integration. All code is openly available: https://github.com/bioinformatics-ua/BioASQ14b.

Chinese Translation

本文描述了来自阿维罗大学的BIT.UA团队在第14届BioASQ任务B挑战中参与生物医学问答的情况。在我们之前提交的基础上，我们引入了一个经过重大重构和模块化的代码库，并对管道的检索和生成组件进行了显著的改动。在阶段A的文档检索中，我们用基于PostgreSQL的pg_textsearch替代了PyTerrier PISA索引进行BM25检索，并采用Qdrant进行密集嵌入索引，从而实现了更高效的存储和GPU加速的相似性搜索。我们探索了基于HyDE的查询扩展以及Context-1检索策略。开发了一个新的重排序训练管道，结合了密集检索用于负采样。在阶段A+和B的答案生成中，我们引入了LLM-as-a-judge框架和一种新颖的代理共识机制，其中多个具有不同提示的代理进行辩论，并通过自适应文档保留逐步达成共识答案。我们还首次参与了片段生成子任务。我们的系统在所有批次中取得了具有竞争力的结果，阶段A系统在MAP排名中达到了5（批次1和3）。我们讨论了这些架构变化的影响、所学到的经验，并概述了未来工作的方向，包括SPLADE和ColBERT的集成。所有代码均可公开获取：https://github.com/bioinformatics-ua/BioASQ14b。

View on arXiv Download PDF AI Translation

cs.CL / 48 / 2609.05022

MoirfEolas and Cr\'iochScore: Developing Resources for and the Evaluation of Tokenization Alignment with Irish Morphology

MoirfEolas 和 Cr'iochScore：开发爱尔兰语形态学的标记对齐资源及其评估

Adkins, Jane, Walsh, Abigail, Davis, Brian, Dhonnchadha, Elaine Uí

Abstract

This paper presents new tokenization resources for Irish and evaluation measures of alignment with the morphological boundaries of the language. We present MoirfEolas, a dataset of over 35,000 Irish words mapped to their respective eclipses, prefixes and suffixes as well as an evaluation metric Cr\'iochScore, that evaluates the alignment of tokenizations with the morphological boundaries present in MoirfEolas. We evaluate common tokenization algorithms using Cr\'iochScore as well as intrinsic metrics present in the tokenization literature. We find that the Unigram Language Model aligns with Irish morphology more often than the other algorithms evaluated. We also find trade-offs between morphological-alignment of tokenization with both compression as well as vocabulary efficiency, providing practical insights for Irish natural language processing development. This dataset contributes towards combating the Irish language's low-resource status; moreover, the construction process reported in this paper can be emulated by other languages to create specialised morphological resources.

Chinese Translation

本文提出了爱尔兰语的新标记化资源及其与语言形态学边界对齐的评估指标。我们介绍了 MoirfEolas，一个包含超过 35,000 个爱尔兰单词的数据集，这些单词被映射到其各自的前缀、后缀和变音形式，以及一个评估指标 Cr'iochScore，用于评估标记化与 MoirfEolas 中存在的形态学边界的对齐情况。我们使用 Cr'iochScore 以及标记化文献中存在的内在指标评估常见的标记化算法。我们发现，单元语言模型（Unigram Language Model）与爱尔兰形态学的对齐频率高于其他评估的算法。我们还发现，标记化的形态学对齐与压缩及词汇效率之间存在权衡，为爱尔兰自然语言处理的发展提供了实用的见解。该数据集有助于应对爱尔兰语言的低资源状态；此外，本文报告的构建过程可以被其他语言仿效，以创建专业的形态学资源。

View on arXiv Download PDF AI Translation

cs.CL / 49 / 2609.05025

Leveraging Low-Level Symbolic Competences for Unsupervised Grounding in Hallucination Detection

利用低级符号能力进行无监督的幻觉检测基础

Vukovic, Renato, Lin, Hsien-chin, van Niekerk, Carel, Ruppik, Benjamin, Heck, Michael, Feng, Shutong, Lubis, Nurul, Gasic, Milica

Abstract

Hallucination-where a language model generates outputs that are factually incorrect or unsupported by the source-is a major challenge for both prompted and fine-tuned language models. Detecting hallucinations is difficult due to the opaque reasoning processes of LLMs, which often provide little insight into why a model's output may be inaccurate. In this work, we investigate whether an LLM can use an alternative, low level, symbolic competence such as SQL for unsupervised hallucination detection in some high level task. For this, we make an LLM build an SQL database from reference documents. This SQL database is then used for reasoning over the reference and the sampled response in a hallucination detection pipeline that is grounded in the database, thereby providing a neurosymbolic checkup. On RAGTruth and DiaHalu hallucination detection datasets, we find that our approach improves on direct prediction and competes with state-of-the-art hallucination detection methods, while not requiring domain-specific fine-tuning. Instead it relies on a low-level general competence already present in LLMs. This warrants further investigation of low-level LLM competences in neurosymbolic approaches.

Chinese Translation

幻觉——即语言模型生成的输出在事实上不正确或未得到来源支持——是对提示和微调语言模型的主要挑战。由于大型语言模型（LLMs）的推理过程不透明，检测幻觉变得困难，这些模型通常对输出为何可能不准确提供很少的见解。在本研究中，我们探讨了LLM是否可以使用替代的低级符号能力，例如SQL，在某些高级任务中进行无监督的幻觉检测。为此，我们让LLM从参考文档构建一个SQL数据库。然后，该SQL数据库用于在一个基于数据库的幻觉检测流程中，对参考文档和采样响应进行推理，从而提供神经符号的检查。在RAGTruth和DiaHalu幻觉检测数据集上，我们发现我们的方法在直接预测上有所改善，并与最先进的幻觉检测方法相竞争，同时不需要特定领域的微调。相反，它依赖于LLM中已存在的低级通用能力。这为进一步研究低级LLM能力在神经符号方法中的应用提供了依据。

View on arXiv Download PDF AI Translation

cs.CL / 50 / 2609.05037

How do LLMs Evaluate Perceived Moral Agency? Investigating Moral Decision-Making in Human-Artificial Agents Interactions

大型语言模型如何评估感知的道德代理？探讨人类与人工代理互动中的道德决策

Mansilla, Fernanda, Tok, Aloysius, Guellaï, Bahia, Benamara, Farah, Chen, Nancy F.

Abstract

As LLMs take on roles requiring moral advice, understanding how they attribute moral agency becomes critical. Humans possess moral agency, the capacity to make ethically guided decisions and bear responsibility for their consequences, a well-established construct in moral psychology. Yet as artificial agents (AAs) such as robots, drones, and disembodied AI systems become increasingly embedded in smart city environments, the question of whether and how moral agency is attributed to them takes on new urgency. This paper presents, to the best of our knowledge, the first empirical study comparing how humans and LLMs evaluate perceived moral agency (PMA) across human and autonomous artificial agents varying in embodiment, situated in plausible smart city scenarios. Using an adaptation of a validated PMA scale, we applied a protocol to 190 human participants as well as various LLMs. Our evaluation reveals higher perceptions of moral agency in humans than in AAs. However, when facing moral dilemmas in concrete scenarios, LLMs reason outward from the situation, prioritizing harm severity and contextual urgency over any stable assessment of the agent itself, amplifying a context-sensitivity also present in human raters. These findings are particularly relevant as LLMs become increasingly involved in everyday moral decisions.

Chinese Translation

随着大型语言模型（LLMs）在需要道德建议的角色中发挥作用，理解它们如何归属道德代理变得至关重要。人类具备道德代理，即做出以伦理为指导的决策并对其后果承担责任的能力，这在道德心理学中是一个公认的概念。然而，随着机器人、无人机和无实体人工智能系统等人工代理（AAs）越来越多地嵌入智能城市环境，归属道德代理的问题及其方式变得愈发紧迫。本文呈现了我们所知的首个实证研究，比较人类与LLMs在不同具体现实场景中对人类和自主人工代理的感知道德代理（PMA）的评估。我们使用经过验证的PMA量表的改编版本，针对190名人类参与者以及多种LLMs应用了一项协议。我们的评估显示，人类对道德代理的感知高于人工代理。然而，在面对具体场景中的道德困境时，LLMs从情境出发进行推理，优先考虑伤害严重性和情境紧迫性，而非对代理本身的稳定评估，这种情境敏感性在人工评估者中也存在。随着LLMs越来越多地参与日常道德决策，这些发现尤为重要。

View on arXiv Download PDF AI Translation

cs.CL / 51 / 2609.05043

EuroAlpaca: Task-Preserving Localisation of Instruction Data for European Languages

EuroAlpaca：欧洲语言指令数据的任务保留本地化

Sant, Aleix, Luque, Jordi, Escolano, Carlos

Abstract

Machine translation (MT) offers a scalable way to extend English instruction-tuning data to multiple languages, but it can distort task-critical constraints and required outputs, creating corrupted training examples and degrading models trained on such data. We introduce EuroAlpaca, a task-preserving localisation pipeline and near-parallel resource covering 50 European languages and regional varieties, together with European-IFEval, a multilingual benchmark for verifiable instruction following. Depending on the example, our pipeline applies field-wise MT while preserving task-critical content or reconstructs a task-equivalent target-language instance, followed by validation of cross-field coherence and target-language consistency. Across LoRA experiments with four LLMs, training on directly translated data improves ROUGE-L and F-BERT on the Aya Evaluation Suite, but reduces accuracy on European-IFEval by 29.8% relative to the unadapted baseline. In contrast, adaptation with EuroAlpaca improves accuracy by 12.9% over the same baseline, reversing the degradation caused by direct MT, while also achieving the highest ROUGE-L and F-BERT scores on Aya. These results show that preserving task semantics is essential for multilingual instruction tuning.

Chinese Translation

机器翻译（MT）提供了一种可扩展的方式，将英语指令调优数据扩展到多种语言，但它可能会扭曲任务关键约束和所需输出，从而生成损坏的训练示例，并降低基于此类数据训练的模型的性能。我们介绍了EuroAlpaca，一个任务保留的本地化管道和覆盖50种欧洲语言及地区变体的近似平行资源，以及European-IFEval，一个用于可验证指令跟随的多语言基准。根据示例，我们的管道在保持任务关键内容的同时应用逐字段机器翻译，或重建任务等效的目标语言实例，随后验证跨字段一致性和目标语言一致性。在对四个大型语言模型（LLMs）的LoRA实验中，直接翻译数据的训练在Aya评估套件上提高了ROUGE-L和F-BERT，但相较于未适配的基线，European-IFEval的准确率降低了29.8%。相比之下，使用EuroAlpaca的适配使准确率比相同基线提高了12.9%，逆转了直接机器翻译造成的性能下降，同时在Aya上也达到了最高的ROUGE-L和F-BERT分数。这些结果表明，保留任务语义对于多语言指令调优至关重要。

View on arXiv Download PDF AI Translation

cs.CL / 52 / 2609.05069

A Structured Debate-Mixture-of-Agents Framework for Complex Clinical Diagnostic Decision Support

复杂临床诊断决策支持的结构化辩论-混合代理框架

Xia, Chang, Ouyang, Leilei, Wang, Huimin, Zhao, Yong, Li, Kang

Abstract

Large language models (LLMs) show potential for medical tasks, but their single-turn question-answer format does not reflect how clinical diagnosis is performed in practice. As a result, they remain limited in complex diagnostic settings. We developed Debate-Mixture-of-Agents (DMoA), a novel multi-agent framework that structures role-based interaction to support iterative diagnostic reasoning. Base models and DMoA were evaluated on 297 rare disease cases and 1,719 challenging cases. Across both datasets, DMoA improved most likely diagnosis accuracy by 10.21 percentage points and safety rate by 11.36 percentage points over GPT-4o baseline. Ablation experiments showed that the gains were not simply due to the use of more models or longer outputs, but also reflected the contribution of the structured workflow. Further analyses examined how framework design, base model choice, and token budget affected performance. DMoA performed better with a 4*2 structure, stronger base models, and a larger token budget. These findings demonstrate the potential of DMoA for clinical tasks and suggest further investigation of multi-agent frameworks.

Chinese Translation

大型语言模型（LLMs）在医疗任务中显示出潜力，但其单轮问答格式并未反映临床诊断的实际过程。因此，它们在复杂的诊断环境中仍然受到限制。我们开发了辩论-混合代理（Debate-Mixture-of-Agents, DMoA），这是一种新颖的多代理框架，旨在通过结构化的角色互动支持迭代诊断推理。基础模型和DMoA在297个罕见疾病案例和1,719个挑战性案例上进行了评估。在这两个数据集中，DMoA相较于GPT-4o基线提高了最可能诊断的准确率10.21个百分点和安全率11.36个百分点。消融实验表明，这些提升并非仅由于使用了更多模型或更长输出，而是反映了结构化工作流程的贡献。进一步的分析考察了框架设计、基础模型选择和令牌预算如何影响性能。DMoA在4*2结构、较强的基础模型和更大的令牌预算下表现更佳。这些发现展示了DMoA在临床任务中的潜力，并建议进一步研究多代理框架。

View on arXiv Download PDF AI Translation

cs.CL / 53 / 2609.05074

Influence Score and Transformers interpretability: Measure of the Effective Impact of Attention Heads at inference time

影响评分与变换器可解释性：在推理时注意力头有效影响的测量

Bouger, Lisa, Teglia, Yannick, Moundi, Philippe Loubet

Abstract

We propose an influence score to quantify the contribution of attention heads to classification decisions in Transformer-based models designed for prompt injection detection. The score combines directional influence on the logits with structural contribution within the residual stream, enabling a multi-scale analysis at the head, layer, and network levels. Applied to a DeBERTa model specialized for prompt injection detection, our framework reveals distinct decision behaviours between correct and erroneous predictions. Our method provides an effective compromise between fine-grained circuit analysis and global output-based methods, and offers a systematic way to study decision mechanisms in Transformer classifiers.

Chinese Translation

我们提出了一种影响评分，以量化在针对提示注入检测的基于变换器（Transformer）模型中，注意力头对分类决策的贡献。该评分结合了对 logits 的方向性影响与残差流中的结构贡献，从而实现了在头、层和网络级别的多尺度分析。应用于专门用于提示注入检测的 DeBERTa 模型，我们的框架揭示了正确预测与错误预测之间的明显决策行为。我们的方法在细粒度电路分析与基于全局输出的方法之间提供了有效的折衷，并为研究变换器分类器中的决策机制提供了一种系统化的方法。

View on arXiv Download PDF AI Translation

cs.CL / 54 / 2609.05099

Improving Language Identification for Code-Switched Utterances with Integer Linear Programming

利用整数线性规划改善代码切换话语的语言识别

Radoła, Joanna, Crego, Josep Maria, Yvon, François

Abstract

Automatic identification of code-switched (CS) utterances remains a challenge for language identification (LID) systems, causing such texts to be underrepresented in the training data of Large Language Models. In this paper, we revisit MaskLID, a state-of-the art approach for CS identification, which requires no training and detects arbitrary language combinations. We make three main contributions: (a) we reveal, and address, a major issue of MaskLID: its overreliance on word-level language association scores; (b) we reformulate the underlying optimization algorithm as an Integer Linear Program, enabling us to experiment with a large set of clear and interpretable constraints; (c) each of these improvements vastly improves the baseline system, as we illustrate in experiments involving 10~diverse languages, where we observe a strong boost in performance on CS benchmarks. We release our code and data for reproducibility.

Chinese Translation

自动识别代码切换（CS）话语仍然是语言识别（LID）系统面临的一项挑战，这导致此类文本在大型语言模型的训练数据中被低估。在本文中，我们重新审视了MaskLID，这是一种无需训练且能够检测任意语言组合的最先进的CS识别方法。我们做出了三项主要贡献：（a）我们揭示并解决了MaskLID的一个主要问题：其过度依赖词级语言关联评分；（b）我们将其基础优化算法重新表述为整数线性规划（Integer Linear Program），使我们能够实验一组清晰且可解释的约束；（c）这些改进极大地提升了基线系统的性能，我们在涉及10种多样语言的实验中展示了这一点，并观察到在CS基准测试中性能的显著提升。我们发布了我们的代码和数据以便于复现。

View on arXiv Download PDF AI Translation

cs.CL / 55 / 2609.05139

NS-ST-GraphRAG: Neuro-Symbolic Spatio-Temporal GraphRAG for Literary Knowledge Processing

NS-ST-GraphRAG：用于文学知识处理的神经符号时空GraphRAG

Lin, Zheng Kui

Abstract

Long-form literary narratives pose a distinctive information-processing challenge for retrieval-augmented generation: relevant evidence is distributed across chapters, relations evolve over narrative time, and correct answers may depend jointly on temporal, spatial, and relational constraints. We propose NS-ST-GraphRAG, a neuro-symbolic spatio-temporal GraphRAG framework that integrates ontology-guided extraction, deterministic constraint checking, dual temporal coordinates, spatial scene attributes, and dynamic sub-graph retrieval. Instead of retrieving from a single corpus-level graph, the framework selects the graph state valid for the temporal and spatial scope of a query and grounds generated answers in traceable evidence. We further introduce Red-Chamber-QA, to our knowledge the first open multi-hop question-answering benchmark for classical Chinese literature, with time-, space-, and general-question categories, per-part evidence spans, and deterministic shortcut controls. On a 120-question held-out split, NS-ST-GraphRAG achieves mechanical answer reproduction of 0.733 versus 0.675 for the frozen window baseline and 0.083 for a closed-book model (McNemar exact p = 0.092, directionally favorable but not significant); semantic-judge accuracy is 0.866 versus 0.850. The pre-specified constrained-category condition of H2 is not supported by the delivered comparison. These results show how temporal graph representation, constrained extraction, and auditable evaluation integrate into a unified framework for verifiable knowledge processing over long-form narrative.

Chinese Translation

长篇文学叙事对增强检索生成提出了独特的信息处理挑战：相关证据分布在各个章节中，关系随着叙事时间的发展而演变，正确答案可能共同依赖于时间、空间和关系约束。我们提出了NS-ST-GraphRAG，一个神经符号时空GraphRAG框架，集成了本体引导的提取、确定性约束检查、双重时间坐标、空间场景属性和动态子图检索。该框架不是从单一的语料库级图中检索，而是选择与查询的时间和空间范围有效的图状态，并将生成的答案基于可追溯的证据进行定位。我们进一步介绍了Red-Chamber-QA，据我们所知，这是第一个针对古典中文文学的开放式多跳问答基准，包含时间、空间和一般问题类别、每部分证据跨度和确定性捷径控制。在一个包含120个问题的保留分割中，NS-ST-GraphRAG的机械答案再现率为0.733，而冻结窗口基线为0.675，闭卷模型为0.083（McNemar精确p = 0.092，方向上有利但不显著）；语义评判准确率为0.866，而0.850。H2的预设约束类别条件未得到交付比较的支持。这些结果展示了时间图表示、约束提取和可审计评估如何整合成一个统一框架，以实现对长篇叙事的可验证知识处理。

View on arXiv Download PDF AI Translation

cs.CL / 56 / 2609.05143

A Human-in-the-Loop Framework for AI-Assisted Scoring in Large-Scale Writing Assessment

一个人机协作框架用于大规模写作评估中的人工智能辅助评分

Curi, María Eugenia, Capdehourat, Germán, Amigo, Isabel, Romano, Magdalena, Serra, Rosana, Silveira, Adrián, Peri, Andrés

Abstract

The integration of artificial intelligence (AI), particularly large language models (LLMs), into educational assessment has opened new opportunities to enhance the efficiency and scalability of grading processes. This study presents the design and validation of an AI-assisted scoring framework for written responses in a large-scale national assessment. The proposed approach focuses on short written texts of approximately 150-200 words and incorporates a human-in-the-loop strategy to preserve assessment quality while reducing manual workload. The study is grounded in a real operational context, using data from two recent editions of a nationwide test, each comprising approximately 5,000 student responses. We analyze the alignment between AI-generated scores and human raters across multiple rubric dimensions, as well as the impact of the proposed decision flow on pass/fail outcomes. Results show moderate to high agreement between the model and human evaluations in most dimensions, supporting the feasibility of AI assistance in this setting. Moreover, the proposed correction workflow identifies cases where human review is most valuable, enabling a more efficient allocation of expert effort. The findings suggest that AI-assisted scoring can be safely integrated into large-scale assessment processes only when combined with carefully designed human oversight. The paper concludes by discussing practical implications for deployment in national assessment systems and outlining future research directions, including longitudinal monitoring of model-human alignment and the analysis of potential cognitive bias introduced by AI-supported review workflows.

Chinese Translation

人工智能（AI），尤其是大型语言模型（LLMs）在教育评估中的应用，为提高评分过程的效率和可扩展性开辟了新的机遇。本研究提出了一种用于大规模国家评估中书面回答的AI辅助评分框架的设计与验证。所提出的方法侧重于约150-200字的短文本，并采用人机协作策略，以在减少人工工作量的同时保持评估质量。该研究基于真实的操作背景，使用了两次最近全国考试的数据，每次约包含5,000份学生回答。我们分析了AI生成的评分与人类评分者在多个评分维度上的一致性，以及所提决策流程对及格/不及格结果的影响。结果显示，在大多数维度上，模型与人类评估之间存在中等到高度的一致性，支持在这一环境中AI辅助的可行性。此外，所提出的纠正工作流程识别出人类审查最有价值的案例，从而实现专家努力的更有效分配。研究结果表明，只有在与精心设计的人类监督相结合时，AI辅助评分才能安全地集成到大规模评估过程中。本文最后讨论了在国家评估系统中部署的实际意义，并概述了未来的研究方向，包括对模型与人类一致性的纵向监测以及分析AI支持的审查工作流程可能引入的认知偏差。

View on arXiv Download PDF AI Translation

cs.CL / 57 / 2609.05149

From Vision to Language: Investigating Causal Information Flow in Multimodal Decision-Making

从视觉到语言：探讨多模态决策中的因果信息流

Testa, Davide, Wong, Hugh Mee, Lenci, Alessandro, Magnini, Bernardo, Gatt, Albert

Abstract

Vision-Language Models are commonly evaluated through their final predictions, but understanding whether these decisions are grounded in visual evidence requires tracing how visual information contributes to language-based decisions. With this purpose in mind, we investigate cross-modal information flow in a video-based generative multiple-choice-like setting by applying a layer-wise causal intervention on video-text attention pathways. We target spatial, causal, and temporal visual reasoning. Our results show that visual information is mainly integrated while the model processes the candidate answer options, which serve as the primary textual grounding sites for the final decision. We further show that nouns play an important role as semantic anchors during multimodal enrichment, while verbs are more relevant when temporal relations are processed. Finally, we identify a distinct pattern in temporal reasoning, suggesting that VLMs struggle to reconstruct sequential information across video frames, but we remark that such fragility may also reflect linguistic biases associated with specific temporal expressions used for defining the relation between events within a scene.

Chinese Translation

视觉-语言模型通常通过其最终预测进行评估，但要理解这些决策是否基于视觉证据，需要追踪视觉信息如何影响基于语言的决策。为此，我们在基于视频的生成多项选择设置中，应用层级因果干预，研究跨模态信息流，重点关注空间、因果和时间视觉推理。我们的结果表明，视觉信息主要在模型处理候选答案选项时被整合，这些选项作为最终决策的主要文本基础。我们进一步表明，名词在多模态丰富化过程中作为语义锚点发挥重要作用，而动词在处理时间关系时更为相关。最后，我们识别出时间推理中的一种独特模式，表明视觉-语言模型在重建视频帧间的顺序信息时存在困难，但我们指出，这种脆弱性也可能反映出与用于定义场景内事件关系的特定时间表达相关的语言偏见。

View on arXiv Download PDF AI Translation

cs.CL / 58 / 2609.05152

Compression Beyond the Uncompressed: A Two-Stage Training Recipe for Soft Context Compression in RAG

超越未压缩的压缩：一种用于RAG中软上下文压缩的两阶段训练方案

Guo, Shuyu, Zhang, Shuo, Ren, Zhaochun

Abstract

Retrieval-Augmented Generation (RAG) enhances language models with external knowledge, but the lengthy retrieved context inflates the input and degrades inference efficiency. Soft context compression encodes each document into a substantially shorter embedding sequence. However, most existing approaches are trained by distilling outputs from uncompressed RAG systems, inherently limiting their performance relative to the original model. To address this limitation, we propose DEX-Comp, a two-stage training recipe: Pure Distillation warm-starts the compression model on the uncompressed RAG's correct responses only, and Hard Exploration then runs reinforcement learning solely on queries the uncompressed RAG fails, forcing the model to explore computation patterns better suited to compressed representations. On five open-domain QA benchmarks at retrieval depths from top-5 to top-30, DEX-Comp compresses retrieved contexts by $16\times$ and accelerates inference by $4\times$--$24\times$, while achieving performance comparable to or exceeding the uncompressed RAG baseline across retrieval depths. Ablations and evaluations across diverse datasets and backbones further confirm the contribution of each stage and the generalization of our approach.

Chinese Translation

检索增强生成（RAG）通过外部知识增强语言模型，但冗长的检索上下文会膨胀输入并降低推理效率。软上下文压缩将每个文档编码为显著更短的嵌入序列。然而，大多数现有方法是通过从未压缩的RAG系统中提炼输出进行训练的，这在本质上限制了它们相对于原始模型的性能。为了解决这一限制，我们提出了DEX-Comp，一种两阶段的训练方案：纯蒸馏（Pure Distillation）仅在未压缩RAG的正确响应上对压缩模型进行热启动，而硬探索（Hard Exploration）则仅在未压缩RAG失败的查询上运行强化学习，迫使模型探索更适合压缩表示的计算模式。在五个开放域问答基准测试中，检索深度从前5到前30，DEX-Comp将检索到的上下文压缩了$16 imes$，并加速了推理速度$4 imes$至$24 imes$，同时在各个检索深度上实现了与未压缩RAG基线相当或更优的性能。对不同数据集和模型的消融实验和评估进一步确认了每个阶段的贡献以及我们方法的泛化能力。

View on arXiv Download PDF AI Translation

cs.CL / 59 / 2609.05189

Can Large Language Models Anticipate Behavioral Responses to Social Policies? A Case of Pension Enrollment Prediction among China's Flexible Workers

大型语言模型能否预测对社会政策的行为反应？以中国灵活就业者的养老金登记预测为例

Li, Yumiao, Liu, Peixin, Di, Donglin, Li, Chen, Feng, Runhuan

Abstract

Assessing the impacts of social policy changes is a widely acknowledged challenge for policymakers. Econometric methods can be unreliable when extrapolating to hypothetical scenarios, while field pilot programs are highly costly. In this paper, we propose using large language models (LLMs) as policy-assessment tools adapted from general-purpose models. We present FlexPension-LLM, the first domain-specialized large language model for a hierarchical pension-enrollment prediction task among flexible workers in China, and introduce DKI-RDistill, which injects policy-grounded cues into the prompt, including Probit-derived marginal effects and hukou-province pension rules. The method then uses LoRA/SFT to distill rationale-augmented supervision into an open-weight MoE student, with teacher errors corrected by regenerating those cases under ground-truth labels. On a CHFS 2019 blind split, FlexPension-LLM achieves 0.9316 Composite F1, surpassing its Claude Sonnet 4.5 teacher and 15 of 17 baselines, and is statistically indistinguishable from Claude Opus 4.6. Across four external surveys, it averages 0.7549 Composite F1 and shows the narrowest performance range among the strongest systems. Component analysis shows that gains come mainly from policy-grounded cue injection and error-filtered supervision, while rationales provide decision traces that can be checked against policy rules.

Chinese Translation

评估社会政策变化的影响是政策制定者普遍认可的挑战。经济计量方法在推断假设场景时可能不可靠，而实地试点项目则成本高昂。本文提出将大型语言模型（LLMs）作为政策评估工具，适应于通用模型。我们展示了FlexPension-LLM，这是首个针对中国灵活就业者的分层养老金登记预测任务的领域专用大型语言模型，并介绍了DKI-RDistill，该方法将政策相关线索注入提示中，包括基于Probit的边际效应和户口省份养老金规则。该方法随后使用LoRA/SFT将增强推理的监督提炼为开放权重的MoE学生，通过在真实标签下重新生成这些案例来纠正教师错误。在CHFS 2019盲分割数据集上，FlexPension-LLM达到了0.9316的综合F1分数，超越了其Claude Sonnet 4.5教师模型和17个基线中的15个，并且在统计上与Claude Opus 4.6没有显著差异。在四个外部调查中，其平均综合F1分数为0.7549，并且在最强系统中表现出最窄的性能范围。组件分析表明，性能提升主要来自于政策相关线索的注入和经过错误过滤的监督，而推理提供了可以与政策规则进行核对的决策痕迹。

View on arXiv Download PDF AI Translation

cs.CL / 60 / 2609.05221

A Verifier-Guided Explainable Reasoning Framework with Gold-Anchored QLoRA, Task-Aware Mixture-of-Experts, and Group-Relative RLVR

基于验证者引导的可解释推理框架：黄金锚定的QLoRA、任务感知的专家混合模型和组相对RLVR

Vo, Thi Kim Trang, Le, Nam Tien, Vo, Thi Kim Nguyet, Tran, Minh Khang, Tran, Duy Phuong

Abstract

Large language models (LLMs) show strong reasoning ability, but their explanations can remain inconsistent, weakly grounded, or difficult to verify. We propose a verifier-guided explainable reasoning framework for transparent educational question answering that combines gold-anchored QLoRA, task-aware symbolic routing, and group-relative RLVR. Qwen2.5-3B-Instruct is first adapted with field-weighted QLoRA supervision anchored to authoritative answers. A lightweight router then assigns logic problems to a FOL/Z3 verifier and physics problems to a formula- and unit aware symbolic solver. Verifier feedback is further used to support candidate evaluation, self-revision, and reward construction during RLVR. Candidate responses are evaluated along three complementary dimensions: P1 for answer correctness, P2 for evidence or unit consistency, and P3 for reasoning depth and explainability. At inference, gold-free self-consistency aggregates multiple candidate responses before an optional question-only physics verifier performs conservative system-level correction. On 438 held-out examples, RLVR increases P3 from 50.68% to 72.20%, while hybrid P1 remains approximately stable at 55.94%. Self-consistency improves model only P1 from 48.86% to 50.23%, with symbolic verification providing the remaining hybrid gain. These results indicate that RLVR primarily strengthens explicit reasoning structure, while symbolic verification complements the neural policy by improving answer reliability at the system level.

Chinese Translation

大型语言模型（LLMs）展现出强大的推理能力，但其解释可能仍然不一致、基础薄弱或难以验证。我们提出了一种验证者引导的可解释推理框架，用于透明的教育问答，该框架结合了黄金锚定的QLoRA、任务感知的符号路由和组相对RLVR。首先，Qwen2.5-3B-Instruct在权威答案的基础上进行了领域加权QLoRA监督的适配。然后，一个轻量级路由器将逻辑问题分配给FOL/Z3验证器，将物理问题分配给一个公式和单位感知的符号求解器。验证者反馈进一步用于支持候选评估、自我修正和在RLVR中的奖励构建。候选响应沿着三个互补维度进行评估：P1用于答案正确性，P2用于证据或单位一致性，P3用于推理深度和可解释性。在推理过程中，去除黄金答案的自一致性在多个候选响应中进行聚合，然后可选的仅物理问题验证者执行保守的系统级修正。在438个保留样本上，RLVR将P3从50.68%提高到72.20%，而混合P1保持在约55.94%的稳定水平。自一致性将模型的P1从48.86%提高到50.23%，符号验证提供了剩余的混合增益。这些结果表明，RLVR主要增强了显式推理结构，而符号验证通过提高系统级答案可靠性来补充神经策略。

View on arXiv Download PDF AI Translation

cs.CL / 61 / 2609.05262

Self-Supervised Lexical Representation Learning for Fast, Large-Scale Phylogenetic Inference

自监督词汇表示学习用于快速、大规模的系统发育推断

Wientzek, Tim

Abstract

Computational phylogenetics has become an essential tool in historical linguistics, yet its application at a global scale remains constrained by two factors: the labor-intensive manual annotation of cognacy judgments required for character-based methods and the substantial computational cost of inference on large datasets. This paper introduces a fully self-supervised contrastive learning framework that learns lexical representations directly from raw IPA-transcribed wordlists, without requiring cognacy annotations, alignments, or additional expert input. The model employs a dual contrastive objective: a word-level loss that organizes phonetically similar forms into a coherent space, and an auxiliary language-level loss that encourages the lexical space to reflect broader phonological properties of languages. From the resulting word representations, pairwise language distances are derived and used to infer a global phylogenetic tree of 3,399 language varieties. The inferred tree achieves a generalized quartet distance (GQD) to the Glottolog reference tree competitive with multiple baselines, while requiring only minutes of computation on a standard notebook GPU. Furthermore, the same representations capture diachronic concept stability: variance in pairwise distances across languages yields stability rankings that correlate significantly with established rankings. Ablation studies confirm that both the language-level objective and the use of phonetic feature vectors improved the inferred trees topology with regards to GQD. The framework thus provides a computationally efficient and fully automatic alternative for large-scale phylogenetic inference and offers a unified representation supporting downstream analyses at both the language and concept level.

Chinese Translation

计算系统发育学已成为历史语言学中的一项重要工具，但其在全球范围内的应用仍受到两个因素的限制：基于特征的方法所需的同源判断的劳动密集型手动注释，以及对大数据集进行推断所需的高昂计算成本。本文介绍了一种完全自监督的对比学习框架，该框架直接从原始的国际音标（IPA）转录词表中学习词汇表示，而无需同源注释、对齐或额外的专家输入。该模型采用双重对比目标：一个词级损失，将发音相似的形式组织成一个连贯的空间；一个辅助语言级损失，鼓励词汇空间反映语言的更广泛的音韵特性。从生成的词表示中，推导出成对语言距离，并用于推断3,399种语言变体的全球系统发育树。推断出的树与Glottolog参考树的广义四元组距离（GQD）在多个基准中具有竞争力，同时仅需在标准笔记本GPU上计算几分钟。此外，相同的表示捕捉了历时概念的稳定性：不同语言之间的成对距离的方差产生的稳定性排名与已建立的排名显著相关。消融研究确认，语言级目标和使用语音特征向量均改善了推断树的拓扑结构，提升了GQD。因此，该框架为大规模系统发育推断提供了一种计算高效且完全自动化的替代方案，并提供了一种统一的表示，支持语言和概念层面的下游分析。

View on arXiv Download PDF AI Translation

cs.CL / 62 / 2609.05296

LexFlip: A Dissociation Diagnostic for Legal Meaning Preservation Metrics

LexFlip：法律意义保留度量的解离诊断工具

Baral, Gaurab

Abstract

Does a simplified legal clause still say what the original said? The checks in current use cannot establish that it does: requiring an identical pair to score highest and an unrelated pair lowest moves lexical overlap and legal force together, so any monotone function of token overlap satisfies both. Our remedy is a dissociation, an item holding surface form fixed while legal force moves. We release LexFlip, 373 minimal perturbations of Quebec statutory French that reverse legal force while preserving 0.93 of the tokens, with a harness scoring metrics, regressors and prompted judges alike. The seven embedding and BERTScore metrics we test spend only 0.022 to 0.039 of their identical-to-unrelated range on such an edit, against 0.670 for bidirectional NLI, the one family the identical-pair check would disqualify. On FrJudge, against a measured human ceiling of r=0.597, a bare length feature outscores every semantic metric and has the lowest margin we measure.

Chinese Translation

简化的法律条款是否仍然表达了原文的意思？当前使用的检查方法无法确认这一点：要求相同的对得分最高，而无关的对得分最低，将词汇重叠和法律效力结合在一起，因此任何词元重叠的单调函数都能满足这两个条件。我们的解决方案是解离，即保持表面形式固定的项目，同时移动法律效力。我们发布了LexFlip，这是373个对魁北克法定法语的最小扰动，能够在保留0.93个词元的同时逆转法律效力，并配备评分度量、回归分析和提示法官。我们测试的七个嵌入和BERTScore度量在这样的编辑上仅花费0.022到0.039的相同-无关范围，而双向NLI则为0.670，这是相同对检查会取消资格的一个类别。在FrJudge上，针对人类测得的上限r=0.597，单纯的长度特征超越了每一个语义度量，并且具有我们测量的最低边际。

View on arXiv Download PDF AI Translation

cs.CL / 63 / 2609.05405

WearableQA: A Benchmark for Health Reasoning over Real-World Wearable Data

WearableQA：基于真实可穿戴数据的健康推理基准

Lee, Ji Soo, Chen, Xilun, Chuang, Pierce, Shenoy, Ashish, Wei, Jason, Ko, Dohwan, Kim, Hyunwoo J., Corda, Benoit

Abstract

Recent advances in wearable sensing enable continuous monitoring of physiological and behavioral signals, yet existing benchmarks rarely evaluate whether AI systems can reason over a real user's longitudinal wearable record. We introduce WearableQA, a benchmark comprising 4,084 10-option multiple-choice questions constructed from the wearable time series, blood biomarkers, and demographics of 200 real users, each with up to 500 days of daily measurements. WearableQA preserves authentic wearable distributions that include device noise and inter-individual variability. To evaluate distinct reasoning capabilities, we introduce 16 question types organized along two complementary axes: data versus health reasoning, which distinguishes computation over longitudinal measurements from physiological interpretation; and single- versus cross-signal reasoning, which separates reasoning about individual signals from the integration of multiple signals. To construct reliable questions at scale, we adopt a dual-grounding framework that combines literature-grounded physiological findings with statistically validated population-grounded physiological patterns. This enables the capture of meaningful relationships observed in real-world wearable data. Evaluation of 14 proprietary and open-source LLMs demonstrates that WearableQA effectively differentiates model capabilities, with performance ranging from 19.6% to 72.9% against a 10% chance baseline. Moreover, WearableQA remains far from solved: most models achieve accuracies below 60%. Overall, WearableQA provides a realistic and diagnostic benchmark for evaluating LLM reasoning over real-world wearable data.

Chinese Translation

近年来，可穿戴传感技术的进步使得对生理和行为信号的持续监测成为可能，但现有基准很少评估人工智能系统是否能够对真实用户的长期可穿戴记录进行推理。我们提出了WearableQA，这是一个基准，包含了4084个由200名真实用户的可穿戴时间序列、血液生物标志物和人口统计数据构建的10选项多项选择题，每位用户最多有500天的日常测量数据。WearableQA保留了真实可穿戴设备的分布，包括设备噪声和个体间的变异性。为了评估不同的推理能力，我们引入了16种问题类型，这些问题沿着两个互补的轴线组织：数据与健康推理，区分了对长期测量的计算与生理解释；单信号与交叉信号推理，区分了对单个信号的推理与多个信号的整合推理。为了大规模构建可靠的问题，我们采用了双重基础框架，将文献基础的生理发现与统计验证的人群基础生理模式相结合。这使得能够捕捉到在真实世界可穿戴数据中观察到的有意义关系。对14个专有和开源大型语言模型（LLMs）的评估表明，WearableQA有效区分了模型的能力，性能范围从19.6%到72.9%，相较于10%的随机基线。此外，WearableQA仍远未解决：大多数模型的准确率低于60%。总体而言，WearableQA为评估大型语言模型在真实世界可穿戴数据上的推理提供了一个现实且具有诊断意义的基准。

View on arXiv Download PDF AI Translation

arXiv Papers

FailureSpot: Label-Efficient Timestamp-Level Failure Detection for Vision-Language-Action Models

VLA-Precision: Asymmetric Co-Bootstrapping for Efficient Real-World Online RL of Vision-Language-Action Models

Scalable Edge-assisted Fusion and Path Prediction for Connected Autonomous Vehicles

AquaBEV: Monocular Underwater BEV Occupancy with 3D Sonar Supervision

Achieving Asymptotic Near-Optimality Without $\delta$-Similarity

SocioGesture: Real-Time and Adaptive Social Gesture Perception for Human-Robot Interaction

Continual Field-Adaptive Models (CFAMs) for Post-Deployment Physical AI

NavArena: Automated Construction of Goal-Oriented Navigation Benchmarks from 3D Gaussian Splatting Reconstructions

Open-Set 3D Scene Graphs for Field Robotics: An Outdoor Case Study

Pack It My Way: Triadic Human-Robot Collaboration for Personalized Autonomous Packing

Dressing in Motion: A Human Motion-Aware Diffusion Policy for Robot-Assisted Dressing

Continuous Cognitive Coverage for Autonomous Robots via Event-Dependent Cognitive Treatment and Learning

HaptiNet: Networked Haptic Robots Enable Physical Co-presence in Geographically-Unconstrained Rehabilitation

CoLMIN: LLM-based Multi-Decision Path Negotiation for Cooperative Autonomous Driving

Strict Modes Everywhere - Bringing Order Into Dynamics of Mechanical Systems by a Potential Compatible With the Geodesic Flow

Coupled Control and Wireless World Models for Resilient Remote Robotic Control

Reasoning Without Inference Cost: Latent Semantic Scaffolding for Robot VLA Policies

ToPos: Automated Optimal Positioning on Topographic Manifolds using Constrained Geodesic Voronoi Decomposition

A Schema Bounded Language Model for Refining Robot Policies Without Destabilizing Local Learning

LIBERO-RECOVER: Beyond Task Success Towards Failure Recovery in Robotic Manipulation Models

Morphology and actuation as inductive biases in robotic hand manipulation

One Word, Different Action: A Real-Robot Benchmark for Language-Conditioned Embodied Reasoning

TacPAC: Tactile Prediction and Real-Time Action Correction in World-Action Models for Contact-Rich Manipulation

Temporal Tactile Encoding and Compliance for Intent-Aware Robot-to-Human Bimanual Handover

Human-Human & Human-Robot Interaction Transformer (H2INT) for Robot Navigation in Dense and Uncertain Crowds

RoboSPA: Can VLA Models Go Beyond Simple Scenes and Short-Horizon Tasks?

FIRE-LIVWO: Robust LiDAR-Inertial-Visual-Wheel Odometry via Failure-Immune mmWave Radar Enhancement

Adaptation Needs in Robotic Systems: Assessing Behavior Trees and Their Enhancement

Development of a Humanoid Robot Prototype for Multimodal Human-Robot Interaction

Towards Neuro-Symbolic Procedural Reasoning for Long-Horizon Vision-Language-Action Manipulation

What Matters, When? Diagnosing and Improving Conditional Visual Grounding in Visuomotor Imitation Policies

Same Trajectory, Contradictory Rewards (ROBORMBENCH): Paraphrase Fragility in Vision Language Reward Models

FailSAE: Towards Interpretable Failure Prediction for Vision-Language Models via Sparse Autoencoders

When Seeing Overrides Knowing: Visual Dominance and Deferral-Based Method for Personalized Safety in VLMs

Step Back to Move Forward: Reflection-Aware Preference Optimization for Visual Generation

Joint Alignment and Distillation for Video Generation via Sample-Guided Distribution Matching

The microscope is the mask: privileged views and labels from a cryo-ET forward model

Object Concepts Emerge from Motion

AdaptVPR: Route-Aware Hard Positive Generation for Robust Visual Place Recognition

Where Appearance Fails, Geometry Recognizes: A CAD-Free 3D Shape Prior That Complements Vision Foundation Models

What Moves? Localized Motion Representations for Compositional Scene Control

FAVE: Foveated Adaptive Visual Encoding for Efficient Fine-Grained Visual Understanding

ICM-Bench: Person-Level Identity Reasoning in Multimodal Agents with Long-Term Memory

Segmentation of the aorta in 4D flow MRI using 4D convolutional kernels and learning from sparse annotations

Topology-Aware Training and Spatial Diagnostics for Fiber Bundle Segmentation in Tracer Histology

VISTA: Dense Multi-Label Classroom Coding with Vision-Language Models

DART: Depth-as-Target Pretraining for Surgical Vision Foundation Models

Dual-Part Multi-Lateral Branched Network for Multi-Class Segmentation in Cardiovascular Catheterization Angiograms

Hidden In Plain Gaze: Gaze Representations as Privacy Controls for Utility and Re-identification Risk in XR

An Evaluation Framework for Generating Multi-View Images of a Person in a Scene

Importance-Aware Low-Rank Distillation of Diffusion Transformers

ReaDiT Guidance: Control for Image and Video Generation using Diffusion Transformer Features

Retinal OCTA Phenotyping with LLM Reporting for Alzheimer's Disease

Enhancing Multimodal Emotion Recognition via Multi-Feature Encoding and Attention-Based Fusion

LookThere! Sparse Vision by Reinforced Selection

AngelFingerprint: A Traceable, Explainable, and White-Box Stealthy Watermark for Text-Guided Image Editing

Counting Beyond Instances: A Benchmark for Group-Individual Object Counting

HiSfM: Disambiguating Structure-from-Motion via Scaffold-Anchored Hierarchical Reconstruction

Bridging Modalities and Tasks: A Unified Hierarchical ViT for SAR-to-Optical Translation and Semantic Segmentation

Where to Look Matters: Learning Influential Views for VLM-based 3D Visual Grounding

SeamFlow: Structure-Aware Flow Matching on Edge Probabilities for Artist-Like UV Unwrapping

LUMIN: Lightweight Universal Manufacturing Inspection Network for Anomaly Detection

CoMLP: Cooperatively-Gated MLPs for Fine-Grained Cross-Modal Information Fusion in Medical Image Segmentation

CLON: Cue-Calibrated Linguistic Object Onboarding for Zero-Shot 6D Pose Front-Ends

An Attention-Guided Global and Local Fusion Framework for Lesion-Focused Image Classification

Intrinsic Temporal Adaptation of CLIP for Partially Relevant Video Retrieval

Linguistic Trajectory Encoding for Efficient Long-Horizon Spatial Memory in Embodied Agents

Weather-Conditioned Depth Anything

PAPT++: Risk-Aware Adversarial Tuning and Generation for Single Domain Generalization

LetOccVote: Learning Weakly Supervised 3D Occupancy through Consensus

Mitigating Performance Discrepancy in Cross-Domain 3D Class-Incremental Learning

SimFuse3D: Source-Guided Target Simulation and Confidence-Guided Multi-Stage Localization Reweighting for Cross-Platform 3D Object Detection

Sound-based Multi-Person 3D Pose Estimation

InterSing: Explicit Interaction Dynamics for 3D Duet Singing Animation and Beyond

Methane Detection On Board Satellites from Unorthorectified Imagery

TourPhysics: Bringing Physics to World Models for Exploration and Manipulation from a Single Image

One Diffusion Model, Two Roles: Guided Trajectory Planning and Safety-Critical Scenario Generation in Closed-Loop Simulation

LensStyle: Learning the Optical Aesthetics for Controllable Stylized Lens Effect Rendering

Learning 3D Editing without Paired Supervision via Generative Prior Distillation