arXiv Daily Digest

240

Papers

Scenario-Independent Criticality Assessment and Prediction for Vulnerable Road Users in Autonomous Driving

自动驾驶中脆弱道路用户的情景独立关键性评估与预测

Gamerdinger, Jörg, Schwarzenberger, Victor, Schmid, Philipp, Teufel, Sven, Bringmann, Oliver

Abstract

Increasing safety is the primary objective of automated vehicles. Achieving this goal requires reliable safety metrics that incorporate safety-relevant factors such as object type, velocity, and criticality. A key capability of such metrics is the distinction between critical and non-critical objects, which is addressed through criticality or relevance estimation. Existing criticality metrics are typically designed for specific scenarios and primarily focus on vehicle-to-vehicle interactions. In this paper, we therefore propose a novel criticality metric tailored to vulnerable road users (VRUs), which require special consideration due to their less predictable motion behavior. Furthermore, to avoid the complexity introduced by scenario-specific metrics, we introduce a scenario-independent criticality prediction framework applicable to all traffic participant classes. The effectiveness of both the proposed VRU-centric criticality metric and the criticality prediction framework is evaluated using the DeepAccident dataset, which contains a diverse set of safety-critical traffic scenarios. The proposed VRU-centric criticality metric improves pedestrian criticality classification performance by up to 50 %. In addition, the proposed criticality prediction framework outperforms state-of-the-art metrics by 275 %, achieving an F1-score of 0.96 and enabling scenario-independent criticality assessment across all object classes. These results demonstrate the strong potential of the proposed approaches to enhance criticality assessment for safety evaluation in automated driving systems.

Chinese Translation

提高安全性是自动化车辆的主要目标。实现这一目标需要可靠的安全指标，这些指标应考虑与安全相关的因素，如物体类型、速度和关键性。此类指标的一个关键能力是区分关键物体与非关键物体，这通过关键性或相关性估计来解决。现有的关键性指标通常针对特定场景设计，主要关注车辆与车辆之间的交互。因此，本文提出了一种针对脆弱道路用户（VRUs）的新型关键性指标，因其运动行为较不易预测而需特别考虑。此外，为了避免场景特定指标带来的复杂性，我们引入了一种适用于所有交通参与者类别的情景独立关键性预测框架。通过使用包含多样化安全关键交通场景的DeepAccident数据集，我们评估了所提出的以VRU为中心的关键性指标和关键性预测框架的有效性。所提出的以VRU为中心的关键性指标使行人关键性分类性能提高了多达50%。此外，所提出的关键性预测框架的表现超越了最先进的指标，提升幅度达到275%，实现了0.96的F1分数，并使得在所有物体类别中进行情景独立的关键性评估成为可能。这些结果展示了所提出方法在增强自动驾驶系统安全评估中的关键性评估潜力。

View on arXiv Download PDF AI Translation

cs.RO / 2 / 2609.12036

Pelican-Sim 1.0: A General World Model Simulator for Embodied Intelligence

Pelican-Sim 1.0：一种用于具身智能的通用世界模型模拟器

Zou, Shilong, Zhang, Shilin, Zhang, Yingji, Huang, Yuhang, Zhang, Yi, Ding, Zeyuan, Dong, Han, Liao, Junwei, Dai, Yong, Tang, Jian, Ju, Xiaozhu

Abstract

In this technical report, we propose Pelican-Sim 1.0, a general world model simulator for embodied intelligence that predicts future observations from visual context and robot actions to support downstream learning and decision making. The model incorporates four key design features: (1) Unified action representation: a 28-dimensional action value space covering most mainstream embodiments, keeping one model valid across heterogeneous devices. (2) Action-visual injection: URDF- and camera-rendered action videos bridge actions and pixels, giving markedly better controllability across embodiments, scenes, and tasks (PSNR +0.904 over alternative fusion baselines). (3) Sparse mixture-of-experts (MoE): sparse MoE layers add capacity for heterogeneous dynamics and absorb the action modality while reducing inter-modality conflict (FVD -6.530 vs. the dense backbone). (4) Efficient rollout generation: causal adaptation and few-step distillation yield a four-step autoregressive simulator, achieving a 5.67-fold speedup over the 35-step model. Benefiting from these designs, we train on approximately one million real-world and simulated trajectories and obtain large gains in action controllability and video quality: PSNR improves over the strongest evaluated baselines by 4.636 on AgiBotWorld Beta, 2.080 on RoboMIND, and 10.343 on RoboTwin, with the adapted EWMBench DYN score up 0.426 on RoboTwin. Relying on this, four downstream applications on RoboTwin succeed: 500 generated trajectories added to 50 demonstrations per task raise policy success from 70% to 93%; policy evaluation reaches a Pearson correlation of 0.994 across five checkpoints; and relative success gains reach 47.7% for action selection and 20.3% for policy improvement. Qualitative generalization across trajectory, scene, object, embodiment, and viewpoint shifts highlights its potential as a general-purpose world model simulator.

Chinese Translation

在本技术报告中，我们提出了Pelican-Sim 1.0，这是一种用于具身智能的通用世界模型模拟器，能够根据视觉上下文和机器人动作预测未来观察，以支持下游学习和决策。该模型包含四个关键设计特征：（1）统一的动作表示：一个28维的动作值空间覆盖大多数主流具身形式，使得在异构设备上保持一个模型有效。（2）动作-视觉注入：URDF和相机渲染的动作视频将动作与像素连接起来，在不同的具身形式、场景和任务之间提供显著更好的可控性（相较于其他融合基线，PSNR提高了0.904）。（3）稀疏专家混合（MoE）：稀疏MoE层增加了对异构动态的容量，并吸收了动作模态，同时减少了模态间的冲突（相较于密集主干，FVD降低了6.530）。（4）高效的回滚生成：因果适应和少步蒸馏生成一个四步自回归模拟器，相较于35步模型实现了5.67倍的加速。得益于这些设计，我们在大约一百万条真实和模拟轨迹上进行训练，获得了在动作可控性和视频质量上的大幅提升：在AgiBotWorld Beta上，PSNR比最强评估基线提高了4.636，在RoboMIND上提高了2.080，在RoboTwin上提高了10.343，适应后的EWMBench DYN分数在RoboTwin上提高了0.426。基于此，四个下游应用在RoboTwin上成功实现：每个任务生成的500条轨迹增加到50条演示，将策略成功率从70%提高到93%；策略评估在五个检查点之间达到0.994的Pearson相关性；相对成功增益在动作选择上达到47.7%，在策略改进上达到20.3%。在轨迹、场景、物体、具身形式和视角变化中的定性泛化突显了其作为通用世界模型模拟器的潜力。

View on arXiv Download PDF AI Translation

cs.RO / 3 / 2609.12081

MoPA: Coordinated Mobile Manipulation via Subsystem-Specific Perception Alignment

MoPA：通过子系统特定感知对齐实现协调移动操控

Chen, Guangyu, Liang, Qiwei, Zhu, Shaolong, Chen, Tianxing, Xiao, Zikuan, Xie, Yifan, Zhang, Lingfeng, Luo, Ping, Xu, Renjing, Ding, Wenbo

Abstract

Mobile manipulation requires perceptual evidence at different spatial scales for base motion and arm control, while the two action modalities remain kinematically coupled. Existing policies often employ specialized action generation for different subsystems but condition heterogeneous action branches on a shared perceptual representation, leaving subsystem-specific perception-action correspondence implicit. We present MoPA, a framework that aligns perceptual conditioning with mobility and manipulation while preserving coordination at the action level. Dual Perceptual Streams employ two mutually masked query banks to extract separate perceptual representations from a shared vision-language context. Perception2Action Adaptation jointly updates each query bank and its corresponding action stream at every layer of a structured Mixture-of-Transformers decoder, while enabling information exchange between the two action streams. Coupled conditional flow matching learns a joint vector field for coordinated generation of both action chunks. On the ManiSkill-HAB benchmark, MoPA achieves state-of-the-art performance across all three task suites. Across four real-world tasks, MoPA achieves a mean full-task success rate of 76.3%, outperforming the best baseline by 12.5 percentage points. Ablation studies and further analyses validate the effectiveness of the proposed design. Website is available at: https://mopa-policy.github.io/.

Chinese Translation

移动操控需要在不同空间尺度上获取感知证据，以支持底盘运动和臂部控制，而这两种动作模式在运动学上仍然是耦合的。现有的策略通常为不同子系统采用专门的动作生成，但将异构动作分支建立在共享的感知表示上，从而使得子系统特定的感知-动作对应关系隐含。我们提出了MoPA，一个将感知条件与移动性和操控对齐的框架，同时在动作层面保持协调。双重感知流利用两个相互屏蔽的查询库，从共享的视觉-语言上下文中提取独立的感知表示。Perception2Action适应在结构化的混合变换器解码器的每一层共同更新每个查询库及其对应的动作流，同时实现两个动作流之间的信息交换。耦合条件流匹配学习一个联合向量场，以协调生成两个动作块。在ManiSkill-HAB基准测试中，MoPA在所有三个任务套件中实现了最先进的性能。在四个真实世界任务中，MoPA的平均全任务成功率为76.3%，比最佳基线高出12.5个百分点。消融研究和进一步分析验证了所提设计的有效性。网站地址为：https://mopa-policy.github.io/

View on arXiv Download PDF AI Translation

cs.RO / 4 / 2609.12095

Uncertainty-Aware Conflict Detection Against Operator-Conditioned Weather Hazards

针对操作员条件下天气危险的基于不确定性的冲突检测

Kandoria, Balram, Kim, Seulki, Samyal, Aryaman Singh

Abstract

Strategic flight plan validation in Advanced Air Mobility (AAM) environments requires robust methods for predicting aircraft state uncertainty and detecting potential conflicts with dynamic airspace hazards. This paper presents a novel framework for uncertainty-conditioned trajectory prediction combined with polyhedra hazard representation for pre-flight conflict detection. We introduce a closed-form uncertainty estimation method that couples non-uniform rational B-spline (NURBS) curve fitting for kinematic trajectory generation with a Kalman Filter for state covariance propagation. Drawing from the Light Propagation Algorithm (LPA) paradigm, we employ a sigmoid-blended measurement noise model that captures the uncertainty reduction behavior of flight management systems approaching the required time of arrival (RTA) for waypoints. The resulting temporal uncertainty bounds are derived through a velocity-to-time variance transformation, enabling probabilistic assessment of arrival time deviations along the flight path. For hazard representation, we develop an operator-conditioned classification scheme that transforms gridded environmental data, specifically weather phenomena, into three-dimensional polyhedra volumes with intensity-based stratification. These hazard polyhedra incorporate aircraft-specific safety buffers computed from vehicle performance characteristics. Conflict detection is performed through mesh intersection algorithms operating on the spatial uncertainty tube surrounding the mean trajectory against the hazard polyhedra and temporal overlap. The framework enables the continuous strategic validation of flight plans throughout the pre-flight planning time horizon as environmental conditions evolve.

Chinese Translation

在先进空中出行（AAM）环境中，战略飞行计划验证需要强大的方法来预测飞机状态的不确定性并检测与动态空域危险的潜在冲突。本文提出了一种新颖的框架，结合不确定性条件的轨迹预测与多面体危险表示，用于飞行前冲突检测。我们引入了一种闭式不确定性估计方法，该方法将非均匀有理B样条（NURBS）曲线拟合用于运动轨迹生成，并与卡尔曼滤波器结合以进行状态协方差传播。借鉴光传播算法（LPA）范式，我们采用了一种sigmoid混合测量噪声模型，捕捉飞行管理系统在接近航点所需到达时间（RTA）时的不确定性减少行为。通过速度到时间的方差变换，得出了结果的时间不确定性界限，从而能够对飞行路径上的到达时间偏差进行概率评估。对于危险表示，我们开发了一种操作员条件的分类方案，将网格化环境数据，特别是天气现象，转化为具有强度基础分层的三维多面体体积。这些危险多面体结合了根据飞行器性能特征计算的特定于飞机的安全缓冲区。冲突检测通过在围绕平均轨迹的空间不确定性管道与危险多面体及时间重叠之间操作的网格交集算法进行。该框架使得在环境条件演变的过程中，能够持续进行飞行计划的战略验证。

View on arXiv Download PDF AI Translation

cs.RO / 5 / 2609.12103

RodForesight: A World Model Enhanced Diffusion Policy for Slender and Material Agnostic Rod Insertion

RodForesight：一种增强世界模型的扩散策略用于细长且材料无关的杆插入

Yu, Chuanbo, Yue, Mingyu, Lyu, Yan, Song, Chuhan, Wang, Peng

Abstract

Slender rod insertion arises in precision manufacturing, where millimetre scale diameter and tight clearances demand accurate perception and control. Conventional peg-in-hole methods assume a rigid object whose tip pose is fixed relative to the gripper. This assumption breaks down for a high aspect ratio rod, which can bend during manipulation, making its tip motion dependent on the rod configuration, grasp, material properties, and contact. We present RodForesight, a learning framework that factorises the task into two stages: 1) coarse approaching, which uses visual servoing to map diverse initial configurations into a compact near hole hand-off region; and 2) predictive insertion, which performs fine alignment and completes the insertion. It is worth noting that the two stages can be wrapped into an end-to-end design. During insertion, a diffusion policy generates candidate action chunks, while an action conditioned world model predicts their effects on rod-hole alignment. This pre-execution evaluation enables RodForesight to select the best action chunk based on predicted tilt and radial errors before execution. Experiments investigate the performance of different stages and the end-to-end setting, where RodForesight improves the success rate from 88.9% to 96.7%, compared to baseline methods such as diffusion policy.

Chinese Translation

细长杆插入出现在精密制造中，其中毫米级直径和紧凑间隙要求准确的感知和控制。传统的 peg-in-hole 方法假设物体是刚性的，其尖端姿态相对于夹持器是固定的。然而，这一假设在高长宽比的杆上失效，因为在操作过程中杆可能会弯曲，使得其尖端运动依赖于杆的配置、抓取、材料特性和接触。我们提出了 RodForesight，一种将任务分解为两个阶段的学习框架：1）粗略接近，使用视觉伺服将多样的初始配置映射到一个紧凑的近孔交接区域；2）预测插入，进行精细对齐并完成插入。值得注意的是，这两个阶段可以整合为一个端到端的设计。在插入过程中，扩散策略生成候选动作块，而基于动作条件的世界模型预测它们对杆-孔对齐的影响。这种执行前评估使得 RodForesight 能够根据预测的倾斜和径向误差选择最佳的动作块。在实验中，我们研究了不同阶段和端到端设置的性能，其中 RodForesight 将成功率从 88.9% 提高到 96.7%，相比于基线方法如扩散策略。

View on arXiv Download PDF AI Translation

cs.RO / 6 / 2609.12108

Multi-Objective Agent-Based Model Predictive Controller for Plug-and-Play Vehicle Control

多目标基于代理的模型预测控制器用于即插即用车辆控制

Zhong, Jiaming, Khoshnevisan, Ladan, Huang, Shucheng, Pirani, Mohammad, Pant, Yash Vardhan, Khajepour, Amir

Abstract

Functional integration is a growing trend in vehicle control, often involving the coordination of multiple controllers to achieve various objectives simultaneously. The need for flexibility and reliability has led to a "plug-and-play" approach in control system design, which presents challenges for traditional integrated model predictive control (MPC). Agent-based model predictive control (AMPC) has recently emerged as a distributed solution that treats controllers as agents, creating a collaborative framework among them to reach a common goal. However, this approach struggles to manage distributed conflicting objectives when agents are coupled or interdependent. To address this, we propose a novel, practical distributed control scheme called multi-objective AMPC, which adapts the alternating direction method of multipliers (ADMM) into a general control strategy that approximates global optimization while decoupling objectives. We systematically develop three formulations that maintain convergence while addressing control regularization and inequality constraints, applying them to complex vehicle control systems for the first time. The proposed method has been tested on two vehicle control scenarios with a multi-objective topology. Different formulations are compared through simulations, and the most computationally efficient one was implemented on an electric vehicle for real-world evaluations. The results demonstrate that the proposed multi-objective AMPC can converge approximately to the same global optimum as integrated MPC with greater flexibility and the potential to reduce computational costs.

Chinese Translation

功能集成在车辆控制中是一种日益增长的趋势，通常涉及多个控制器的协调，以同时实现各种目标。对灵活性和可靠性的需求促使控制系统设计采用“即插即用”的方法，这给传统的集成模型预测控制（MPC）带来了挑战。基于代理的模型预测控制（AMPC）最近作为一种分布式解决方案出现，将控制器视为代理，创建了一个协作框架以实现共同目标。然而，当代理之间存在耦合或相互依赖时，这种方法在管理分布式冲突目标方面面临困难。为了解决这一问题，我们提出了一种新颖的实用分布式控制方案，称为多目标AMPC，该方案将交替方向乘子法（ADMM）应用于一种通用控制策略，以在解耦目标的同时近似全局优化。我们系统地开发了三种保持收敛性的公式，同时处理控制正则化和不等式约束，并首次将其应用于复杂的车辆控制系统。所提出的方法在具有多目标拓扑的两个车辆控制场景中进行了测试。通过仿真比较不同的公式，并在电动车上实施了计算效率最高的公式进行实际评估。结果表明，所提出的多目标AMPC能够以更大的灵活性和降低计算成本的潜力，近似收敛到与集成MPC相同的全局最优解。

View on arXiv Download PDF AI Translation

cs.RO / 7 / 2609.12142

A Data-Driven Distributed Control Scheme: Learning Multi-Objective Agent-Based MPC for Path-Tracking

基于数据驱动的分布式控制方案：学习多目标基于代理的模型预测控制用于路径跟踪

Zhong, Jiaming, Mehrizi, Reza Valiollahi, Pant, Yash Vardhan, Khajepour, Amir

Abstract

Agent-based model predictive control (AMPC) has recently been proposed for vehicle systems with various controllers, such as differential braking and torque vectoring, where controllers are regarded as distributed agents contributing to the same objective. However, this scheme is challenging in handling multiple conflicting objectives with coupled agents. A common approach for such tasks is the integrated MPC, where all objectives and agents are stacked together in one optimization. Nevertheless, as more agents and objectives are involved, the integrated MPC will face challenges like computational burdens and maintenance difficulties in practice. To this end, this paper proposes a learning multi-objective AMPC that can improve design flexibility and computing efficiency. First, under the assumption of information exchange, a multi-objective AMPC tailored from the alternating direction method of multipliers (ADMM) is proposed to decouple the system and achieve the same performance as the integrated scheme iteratively. Second, a learning-based method for initializing iterations is proposed to accelerate convergence. In addition, a data management method is proposed for real-time efficiency, and an authentication module is designed for learning reliability. We compare the proposed scheme against the integrated scheme via a combined path-tracking simulation for autonomous vehicles with various controllers. The proposed scheme achieves the same control performance as the integrated one while reducing the computational time by 43.5%. Furthermore, the learning-based method saves 88.6% more computational time than without learning, making it suitable for real-time implementation.

Chinese Translation

基于代理的模型预测控制（AMPC）最近被提出用于具有各种控制器的车辆系统，例如差速制动和扭矩矢量控制，其中控制器被视为为实现相同目标而贡献的分布式代理。然而，这种方案在处理多个相互冲突的目标和耦合代理时面临挑战。对于此类任务，一种常见的方法是集成模型预测控制（MPC），其中所有目标和代理被堆叠在一个优化中。然而，随着涉及的代理和目标数量的增加，集成MPC在实践中将面临计算负担和维护困难等挑战。为此，本文提出了一种学习多目标AMPC，旨在提高设计灵活性和计算效率。首先，在信息交换的假设下，提出了一种基于交替方向乘子法（ADMM）定制的多目标AMPC，以解耦系统并迭代实现与集成方案相同的性能。其次，提出了一种基于学习的方法用于初始化迭代，以加速收敛。此外，提出了一种数据管理方法以提高实时效率，并设计了一个认证模块以确保学习的可靠性。我们通过对具有各种控制器的自主车辆进行的综合路径跟踪仿真，将所提方案与集成方案进行了比较。所提方案在实现与集成方案相同的控制性能的同时，将计算时间减少了43.5%。此外，基于学习的方法比不使用学习节省了88.6%的计算时间，使其适合实时实施。

View on arXiv Download PDF AI Translation

cs.RO / 8 / 2609.12159

A Physics-Based Closed-Loop Robotic Bioprinting Framework Towards Volumetric Muscle Loss Treatment

基于物理的闭环机器人生物打印框架用于体积肌肉损失治疗

Rezayof, Omid, Andrews, Jerin T., Zobeidi, Ehsan, Ghasemkhani, Ali, Kamaraj, Meenakshi, Tilton, Maryam, John, Johnson V., Alambeigi, Farshid

Abstract

Robotic bioprinting and Direct Ink Writing (DIW) are being explored towards the treatment of Volumetric Muscle Loss (VML). While previous studies have shown the importance of proper parameter selection on the print outcome, existing approaches often rely on time- and material-intensive design of experiments methods, or require large, well-curated datasets for training machine learning models. In this paper, we propose a physics-based closed-loop robotic bioprinting system capable of near real-time parameter adaptation. The system integrates a 3D point cloud camera and fully autonomous vision-based algorithms to provide quantitative evaluation of printed constructs. This evaluation is fed into a controller that adjusts printing parameters to achieve a desired bead thickness. To assess the framework's performance, four experimental configurations were tested, each repeated three times. In these tests, printing began from an arbitrary initial parameter value, and the controller was tasked with adjusting the parameters to reach the desired thickness. The system converged in all trials, achieving a tracking error below 0.5 mm within an average of 5.2 seconds from the start of printing. The low standard deviation of the converged pressure over different tests (0.04 bar on average) demonstrates robustness and repeatability. Additional experiments were conducted with the controller turned off, enabling direct comparison with open-loop DIW bioprinting, further confirming the effectiveness of the proposed closed-loop framework in achieving the desired bead geometry.

Chinese Translation

机器人生物打印和直接墨水书写（Direct Ink Writing, DIW）正被探索用于体积肌肉损失（Volumetric Muscle Loss, VML）的治疗。尽管以往研究已显示适当参数选择对打印结果的重要性，但现有方法通常依赖于耗时且材料密集的实验设计方法，或需要大量经过精心整理的数据集来训练机器学习模型。本文提出了一种基于物理的闭环机器人生物打印系统，能够实现近实时的参数适应。该系统集成了3D点云相机和完全自主的基于视觉的算法，以提供打印构件的定量评估。该评估结果被反馈至控制器，调整打印参数以达到期望的珠子厚度。为了评估框架的性能，测试了四种实验配置，每种配置重复三次。在这些测试中，打印从任意初始参数值开始，控制器负责调整参数以达到期望的厚度。所有试验均实现了收敛，打印开始后平均在5.2秒内跟踪误差低于0.5毫米。不同测试中收敛压力的低标准偏差（平均为0.04 bar）显示了系统的稳健性和重复性。还进行了关闭控制器的额外实验，以便与开放循环的DIW生物打印进行直接比较，进一步确认了所提闭环框架在实现期望珠子几何形状方面的有效性。

View on arXiv Download PDF AI Translation

cs.RO / 9 / 2609.12188

Battery-Aware Predictive Trajectory Planning and Control for Multirotors Under Disturbances

考虑电池的多旋翼飞行器在干扰下的预测轨迹规划与控制

Kidambi, Krishna Bhavithavya

Abstract

This paper presents a battery-aware predictive trajectory-planning and control framework for multirotors operating under spatially localized disturbances. Candidate trajectories are evaluated through closed-loop vehicle--motor--battery propagation, allowing disturbance-induced control demand, electrical energy, battery evolution, and terminal-voltage-dependent actuator capability to enter the planning process. % A reduced-order battery model is numerically benchmarked against an independently implemented Simscape equivalent-circuit reference, with a power NRMSE of $0.64\%$ and a cumulative-energy discrepancy below $0.7\%$. % In a $150$-s, $640$-m mission containing three disturbance regions, the selected trajectory reduces electrical energy consumption by $7.46\%$ and position-tracking RMSE by approximately $72\%$ relative to the disturbance-aware fixed-reference baseline. % Planner ablations show that battery-dependent terms are nonbinding at nominal SOC but alter the selected trajectory under a depleted-battery stress condition. % Execution with multiple feedback controllers further demonstrates that controller selection changes the tradeoff among tracking accuracy, energy consumption, and actuator utilization. % The results demonstrate the benefit of accounting for predicted closed-loop energetic and battery--actuator consequences during trajectory selection.

Chinese Translation

本文提出了一种考虑电池的预测轨迹规划与控制框架，适用于在空间局部干扰下运行的多旋翼飞行器。候选轨迹通过闭环的飞行器-电机-电池传播进行评估，使得干扰引起的控制需求、电能、电池演变以及依赖于终端电压的执行器能力得以纳入规划过程。一个简化的电池模型与独立实现的Simscape等效电路参考模型进行了数值基准测试，功率的归一化均方根误差（NRMSE）为0.64%，累计能量差异低于0.7%。在一个包含三个干扰区域的150秒、640米的任务中，所选轨迹相较于考虑干扰的固定参考基线，电能消耗减少了7.46%，位置跟踪均方根误差（RMSE）约降低了72%。规划器的消融实验表明，在名义状态下电池相关项并不约束轨迹选择，但在电池耗尽的压力条件下会改变所选轨迹。使用多个反馈控制器的执行进一步表明，控制器的选择改变了跟踪精度、能量消耗和执行器利用率之间的权衡。结果表明，在轨迹选择过程中考虑预测的闭环能量和电池-执行器后果具有显著的优势。

View on arXiv Download PDF AI Translation

cs.RO / 10 / 2609.12206

An Automated Thickness Evaluation Procedure Using an Integrated Structured Light 3D Camera in a Robotic Bioprinting Framework

基于集成结构光3D相机的机器人生物打印框架中的自动厚度评估程序

Zobeidi, Ehsan, Rezayof, Omid, Alambeigi, Farshid

Abstract

Bioprinting is emerging as a tissue engineering technique to replace common treatment methods for large scale injuries. While thickness of the BioPrinted Constructs (BPCs) have shown to be of importance in the cell maturation and integration, the literature lacks a robust, automated, and quantitative method for measuring these metrics. In this paper, we propose a fully automated vision-based method for measuring the thickness of the BPCs with complex geometries. Leveraging the point cloud and RGB images of a structured light 3D camera, our proposed method performs an image-based segmentation for delineating the BPCs from the RGB images, accompanied by novel geometry-based thickness measurement algorithms performed on the point cloud scans. These algorithms combine the segmentation mask with the robot's forward kinematics data and a 3D point cloud scan to precisely measure the aforementioned metrics for complex-shaped BPCs. The proposed method was evaluated in simulation and experimental studies. In simulation studies, the algorithms were used to measure the thickness of some virtually created BPCs with known thickness. The comparison between the measured and true thicknesses demonstrates the high accuracy of the proposed method, achieving mean absolute errors between 0.025 mm and 0.057 mm in simulation at a spatial resolution of 0.1 mm x 0.1 mm per pixel. Furthermore, we successfully deployed the algorithms on our robotic bioprinting setup utilizing a structure light 3D camera, where complex patterns were printed and the developed methods utilized to accurately measure the thickness of printed BPCs.

Chinese Translation

生物打印作为一种组织工程技术，正在逐渐取代大规模损伤的常见治疗方法。虽然生物打印构件（BioPrinted Constructs, BPCs）的厚度在细胞成熟和整合中具有重要意义，但文献中缺乏一种稳健、自动化和定量的方法来测量这些指标。本文提出了一种完全自动化的基于视觉的方法，用于测量具有复杂几何形状的BPCs的厚度。我们的方法利用结构光3D相机的点云和RGB图像，执行基于图像的分割，以从RGB图像中划定BPCs，并结合在点云扫描上执行的新型基于几何的厚度测量算法。这些算法将分割掩膜与机器人的前向运动学数据和3D点云扫描结合起来，以精确测量复杂形状BPCs的上述指标。所提方法在仿真和实验研究中进行了评估。在仿真研究中，算法用于测量一些虚拟创建的已知厚度的BPCs。测量厚度与真实厚度之间的比较表明，所提方法具有高精度，在空间分辨率为每像素0.1 mm x 0.1 mm的仿真中，平均绝对误差在0.025 mm到0.057 mm之间。此外，我们成功地在利用结构光3D相机的机器人生物打印设备上部署了这些算法，打印了复杂图案，并利用所开发的方法准确测量了打印BPCs的厚度。

View on arXiv Download PDF AI Translation

cs.RO / 11 / 2609.12216

Guardrailed Meta-Agent Loops: Stress-Testing Policy Pinning, Budget Bounds, and Crash Recovery

有护栏的元代理循环：压力测试政策固定、预算边界和崩溃恢复

Ma, Qinzhen, Wu, Jialin

Abstract

Self-improving agent workflows create an audit problem when the same controller can change both its behavior and the conditions under which that behavior is judged. We present GuardrailLoop, a simulation-based testbed that makes three operational contracts jointly testable: preservation of human-defined policy, compute accounting at every recorded execution prefix, and recovery of a specified scientific state after crashes. A hash-pinned policy fixes goals, scope, evaluation identity, budget, and release conditions; machine-directed evolution is restricted to a code-owned feature catalog and bounded knobs. The contribution is an executable boundary and an evaluation protocol that separates useful adaptation, state recovery, and repeated execution. In a paired 50-seed 2 x 2 study, round-stage growth changes target attainment by +1.00 and restricted mean compute to target by -56.97 simulated GPU-hours (95% paired-bootstrap interval [-58.91,-54.70]); idle growth has zero measured utility effect. Across 240 enumerated crash injections, all runs recover the defined outcome, but only 210 preserve the normalized trace: 30 pre-commit crashes repeat a planner call. Resource-drift, kill-switch, integrity, and output-guard matrices satisfy their specified checks. These findings show why successful outcome recovery is insufficient evidence of exactly-once execution. They establish conformance within one calibrated deterministic testbed, rather than general safety or real-world self-improvement.

Chinese Translation

自我改进的代理工作流在同一控制器能够同时改变其行为和评判该行为的条件时，产生了审计问题。我们提出了GuardrailLoop，一个基于模拟的测试平台，使三个操作契约可以联合测试：人定义政策的保留、每个记录的执行前缀的计算核算，以及在崩溃后恢复指定的科学状态。哈希固定的政策确定了目标、范围、评估身份、预算和发布条件；机器导向的演化被限制在代码拥有的功能目录和有界的调节器中。我们的贡献是一个可执行的边界和一个评估协议，区分有用的适应、状态恢复和重复执行。在一项配对的50种种子2 x 2研究中，轮阶段增长使目标达成率提高了+1.00，并将目标的平均计算限制在-56.97个模拟GPU小时（95%配对自助区间[-58.91,-54.70]）；闲置增长的测量效用效果为零。在240次枚举的崩溃注入中，所有运行均恢复了定义的结果，但仅有210次保留了标准化的轨迹：30次预提交崩溃重复了规划器调用。资源漂移、杀开关、完整性和输出保护矩阵满足其指定检查。这些发现表明，成功的结果恢复不足以证明确切的一次执行。它们在一个经过校准的确定性测试平台内建立了一致性，而不是一般安全性或现实世界的自我改进。

View on arXiv Download PDF AI Translation

cs.RO / 12 / 2609.12221

Chain-SLAM: Globally Consistent Backend for Multi-Session LiDAR SLAM via Chained Loop Closure

链式SLAM：通过链式闭环实现多会话LiDAR SLAM的全局一致性后端

Li, Zhiheng, Liu, Xinhao, Zhang, Juexiao, Liang, Yongqing, Feng, Chen

Abstract

Maintaining consistency over long spatial and temporal horizons remains a fundamental challenge in large-scale LiDAR SLAM, particularly when integrating maps collected across multiple sessions. We present Chain-SLAM, a LiDAR SLAM backend enabling online multi-session map alignment and reuse with global consistency at large scale. We implement a chained loop closure mechanism that efficiently propagates geometric constraints across inter-session keyframes through an adjacency graph, enabling robust long-horizon consistency triggered by reliable short-horizon loop closures. The system initializes inter-session alignment with GNSS-proximity place recognition, then performs on-the-fly loop closure detections and joint optimization of loaded maps and newly acquired trajectories within a unified factor graph, maintaining both inter- and intra-session geometric consistency without dynamic object removal, and cross-platform robustness with minimal hyperparameter tuning. Experimental results show improved trajectory accuracy and robust multi-session integration on large-scale datasets. We release our source code to support reproducible research in large-scale multi-session LiDAR SLAM. Project site: https://ai4ce.github.io/Chain-SLAM/

Chinese Translation

在大规模LiDAR SLAM中，保持长时间和空间范围内的一致性仍然是一个基本挑战，特别是在整合跨多个会话收集的地图时。我们提出了Chain-SLAM，这是一种LiDAR SLAM后端，能够在大规模下实现在线多会话地图对齐和重用的全局一致性。我们实现了一种链式闭环机制，通过邻接图有效地传播跨会话关键帧的几何约束，从而实现由可靠的短范围闭环触发的稳健长范围一致性。该系统通过GNSS接近位置识别初始化会话间对齐，然后在统一的因子图中实时执行闭环检测和加载地图与新获取轨迹的联合优化，保持会话间和会话内的几何一致性，而无需动态物体移除，并且在最小超参数调优的情况下实现跨平台的鲁棒性。实验结果表明，在大规模数据集上，轨迹精度和多会话集成的鲁棒性得到了改善。我们发布了源代码，以支持大规模多会话LiDAR SLAM的可重复研究。项目网站：https://ai4ce.github.io/Chain-SLAM/

View on arXiv Download PDF AI Translation

cs.RO / 13 / 2609.12245

DIA: Denoising Intermediate Advantage for Diffusion Policy Optimization

DIA：用于扩散策略优化的去噪中间优势

Sohal, Arjun, Zhao, Yuchi, Bogdanovic, Miroslav, Aspuru-Guzik, Alan

Abstract

Diffusion-based robot policies have become widely used in robotic manipulation, where they are typically trained with behavior cloning. However, policies trained purely from demonstrations are limited by the quality and coverage of the available data. Reinforcement learning can further improve the performance of these pretrained policies through interaction. A common approach is to use policy-gradient methods that formulate diffusion-policy fine-tuning as an outer environment MDP together with an inner denoising MDP. However, existing methods typically assign the same environment-level credit to all denoising steps used to construct an action chunk, without distinguishing which intermediate decisions contributed most to the final return. We introduce Denoising Intermediate Advantage (DIA), a policy-gradient method that learns a value function over partially denoised actions and uses it to construct a denoising- level advantage for each step of the generative process. DIA com- bines this inner credit signal with the standard environment-level PPO advantage, providing state-dependent credit throughout the denoising chain. Across Robomimic, FurnitureBench, Franka Kitchen, and D3IL, DIA consistently improves final performance over existing diffusion-policy fine-tuning methods. Beyond final reward, DIA reaches successful states more efficiently and can shift farther from the pretrained behavior distribution, enabling it to discover more effective and efficient task-level strategies and subtask sequences that baseline methods fail to reach.

Chinese Translation

基于扩散的机器人策略在机器人操作中得到了广泛应用，通常通过行为克隆进行训练。然而，纯粹从示范中训练的策略受到可用数据的质量和覆盖范围的限制。强化学习可以通过交互进一步提高这些预训练策略的性能。一种常见的方法是使用策略梯度方法，将扩散策略的微调形式化为外部环境马尔可夫决策过程（MDP），并结合内部去噪MDP。然而，现有方法通常对构建动作块所使用的所有去噪步骤分配相同的环境级别信用，而不区分哪些中间决策对最终回报的贡献最大。我们引入了去噪中间优势（Denoising Intermediate Advantage, DIA），这是一种策略梯度方法，通过部分去噪动作学习价值函数，并利用该函数为生成过程的每一步构建去噪级别的优势。DIA将这一内部信用信号与标准环境级别的PPO优势相结合，在去噪链中提供状态依赖的信用。在Robomimic、FurnitureBench、Franka Kitchen和D3IL上，DIA始终在现有的扩散策略微调方法中提高最终性能。除了最终奖励外，DIA更高效地达到成功状态，并能够更远离预训练的行为分布，从而使其能够发现基线方法无法达到的更有效和高效的任务级策略和子任务序列。

View on arXiv Download PDF AI Translation

cs.RO / 14 / 2609.12248

Trajectory Bundle Method in SE(3) for Black-Box Fixed-Wing Aircraft Trajectory Optimization

基于 SE(3) 的轨迹束方法用于黑箱固定翼飞机轨迹优化

Osburn, Matthew D., Peterson, Cameron K., Salmon, John L.

Abstract

Dynamically feasible trajectory optimization for rigid-body systems is naturally formulated on the special Euclidean group SE(3) but is challenging when dynamics are available only as black-box computations without derivatives. This paper formulates the Trajectory Bundle Method (TBM) for motion planning implicitly on SE(3). Bundles are constructed in the Lie algebra and propagated through nonlinear rigid-body dynamics using exponential and logarithmic maps, enabling derivative-free planning of non-Euclidean trajectories. We show that Euclidean TBM interpolation error is bounded quadratically by bundle diameter and extend this result to SE(3), where the bound additionally depends on a local Lipschitz constant of the Log map. Numerical experiments corroborate these bounds. Finally, we demonstrate SE(3) TBM by optimizing an acrobatic, collision-free fixed-wing maneuver through a rotated aperture without explicit models or derivatives of the vehicle dynamics, aerodynamics, or collision model.

Chinese Translation

刚体系统的动态可行轨迹优化自然地在特殊欧几里得群 SE(3) 上进行建模，但当动态信息仅以黑箱计算形式存在且没有导数时，这一过程会变得具有挑战性。本文在 SE(3) 上隐式地构建了轨迹束方法（Trajectory Bundle Method, TBM）用于运动规划。束在李代数中构建，并通过非线性刚体动力学利用指数和对数映射进行传播，从而实现无导数的非欧几里得轨迹规划。我们证明了欧几里得 TBM 插值误差与束直径的平方成正比，并将该结果扩展到 SE(3)，其中界限还依赖于对数映射的局部利普希茨常数。数值实验验证了这些界限。最后，我们通过优化一个穿过旋转孔径的特技、无碰撞的固定翼机动，展示了 SE(3) TBM，而无需显式的模型或车辆动力学、空气动力学或碰撞模型的导数。

View on arXiv Download PDF AI Translation

cs.RO / 15 / 2609.12258

Pneumatic neurons for soft robots enable inflate-and-fire networks for rhythmic motion

气动神经元为软机器人提供了充气-发射网络以实现节律运动

Li, Dongting, Tolley, Michael, Gravish, Nick

Abstract

Animals coordinate their movements through distributed neural circuits, but soft robots still typically depend on external, centralized electronics for control. Building soft robots that operate without centralized electronic controllers while remaining responsive to their environment remains a frontier challenge in soft robotics. In this work we introduce a soft-robot control architecture inspired by leaky integrate-and-fire models of biological neural circuits. The Pneumatic neuron (Pneu-ron) is a soft actuator that unifies energy conversion, logic, and actuation in one component. Each module combines a low-boiling-point fluid (LBF), a heater, and a mechanical switch into a self-excitable unit. Boiling the LBF inflates the module and triggers excitation and inhibition of adjacent modules in a process we call "inflate-and-fire". When interconnected into excitatory-inhibitory rings, Pneu-rons generate stable, sequential oscillations whose frequency emerges from the material dynamics and environmental conditions. By harnessing the inflation of Pneu-rons for actuation these networks can drive oscillatory locomotion of soft robots. Pneu-ron networks sustain oscillation under mechanical load and thermal variations, adapting through material physics rather than computation. Dynamical modeling of these networks reveals a dimensionless bifurcation diagram that dictates the network's oscillatory behavior. Encoding logic and actuation into material-level modules presents a new avenue for adaptive, electronics controller-free, soft robots.

Chinese Translation

动物通过分布式神经电路协调其运动，但软机器人通常仍依赖于外部集中电子设备进行控制。构建在没有集中电子控制器的情况下仍能对环境做出响应的软机器人，仍然是软机器人领域面临的前沿挑战。在本研究中，我们介绍了一种受生物神经电路的漏积分-发射模型启发的软机器人控制架构。气动神经元（Pneu-ron）是一种软执行器，将能量转换、逻辑和驱动整合为一个组件。每个模块将低沸点液体（LBF）、加热器和机械开关结合成一个自激励单元。加热LBF使模块充气，并触发相邻模块的兴奋和抑制，这一过程我们称之为“充气-发射”。当相互连接成兴奋-抑制环时，Pneu-rons产生稳定的顺序振荡，其频率源于材料动力学和环境条件。通过利用Pneu-rons的充气进行驱动，这些网络能够推动软机器人的振荡运动。Pneu-ron网络在机械负载和温度变化下维持振荡，通过材料物理而非计算进行适应。对这些网络的动态建模揭示了一个无量纲分岔图，决定了网络的振荡行为。将逻辑和驱动编码到材料级模块中，为自适应、无电子控制器的软机器人提供了一条新途径。

View on arXiv Download PDF AI Translation

cs.RO / 16 / 2609.12285

AnchorVLN: Geometry-Anchored Vision-Language Grounding Reasoning for Open-Vocabulary Navigation

AnchorVLN：基于几何锚定的开放词汇视觉-语言定位推理用于导航

Vu, Long Giang, Yao, Chengkai, Liu, Yuxin, Aryan, FNU, Aralikatti, Rajath Chandrashekar

Abstract

Vision-Language Navigation (VLN) in unseen indoor environments is useful in real-world robotics, where an agent must follow natural-language instructions, locate objects, and answer spatial questions without a pre-built map or fixed object vocabulary. Multimodal vision-language models (VLMs) provide strong open-vocabulary grounding and zero-shot reasoning, but struggle to emit reliable metric quantities such as range, bearing, and comparative spatial relations directly from images. Existing approaches address this by folding geometry into hand-engineered pipelines or asking models to output waypoints, requiring changes to the control stack for different robots, tasks, or vocabularies. We introduce AnchorVLN, an open-vocabulary VLN system built on a simple rule: the VLM proposes semantics; geometry decides metrics. It is realised as EMBODIED-NAV-MCP, a Model Context Protocol (MCP) server driven by a VLM agent through a compact set of callable tools. Since no tool accepts distance in metres or bearing in radians, the schema enforces the semantic-geometry boundary without modifying the downstream autonomy stack. We benchmark both tasks of the CMU Vision-Language Navigation Challenge 2026: 30 instruction-following questions over 15 scenes and a frozen 45-question object-reference set. The full system achieves 64.4 percent on instruction following, dropping by 13.3 percentage points without controller modeling (t = 2.77). On object reference, geometric anchoring clears the challenge overlap threshold on 10 of 45 questions, versus 0 of 45 for direct coordinate estimation, reducing median center error from 3.37 m to 2.48 m.

Chinese Translation

在未知室内环境中进行视觉-语言导航（VLN）对现实世界的机器人技术具有重要意义，代理必须遵循自然语言指令，定位物体，并在没有预先构建的地图或固定物体词汇的情况下回答空间问题。多模态视觉-语言模型（VLMs）提供了强大的开放词汇锚定和零样本推理能力，但在直接从图像中输出可靠的度量量（如范围、方位和比较空间关系）方面存在困难。现有方法通过将几何信息融入手工设计的管道或要求模型输出路径点来解决这一问题，这需要针对不同的机器人、任务或词汇对控制堆栈进行更改。我们提出了AnchorVLN，一个基于简单规则的开放词汇VLN系统：VLM提出语义，几何决定度量。该系统实现为EMBODIED-NAV-MCP，一个由VLM代理驱动的模型上下文协议（MCP）服务器，通过一组紧凑的可调用工具进行操作。由于没有工具接受以米为单位的距离或以弧度为单位的方位，该架构在不修改下游自主堆栈的情况下强制执行语义-几何边界。我们对2026年CMU视觉-语言导航挑战的两个任务进行了基准测试：在15个场景中进行30个指令跟随问题和一个冻结的45个问题的物体引用集。完整系统在指令跟随任务中达到了64.4%的准确率，而在没有控制器建模的情况下下降了13.3个百分点（t = 2.77）。在物体引用任务中，几何锚定在45个问题中有10个超过了挑战重叠阈值，而直接坐标估计则为0，减少了中位数中心误差从3.37米降至2.48米。

View on arXiv Download PDF AI Translation

cs.RO / 17 / 2609.12292

Mission Performance: Automatic and Adaptive Race Pace Progression for Autonomous Racing

任务性能：自动和自适应的自主赛车赛道进度管理

Lambertini, Giovanni, Pini, Matteo, Musiu, Nicola, Raji, Ayoub, Iacovacci, Francesco, Bertogna, Marko

Abstract

In this paper, we describe the Mission Performance module implemented for a fully autonomous racing car to automatically manage the longitudinal, lateral, and combined performances, aiming to speedup the laptime progression while assuring safety. Motivated by the difficulty and risks of applying the real-time estimation of the grip to critical modules like the motion planner and controller, the Mission Performance guides these modules adapting their target performance instead of changing the vehicle model parameters. The module is formed by pre-defined progressions to warm up the tires at the beginning of a run. Then, the system continuously monitors safety and vehicle dynamics metrics on a per-sector basis to adaptively reduce, maintain, or increase the performance levels for each sector, progressively converging toward the maximum allowed value. The solution's effectiveness is demonstrated on the EAV-25, a fully autonomous Dallara Superformula, at the Yas Marina Circuit during the Abu Dhabi Autonomous Racing League (A2RL) Season 2.

Chinese Translation

本文描述了为完全自主赛车实现的任务性能模块，该模块旨在自动管理纵向、横向和综合性能，以加快圈速进度，同时确保安全。由于实时估计抓地力对运动规划器和控制器等关键模块的应用存在困难和风险，任务性能模块指导这些模块调整其目标性能，而不是改变车辆模型参数。该模块由预定义的进度组成，以在比赛开始时为轮胎预热。随后，系统持续监测每个赛段的安全性和车辆动态指标，以自适应地降低、维持或提高每个赛段的性能水平，逐步接近允许的最大值。该解决方案的有效性在阿布扎比自主赛车联赛（A2RL）第二季期间于雅斯码头赛道上的完全自主Dallara Superformula赛车EAV-25上得到了验证。

View on arXiv Download PDF AI Translation

cs.RO / 18 / 2609.12316

DATAFARM: Distribution-Aligned Task and Motion Planning for Fine-Tuning Vision-Language-Action Models

DATAFARM：用于微调视觉-语言-动作模型的分布对齐任务与运动规划

Sahoo, Samrat, Huang, Yixuan, Silver, Tom

Abstract

Collecting high-quality robot data remains a fundamental challenge for training robot foundation models. Task and motion planning (TAMP) offers a scalable way to generate demonstrations, but our experiments show that raw TAMP trajectories provide surprisingly little benefit when used to fine-tune pretrained vision-language-action (VLA) models, despite successfully solving the target tasks. We hypothesize that this failure arises from a behavioral distribution mismatch between planner-generated trajectories and the data used to pretrain the VLA. To address this mismatch, we introduce DATAFARM: Distribution-Aligned Task And motion planning for Fine-tuning A Robot foundation Model, an approach that incorporates the pretraining distribution directly into TAMP trajectory generation. DATAFARM aligns generated trajectories with the pretraining data in robot joint configurations, motion style, and temporal execution profiles. We evaluate DATAFARM on three tabletop manipulation tasks that TAMP can perform and a cloth-folding task beyond the capability of TAMP. DATAFARM achieves an average success rate of 56.7%, substantially outperforming raw TAMP (8.3%) while approaching human teleoperation (61.7%). On Deformable Object Manipulation, which is outside the fine-tuning distribution, the fine-tuned model retains 85% success, compared with 90% for the pretrained model. These results show that aligning planner-generated demonstrations with the pretraining distribution can make TAMP an effective source of data for VLA fine-tuning. Website and code: https://prpl-group.com/datafarm/

Chinese Translation

收集高质量的机器人数据仍然是训练机器人基础模型的一项基本挑战。任务与运动规划（TAMP）提供了一种可扩展的生成演示的方法，但我们的实验表明，尽管成功解决了目标任务，原始的 TAMP 轨迹在微调预训练的视觉-语言-动作（VLA）模型时却提供了意想不到的少量益处。我们假设，这一失败源于规划者生成的轨迹与用于预训练 VLA 的数据之间的行为分布不匹配。为了解决这一不匹配，我们引入了 DATAFARM：用于微调机器人基础模型的分布对齐任务与运动规划，这是一种将预训练分布直接融入 TAMP 轨迹生成的方法。DATAFARM 在机器人关节配置、运动风格和时间执行特征上将生成的轨迹与预训练数据对齐。我们在 TAMP 可以执行的三个桌面操作任务和一个超出 TAMP 能力的折布任务上评估了 DATAFARM。DATAFARM 的平均成功率为 56.7%，显著优于原始 TAMP（8.3%），接近人类遥控操作（61.7%）。在可变形物体操作任务中，该任务超出了微调分布，微调后的模型保持了 85% 的成功率，而预训练模型为 90%。这些结果表明，将规划者生成的演示与预训练分布对齐可以使 TAMP 成为 VLA 微调的有效数据来源。网站和代码：https://prpl-group.com/datafarm/

View on arXiv Download PDF AI Translation

cs.RO / 19 / 2609.12347

DWMP: Leveraging Dual World Models for Humanoid Obstacle Traversal

DWMP：利用双重世界模型进行类人机器人障碍物穿越

Jin, Rongjun, Ma, Jianming, Gao, Yue

Abstract

Humanoid robots must traverse cluttered obstacle fields using onboard proprioceptive and visual observations, yet existing methods usually process multimodal observations without explicitly considering their different characteristics: proprioceptive observations are low-dimensional but governed by highly nonlinear robot dynamics, while egocentric visual observations are high-dimensional, noisy, and redundant. We propose DWMP (Dual World Model Policy), a framework that provides the actor with separate but complementary world-model representations for humanoid obstacle traversal. A Koopman-based dynamics world model lifts proprioceptive observations into a latent space where their temporal evolution is approximately linear, making the dynamics features easier for the actor to learn from. An RSSM-based visual world model compresses egocentric depth observations into compact stochastic states while preserving obstacle-related geometry. The student policy receives the fused latent representation for action generation, combining linearized proprioceptive dynamics with compressed visual perception. Experiments in simulation and on a Unitree G1 humanoid robot show that DWMP improves obstacle traversal performance over baselines and supports real-world deployment under randomized obstacle layouts.

Chinese Translation

类人机器人必须利用机载的本体感知和视觉观察穿越杂乱的障碍物场，但现有方法通常在处理多模态观察时未明确考虑它们的不同特性：本体感知观察是低维的，但受高度非线性的机器人动力学控制，而自我中心的视觉观察则是高维的、噪声较大且冗余。我们提出了DWMP（双重世界模型策略），这是一个为类人机器人障碍物穿越提供分离但互补的世界模型表示的框架。基于Koopman的动力学世界模型将本体感知观察提升到一个潜在空间，在该空间中，它们的时间演变近似线性，使得动力学特征更易于被执行者学习。基于RSSM的视觉世界模型将自我中心的深度观察压缩为紧凑的随机状态，同时保留与障碍物相关的几何信息。学生策略接收融合的潜在表示以生成动作，将线性化的本体感知动力学与压缩的视觉感知相结合。在仿真和Unitree G1类人机器人上的实验表明，DWMP在障碍物穿越性能上优于基线，并支持在随机障碍物布局下的实际部署。

View on arXiv Download PDF AI Translation

cs.RO / 20 / 2609.12349

A Deployable Architecture for Robot-Mediated Tasks (DART): Evaluation in Socially Assistive Robot-Guided Cognitive Behavioral Therapy Exercises

可部署的机器人介导任务架构 (DART)：在社会辅助机器人指导的认知行为疗法练习中的评估

Kian, Mina, Ignatova, Lydia, Wang, Jiong, Lee, Ji Min, Li, Jiancheng, Guo, Qianwei, Weiss, Emily, O'Connell, Amy, Zareno, Kaitlin, Li, Jiani, Patel, Reyna, George, Leyaa, Huang, Minyu, Yang, Justin, Matarić, Maja J.

Abstract

Socially assistive robots (SARs) can support structured health and well-being interventions, but hardware and cost constraints limit interaction complexity and longitudinal real-world deployments. We present DART: Deployable Architecture for Robot-Mediated Tasks, an architecture that extends SARs through a web application and cloud infrastructure, enabling visual content, user input, remote computation, and persistent data storage synergistically with the robot's physical embodiment, speech, and movement. We evaluated DART by instantiating it in an interatively-developed full-stack HRI system for helping university students with elevated generalized anxiety to complete cognitive behavioral therapy (CBT) homework exercises. The resulting system, which used the low-cost open-source Blossom robot platform, was refined and evaluated through a participatory design process and multiple user studies, and finally evaluated in an in-lab study with 103 participants, and then a six-week in-home deployment with four participants. In the in-lab evaluation, participants showed significant within-session reductions in stress, state anxiety, and negative affect, and gave the platform a mean System Usability Scale score of 78.89. In the home deployment, the mean System Usability Scale score was 87.5, with positive qualitative feedback on usability. Participants across both groups identified speech input, visual presentation, and web-robot synchronization as priorities for improvement. These findings validate DART as an effective architecture for extending the capabilities of a low-cost SAR in both in-lab single-session and in real-world longitudinal deployments.

Chinese Translation

社会辅助机器人 (SARs) 可以支持结构化的健康和福祉干预，但硬件和成本限制了交互的复杂性和长期的实际部署。我们提出了 DART：可部署的机器人介导任务架构，这是一种通过网络应用程序和云基础设施扩展 SAR 的架构，能够与机器人的物理体现、语音和运动协同工作，实现视觉内容、用户输入、远程计算和持久数据存储。我们通过在一个迭代开发的全栈人机交互 (HRI) 系统中实例化 DART，对其进行了评估，旨在帮助大学生完成认知行为疗法 (CBT) 作业练习，特别是那些具有较高广泛性焦虑的学生。最终的系统使用了低成本的开源 Blossom 机器人平台，通过参与式设计过程和多次用户研究进行了优化和评估，并最终在一个包含 103 名参与者的实验室研究中进行了评估，随后在一个为期六周的家庭部署中进行了测试。在实验室评估中，参与者在会话内表现出显著的压力、状态焦虑和负面情绪的减少，并给予该平台平均 78.89 的系统可用性评分。在家庭部署中，平均系统可用性评分为 87.5，用户对可用性给予了积极的定性反馈。两个组的参与者均认为语音输入、视觉呈现和网络-机器人同步是改进的优先事项。这些发现验证了 DART 作为一种有效架构的能力，能够在实验室单次会话和实际长期部署中扩展低成本 SAR 的功能。

View on arXiv Download PDF AI Translation

cs.RO / 21 / 2609.12371

READ: Learning Risk-Informed Fields for End-to-End Autonomous Driving

READ：学习风险信息场以实现端到端自动驾驶

Liu, Zhiyuan, Tian, Yuanxin, Ke, Zehong, Li, Jinhao, Cheng, Hao, Xu, Zhenhua, Yu, Wenhao, Wang, Jianqiang

Abstract

Autonomous driving requires more than recognizing what is present in a scene: a planner must determine how road structure, surrounding agents, and their motion states should influence a future maneuver. Existing learning-based planners can capture these influences through latent scene features and trajectory decoders, but the relationship between environmental factors and candidate actions often remains implicit. This limits the ability to inspect, diagnose, or refine how scene context affects the safety of a predicted trajectory. Classical safety fields provide an explicit spatial representation of this relationship, but their risk shapes and relative weights are prescribed in advance and do not adapt to each scene. We introduce READ, a framework that learns an explicit, planning-aligned risk representation from complementary geometric and behavioral constraints. READ instantiates this representation as a continuous spatiotemporal field, enabling differentiable queries along candidate trajectories. The learned field connects scene understanding with action selection by encouraging predicted trajectories to align with low-risk regions, while retaining a differentiable interface for trajectory evaluation and refinement. READ integrates with both end-to-end planners and Vision-Language-Action models. Experiments on NAVSIM show consistent gains across matched end-to-end backbones and strong performance in a VLA setting; READ also achieves competitive results on NAVSIM v2. These results establish learned spatial risk as an explicit, adaptable representation for safe planning.

Chinese Translation

自动驾驶不仅需要识别场景中的物体：规划者还必须确定道路结构、周围代理及其运动状态如何影响未来的操作。现有的基于学习的规划器可以通过潜在场景特征和轨迹解码器捕捉这些影响，但环境因素与候选动作之间的关系往往仍然是隐含的。这限制了检查、诊断或优化场景上下文如何影响预测轨迹安全性的能力。经典的安全场提供了这种关系的显式空间表示，但它们的风险形状和相对权重是预先规定的，无法适应每个场景。我们提出了READ，一个从互补的几何和行为约束中学习显式、与规划对齐的风险表示的框架。READ将这种表示实例化为一个连续的时空场，使得沿候选轨迹的可微查询成为可能。学习到的场将场景理解与动作选择连接起来，鼓励预测轨迹与低风险区域对齐，同时保留用于轨迹评估和优化的可微接口。READ可以与端到端规划器和视觉-语言-动作（Vision-Language-Action，VLA）模型集成。在NAVSIM上的实验显示，在匹配的端到端骨干网络中取得了一致的提升，并在VLA设置中表现出强大的性能；READ在NAVSIM v2上也取得了竞争性的结果。这些结果确立了学习的空间风险作为安全规划的显式、可适应的表示。

View on arXiv Download PDF AI Translation

cs.RO / 22 / 2609.12384

Understanding Whole-Body Robot Teleoperation Strategies Under Diverse Task Objectives and Constraints

理解多样任务目标和约束下的全身机器人遥操作策略

Lin, Tsung-Chi, Chen, Juo-Tung, Huang, Chien-Ming

Abstract

This work investigates the control strategies of complex whole-body robot teleoperation that coordinate active perception, bimanual manipulation, and navigation. We developed a hybrid control framework, combining the free-form and constrained control, for the whole-body teleoperation of the TIAGo mobile manipulator. We conducted a user study to explore people's control strategies under different task constraints such as limited time and low tolerance of errors. Our results highlight the effective use of coordinated control in improving task efficiency and reducing the risk of reaching individual joint limits. We discuss our results and their implications for designing future whole-body robot teleoperation systems.

Chinese Translation

本研究探讨了复杂全身机器人遥操作的控制策略，该策略协调主动感知、双手操作和导航。我们开发了一种混合控制框架，结合了自由形式控制和约束控制，用于TIAGo移动操纵器的全身遥操作。我们进行了用户研究，以探索在有限时间和低容错率等不同任务约束下，人们的控制策略。我们的结果强调了协调控制在提高任务效率和降低达到各个关节极限风险方面的有效性。我们讨论了我们的结果及其对未来全身机器人遥操作系统设计的影响。

View on arXiv Download PDF AI Translation

cs.RO / 23 / 2609.12433

FoldNet++: a Large-Scale Synthetic Dataset for Robotic T-Shirt Folding and Unfolding

FoldNet++：用于机器人T恤折叠和展开的大规模合成数据集

Chen, Yuxing, Wei, Zhiyuan, Xiao, Bowen, Zhang, Zhizheng, Wang, He

Abstract

Due to the highly deformable nature of garments, training a generalizable policy for robotic T-shirt folding and unfolding remains a significant challenge. In this work, we present a large-scale synthetic dataset for robotic T-shirt folding and unfolding, covering 6 robotic embodiments, 1K T-shirts, 1K environmental assets, and 120K episodes with rich annotations, which can be used to train a wide range of manipulation policies. We first follow the FoldNet pipeline to generate a large-scale dataset of physically simulatable T-shirts with diverse appearances and annotated semantic keypoints. Based on these semantic keypoints, we then generate manipulation demonstrations for different robotic embodiments through a unified rule-based framework. We use these demonstrations to train visuomotor policies, and experimental results demonstrate that models trained solely on our synthetic data can achieve over 90\% end-to-end task success rates when directly deployed to unseen real-world environments and previously unseen T-shirts from arbitrary initial configurations. Project URL: https://pku-epic.github.io/FoldNetXX/.

Chinese Translation

由于服装具有高度可变形的特性，为机器人T恤折叠和展开训练一个可泛化的策略仍然是一个重大挑战。在本研究中，我们提出了一个大规模的合成数据集，用于机器人T恤的折叠和展开，涵盖了6种机器人形态、1000件T恤、1000个环境资产和120,000个富含注释的实验场景，可用于训练广泛的操作策略。我们首先遵循FoldNet流程生成一个具有多样外观和注释语义关键点的物理可模拟T恤的大规模数据集。基于这些语义关键点，我们通过统一的基于规则的框架生成不同机器人形态的操作演示。我们使用这些演示来训练视觉运动策略，实验结果表明，单独基于我们的合成数据训练的模型在直接部署到未见过的真实环境和之前未见过的T恤时，能够实现超过90%的端到端任务成功率，且适用于任意初始配置。项目网址：https://pku-epic.github.io/FoldNetXX/

View on arXiv Download PDF AI Translation

cs.RO / 24 / 2609.12441

IMPLY: Physically Anchored Consistency for World-Model Rollouts

IMPLY：物理锚定的一致性用于世界模型的展开

Mehta, Aman, Baviskar, Riya

Abstract

A world model asked what happens if an object is pushed at several speeds produces several futures. If the model has the object in mind, those futures agree about it: each implies the same mass and friction. The consistency checks now used to vet world-action models ask whether a model's futures agree with each other, and none of them knows any physics. We show that this is not enough, and what to do instead. IMPLY reads the physics each rollout implies by inverting a simulator and scores a set of rollouts by how well one object explains all of them, anchored to two calibration pushes the model has observed. In a controlled setting, self-consistency gives a perfect score to a model that ignores the object and always predicts a typical push; anchoring exposes it (AUROC 0.70 versus 1.00). On a real model, V-JEPA 2-AC adapted to the scene, the same thing happens. Given its own calibration pushes the model tracks the object (per-object correlation with the truth 0.91); given another object's, it does not (0.05). Self-consistency cannot tell these apart, preferring the right evidence on 52% of objects, chance level, while anchored disagreement prefers it on 73% and correlates 0.92-0.99 with the rollouts' error. Used to choose among candidate rollout sets, it comes within 0.003 of an oracle that sees the truth. A model that has internalised the wrong object is exactly as self-consistent as one that has internalised the right one; consistency has to be anchored to evidence.

Chinese Translation

一个世界模型在多个速度下推送物体时，会产生多个未来。如果模型考虑到了该物体，这些未来之间是一致的：每个未来都暗示相同的质量和摩擦力。目前用于验证世界行动模型的一致性检查只关注模型的未来是否彼此一致，而这些模型并不具备物理知识。我们展示了这并不够，以及该如何改进。IMPLY通过反转一个模拟器来读取每个展开所暗示的物理，并根据一个物体对所有展开的解释能力进行评分，锚定于模型观察到的两个校准推力。在一个受控环境中，自我一致性为一个忽略物体并始终预测典型推力的模型打出了完美分数；而锚定则揭示了这一点（AUROC 0.70 对比 1.00）。在一个真实模型中，适应于场景的 V-JEPA 2-AC 也出现了相同的情况。考虑到自身的校准推力，模型能够跟踪物体（与真实值的每个物体相关性为 0.91）；而在考虑另一个物体的情况下，则无法做到（0.05）。自我一致性无法区分这两者，偏好正确证据的比例为 52%，相当于随机水平，而锚定的不一致性在 73% 的情况下偏好正确证据，并与展开的误差相关性为 0.92-0.99。用于在候选展开集中进行选择时，它的结果与一个能够看到真实情况的神谕相差仅 0.003。一个内化了错误物体的模型与一个内化了正确物体的模型在自我一致性上是完全相同的；一致性必须锚定于证据。

View on arXiv Download PDF AI Translation

cs.RO / 25 / 2609.12456

PATH: Continuous Target Sensing among Autonomous Cooperative Drones

PATH：自主协作无人机间的连续目标感知

Kim, Heegyeong, James, Alice, Seth, Avishkar, Kuantama, Endrowednes, Williamson, Jane, Feng, Yimeng, Han, Richard

Abstract

Continuous target sensing by uncrewed aerial vehicles (UAVs) is constrained by limited flight endurance, motivating the transfer of tracking responsibility between cooperating UAVs. Such a handoff requires the receiver to identify the same physical target currently tracked by the sender despite differences in viewpoint, scale, and target appearance. Existing approaches based on global target localization or appearance-based cross-view association are limited by positioning uncertainty or ambiguous visual features. This paper presents Perspective Alignment \& Tracking Handoff (\textbf{PATH}), a platform-agnostic, geometry-assisted sensing and verification framework for target handoff between two moving UAVs. The sender reconstructs the tracked target as a metric 3D point using RGB-D sensing, while the receiver estimates its relative pose from a fiducial observation and projects the transmitted target point into its own image as a spatial prior for target acquisition. The receiver-generated candidate is then returned to the sender and verified through a cross-view Mutual Agreement Handshake before tracking responsibility is transferred. Real-world UAV experiments show mean relative-position and target-position errors of 0.047~m and 0.030~m, respectively. Under visually ambiguous conditions, PATH achieves 96.0\% frame-level receiver-side target acquisition accuracy, with 2.0\% false-positive and 2.0\% false-negative rates. A sensor-error sensitivity analysis shows that relative-pose uncertainty is the dominant contributor to receiver-view projection error. The implementation operates at video rate with compact inter-UAV communication below 16~kB/s at 60~Hz, demonstrating the feasibility of lightweight geometry-assisted target handoff on resource-constrained UAV platforms.

Chinese Translation

无人机（UAV）进行连续目标感知受到有限飞行耐力的限制，这促使在协作无人机之间转移跟踪责任。这种交接要求接收方在视角、尺度和目标外观存在差异的情况下，识别当前由发送方跟踪的同一物理目标。基于全球目标定位或基于外观的跨视图关联的现有方法受到定位不确定性或模糊视觉特征的限制。本文提出了视角对齐与跟踪交接（Perspective Alignment & Tracking Handoff，简称PATH），这是一个平台无关的、几何辅助的目标交接感知与验证框架，适用于两架移动无人机之间的目标交接。发送方使用RGB-D传感器重建被跟踪目标为一个度量3D点，而接收方则从一个基准观察中估计其相对姿态，并将传输的目标点投影到其自身图像中，作为目标获取的空间先验。接收方生成的候选目标随后返回给发送方，并通过跨视图的相互一致握手进行验证，之后再转移跟踪责任。实际无人机实验显示，平均相对位置误差和目标位置误差分别为0.047米和0.030米。在视觉模糊条件下，PATH实现了96.0%的帧级接收方目标获取准确率，假阳性率和假阴性率均为2.0%。传感器误差敏感性分析表明，相对姿态不确定性是接收视图投影误差的主要贡献因素。该实现以视频速率运行，且在60Hz下的无人机间通信量低于16kB/s，展示了在资源受限的无人机平台上轻量级几何辅助目标交接的可行性。

View on arXiv Download PDF AI Translation

cs.RO / 26 / 2609.12498

ArtManip: Category-Level Articulated In-Hand Manipulation

ArtManip：类别级别的关节式手中操控

Yang, Yang, Liu, Tengyu, Li, Puhao, Chen, Zeyuan, Li, Yuyang, Wang, Xingwan, Wu, Yingying, Cui, Zhaopeng, Huang, Siyuan

Abstract

Category-level in-hand manipulation of articulated objects is a formidable yet underexplored challenge for dexterous robotic hands. This difficulty stems from two core bottlenecks: first, controlling an object's internal degrees of freedom is tightly coupled with maintaining grasp stability on a free-floating base; second, acquiring diverse object models and functional grasps at scale is highly labor-intensive, yet vital for generalization given the system's sensitivity to initial configurations. In this work, we present ArtManip, the first category-level articulated in-hand manipulation method that generalizes across object instances and diverse initial grasps. For initial configuration construction, we develop an automated pipeline that procedurally generates diverse articulated objects and synthesizes task-oriented functional grasps. For policy learning, we propose a robust two-stage training strategy that incorporates articulation physics randomization, reward curriculum, and latent representation distillation to handle complex contact and joint dynamics during deployment. Extensive experiments across four object categories demonstrate that our policy generalizes to unseen instances and varied configurations in simulation, and achieves zero-shot transfer to 12 real-world objects featuring diverse shapes and joint mechanics.

Chinese Translation

关节物体的类别级别手中操控是灵巧机器人手面临的一项艰巨但尚未深入探索的挑战。这一困难源于两个核心瓶颈：首先，控制物体的内部自由度与维持自由浮动基座上的抓取稳定性紧密相关；其次，大规模获取多样的物体模型和功能性抓取是高度劳动密集型的，但鉴于系统对初始配置的敏感性，这对于泛化至关重要。在本研究中，我们提出了ArtManip，这是首个能够在物体实例和多样初始抓取之间进行泛化的类别级别关节式手中操控方法。为了构建初始配置，我们开发了一个自动化流程，程序性地生成多样的关节物体并合成面向任务的功能性抓取。在策略学习方面，我们提出了一种稳健的两阶段训练策略，该策略结合了关节物理随机化、奖励课程和潜在表示蒸馏，以处理部署过程中复杂的接触和关节动力学。在四个物体类别上的广泛实验表明，我们的策略能够在模拟中对未见实例和不同配置进行泛化，并成功实现对12个具有多样形状和关节力学的真实物体的零样本迁移。

View on arXiv Download PDF AI Translation

cs.RO / 27 / 2609.12502

Communication-Constrained Multi-Robot Exploration With Adaptive Communication Windows

具有自适应通信窗口的通信受限多机器人探索

Rossano, Ben, Lim, Jaein, How, Jonathan P.

Abstract

Exploring unknown environments with multi-robot teams can improve efficiency by allowing robots to explore in parallel. However, realizing these gains requires effective information sharing. When communication is intermittent, robots must balance the benefits of sharing information against the cost of diverting from exploration to establish communication. This paper introduces MACE, a decentralized exploration framework that actively evaluates whether establishing communication is worthwhile. At scheduled communication windows, robots estimate the cost of reaching previously identified communication locations. By formulating this decision as a variant of the Vehicle Orienteering Problem, robots evaluate routes based on the travel required to establish communication and the exploration that can be completed along the way. This approach enables robots to communicate more frequently than under purely opportunistic strategies while reducing the unnecessary travel associated with fixed rendezvous strategies. Across a set of simulated environments with varying size and geometry, we demonstrate that MACE reduces the total exploration time by up to 23% compared to existing communication-constrained exploration strategies.

Chinese Translation

利用多机器人团队探索未知环境可以通过允许机器人并行探索来提高效率。然而，实现这些收益需要有效的信息共享。当通信不稳定时，机器人必须在共享信息的好处与为了建立通信而偏离探索的成本之间进行权衡。本文介绍了MACE（多机器人自适应通信框架），这是一个去中心化的探索框架，能够主动评估建立通信是否值得。在预定的通信窗口期，机器人估算到达先前识别的通信位置的成本。通过将这一决策形式化为车辆定向问题（Vehicle Orienteering Problem）的变体，机器人根据建立通信所需的旅行和沿途可以完成的探索来评估路线。这种方法使机器人能够比纯粹的机会主义策略更频繁地进行通信，同时减少与固定会合策略相关的不必要旅行。在一组具有不同大小和几何形状的模拟环境中，我们展示了MACE相比于现有的通信受限探索策略，能够将总探索时间减少多达23%。

View on arXiv Download PDF AI Translation

cs.RO / 28 / 2609.12530

Autonomous Precision Milling of Biological Structures via Generic Anatomical Priors and Active Boundary Perception

通过通用解剖先验和主动边界感知实现生物结构的自主精密铣削

Zhao, Enduo, Lin, Xiaofeng, Wang, Yifan, Song, Yuhan, Li, Weihan, Perez, Saul Alexis Heredia, Harada, Kanako

Abstract

Autonomous precision milling of biological structures is challenged by incomplete knowledge of target geometry, local material thickness, and critical internal boundaries. Subject-specific preoperative models can address geometric and thickness variations, but static models cannot determine boundary status encountered during execution, while repeated target-specific imaging limits scalability. This article presents an uncertainty-aware autonomous milling framework that assigns complementary roles to generic anatomical priors and active boundary perception. A generic anatomical prior provides conservative global guidance and is transformed through semantic-guided registration and hybrid vision-force calibration into robot-executable guidance for individual targets. As milling approaches uncertain boundaries, the robot actively probes the remaining structure and uses relative stiffness changes to estimate boundary status and structural detachability. A state-adaptive controller governs transitions between active perception and spatially selective incremental refinement, repeating this cycle until the termination criterion is satisfied. Hierarchical experiments on biological surrogates and in vivo mouse cranial window creation demonstrate accurate anatomical prior transfer, reliable boundary adaptation, and autonomous precision milling of biological structures.

Chinese Translation

生物结构的自主精密铣削面临着目标几何形状、局部材料厚度和关键内部边界知识不完全的挑战。特定于个体的术前模型可以解决几何形状和厚度的变化，但静态模型无法确定执行过程中遇到的边界状态，而重复的目标特定成像限制了可扩展性。本文提出了一种基于不确定性的自主铣削框架，该框架为通用解剖先验和主动边界感知分配互补角色。通用解剖先验提供保守的全局指导，并通过语义引导配准和混合视觉-力校准转化为可供机器人执行的个体目标指导。当铣削接近不确定边界时，机器人主动探测剩余结构，并利用相对刚度变化来估计边界状态和结构可分离性。状态自适应控制器管理主动感知与空间选择性增量细化之间的过渡，重复这一循环直到满足终止标准。在生物替代物和体内小鼠颅窗创建的分层实验中，展示了准确的解剖先验转移、可靠的边界适应以及生物结构的自主精密铣削。

View on arXiv Download PDF AI Translation

cs.RO / 29 / 2609.12549

STAR: Sparse Tactile Representation Learning in Vision-Tactile-Language-Action Models for Dexterous Manipulation

STAR：视觉-触觉-语言-动作模型中稀疏触觉表征学习用于灵巧操作

Liu, Xiangcheng, Wu, Tianhao, Zheng, Le, Wang, Yidong, Jiang, Bowen, Pan, Mingjie, Ren, Xinlin, Liu, Yi, Luo, Jianlan

Abstract

Dexterous manipulation requires coordinated multi-finger control and effective tactile feedback, yet learning these capabilities remains challenging due to the lack of large-scale real-world data and the difficulty of extracting effective representations from sparse tactile signals. We build a robot platform and teleoperation system to collect a 200-hour bimanual dexterous manipulation dataset with synchronized visual, tactile, and language annotations, comprising 10,576 trajectories across 65 tasks, 69.5% of which involve dexterous multi-finger manipulation. We further propose STAR, an integrated training recipe for vision-tactile-language-action (VTLA) models that addresses the spatial, temporal, and informational sparsity of tactile signals through visual-tactile joint pre-training, sparse-global tactile token representation, and sparse future tactile prediction. Trained on this dataset, STAR achieves a 61% average success rate across four real-world tasks with 100 post-training trajectories per task, demonstrating dexterous performance under task-specific post-training.

Chinese Translation

灵巧操作需要协调的多指控制和有效的触觉反馈，但由于缺乏大规模的真实世界数据以及从稀疏触觉信号中提取有效表征的困难，学习这些能力仍然具有挑战性。我们构建了一个机器人平台和遥操作系统，以收集一个包含同步视觉、触觉和语言注释的200小时双手灵巧操作数据集，该数据集涵盖65个任务的10,576条轨迹，其中69.5%的任务涉及灵巧的多指操作。我们进一步提出了STAR，这是一个针对视觉-触觉-语言-动作（VTLA）模型的综合训练方案，通过视觉-触觉联合预训练、稀疏-全局触觉标记表征和稀疏未来触觉预测，解决了触觉信号的空间、时间和信息稀疏性。在该数据集上训练后，STAR在四个真实世界任务中实现了61%的平均成功率，每个任务有100条后训练轨迹，展示了在特定任务后训练下的灵巧表现。

View on arXiv Download PDF AI Translation

cs.RO / 30 / 2609.12595

A Hierarchical Coverage Path Planning Algorithm for Unknown Environments

一种针对未知环境的分层覆盖路径规划算法

Shen, Zongyuan, Liu, Haodong, Wang, Gao, Ma, Hongbin, Ou, Yaming, Zhao, Shancheng, Zhou, Dehua

Abstract

This paper presents an online coverage path planning algorithm for unknown environments. During navigation, the initially unknown search area is progressively decomposed into disconnected subareas as new obstacle information is acquired and coverage proceeds. These subareas are organized in an incrementally constructed decomposition tree that preserves their hierarchical parent-child relationships. Based on this tree, a global coverage tour is maintained and updated online by prioritizing newly generated child subareas according to their exploration states and distances from the robot. A local planner then generates coverage motions within each selected subarea, allowing the robot to adapt its trajectory as the environment is gradually revealed. Its performance is evaluated via high-fidelity simulations in complex scenarios. The results show improved coverage efficiency in terms of path length and overlap ratio in comparison to three baseline algorithms.

Chinese Translation

本文提出了一种针对未知环境的在线覆盖路径规划算法。在导航过程中，最初未知的搜索区域随着新障碍信息的获取和覆盖的进行，逐步被分解为不相连的子区域。这些子区域在增量构建的分解树中组织，保持其层次化的父子关系。基于该树，全球覆盖巡航路径得以维护并在线更新，通过根据子区域的探索状态和与机器人之间的距离优先考虑新生成的子区域。然后，局部规划器在每个选定的子区域内生成覆盖运动，使机器人能够随着环境的逐步揭示而调整其轨迹。通过在复杂场景中的高保真模拟评估其性能。结果显示，与三种基线算法相比，该算法在路径长度和重叠比方面提高了覆盖效率。

View on arXiv Download PDF AI Translation

cs.RO / 31 / 2609.12609

Quantifying Spectral Differences in Vehicle Between Production Autonomous and Human-Driven Vehicles Across Driving Scenarios

量化生产型自主车辆与人类驾驶车辆在不同驾驶场景下的光谱差异

Fang, Peiyi, Li, Xiangyu, Weng, Yonglin, Ma, Ke

Abstract

Differences in vehicle kinematic characteristics between production autonomous vehicles (PAVs) and human-driven vehicles (HVs) have been limitedly investigated by empirical studies. Most recent studies rely on simulation-based models, while some further investigate low-level adaptive cruise control (ACC) systems in controlled experiments. These methods commonly adapt some time-domain metrics to characterize PAV-HV differences across limited driving conditions. However, current PAVs equipped with high-level autonomous driving systems generate driving behaviors in a black box using data-driven models. These fundamentally different mechanisms for generating behaviors may produce distinct kinematic characteristics in traffic. More importantly, these time-domain metrics cannot reflect frequency-related traffic dynamics across different driving scenarios. Thus, this study adapted a real-world PAV dataset with four PAV platforms and developed a frequency-domain framework to quantify kinematic differences between PAVs and HVs across diverse driving scenarios, including varying driving states, lighting, weather, and vehicle densities. The framework transforms kinematic signals into the frequency domain and extracts spectral features, and then compares these features between PAVs and HVs based on kernel density estimation and Wasserstein distance. The results reveal clear scenario-dependent PAV-HV spectral differences. Specifically, speed-related differences were consistently smaller during car-following than cruising, while rainy conditions consistently enlarged acceleration-related differences compared with clear conditions. These findings highlight the necessity of multi-scenario evaluations and demonstrate the value of frequency-domain analysis for characterizing PAV-HV kinematic differences under real-world conditions.

Chinese Translation

生产型自主车辆（PAVs）与人类驾驶车辆（HVs）之间的运动学特征差异在实证研究中受到的关注有限。最近的研究大多依赖于基于仿真的模型，而一些研究则在受控实验中进一步探讨低级自适应巡航控制（ACC）系统。这些方法通常采用一些时域指标来表征PAV与HV之间的差异，且仅限于有限的驾驶条件。然而，当前配备高水平自主驾驶系统的PAVs通过数据驱动模型在黑箱中生成驾驶行为。这些生成行为的根本不同机制可能在交通中产生独特的运动学特征。更重要的是，这些时域指标无法反映不同驾驶场景下与频率相关的交通动态。因此，本研究采用了一个包含四种PAV平台的真实世界PAV数据集，并开发了一个频域框架，以量化PAV与HV在多种驾驶场景下的运动学差异，包括不同的驾驶状态、光照、天气和车辆密度。该框架将运动学信号转换为频域并提取光谱特征，然后基于核密度估计和Wasserstein距离比较PAV与HV之间的这些特征。结果揭示了明显依赖场景的PAV-HV光谱差异。具体而言，在跟车时速度相关的差异始终小于巡航时，而在雨天条件下，与晴天条件相比，加速度相关的差异始终增大。这些发现强调了多场景评估的必要性，并展示了频域分析在表征真实世界条件下PAV-HV运动学差异中的价值。

View on arXiv Download PDF AI Translation

cs.RO / 32 / 2609.12614

ProClosure: Hierarchical Room-Object Assignment using Progressive Boundary Closure from Monocular Video

ProClosure：基于单目视频的渐进边界闭合的层次房间-物体分配

Muthuraj, Vinoth Kumar, Banik, Soumyadeep, Sharma, Kushal, Jain, Hardik

Abstract

A 3D scene graph groups objects into rooms. When a robot is asked to fetch an object from the kitchen, that grouping is what tells it where to look. An object recorded in the wrong room is not retrievable by a query naming the correct room. We introduce Progressive Boundary Closure, which recovers room layer from a monocular RGB video. A SLAM front end and an open-vocabulary segmenter supply a structural point cloud, camera trajectory and object tracks. The cloud is rasterised into a top-down map, rooms are recovered from it, and each object takes the room holding most of its extent. The difficulty lies in the map itself. Walls are recorded only where the camera looked, so a gap in the boundary may be a doorway or a stretch of wall that was never observed; nothing distinguishes the two. Prior methods treat both as passages, merging rooms that should remain separate. We observe that both require the same treatment: a room should not extend across either, so both are closed and need not be distinguished. Such an opening closes under a small amount of boundary growth, and few sightlines cross it, so points in different rooms rarely see one another. We use the first to recover rooms and the second to assign objects to them. Rooms are obtained by Progressively thickening the boundary inward and freezing each free-space region once it becomes enclosed, so every opening seals at its own scale rather than at a radius fixed in advance. Camera poses are used as seeds, which removes the sampling heuristic and makes the segmentation deterministic. Over 10 floors of 6 HM3D-Semantics scenes, scored against HOV-SG on identical top-down maps, we recover 74 rooms for 72 annotated regions (HOV-SG: 44), raising room F_1 from 0.741 to 0.890 at IoU 0.25 at some cost in precision, and object-to-room ARI from 0.488 to 0.696 (p=0.002, ahead on every floor).

Chinese Translation

3D场景图将物体分组到房间中。当机器人被要求从厨房取回物体时，这种分组告诉它该在哪里寻找。在错误的房间中记录的物体无法通过命名正确房间的查询进行检索。我们提出了渐进边界闭合（Progressive Boundary Closure），该方法从单目RGB视频中恢复房间层。SLAM前端和开放词汇分割器提供了结构点云、相机轨迹和物体轨迹。点云被栅格化为自上而下的地图，从中恢复房间，每个物体被分配到占据其大部分范围的房间。困难在于地图本身。墙壁仅在相机观察到的地方被记录，因此边界中的间隙可能是一个门口或一段从未观察到的墙壁；两者之间没有区别。先前的方法将两者视为通道，合并应保持分开的房间。我们观察到两者需要相同的处理：房间不应跨越任一边界，因此两者都被闭合且无需区分。这样的开口在小量边界增长下闭合，且很少有视线穿过它，因此不同房间中的点很少能相互看到。我们利用第一个来恢复房间，利用第二个来将物体分配给它们。房间是通过向内逐渐加厚边界并在每个自由空间区域被封闭后冻结而获得的，因此每个开口在其自身尺度上密封，而不是在预先固定的半径下。相机位姿被用作种子，这消除了采样启发式，使分割变得确定性。在6个HM3D-Semantics场景的10层楼中，与相同的自上而下地图上的HOV-SG进行评分，我们为72个注释区域恢复了74个房间（HOV-SG：44），在IoU 0.25时将房间F_1从0.741提高到0.890，尽管在精度上有所损失，物体与房间的ARI从0.488提高到0.696（p=0.002，在每层楼上均领先）。

View on arXiv Download PDF AI Translation

cs.RO / 33 / 2609.12634

Online Material Estimation for Conditioned Diffusion Policy in Shaping Deformable Linear Objects

针对变形线性物体的条件扩散策略的在线材料估计

Yamada, Ryunosuke, Motoda, Tomohiro, Domae, Yukiyasu, Tsuji, Tokuo

Abstract

Shape control of deformable linear objects (DLOs) is challenging for imitation learning because deformation behavior varies with material properties such as stiffness and elasticity, so a single policy must generate different action sequences for different objects even when the goal shape is identical. We propose a diffusion policy conditioned on material labels that are estimated online during manipulation. A recurrent estimation network predicts the material label of the grasped object from the time series of multi-view images and robot joint states, and the predicted label conditions the diffusion policy at every inference step. We collected 480 real-robot demonstrations covering four DLO materials and three groove-placement tasks, and compared per-material specialist policies, a task-conditioned policy without material labels, a policy conditioned on ground-truth material labels, and the proposed policy. Conditioning on ground-truth material labels improved the average success rate from 45.8% to 60.0% over the task-only policy, and the proposed policy reached 60.8% without any prior material information, matching the policy given ground-truth labels. A post-hoc analysis shows that the estimator extracts material-related information from the manipulation observations and that the diffusion policy responds to the resulting conditioning signal, while the one pronounced failure case is associated with persistent confusion between two similar materials.

Chinese Translation

变形线性物体（DLOs）的形状控制对于模仿学习来说具有挑战性，因为变形行为会随着材料属性（如刚度和弹性）的不同而变化，因此即使目标形状相同，单一策略也必须为不同物体生成不同的动作序列。我们提出了一种基于材料标签的扩散策略，该标签在操作过程中在线估计。一个递归估计网络从多视角图像和机器人关节状态的时间序列中预测被抓取物体的材料标签，并在每次推理步骤中将预测的标签作为扩散策略的条件。我们收集了480个真实机器人演示，涵盖四种DLO材料和三种槽位放置任务，并比较了每种材料的专家策略、没有材料标签的任务条件策略、基于真实材料标签的策略以及所提出的策略。基于真实材料标签的条件使得平均成功率从45.8%提高到60.0%，而所提出的策略在没有任何先前材料信息的情况下达到了60.8%，与基于真实标签的策略相匹配。后续分析表明，估计器从操作观察中提取了与材料相关的信息，而扩散策略对由此产生的条件信号作出了响应，唯一的明显失败案例与两种相似材料之间的持续混淆有关。

View on arXiv Download PDF AI Translation

cs.RO / 34 / 2609.12641

Breaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Models

打破视觉-动作捷径：用于通用机器人基础模型的潜在接口训练

Lin, Jianman, Shailesh, Shailesh, Luo, Zhongyi, Duan, Jiafei

Abstract

Robot foundation models achieve strong in-distribution performance but often degrade under visual distribution shifts. When learning to generate actions from pretrained visual representations, models may exploit task-irrelevant visual cues that correlate with demonstrated actions within the training distribution. Such vision-action shortcuts can undermine generalization when these correlations change under distribution shifts. Mitigating these shortcuts requires constraining how visual information is used for action generation while preserving task-relevant spatial information. We propose Latent Interface Training (LIT), a framework-agnostic two-stage strategy that first establishes a spatial-goal-conditioned action prior without images, then constrains visual conditioning through a pose-supervised latent interface. Stage 1 trains the action expert to generate action chunks conditioned on language, robot state, and each demonstrated chunk's terminal SE(3) end-effector pose, learning goal-directed action generation independently of visual cues. Stage 2 introduces a latent interface that aggregates visual and semantic representations and serves as the pretrained action expert's only visual conditioning pathway. The interface is supervised to reconstruct the terminal pose previously used to condition Stage 1, encouraging it to retain the goal-relevant spatial information needed for action generation. Across four vision-language-action and world-action architectures (Pi0.5, MolmoAct2, FAST-WAM, and ImageWAM), LIT improves overall LIBERO-Plus success by 3.87-10.70 percentage points while preserving or improving average LIBERO success. Real-world evaluations show 13.30-16.70 percentage-point gains in success aggregated across three tasks under unseen camera configurations, lighting variations, and distractors.

Chinese Translation

机器人基础模型在训练分布内表现优异，但在视觉分布变化下性能常常下降。在从预训练视觉表征学习生成动作时，模型可能利用与训练分布内示范动作相关的任务无关视觉线索。这种视觉-动作捷径在分布变化时会削弱模型的泛化能力。缓解这些捷径需要在保留任务相关空间信息的同时，限制视觉信息在动作生成中的使用。我们提出了潜在接口训练（Latent Interface Training，LIT），这是一种与框架无关的两阶段策略：首先建立一个不依赖图像的空间-目标条件动作先验，然后通过姿态监督的潜在接口约束视觉条件。第一阶段训练动作专家基于语言、机器人状态及每个示范动作片段的终端SE(3)末端执行器姿态生成动作片段，独立于视觉线索学习目标导向的动作生成。第二阶段引入潜在接口，聚合视觉和语义表征，作为预训练动作专家唯一的视觉条件路径。该接口通过重构第一阶段用于条件的终端姿态进行监督，促使其保留动作生成所需的目标相关空间信息。在四种视觉-语言-动作及世界-动作架构（Pi0.5、MolmoAct2、FAST-WAM和ImageWAM）上，LIT使整体LIBERO-Plus成功率提升3.87至10.70个百分点，同时保持或提升平均LIBERO成功率。真实环境评估显示，在未见过的摄像机配置、光照变化及干扰物条件下，三项任务的成功率提升13.30至16.70个百分点。

View on arXiv Download PDF AI Translation

cs.RO / 35 / 2609.12660

Driving Context-guided Model Predictive Planning and Control for Autonomous Car Racing at the Limit and Beyond

基于驾驶情境引导的模型预测规划与控制在极限及超越极限的自主赛车中的应用

Raji, Ayoub, Sacco, Federico, Musiu, Nicola, Bertogna, Marko

Abstract

This paper presents a Model Predictive Control-based motion planning and control pipeline for autonomous car racing capable of adapting to different driving contexts, such as overtaking, nominal driving, and countersteering. A Cost Blending state machine manages the identification of different driving contexts and the selection of their predefined weights to be applied to the Model Predictive Planning (MPP) and Control (MPC) modules. The two optimization-based solutions share the same problem formulation and model, differing only in horizon length, rate, tuning, and in their open-loop versus closed-loop approach to maximize the effectiveness of their interaction. The work is validated on the fully autonomous open-wheel racecar Superformula EAV-25, with a lap time achieved that is within 2% of the best human driver reference. The results demonstrate the capability of the solution in driving at the limit of handling, smoothly executing overtaking maneuvers, and quickly reacting to high oversteering conditions to recover the vehicle stability.

Chinese Translation

本文提出了一种基于模型预测控制（Model Predictive Control, MPC）的运动规划与控制管道，适用于自主赛车，能够适应不同的驾驶情境，如超车、正常驾驶和反打方向。一个成本混合状态机（Cost Blending state machine）管理不同驾驶情境的识别及其预定义权重的选择，这些权重将应用于模型预测规划（Model Predictive Planning, MPP）和控制模块。两种基于优化的解决方案共享相同的问题表述和模型，仅在时间范围、速率、调节和开放环路与闭环方法上有所不同，以最大化它们交互的有效性。该工作在完全自主的开放轮赛车Superformula EAV-25上进行了验证，所实现的圈速在最佳人类驾驶员参考值的2%以内。结果表明，该解决方案能够在极限操控下驾驶，平稳执行超车操作，并快速应对高过度转向条件以恢复车辆稳定性。

View on arXiv Download PDF AI Translation

cs.RO / 36 / 2609.12677

Size Doesn't Matter: Material-State Reinforcement Learning for Excavator Transferable Soil Manipulation

大小无关紧要：用于挖掘机可转移土壤操作的材料状态强化学习

Werner, Lennart, Eyschen, Pol, Costello, Sean, Micarelli, Pierluigi, Cramariuc, Andrei, Hutter, Marco

Abstract

Earthmoving tasks such as excavation, backfilling, or embankment construction require deliberate repositioning of deformable soil. For these tasks, human operators use all shovel faces, while autonomous systems so far are limited to excavation and dumping. Current methods often rely on heuristic models but do not incorporate soil mechanics. We address this shortcoming by using Reinforcement Learning in a GPU-parallelized Material Point Method particle simulation. Our controllers are conditioned on material state such as shape and compactness, enabling skills that use multiple contact faces of the tool and displace material both inside and outside of the shovel. To use the same learned weights across machines, our policies operate in a normalized end-effector space and are deployed through a calibrated machine interface. We evaluate this calibrated transfer on an 11.5t hydraulic excavator and a 500g tabletop robot. We validate performance through autonomous construction of a 42m long, 2.1m high embankment in 45min, executing 201 individual policy strokes without failure, retry, or operator intervention. In a direct comparison, the autonomous controller matches an expert operator's progression speed and produces a higher, more consistent embankment. Additional qualitative backfilling and compaction experiments demonstrate the material-state awareness and calibrated transfer across machines.

Chinese Translation

土方作业如挖掘、回填或堤坝建设需要对可变形土壤进行精确的重新定位。在这些任务中，人类操作员使用铲斗的所有面，而目前的自主系统仅限于挖掘和倾倒。当前的方法通常依赖启发式模型，但未能考虑土壤力学。我们通过在GPU并行化的材料点法（Material Point Method）粒子模拟中使用强化学习来解决这一不足。我们的控制器以材料状态（如形状和密实度）为条件，使得工具的多个接触面能够使用，并在铲斗内外移动材料。为了在不同机器上使用相同的学习权重，我们的策略在归一化的末端执行器空间中运行，并通过校准的机器接口进行部署。我们在一台11.5吨的液压挖掘机和一台500克的桌面机器人上评估了这种校准转移。我们通过在45分钟内自主构建一条42米长、2.1米高的堤坝，执行201次独立的策略操作而没有失败、重试或操作员干预来验证性能。在直接比较中，自主控制器的进度速度与专家操作员相匹配，并且产生了更高、更一致的堤坝。额外的定性回填和压实实验展示了材料状态意识和跨机器的校准转移。

View on arXiv Download PDF AI Translation

cs.RO / 37 / 2609.12721

Improving Imitation Learning Efficiency for Manipulation through Geometric Prior Pretraining

通过几何先验预训练提高操控的模仿学习效率

Iwakata, Shogo, Motoda, Tomohiro, Yamada, Ryosuke, Makihara, Koshi, Nakajo, Ryoichi, Tanaka, Keitaro, Murooka, Masaki, Mykhailyshyn, Roman, Kataoka, Hirokatsu, Morishima, Shigeo, Domae, Yukiyasu

Abstract

Applying an imitation learning policy to a new manipulation task usually requires collecting new demonstrations and retraining the model, which makes sample efficiency a practical concern. Pretraining on large-scale robot datasets is effective in this respect, but such datasets are costly to collect and train on, while data augmentation techniques typically require a new round of data generation and retraining for each task. A complementary question is what useful prior can be provided to a policy at negligible cost before any task-specific data are collected. In this study, we construct a geometric visual pretraining dataset in which each scene contains only a plane, an object, and a hand, and trajectories are generated automatically. The scenes contain neither textures nor backgrounds; pretraining primarily exposes the policy to the geometric relationship between the hand and the object. Furthermore, representing the hand as a cube avoids tailoring the dataset to a specific robot morphology. We evaluate this geometric prior using ACT on three simulated robots across five manipulation tasks each, as well as on three real-world robot tasks. Across many of these robot--task combinations, fine-tuning from the geometric prior achieves higher success rates in the early stages of training than training from scratch while using only a small number of task demonstrations. These results suggest that even highly simplified geometric scenes can provide a useful initialization that transfers across robots and to real-world tasks when task data are limited.

Chinese Translation

将模仿学习策略应用于新的操控任务通常需要收集新的示范并重新训练模型，这使得样本效率成为一个实际问题。在这方面，对大规模机器人数据集的预训练是有效的，但此类数据集的收集和训练成本高昂，而数据增强技术通常需要为每个任务进行新一轮的数据生成和重新训练。一个互补的问题是，在收集任何特定任务数据之前，可以以微不足道的成本向策略提供哪些有用的先验知识。在本研究中，我们构建了一个几何视觉预训练数据集，其中每个场景仅包含一个平面、一个物体和一只手，轨迹是自动生成的。这些场景既没有纹理也没有背景；预训练主要使策略接触到手与物体之间的几何关系。此外，将手表示为一个立方体避免了将数据集定制为特定机器人形态。我们使用 ACT 在三个模拟机器人和五个操控任务上评估这一几何先验，同时也在三个真实世界的机器人任务上进行评估。在许多这些机器人-任务组合中，从几何先验进行微调在训练的早期阶段实现了比从头开始训练更高的成功率，同时仅使用少量任务示范。这些结果表明，即使是高度简化的几何场景也可以提供有用的初始化，能够在任务数据有限时跨机器人和真实世界任务进行迁移。

View on arXiv Download PDF AI Translation

cs.RO / 38 / 2609.12737

Control Architecture for Safe Grasping of Fragile Objects Using a Coarse Position-Controlled Gripper

基于粗位置控制夹持器的脆弱物体安全抓取控制架构

Pavlic, Marko, Geier, Moritz, Markert, Timo, Burschka, Darius

Abstract

Robots are increasingly used in unstructured environments. The need for them to safely grasp unknown objects without damaging them becomes crucial. Humans achieve this by sensing and quickly responding by adjusting their grasping force. Similarly, effective grasp acquisition in robots requires compliant interaction strategies that can adapt to uncertain object properties and adjust to any instabilities during manipulation. We present a geometry-aware force/torque-based contact estimation method for a coarse position-controlled gripper, combined with an adaptive admittance controller for safe grasp acquisition. The desired contact forces are estimated online to keep stable contact with objects of unknown properties. This enables compliant and stable grasps while avoiding excessive forces. Experiments with objects of different sizes, shapes, stiffnesses, and weights show that the proposed algorithm not only prevents slippage but also applies minimal force to safely grasp an object without causing excessive deformation.

Chinese Translation

机器人在非结构化环境中的应用日益增多。安全抓取未知物体而不造成损坏的需求变得至关重要。人类通过感知并快速反应来调整抓取力度，从而实现这一目标。同样，机器人有效的抓取获取也需要能够适应不确定物体特性并在操作过程中调整不稳定性的柔性交互策略。我们提出了一种基于几何感知的力/扭矩接触估计方法，适用于粗位置控制的夹持器，并结合自适应导纳控制器以实现安全抓取。所需的接触力在线估计，以保持与未知特性的物体的稳定接触。这使得抓取过程既柔性又稳定，同时避免施加过大的力。对不同尺寸、形状、刚度和重量的物体进行的实验表明，所提出的算法不仅能够防止滑动，还能以最小的力安全抓取物体，而不会造成过度变形。

View on arXiv Download PDF AI Translation

cs.RO / 39 / 2609.12795

High-Fidelity Multi-Body Simulator for Autonomous Racing

高保真多体仿真器用于自主赛车

Musiu, Nicola, Iacovacci, Francesco, Lupo, Fausto, Pini, Matteo, Scapicchi, Giovanni, Moretti, Francesco, Mascaro, Eugenio, Musso, Pietro, Raji, Ayoub, Bertogna, Marko, Arricale, Vincenzo Maria, Sapio, Angelo Lo, Piccarelli, Alessandro, Fish, Garron

Abstract

We present a custom high-fidelity vehicle dynamics simulation environment for testing and validation of Autonomous Racing software. The digital twin of the autonomous vehicle is developed in Dymola, using racecar dynamics modeling libraries to build a complete multi-body model. A 3D road surface, including elevation profiles and curbs, is implemented using the Curved Regular Grid (CRG) standard. The model is exported from Dymola as a Functional Mock-up Unit (FMU) and integrated into a custom software-in-the-loop simulator, where communication interfaces with the autonomous racing stack were developed in C++. A calibration procedure based on experimental data is also presented, along with a validation study to further support the quality of the proposed framework. The simulator runs in real time on a portable computer and provides reliable ground truth for algorithms validation prior to real-world deployment.

Chinese Translation

我们提出了一种定制的高保真车辆动力学仿真环境，用于测试和验证自主赛车软件。自主车辆的数字双胞胎在Dymola中开发，利用赛车动力学建模库构建完整的多体模型。采用曲面规则网格（Curved Regular Grid, CRG）标准实现了包含高程轮廓和路缘的3D路面。该模型从Dymola导出为功能模型单元（Functional Mock-up Unit, FMU），并集成到一个定制的软件在环仿真器中，其中与自主赛车堆栈的通信接口是用C++开发的。还提出了一种基于实验数据的校准程序，以及一项验证研究，以进一步支持所提框架的质量。该仿真器在便携式计算机上实时运行，并为算法验证提供可靠的真实数据，以便在实际部署之前进行验证。

View on arXiv Download PDF AI Translation

cs.RO / 40 / 2609.12831

VertexCBF: Improving Neural Control Barrier Functions via Vertex-Restricted Control Search

VertexCBF：通过顶点限制控制搜索改进神经控制屏障函数

Derajić, Bojan, Bernhard, Sebastian, Hönig, Wolfgang

Abstract

As the number of autonomous robots continues to grow, safety becomes increasingly important. Control barrier functions (CBFs) provide a theoretically grounded framework for ensuring safety, but existing design methods often face limitations in effectiveness, scalability, or interpretability, and may result in overly conservative safe sets. In this paper, we propose \emph{VertexCBF}, a framework for learning neural CBFs in a scalable, systematic, and explainable way. We approximate the stationary Hamilton--Jacobi value function using a neural network trained via a combination of physics-informed and sparsely supervised learning. By exploiting control-affine dynamics and a convex polytope control set, under which the Hamiltonian is maximized at the control vertices, we efficiently generate supervision points via GPU-parallel vertex-restricted tree search, while a residual architecture guarantees that the learned CBF is never larger than the specified constraint function. We evaluate the method on 15 systems and compare it against relevant baselines, showing that it reliably recovers large safe sets where the baselines are conservative or fail completely. In addition, we perform a hardware experiment in which a mobile robot safely avoids pedestrians using a neural CBF trained with our method.

Chinese Translation

随着自主机器人数量的不断增加，安全性变得愈发重要。控制屏障函数（CBFs）提供了一个理论基础框架来确保安全，但现有的设计方法往往在有效性、可扩展性或可解释性方面存在局限，可能导致过于保守的安全集合。在本文中，我们提出了 extit{VertexCBF}，一个以可扩展、系统化和可解释的方式学习神经控制屏障函数的框架。我们通过结合物理信息和稀疏监督学习训练的神经网络来近似静态哈密顿-雅可比值函数。通过利用控制仿射动态和凸多面体控制集，在该控制集下哈密顿量在控制顶点处达到最大值，我们通过GPU并行的顶点限制树搜索高效生成监督点，同时残差架构保证学习到的控制屏障函数不会超过指定的约束函数。我们在15个系统上评估该方法，并与相关基线进行比较，结果表明它能够可靠地恢复大型安全集合，而基线则显得保守或完全失效。此外，我们还进行了一个硬件实验，其中一台移动机器人使用我们的方法训练的神经控制屏障函数安全地避开行人。

View on arXiv Download PDF AI Translation

cs.RO / 41 / 2609.12837

Parameter Sensitivity Analysis for Aerial LiDAR-Inertial Odometries in low-altitude flights

低空飞行中空中激光雷达惯性里程计的参数敏感性分析

Milijas, Robert, Dios, Jose Ramiro Martinez-de, Bogdan, Stjepan

Abstract

LiDAR-based SLAM (Simultaneous Localization and Mapping) and LIO (LiDAR-inertial odometry) algorithms are often used for precise navigation of unmanned aerial vehicles, especially during interactions with the aerial robot's environment. However, the performance of these algorithms is greatly dependent on the scenario, LiDAR, and robot motion characteristics, often requiring an intensive tuning process to achieve the desired performance. To aid these tuning efforts, this paper analyzes the influence on performance of the parameters of an EKF-based LIO algorithm (FAST-LIO2) and the LIO module of a graph-based SLAM algorithm (Cartographer) on aerial LiDAR SLAM datasets recorded using different LiDARs in low-to-moderate-altitude flights in diverse environments. The analysis is conducted on the absolute trajectory error (ATE) resulting from processing the datasets with the LIO algorithms configured with each combination of parameters obtained in an exhaustive grid search. The relationship between individual parameters and the ATE results is assessed using Pearson's correlation, while the influence of each parameter is assessed using random forest permutation importance analyses with random forest models trained to predict the resulting ATE values based on the choice of parameters. The performed analysis obtains for Cartographer and FAST-LIO2: i) the identification of parameters with stronger influence in performance, ii) a simplified tuning procedure, and iii) tuning recommendations. Using the proposed tuning recommendations, both algorithms obtain on the analyzed datasets ATE values within 5 cm to the optimal performance found in the grid search procedure in 94% of the analyzed cases.

Chinese Translation

基于激光雷达的SLAM（同步定位与地图构建）和LIO（激光雷达惯性里程计）算法通常用于无人机的精确导航，特别是在与空中机器人环境交互时。然而，这些算法的性能在很大程度上依赖于场景、激光雷达和机器人运动特性，通常需要进行密集的调优过程以实现期望的性能。为帮助这些调优工作，本文分析了基于EKF的LIO算法（FAST-LIO2）和基于图的SLAM算法（Cartographer）中LIO模块的参数对性能的影响，研究对象为在不同环境中以低至中等高度飞行时使用不同激光雷达记录的空中激光雷达SLAM数据集。分析基于使用每种参数组合配置的LIO算法处理数据集所产生的绝对轨迹误差（ATE）。通过皮尔逊相关系数评估单个参数与ATE结果之间的关系，同时使用随机森林置换重要性分析评估每个参数的影响，随机森林模型经过训练以根据参数选择预测结果ATE值。分析结果为Cartographer和FAST-LIO2提供了：i）识别对性能影响较大的参数，ii）简化的调优程序，以及iii）调优建议。根据所提出的调优建议，这两种算法在分析的数据集中获得的ATE值在94%的案例中均在5厘米以内，接近在网格搜索过程中找到的最佳性能。

View on arXiv Download PDF AI Translation

cs.RO / 42 / 2609.12871

A Multi-Vehicle Dataset with Camera, LiDAR, and Radar Sensors and Scanned 3D Models for Custom Auto-Annotation using RTK-GNSS

一个包含摄像头、激光雷达和雷达传感器的多车辆数据集及用于自定义自动标注的扫描3D模型，采用RTK-GNSS

Berthold, Philipp, Forkel, Bianca, Maehlisch, Mirko

Abstract

Datasets are a crucial element in the development of perception algorithms. They relate sensor measurement data to annotated reference information and allow for the deduction of sensor and object characteristics. In autonomous driving, the reference data commonly consist of semantic image segmentation, point-wise associations, or bounding box annotations. The dataset proposed in this work, however, aims to dig deeper into the evaluation of measurement principles and provides scanned 3D models of all vehicles together with a pose and continuous kinematics reference obtained by RTK-GNSS. Combined, the state of the complete dynamic surrounding of the sensor vehicle is known for any point in time. Subsequent reference formats can be easily computed in user-defined granularity. This dataset involves single-object and multi-object recordings with seven target vehicles. In particular, measurement effects such as occlusion, as well as reflections, can be evaluated, as the normals of the shape of the target vehicles are known. We describe the dataset, discuss the technical background of its development, and briefly present exemplary evaluations.

Chinese Translation

数据集是感知算法开发中的关键要素。它们将传感器测量数据与标注的参考信息关联起来，并允许推导传感器和物体的特性。在自动驾驶中，参考数据通常包括语义图像分割、逐点关联或边界框标注。然而，本研究提出的数据集旨在深入评估测量原理，并提供所有车辆的扫描3D模型，以及通过RTK-GNSS获得的姿态和连续运动学参考。结合起来，传感器车辆在任何时刻的完整动态环境状态都是已知的。后续的参考格式可以在用户定义的粒度下轻松计算。该数据集涉及七辆目标车辆的单对象和多对象录制。特别地，可以评估诸如遮挡和反射等测量效应，因为目标车辆的形状法线是已知的。我们描述了该数据集，讨论了其开发的技术背景，并简要介绍了示例评估。

View on arXiv Download PDF AI Translation

cs.RO / 43 / 2609.12883

From Transportation to Manipulation: Enabling Grasping in Magnetic Robotics

从运输到操作：在磁性机器人中实现抓取

Bergmann, Lara, Greis, Noah, Grothues, Cedric, Weigelt, Lisa-Marie, Neumann, Klaus

Abstract

Magnetic levitation (MagLev) systems have great potential for application in high-mix, low-volume manufacturing due to their scalability and flexibility, enabling highly reconfigurable in-machine material flow. However, their manipulation capabilities remain largely unexploited, as current applications almost exclusively focus on transportation. To enable grasping and manipulation directly on MagLev systems without requiring additional costly handling equipment, such as industrial robot arms, we present the Gripper MagBot, a low-cost parallel 6-DoF manipulator with an integrated 1-DoF gripper that mechanically couples three MagLev movers. The Gripper MagBot supports two operating configurations: a default mode and a single-track mode, selectable depending on the required stability and workspace footprint. To reconfigure a machine, the MagBot can be autonomously dropped off and picked up using a docking station. We showcase pick-and-place examples in simulation, as well as with the real Gripper MagBot using our inverse kinematics controller. CAD files, assembly instructions, a component list, and videos are available at https://sites.google.com/view/gripper-magbot.

Chinese Translation

磁悬浮（MagLev）系统因其可扩展性和灵活性，在高混合、低批量制造中具有很大的应用潜力，能够实现高度可重构的机内物料流动。然而，它们的操作能力仍然未得到充分利用，目前的应用几乎完全专注于运输。为了在MagLev系统上直接实现抓取和操作，而无需额外昂贵的处理设备，如工业机器人手臂，我们提出了Gripper MagBot，这是一种低成本的并联6自由度（6-DoF）操纵器，配备一个集成的1自由度（1-DoF）抓手，机械耦合了三个MagLev移动器。Gripper MagBot支持两种操作配置：默认模式和单轨模式，可根据所需的稳定性和工作空间足迹进行选择。为了重新配置机器，MagBot可以通过对接站自主放置和拾取。我们展示了在仿真中的拾取与放置示例，以及使用我们的逆向运动学控制器的真实Gripper MagBot。CAD文件、组装说明、组件清单和视频可在https://sites.google.com/view/gripper-magbot获取。

View on arXiv Download PDF AI Translation

cs.RO / 44 / 2609.12894

Before the Tipping Point: Force-Guided Active Perception for Shape-Agnostic Estimation of 3D Centers of Mass

临界点之前：基于力的主动感知用于形状无关的三维质心估计

Hyland, Steven M., Xiao, Jing, Onal, Cagdas D.

Abstract

Estimating the 3D center of mass of unknown objects is challenging when grasping is infeasible, geometry is irregular, or mass distribution is uneven. We present a force-based method that estimates CoM height and mass from a single sub-critical tipping experiment by a robot manipulator. The robot applies a quasistatic elevated push and retract motion, using force-angle measurements recorded during tipping to identify parameters from the object trajectory. Our proposed push-retract cycle mitigates frictional bias, enabling generalized fitting. We experimentally validate our method using a robot manipulator with a six-axis force torque sensor on varying types of objects without prior shape information and without specific models. We also propose a method to prevent toppling, keeping the object in a sub-critical tipping regime by leveraging a safety margin. In experimental studies, our method recovers mass, CoM height, and toppling angle with relative errors below 5.0 percent across all unknown objects. This work demonstrates reliable 3D inertial parameter estimation under proper safety thresholds in tipping. Our proposed method informs and enables reliable non-prehensile manipulation and robotic grasping of challenging objects that were previously infeasible.

Chinese Translation

在抓取不可行、几何形状不规则或质量分布不均的情况下，估计未知物体的三维质心是具有挑战性的。我们提出了一种基于力的方法，通过机器人操纵器在单次亚临界倾斜实验中估计质心高度和质量。机器人施加一种准静态的上推和收回运动，利用在倾斜过程中记录的力-角度测量来识别物体轨迹中的参数。我们提出的推-收回周期减轻了摩擦偏差，使得参数拟合更加通用。我们通过在不同类型的物体上使用配备六轴力矩传感器的机器人操纵器进行实验验证，未使用先前的形状信息和特定模型。我们还提出了一种防止倾覆的方法，通过利用安全边际保持物体在亚临界倾斜状态。实验研究表明，我们的方法在所有未知物体上恢复质量、质心高度和倾覆角的相对误差均低于5.0%。这项工作展示了在倾斜过程中在适当安全阈值下可靠的三维惯性参数估计。我们提出的方法为可靠的非抓取操控和机器人抓取以前不可行的复杂物体提供了信息和支持。

View on arXiv Download PDF AI Translation

cs.RO / 45 / 2609.12927

Robust Underwater Grasping of Sloped Objects with a Waterproof Passive Adaptive Gripper

具有防水被动自适应抓手的斜面物体稳健水下抓取

Hong, Jooyoung, Hong, Daewon, Kim, Joohyung

Abstract

Robust grasping of everyday objects remains challenging for parallel-jaw grippers, particularly when handling sloped or asymmetric items that induce torque-driven rolling and shear slip. These challenges become even more severe in domestic environments such as kitchens, where objects are often wet or submerged, drastically reducing friction between the gripper and the object. To address these issues, we present a waterproof passive adaptive gripper that combines local and global adaptability for high-performance grasping in the submerged environments. The proposed gripper features a fully waterproof design that integrates a passive rotational joint for global self-alignment on sloped surfaces and passive variable-stiffness pads for local surface adaptation. Experiments in both dry and underwater conditions on various cylindrical and conical objects demonstrate superior holding capability and grasp robustness compared to rigid-pad and fixed-joint baselines. The proposed design offers a practical and robust solution that successfully enables stable underwater manipulation of diverse everyday objects, effectively addressing a critical gap in current underwater robotic grasping for daily-life applications.

Chinese Translation

日常物体的稳健抓取对于平行夹爪仍然具有挑战性，尤其是在处理倾斜或不对称物体时，这些物体会引发扭矩驱动的滚动和剪切滑移。在家庭环境中，例如厨房，这些挑战变得更加严重，因为物体通常是湿的或浸没在水中，极大地降低了夹具与物体之间的摩擦。为了解决这些问题，我们提出了一种防水的被动自适应抓手，它结合了局部和全局的适应性，以实现水下环境中的高性能抓取。所提出的抓手具有完全防水的设计，集成了被动旋转关节，以便在倾斜表面上实现全局自对准，并配备被动可变刚度垫，以实现局部表面适应性。在干燥和水下条件下对各种圆柱形和锥形物体的实验表明，与刚性垫和固定关节的基线相比，所提出的抓手具有更优越的保持能力和抓取稳健性。该设计提供了一种实用且稳健的解决方案，成功实现了对多种日常物体的稳定水下操作，有效填补了当前日常生活应用中水下机器人抓取的关键空白。

View on arXiv Download PDF AI Translation

cs.RO / 46 / 2609.12932

ARC: Autonomous Robotics Compliance A Three-Layer Governance Architecture for Deployed Autonomous Systems

ARC：自主机器人合规性三层治理架构用于部署的自主系统

Eide, Tord, Holt, Einar

Abstract

Proposed governance framework for autonomous robotic systems, introducing a three-layer compliance architecture (ARC) instantiated through model safety validation, cognitive certification benchmarks, and operational authorization standards.

Chinese Translation

提出了一种针对自主机器人系统的治理框架，介绍了一种通过模型安全验证、认知认证基准和操作授权标准实例化的三层合规架构（ARC）。

View on arXiv Download PDF AI Translation

cs.RO / 47 / 2609.12937

A Robot Among People:From Social Imitation to the Social Becoming of Human Groups

人与机器人：从社会模仿到人类群体的社会生成

Pham, Victor Tuan Vu, Dörrenbächer, Judith, Weisswange, Thomas H., Hassenzahl, Marc

Abstract

Robots designed to mediate human groups often fall into the solutionist trap: they are framed as sociable agents that fix problems such as conflict, disengagement, or lack of coordination. We suggest a different way of thinking. Rather than discrete agents, robots can be understood as situated elements of shared environments; catalysts and carriers of group experience whose meaning emerges through how people position, interpret, and interact with them. From this perspective, robots are not there to repair some ostensible dysfunctionality, but to enable group-level sense-making around care, norms, and identity. Our prior work on robotic street furniture suggests that this does not happen by imitating human sociality but by taking the shape of deliberately constrained, group-facing entities that happen and act for \textit{us} without being socially entangled as one of us. We thus understand robots in public spaces not in terms of autonomy or intelligence, but as a relational capacity. This implies designing robots not in our image or for our utility, but grounded in our needs in being and becoming together.

Chinese Translation

旨在调解人类群体的机器人常常陷入解决主义的陷阱：它们被视为能够解决冲突、脱离或缺乏协调等问题的社交代理。我们建议一种不同的思考方式。机器人可以被理解为共享环境中的情境元素；作为群体经验的催化剂和承载者，其意义通过人们如何定位、解释和与之互动而显现。从这个角度来看，机器人并不是用来修复某种表面上的功能失调，而是为了促进围绕关怀、规范和身份的群体层面的意义构建。我们之前关于机器人街头家具的研究表明，这并不是通过模仿人类的社交行为来实现的，而是通过形成有意限制的、面向群体的实体，这些实体为我们发生和行动，而不与我们作为社会的一部分纠缠在一起。因此，我们在公共空间中理解机器人时，不是从自主性或智能的角度，而是作为一种关系能力。这意味着设计机器人时，不是以我们的形象或为了我们的效用，而是基于我们共同存在和发展的需求。

View on arXiv Download PDF AI Translation

cs.RO / 48 / 2609.12959

Distributed Stochastic Optimal Control for Pattern-Oriented Swarms

面向模式的群体分布式随机最优控制

Zhang, Qingrui, Yu, Chenghao, Xue, Feng, Wang, Xintong

Abstract

While offering significant promise for diverse applications, pattern-oriented swarms encounter multifaceted challenges in geometric control, self-organization, and safe navigation through dynamic environments. In this paper, we present a GRF-based stochastic optimal control framework to address these challenges within a unified probabilistic architecture. By extending the GRF into the temporal domain, the proposed framework casts collective coordination as a Bayesian inference task, enabling swarms to accommodate environmental uncertainty, satisfy non-convex constraints, and reconcile heterogeneous dynamics across diverse platforms. We develop an uncertainty- and safety-aware collision avoidance module for navigation in the presence of stochastic obstacle motion. The unscented transform is employed to propagate state uncertainty for both dynamic obstacles and neighboring agents, yielding principled confidence bounds for collision avoidance. In addition, density-guided pattern control is introduced, which encodes geometric patterns as implicit density fields. This representation decouples pattern specification from explicit agent-to-target assignments, thereby facilitating intrinsic self-healing and elastic reconfiguration in a distributed manner. The proposed framework is extensively evaluated through Monte Carlo simulations across diverse scenarios. Its model-agnostic nature is demonstrated on both quadrotor and fixed-wing UAV swarms, highlighting its generalizability across platforms with heterogeneous dynamics. Finally, the efficacy and robustness of the proposed method are validated through indoor experiments with a 15-quadrotor swarm and outdoor deployments involving 4 custom-built autonomous quadrotors. These experiments substantiate the proposed framework's capacity to maintain reliable geometric pattern transitions and safety-aware navigation within real-world environments.

Chinese Translation

尽管面向模式的群体在多种应用中展现出显著的潜力，但在几何控制、自组织以及在动态环境中安全导航方面仍面临多方面的挑战。本文提出了一种基于GRF（Gaussian Random Field）的随机最优控制框架，以在统一的概率架构中解决这些挑战。通过将GRF扩展到时间域，所提出的框架将集体协调视为贝叶斯推断任务，使群体能够适应环境的不确定性，满足非凸约束，并协调不同平台之间的异质动态。我们开发了一个考虑不确定性和安全性的碰撞避免模块，以应对随机障碍物运动下的导航问题。采用无迹变换（Unscented Transform）来传播动态障碍物和邻近代理的状态不确定性，从而为碰撞避免提供有原则的置信界。此外，引入了基于密度引导的模式控制，将几何模式编码为隐式密度场。这种表示法将模式规范与显式的代理到目标的分配解耦，从而以分布式方式促进内在的自愈和弹性重构。通过在多种场景下进行的蒙特卡洛模拟，广泛评估了所提出的框架。其模型无关的特性在四旋翼和固定翼无人机群体上得到了验证，突显了其在异质动态平台上的普适性。最后，通过在室内实验中使用15个四旋翼群体以及在户外部署中涉及4个定制自主四旋翼的实验，验证了所提出方法的有效性和鲁棒性。这些实验证实了所提出框架在现实环境中维持可靠的几何模式转变和安全意识导航的能力。

View on arXiv Download PDF AI Translation

cs.RO / 49 / 2609.12971

Tuning ROS 2 for Energy-Efficient Navigation: Empirical Insights from Costmap 2D Configurations

为能效导航调优 ROS 2：来自 Costmap 2D 配置的实证见解

Albonico, Michel, Wortmann, Andreas, Malavolta, Ivano

Abstract

Robots are increasingly used in diverse application areas, where autonomous navigation plays a central role. As these systems become more widespread, improving their energy efficiency is critical to extending operational time and reducing environmental impact. The Robot Operating System (ROS) is a widely adopted middleware for robotics, offering a rich set of configurable packages. However, this flexibility can result in suboptimal software configurations in dynamic environments, negatively affecting both performance and energy consumption. This paper investigates the impact of ROS 2 package reconfigurations on the energy efficiency of mobile robot navigation. We conduct a controlled experiment in two warehouse-like scenarios (small and large) with varying obstacle layouts and Costmap 2D configurations (essential to the Nav2 stack). Through repeated trials, we measure energy usage, power profile, CPU load, memory consumption, and navigation performance. Results show that configurations must be carefully chosen for the specific robotic environment, and we were able to identify critical settings that lead to good and poor performance and energy consumption.

Chinese Translation

机器人在多种应用领域中的使用日益广泛，其中自主导航发挥着核心作用。随着这些系统的普及，提高其能效对于延长操作时间和减少环境影响至关重要。机器人操作系统（Robot Operating System, ROS）是一种广泛采用的机器人中间件，提供了一套丰富的可配置包。然而，这种灵活性可能导致在动态环境中软件配置的次优，进而对性能和能耗产生负面影响。本文研究了 ROS 2 包重配置对移动机器人导航能效的影响。我们在两个类似仓库的场景（小型和大型）中进行了一项受控实验，场景中障碍物布局和 Costmap 2D 配置（对 Nav2 堆栈至关重要）各不相同。通过重复试验，我们测量了能量使用、功率曲线、CPU 负载、内存消耗和导航性能。结果表明，配置必须根据特定的机器人环境仔细选择，我们能够识别出导致良好和较差性能及能耗的关键设置。

View on arXiv Download PDF AI Translation

cs.RO / 50 / 2609.13011

Comfort by Construction: Adaptive, Comfort-Bounded Action Spaces for Learned Driving Policies

通过构建实现舒适性：适应性、舒适性约束的动作空间用于学习驾驶策略

Rothenhäusler, Anna, Jost, Daniel, Rajan, Raghu, Janjos, Faris, Scheel, Oliver, Look, Andreas, Boedecker, Joschka

Abstract

Data-driven driving simulators command accelerations and steering rates from a fixed grid without constraining the realized accelerations and jerks. As a result, reinforcement-learning policies inflate safety metrics through abrupt, last-second maneuvers that lie far outside the range of human driving and would be unacceptable to occupants of a real vehicle, so the metrics measure simulator permissiveness rather than policy quality. Enforcing comfort bounds naively is not enough: lateral limits shrink quadratically with speed, so clamping a static grid saturates it and destroys fine-grained control ("grid collapse"). We propose an adaptive action parameterization that rediscretizes the grid at every step to span exactly the per-step feasible control set, via closed-form inversion of the lateral-jerk constraint. We further present PufferDrive-Editor, a browser-based tool to audit realized kinematics and author kinematically challenging scenes. On the Waymo Open Motion Dataset and a hand-authored slalom, our adaptive model holds comfort violations below 1% while outperforming clipped-grid and direct-jerk baselines in navigability.

Chinese Translation

数据驱动的驾驶模拟器从固定网格中指挥加速度和转向速率，而不限制实际的加速度和冲击。因此，强化学习策略通过突然的、最后一刻的操控来膨胀安全指标，这些操控远远超出了人类驾驶的范围，并且对于真实车辆的乘员来说是不可接受的，因此这些指标衡量的是模拟器的宽容性而非策略的质量。简单地强制执行舒适性约束是不够的：横向限制随着速度的增加而呈二次方缩小，因此固定网格的限制会使其饱和并破坏细粒度控制（“网格崩溃”）。我们提出了一种自适应的动作参数化方法，通过横向冲击约束的封闭形式反演，在每一步重新离散化网格，以准确涵盖每一步的可行控制集。我们进一步介绍了PufferDrive-Editor，这是一个基于浏览器的工具，用于审计实际运动学并创作运动学挑战场景。在Waymo开放运动数据集和手动创作的障碍赛中，我们的自适应模型将舒适性违规率保持在1%以下，同时在可导航性方面优于剪切网格和直接冲击基线。

View on arXiv Download PDF AI Translation

cs.RO / 51 / 2609.13015

Global Path Planner with Multi-Model Switching

具有多模型切换的全球路径规划器

Gori, Pietro, Iotti, Francesco, Zelenay, Eduard, Marko, Rastislav, Pierallini, Michele, Angelini, Franco, Pannocchia, Gabriele, Garabini, Manolo

Abstract

This work enhances global path planning via a pure-pursuit controller with multi-model kinematic switching that sustains plan fidelity across diverse terrains. The system includes a traversability graph for terrain analysis, a Heading-Aware A* algorithm for generating feasible paths, and a multi-model Pure Pursuit controller for dynamic tracking. A core innovation is adaptive kinematic modeling, enabling real-time switching between kinematic models based on terrain features and robot states. This adaptability optimizes path efficiency and energy use in challenging scenarios. We validate the approach in simulation on different platforms, namely the Artaban quadruped and the X3 quadrotor drone, showcasing improved performance, robustness, and adaptability over standard baselines.

Chinese Translation

本研究通过一种纯追踪控制器增强了全球路径规划，该控制器具备多模型运动学切换功能，能够在多样地形中维持规划的准确性。该系统包括用于地形分析的可 traversability 图、用于生成可行路径的朝向感知 A* 算法，以及用于动态跟踪的多模型纯追踪控制器。核心创新在于自适应运动学建模，能够根据地形特征和机器人状态实时切换运动学模型。这种适应性在挑战性场景中优化了路径效率和能量使用。我们在不同平台上进行了仿真验证，包括 Artaban 四足机器人和 X3 四旋翼无人机，展示了相较于标准基线的性能、鲁棒性和适应性的提升。

View on arXiv Download PDF AI Translation

cs.RO / 52 / 2609.13053

Dynin-Robotics: Omnimodal Unified Diffusion Vision-Language-Action Model

Dynin-Robotics：全模态统一扩散视觉-语言-动作模型

Lee, Hoeun, Kim, Jaeik, Oh, Jusang, Kim, Jinhyeok, Choi, Geon, Kim, Hyeonggeun, Do, Jaeyoung

Abstract

Visual goal and dynamics prediction can provide language-conditioned robot policies with both a target outcome and a representation of action-dependent scene changes. We bring these predictions into action generation and selection through a shared trajectory model. Dynin-Robotics implements this formulation on Dynin-Omni, an omnimodal masked-diffusion backbone, representing language, visual observations, goals, and actions as discrete tokens. By varying conditioning and target spans, the same model learns action prediction, action-conditioned next-observation prediction, terminal goal-state prediction, and trajectory-to-instruction reconstruction. These interfaces support test-time scaling through goal prediction, action-candidate evaluation, and joint refinement of action and future-state predictions. We continually pretrain the model on approximately 1.33 million trajectories from 48 Open X-Embodiment datasets and adapt it separately to downstream domains. On two VLABench tasks, robot pretraining improves adaptation within a fixed Stage-2 step budget, and the full objective mixture improves shifted-instruction success over Policy-only post-training under the same coupled decoder. Combining goal guidance with joint action-next-state denoising further improves shifted-instruction success over action-only decoding; the benefit depends on how the predictions are composed. Dynin-Robotics achieves competitive performance on LIBERO and zero-shot LIBERO-Plus, together with a 78.4% average success rate across four manipulation conditions on a Franka Research 3 robot. An optimized block-parallel implementation accelerates model-side action decoding by up to 29.2x relative to the base implementation under the reported profiling setup. These results support shared trajectory modeling as a common interface for learning complementary robot objectives and composing their predictions during control.

Chinese Translation

视觉目标和动态预测可以为语言条件下的机器人策略提供目标结果以及与动作相关的场景变化的表征。我们通过共享轨迹模型将这些预测引入动作生成和选择。Dynin-Robotics 在 Dynin-Omni 上实现了这一公式，Dynin-Omni 是一个全模态的掩蔽扩散骨干网络，将语言、视觉观察、目标和动作表示为离散标记。通过变化条件和目标跨度，同一模型学习动作预测、动作条件下的下一观察预测、终端目标状态预测和轨迹到指令的重建。这些接口通过目标预测、动作候选评估和动作与未来状态预测的联合优化支持测试时的扩展。我们持续在来自 48 个 Open X-Embodiment 数据集的约 133 万条轨迹上对模型进行预训练，并分别适应下游领域。在两个 VLABench 任务中，机器人预训练在固定的第二阶段步骤预算内改善了适应性，而完整目标混合在相同耦合解码器下提高了相较于仅政策后训练的转移指令成功率。将目标指导与联合动作-下一状态去噪结合进一步提高了转移指令的成功率，相较于仅动作解码的情况；这一收益取决于预测的组合方式。Dynin-Robotics 在 LIBERO 和零-shot LIBERO-Plus 上实现了竞争性能，并在 Franka Research 3 机器人上在四种操作条件下获得了 78.4% 的平均成功率。优化的块并行实现相较于报告的分析设置下的基础实现加速了模型侧的动作解码，提升幅度高达 29.2 倍。这些结果支持共享轨迹建模作为学习互补机器人目标和在控制过程中组合其预测的共同接口。

View on arXiv Download PDF AI Translation

cs.RO / 53 / 2609.13083

ASTRIL-MPC: Autonomous Traversal Framework of Articulated Tracked Robots with Language-Guided Neural-Kinematic MPC

ASTRIL-MPC：具有语言引导的神经运动学模型预测控制的关节式履带机器人自主穿越框架

Gan, Zhenfeng, Chen, Yanbo, Che, Lirong, Tan, Junbo, Wang, Xueqian

Abstract

In urban search and rescue, articulated tracked robots (ATRs) must traverse structured but contact-rich environments such as stairwells and cluttered building interiors. Reliable autonomy remains challenging because robot-terrain interaction (RTI) is hybrid and discontinuous, and effective flipper-track coordination is difficult to model analytically. We present ASTRIL-MPC, a language-guided neural kinematics model predictive control (MPC) framework for autonomous traversal. A learned kinematics model predicts short-horizon task-state increments from a height sequence and recent trajectories; NMPC plans with multi-objective costs and strict feasibility constraints; and a large language model (LLM) proposes bounded updates to selected weights and bounds through a safety-checked interface with range clipping, rate limiting, and consistency checks. The compiled predictor enables a full control cycle within 100 ms. Across three traversal tasks and a multi-height generalization setting, ASTRIL-MPC improves an aggregate traversal-quality score by up to 71% over a non-adaptive NMPC and by 67% over a PPO baseline, while eliminating measurable collision impacts during descent. These results indicate that combining learned kinematics, optimization-based planning, and language-guided retuning yields data-efficient and robust autonomy for articulated tracked robots.

Chinese Translation

在城市搜索与救援中，关节式履带机器人（ATR）必须穿越结构化但接触丰富的环境，如楼梯间和杂乱的建筑内部。由于机器人与地形的交互（RTI）是混合且不连续的，可靠的自主性仍然面临挑战，而有效的翻转-履带协调也难以进行解析建模。我们提出了ASTRIL-MPC，这是一种用于自主穿越的语言引导神经运动学模型预测控制（MPC）框架。学习的运动学模型从高度序列和最近的轨迹中预测短期任务状态增量；NMPC在多目标成本和严格可行性约束下进行规划；大型语言模型（LLM）通过安全检查的接口提出对选定权重和界限的有界更新，接口包括范围裁剪、速率限制和一致性检查。编译的预测器使得完整控制周期在100毫秒内完成。在三个穿越任务和一个多高度泛化设置中，ASTRIL-MPC相较于非自适应NMPC提升了高达71%的综合穿越质量评分，相较于PPO基线提升了67%，同时在下降过程中消除了可测量的碰撞影响。这些结果表明，将学习的运动学、基于优化的规划和语言引导的再调优相结合，可以为关节式履带机器人提供数据高效且稳健的自主性。

View on arXiv Download PDF AI Translation

计算机视觉 (Computer Vision)

cs.CV / 1 / 2609.12078

Feature Recovery for Object Understanding After Irreversible Fire Damage

不可逆火灾损伤后的物体理解特征恢复

Tiwari, Aditi, Stoica, Sofia, Khosla, Savya, Forsyth, David, Ji, Heng

Abstract

Objects in post-fire environments often undergo irreversible physical transformations that change their geometry, material state, and visual appearance. Detecting and identifying these remnants is critical for locating hazards, reconstructing pre-incident contents, and inventorying losses. Unlike standard image corruptions, these degradations affect the physical structure of the object itself. To study this setting, we introduce TRACE, a transformation-aware benchmark for post-fire object understanding. TRACE contains 21.4K real-image-grounded synthetic scenes and paired object-level pristine-to-degraded progressions spanning 499 object identities across 189 categories. We define five tasks targeting localization and pre-degradation understanding: degraded-object detection, pristine-state recovery and retrieval, original material recovery, pristine description generation, and functional reasoning. Existing models degrade sharply with severity. From the least to the most severe level, RF-DETR mAP decreases by 71% relative, while InternVL3.5 retrieval R@1 falls from 93.85 to 28.11. To address this, we propose the Feature Recovery Module (FRM), a plug-and-play module that maps degraded encoder features to pristine-aligned representations while keeping the host frozen. Trained only with paired feature supervision, FRM improves scene-level detection, CLIP/SigLIP2 feature recovery, and all four object-level VLM tasks, with larger gains under more severe degradation. Across VLM hosts and severity levels, relative gains average 12.5% for retrieval, 20.1% for material recovery, 13.2% for description generation, and 12.4% for functional reasoning.

Chinese Translation

在火灾后环境中，物体通常经历不可逆的物理变形，这些变形改变了它们的几何形状、材料状态和视觉外观。检测和识别这些残余物对定位危险、重建事故前的内容以及清点损失至关重要。与标准图像损坏不同，这些退化影响的是物体本身的物理结构。为了研究这一情境，我们引入了TRACE，这是一个针对火灾后物体理解的变换感知基准。TRACE包含21.4K个真实图像基础的合成场景和配对的物体级别从完好到退化的进展，涵盖499个物体身份，分属于189个类别。我们定义了五个任务，针对定位和退化前理解：退化物体检测、完好状态恢复与检索、原始材料恢复、完好描述生成和功能推理。现有模型在退化严重程度上急剧下降。从最低到最高的严重程度，RF-DETR的mAP相对下降了71%，而InternVL3.5的检索R@1从93.85降至28.11。为了解决这个问题，我们提出了特征恢复模块（Feature Recovery Module，FRM），这是一个即插即用的模块，可以将退化的编码器特征映射到完好对齐的表示，同时保持主机不变。FRM仅通过配对特征监督进行训练，提升了场景级检测、CLIP/SigLIP2特征恢复以及所有四个物体级别的VLM任务，在更严重的退化下获得更大的提升。在不同的VLM主机和严重程度下，检索的相对增益平均为12.5%，材料恢复为20.1%，描述生成为13.2%，功能推理为12.4%。

View on arXiv Download PDF AI Translation

cs.CV / 2 / 2609.12090

Does Video Memory Use What It Retrieves? A Causal Audit of Memory Specificity

视频记忆是否利用其检索内容？记忆特异性的因果审计

Tiwari, Aditi, Bhalla, Akshit, Prasad, Darshan, Ji, Heng

Abstract

Video models increasingly use memory to preserve information over long sequences, with the assumption that gains come from retrieving and using the correct past content. Standard memory ablations test whether memory helps, but not whether the retrieved content is responsible. We test this directly with read-time memory substitution, which replaces the consumed memory value while leaving the rest of the computation unchanged. This separates memory benefit from memory specificity, the extent to which the gain depends on retrieved content. Across frozen video world models, identity-free controls containing no evaluation-specific content recover essentially the full benefit on Ego-Exo4D and 7-Scenes and about 70% on TUM. In the Ego-Exo4D dose response, recovery falls from 102% to 1% as these values move away from observed training-memory representations, supporting representation repair as the best-supported explanation in this setting. WorldMem shows graded dependence. A wrong memory from the same trajectory recovers 94.1% of the PSNR benefit relative to zero content, while a donor from a disjoint trajectory and biome recovers 43.7%. SAM 2 shows strong content dependence. On DAVIS, replacing the correct spatial memory with a valid wrong memory reduces mean region and boundary score from 0.926 to 0.182. At MOSEv2 reappearance, it falls from 0.459 to 0.000. These results show that memory gains can depend on generic representation support, broader context, or exact episodic content. Read-time substitution provides a direct way to distinguish them.

Chinese Translation

视频模型越来越多地使用记忆来在长序列中保留信息，假设收益来自于检索和使用正确的过去内容。标准的记忆切除测试检验记忆是否有帮助，但并不检验检索到的内容是否负责。我们通过读取时间的记忆替代直接测试这一点，该方法在不改变其余计算的情况下替换已消耗的记忆值。这将记忆效益与记忆特异性区分开来，后者是指收益在多大程度上依赖于检索到的内容。在冻结的视频世界模型中，包含无评估特定内容的身份无关控制在Ego-Exo4D和7-Scenes上几乎恢复了全部效益，而在TUM上恢复了约70%。在Ego-Exo4D的剂量反应中，随着这些值远离观察到的训练记忆表示，恢复从102%降至1%，支持在这种情况下代表修复作为最佳支持的解释。WorldMem显示出分级依赖性。来自同一路径的错误记忆相对于零内容恢复了94.1%的PSNR收益，而来自不相交路径和生物群落的捐赠者恢复了43.7%。SAM 2显示出强烈的内容依赖性。在DAVIS上，用有效的错误记忆替换正确的空间记忆使得平均区域和边界得分从0.926降至0.182。在MOSEv2重现时，它从0.459降至0.000。这些结果表明，记忆收益可能依赖于通用表示支持、更广泛的上下文或确切的情节内容。读取时间替代提供了一种直接区分它们的方法。

View on arXiv Download PDF AI Translation

cs.CV / 3 / 2609.12099

Beyond Argmax: A Mechanistic Study of Semantic Retention in Frozen Foundation-Model Composition for Generalized Few-Shot 3D Segmentation

超越 Argmax：对冻结基础模型组合中语义保留机制的研究，针对广义少样本 3D 分割

Gah, Silas Kwabla, Owusu, Ebenezer

Abstract

Classical classifier-combination work distinguishes score-level fusion from hard decision-level voting. We revisit this distinction where independently pretrained, frozen foundation models are composed at inference time for generalized few-shot 3D segmentation. We ask: how much useful semantic information is lost when heterogeneous sources are collapsed to a single class before they can interact? We answer with a same-input semantic-retention intervention. Dense RegionPLC and sparse cross-view SAM3 evidence, model weights, masks, geometry, vocabularies, and fusion rules are frozen; only the number of semantic alternatives retained before interaction is varied via a matched top-k ladder. On 156 held-out ScanNet200 scenes, top-1 reaches 28.47 harmonic-mean (HM) IoU while full distribution fusion reaches 34.87 HM (+6.40, 95% CI [+5.24,+7.64]). The pattern replicates on 50 ScanNet++ scenes: 23.02 vs. 26.50 HM (+3.48, 95% CI [+1.64,+5.93]). The conclusion is robust: full-distribution HM is stable across sparse-source weights 0.3--0.7; alternative operators (max, geometric pooling) also outperform top-1; and a GroundingDINO--SAM2.1 source-replacement diagnostic shows monotonic HM increase from 14.77 to 18.75 with full retention. Calibration diagnostics reveal opposite miscalibration of the two sources, yet correcting calibration does not eliminate the retention advantage. Across datasets and source stacks, most information is recovered by retaining a compact set of plausible alternatives. The contribution is a controlled diagnosis of premature semantic collapse as a repeatable information bottleneck in heterogeneous frozen-model composition.

Chinese Translation

经典的分类器组合研究区分了得分级融合与硬决策级投票。我们重新审视这一区分，在推理时独立预训练的冻结基础模型被组合用于广义少样本 3D 分割。我们提出问题：在异构源被合并为单一类别以便相互作用之前，有多少有用的语义信息被丢失？我们通过同输入的语义保留干预来回答这一问题。Dense RegionPLC 和稀疏跨视图 SAM3 证据、模型权重、掩码、几何形状、词汇和融合规则均被冻结；仅通过匹配的 top-k 梯度变化保留的语义选择数量被改变。在 156 个保留的 ScanNet200 场景中，top-1 达到 28.47 的调和均值 (HM) IoU，而完整分布融合达到了 34.87 HM (+6.40, 95% CI [+5.24,+7.64])。这一模式在 50 个 ScanNet++ 场景中得到了重复验证：23.02 对比 26.50 HM (+3.48, 95% CI [+1.64,+5.93])。结论是稳健的：完整分布的 HM 在稀疏源权重 0.3--0.7 范围内稳定；其他操作符（最大值、几何池化）也优于 top-1；而 GroundingDINO--SAM2.1 源替换诊断显示，在完全保留的情况下，HM 从 14.77 单调增加到 18.75。校准诊断揭示了两个源的相反失校准，但校准的修正并未消除保留优势。在不同数据集和源堆栈中，通过保留一组紧凑的合理替代方案，大部分信息得以恢复。该研究的贡献在于对异构冻结模型组合中早期语义崩溃的可控诊断，作为可重复的信息瓶颈。

View on arXiv Download PDF AI Translation

cs.CV / 4 / 2609.12151

HSI-Road Relabeled: Surface-Aware Road-Scene Segmentation

HSI-Road重新标注：表面感知道路场景分割

Shah, Imad Ali, Mehmood, Imran, Ward, Enda, Glavin, Martin, Jones, Edward, Deegan, Brian

Abstract

The HSI-Road dataset provides paired RGB and 25-channel NIR (600--960~nm) images with binary masks but no surface-level labels.~This paper introduces a manually labeled six-class taxonomy: Background, Asphalt, Concrete, Dirt, Water, and Grass, and an RGB-to-NIR registration pipeline with corresponding annotations. Six semantic-segmentation models (SSMs) are evaluated under four input configurations: original-resolution RGB (RGB$_{\text{ori}}$), registered low-resolution RGB (RGB$_{\text{reg}}$), NIR, and channel-stacked RGB$_{\text{reg}}$--NIR (RGBN$_{\text{stk}}$). The comparison quantifies the effect of spatial-resolution reduction on RGB, along with evaluation of NIR and RGBN$_{\text{stk}}$, with results reported using per-class and mean IoU and F1 scores. RGB$_{\text{ori}}$ achieves the highest overall performance but contains 12$\times$ more pixels than the matched-resolution inputs. At the matched 192$\times$384 resolution, RGBN$_{\text{stk}}$ outperforms NIR for all six SSMs and RGB$_{\text{reg}}$ for five of six, with the most consistent gains for the Water class. These results highlight the importance of spatial resolution while showing that NIR provides complementary information to RGB.

Chinese Translation

HSI-Road数据集提供了配对的RGB和25通道近红外（NIR，600--960 nm）图像及其二进制掩膜，但缺乏表面级标签。本文引入了一种手动标注的六类分类法：背景、沥青、混凝土、泥土、水和草地，以及相应注释的RGB到NIR注册流程。评估了六种语义分割模型（SSMs）在四种输入配置下的表现：原始分辨率RGB（RGB$_{ ext{ori}}$）、注册低分辨率RGB（RGB$_{ ext{reg}}$）、NIR和通道堆叠的RGB$_{ ext{reg}}$--NIR（RGBN$_{ ext{stk}}$）。比较量化了空间分辨率降低对RGB的影响，同时评估了NIR和RGBN$_{ ext{stk}}$，结果通过每类和平均IoU及F1分数进行报告。RGB$_{ ext{ori}}$实现了最高的整体性能，但其像素数量是匹配分辨率输入的12倍。在匹配的192$ imes$384分辨率下，RGBN$_{ ext{stk}}$在所有六种SSMs中均优于NIR，并且在六种中有五种优于RGB$_{ ext{reg}}$，其中水类的增益最为一致。这些结果突显了空间分辨率的重要性，同时表明NIR为RGB提供了互补信息。

View on arXiv Download PDF AI Translation

cs.CV / 5 / 2609.12155

Single-Query Person-Centric Bimanual Hand-Object Interaction Detection

单查询以人为中心的双手物体交互检测

Kim, Jonghyun, Roh, Junho, Yoon, Yubin, Lee, Hyotae, Park, Jongkuk, Hwang, Taehwan, Kim, Jaechul, Lee, Jungho

Abstract

Understanding person-level bi-manual interactions requires not only detecting hands, but also identifying which two hands belong to the same person and what each hand interacts with. Existing hand--object interaction methods are mostly hand-centric: they treat each hand as an independent instance, which can lead to ambiguous ownership in multi-person scenes. We propose a person-centric formulation in which a single query predicts a structured output for one person, including the human box, body pose, hand boxes and states, and interaction targets. We introduce part-aware deformable attention to allocate attention across human, hand, and pose-specific reference regions, enabling one query to capture the full person structure. We further unify detection and interaction reasoning with a hand-to-query relationship matrix, where each hand selects its interaction target from the detected query set plus a learnable off token, directly recovering the target's box and class without separate object regression. We build a COCO-based dataset with person-centric bi-manual interaction annotations and define structured metrics for evaluating hand states and complete hand--object tuples. Experiments with a transformer-based detector show that our formulation improves person-level bi-manual interaction parsing and provides an effective unified framework for joint detection, pose estimation, and hand reasoning.

Chinese Translation

理解以人为中心的双手交互不仅需要检测手部，还需要识别哪两只手属于同一个人以及每只手与什么物体进行交互。现有的手-物体交互方法大多是以手为中心的：它们将每只手视为独立的实例，这可能导致在多人的场景中出现模糊的归属关系。我们提出了一种以人为中心的表述，其中单个查询预测一个人的结构化输出，包括人框、身体姿态、手框和状态以及交互目标。我们引入了部分感知的可变形注意力机制，以在人体、手部和姿态特定的参考区域之间分配注意力，使得一个查询能够捕捉完整的人体结构。我们进一步通过手与查询关系矩阵统一检测和交互推理，其中每只手从检测到的查询集中选择其交互目标，外加一个可学习的离散标记，直接恢复目标的框和类别，而无需单独的物体回归。我们构建了一个基于COCO的数据集，包含以人为中心的双手交互注释，并定义了用于评估手部状态和完整手-物体元组的结构化指标。基于变压器的检测器的实验表明，我们的表述改善了以人为中心的双手交互解析，并提供了一个有效的统一框架，用于联合检测、姿态估计和手部推理。

View on arXiv Download PDF AI Translation

cs.CV / 6 / 2609.12168

USPLIT-VQA: U-Shaped Split Learning for Visual Question Answering with Contribution-Aware Weighted Aggregation

USPLIT-VQA：一种基于U形分割学习的视觉问答隐私保护框架，结合贡献感知加权聚合

Syfullah, Md Khalid, Khalil, Alvi Ataur

Abstract

Visual Question Answering (VQA) systems, jointly interpreting images and natural language queries, hold significant promise across many domains, yet the privacy-sensitive nature of user data creates a fundamental barrier. Centralized training requires access to all data, while federated learning requires each client to host the full model. We propose USPLIT-VQA, a U-shaped split learning framework for privacy-preserving VQA in which each client retains the initial layers and the classification head while the server hosts the computationally heavy intermediate layers, keeping raw inputs and labels on the client device. We further introduce Contribution-Aware Weighted Aggregation (CAWA), a gradientsimilarity-based client scoring mechanism designed to reduce the influence of malicious updates. Experiments on four VQA datasets (VQA-RAD, SLAKE, PathVQA, and VizWiz) with two backbones show accuracy gains over Federated Learning for the Custom model and reduced accuracy for BiomedCLIP under the evaluated fixed split, alongside client memory reductions of up to 5.8X and communication reductions of up to 10.8X. With one malicious client, CAWA reduces the attacker's influence by over 98%, while experiments at higher corruption levels identify its limitations. Reconstruction experiments further show lower inversion quality under the evaluated attacks.

Chinese Translation

视觉问答（VQA）系统能够共同解读图像和自然语言查询，在多个领域展现出巨大的潜力。然而，用户数据的隐私敏感性构成了一个根本性的障碍。集中式训练需要访问所有数据，而联邦学习要求每个客户端托管完整模型。我们提出了USPLIT-VQA，一种用于隐私保护的VQA的U形分割学习框架，其中每个客户端保留初始层和分类头，而服务器则托管计算量大的中间层，保持原始输入和标签在客户端设备上。我们进一步引入了贡献感知加权聚合（Contribution-Aware Weighted Aggregation, CAWA），这是一种基于梯度相似性的客户端评分机制，旨在减少恶意更新的影响。在四个VQA数据集（VQA-RAD、SLAKE、PathVQA和VizWiz）上进行的实验显示，与定制模型的联邦学习相比，USPLIT-VQA在准确性上有所提升，而在评估的固定分割下，BiomedCLIP的准确性有所降低，同时客户端内存减少了最多5.8倍，通信量减少了最多10.8倍。在一个恶意客户端的情况下，CAWA将攻击者的影响降低了超过98%，而在更高的腐败水平下的实验则揭示了其局限性。重建实验进一步表明，在评估的攻击下，反演质量较低。

View on arXiv Download PDF AI Translation

cs.CV / 7 / 2609.12169

When Ground-Truth Fidelity Matters: An Orchestrated UAS Framework for Wheat Streak Mosaic Virus Detection Using Vision Transformers and Machine Learning

当真实标签的准确性至关重要：基于视觉变换器和机器学习的协调无人机框架用于小麦条纹马赛克病毒检测

Kharismawati, Dewi Endah, Dhakal, Sandeep, McCusker, Courtney E., Wilson, Jennifer R., Ohlson, Erik W., Khanal, Sami

Abstract

Wheat streak mosaic virus (WSMV) is a destructive pathogen of sweet corn and other cereal crops, causing yield losses and complicating early detection because symptoms are spatially variable and subtle. In sweet corn seed production, WSMV also has regulatory importance, as phytosanitary regulations from countries such as New Zealand and Chile require seed lots to be certified virus-free. Visual scouting is unreliable because symptoms can resemble abiotic stress, while enzyme-linked immunosorbent assay (ELISA) is accurate but expensive, labor-intensive, and difficult to scale. We present an automated pipeline for plant-level WSMV detection using unmanned aircraft systems (UAS) multispectral imagery. The framework integrates orthomosaic reconstruction, geospatial alignment, plant extraction, and classification using a Vision Transformer with seven-channel inputs (five spectral bands, NDVI, and NDRE). Using treatment-based labels, the model achieved 89% accuracy on over 6,500 test patches across multiple growth stages. However, ELISA-based ground truth revealed substantial label noise: only a small fraction of sampled plants in inoculated plots were infected. Treatment labels therefore did not reliably represent infection status, and the high accuracy was largely driven by label bias rather than disease detection. Performance decreased markedly against row-level symptom severity and plant-level ELISA labels. Under these higher-fidelity but smaller-sample conditions, both deep learning and classical machine learning showed limited generalization and weak separability between ELISA-confirmed mock-inoculated and infected plants. These results show that UAS-based disease detection is constrained by label fidelity and data availability, emphasizing biologically grounded labels and models aligned with real-world conditions.

Chinese Translation

小麦条纹马赛克病毒（WSMV）是甜玉米和其他谷物作物的破坏性病原体，导致产量损失，并因症状空间变异性和微妙性而使早期检测复杂化。在甜玉米种子生产中，WSMV 还具有监管重要性，因为新西兰和智利等国家的植物检疫法规要求种子批次必须经过认证为无病毒。视觉侦察不可靠，因为症状可能类似于非生物胁迫，而酶联免疫吸附测定（ELISA）虽然准确，但成本高、劳动密集且难以扩展。我们提出了一种基于无人机系统（UAS）多光谱图像的植物级 WSMV 检测自动化流程。该框架整合了正射影像重建、地理空间对齐、植物提取和使用七通道输入（五个光谱波段、NDVI 和 NDRE）的视觉变换器进行分类。使用基于处理的标签，该模型在多个生长阶段的 6,500 多个测试样本上达到了 89% 的准确率。然而，基于 ELISA 的真实标签揭示了显著的标签噪声：在接种地块中，只有一小部分采样植物被感染。因此，处理标签并未可靠地代表感染状态，而高准确率在很大程度上是由标签偏差驱动，而非疾病检测。针对行级症状严重性和植物级 ELISA 标签的表现显著下降。在这些更高保真度但样本量较小的条件下，深度学习和经典机器学习在 ELISA 确认的假接种和感染植物之间显示出有限的泛化能力和弱分离性。这些结果表明，基于 UAS 的疾病检测受到标签准确性和数据可用性的限制，强调了生物学基础标签和与现实条件对齐的模型的重要性。

View on arXiv Download PDF AI Translation

cs.CV / 8 / 2609.12181

Physics as the label for measuring and correcting materials reasoning in multimodal models

物理作为多模态模型中测量和校正材料推理的标签

Kurban, Hasan, Khanbayov, Rasul, Kurban, Mustafa

Abstract

Vision-language and language models increasingly interpret materials data, yet benchmarks report that they hallucinate invalid properties and violate physical law. Evaluation matches final answers to scarce human labels, while discovery agents verify final proposals or density functional theory (DFT) execution. Neither measures the physical consistency of a model's reasoning chain. Materials data carries its own physics, making a large class of materials reasoning verifiable without annotation. We introduce MatPCR, a label-free benchmark whose programmatic oracles check diffraction geometry through Bragg's law, scale bars, spectral peaks, and Materials Project-grounded checks of near-hull stability, computed band-gap class, and net magnetization. We define the Physical-Consistency Rate over image and structure inputs; introduce Constraint-Grounded Self-Verification, an agentic loop whose gain survives self-refinement and equal-compute re-prompting controls; release an open verifier useful in distribution but near chance on all six held-out constraint types; and derive an exact identity for how oracle error displaces the reported rate.

Chinese Translation

视觉-语言和语言模型越来越多地解释材料数据，但基准测试报告它们会幻觉出无效属性并违反物理法则。评估将最终答案与稀缺的人类标签进行匹配，而发现代理则验证最终提案或密度泛函理论（DFT）的执行。两者都未能衡量模型推理链的物理一致性。材料数据自带物理特性，使得一大类材料推理在没有注释的情况下可验证。我们引入了MatPCR，一个无标签的基准，其程序化神谕通过布拉格定律、比例尺、光谱峰值以及基于材料项目的近壳稳定性、计算带隙类别和净磁化强度进行检查。我们定义了基于图像和结构输入的物理一致性率；引入约束基础自我验证，这是一种代理循环，其增益在自我优化和相等计算重新提示控制下得以保留；发布一个在分布中有用但在所有六种保留约束类型上接近随机的开放验证器；并推导出神谕错误如何影响报告率的确切关系。

View on arXiv Download PDF AI Translation

cs.CV / 9 / 2609.12202

QuPAINT: Physics-Aware Multimodal Reasoning for Quantum Material Characterization

QuPAINT：面向物理的多模态推理用于量子材料表征

Pandey, Sankalp, Nguyen, Xuan-Bac, Nguyen, Hoang-Quan, Faltermeier, Tim, Borys, Nicholas, Churchill, Hugh, Luu, Khoa

Abstract

Characterizing two-dimensional (2D) quantum materials by optical microscopy requires localizing exfoliated flakes and determining their layer thickness from subtle optical contrast and interference color to select suitable flakes for device fabrication. However, models face synthetic-to-real domain shifts and variation across materials, substrates, laboratories, and imaging conditions. We present QuPAINT, a physics-aware multimodal framework for transferable quantum flake characterization. The Synthetic Materials Framework (Synthia) generates diverse synthetic microscopy images while preserving layer-dependent optical behavior. Using these images, we construct QMat-Instruct, a multimodal instruction dataset with image-specific reasoning traces generated from verified annotations and constrained to observable optical cues. QuPAINT integrates these signals through Physics-Informed Attention (PIA), which injects substrate-relative optical priors into the visual representation to support grounded multimodal reasoning. For evaluation, we introduce QF-Bench, to our knowledge, the largest real-world benchmark for this problem, spanning diverse microscopy and substrate conditions. Using its verified annotations, we study counting, visual grounding, reasoning quality, confidence calibration, and transfer to an unseen material. QuPAINT-8B substantially outperforms prior methods and establishes state-of-the-art performance for both general and monolayer flake detection. Additional experiments show that image-grounded supervision improves strict spatial grounding and confidence calibration while preserving robust general flake detection on the unseen material.

Chinese Translation

通过光学显微镜表征二维（2D）量子材料需要定位剥离的薄片，并通过微妙的光学对比和干涉颜色确定其层厚度，以选择适合器件制造的薄片。然而，模型面临合成与真实领域的转变，以及材料、基底、实验室和成像条件之间的变化。我们提出了QuPAINT，一种面向物理的多模态框架，用于可转移的量子薄片表征。合成材料框架（Synthia）生成多样的合成显微图像，同时保留层依赖的光学行为。利用这些图像，我们构建了QMat-Instruct，这是一个多模态指令数据集，包含从经过验证的注释生成的图像特定推理轨迹，并限制于可观察的光学线索。QuPAINT通过物理信息注意力（Physics-Informed Attention, PIA）整合这些信号，该方法将基底相关的光学先验注入视觉表示中，以支持基于实证的多模态推理。为了评估，我们引入了QF-Bench，据我们所知，这是针对该问题的最大真实世界基准，涵盖了多样的显微镜和基底条件。利用其经过验证的注释，我们研究了计数、视觉定位、推理质量、置信度校准以及对未见材料的转移。QuPAINT-8B显著优于先前的方法，并在一般和单层薄片检测中建立了最先进的性能。额外实验表明，基于图像的监督改善了严格的空间定位和置信度校准，同时在未见材料上保持了稳健的薄片检测能力。

View on arXiv Download PDF AI Translation

cs.CV / 10 / 2609.12261

Revisiting Multi-Object Tracking Baselines: Hyperparameter Optimization with Multi-Fidelity Greedy Coordinate Search

重新审视多目标跟踪基准：使用多保真度贪婪坐标搜索进行超参数优化

Adžemović, Momir

Abstract

Multi-object tracking (MOT) is dominated by the tracking-by-detection paradigm, whose methods typically rely on a small set of hyperparameters that are conventionally chosen by hand. Tuning them requires repeated expert-guided experimentation, while the procedures used to select reported values are often not systematically evaluated or fully documented. Hyperparameter optimization (HPO) automates this process, yet it remains rarely used in MOT, and existing studies applying HPO to MOT predate modern deep-detector-based trackers and HOTA evaluation. We systematically apply HPO across two datasets and four tracking-by-detection methods. We also propose Multi-Fidelity Greedy Coordinate Search (MFGCS), which optimizes one hyperparameter at a time by first evaluating candidate values on a small subset of scenes and re-evaluating only promising candidates on the full dataset. Across all eight tracker-dataset combinations, the Tree-structured Parzen Estimator (TPE) and MFGCS outperform both our hand-tuned configurations and the corresponding published results, with improvements of up to 4.38 and 16.05 HOTA points, respectively. MFGCS also reaches a predefined HOTA target faster than TPE in seven of the eight combinations. Within each tracker-dataset pair, all optimizers share the same search space and evaluation pipeline, isolating the effect of the search strategy. We release the code and tuned configurations to enable future work to compare against systematically optimized rather than default or manually tuned baselines.

Chinese Translation

多目标跟踪（MOT）主要采用基于检测的跟踪范式，其方法通常依赖于一小组手动选择的超参数。调整这些超参数需要反复进行专家指导的实验，而用于选择报告值的程序往往没有系统评估或充分记录。超参数优化（HPO）自动化了这一过程，但在MOT中仍然很少使用，现有将HPO应用于MOT的研究早于现代基于深度检测器的跟踪器和HOTA评估。我们在两个数据集和四种基于检测的跟踪方法中系统地应用HPO。我们还提出了多保真度贪婪坐标搜索（MFGCS），该方法通过首先在小部分场景上评估候选值，并仅在完整数据集上重新评估有前景的候选者，从而一次优化一个超参数。在所有八个跟踪器-数据集组合中，树结构的帕尔岑估计器（TPE）和MFGCS的表现均优于我们手动调整的配置和相应的已发布结果，分别提高了最多4.38和16.05个HOTA点。MFGCS在八个组合中的七个中比TPE更快地达到预定义的HOTA目标。在每个跟踪器-数据集对中，所有优化器共享相同的搜索空间和评估流程，从而隔离搜索策略的影响。我们发布了代码和调优配置，以便未来的工作能够与系统优化的基准进行比较，而不是默认或手动调优的基准。

View on arXiv Download PDF AI Translation

cs.CV / 11 / 2609.12342

UniMo: Unifying Human and Animal Motion Generation

UniMo：统一人类与动物运动生成

Zhang, Zeyu, Zhang, Zhiyuan, Wang, Siheng, Wang, Yiran, Li, Danning, Reid, Ian, Hartley, Richard

Abstract

The conditional generation of 3D motion has emerged as a key research topic due to its wide applicability across robotics, AR/VR, gaming, and content creation. However, extending recent advances in text-driven human motion generation to the animal domain remains challenging due to two core limitations. First, animals exhibit highly diverse skeletal topologies, unlike the standard human structure, making unified modeling across species difficult and leading to inefficient per-species models. Second, existing animal motion datasets suffer from limited scale and annotation quality, constraining model performance. To address these challenges, we propose UniMo, a unified point cloud-based motion generation framework that bypasses topological discrepancies by converting parametric skeletons into unparametric representations, further enhanced by dynamic sampling that allocates more points to active joints. Additionally, we present UniML3D, a large-scale motion-language dataset spanning both human and animal categories, containing 145,907 motion sequences and 433,388 captions-over 102x larger than existing animal datasets. Our method achieves state-of-the-art results on UniML3D and three public benchmarks including HumanML3D, KIT-ML, and AnimalML3D, demonstrating the feasibility and effectiveness of unified human-animal motion generation. Website: https://steve-zeyu-zhang.github.io/UniMo.

Chinese Translation

3D运动的条件生成已成为一个关键研究主题，因为它在机器人技术、增强现实/虚拟现实、游戏和内容创作等领域具有广泛的应用。然而，将近期在文本驱动的人类运动生成方面的进展扩展到动物领域仍然面临挑战，主要有两个核心限制。首先，动物的骨骼拓扑结构高度多样化，与标准的人类结构不同，这使得跨物种的统一建模变得困难，并导致每个物种模型的效率低下。其次，现有的动物运动数据集在规模和注释质量上都存在限制，制约了模型的性能。为了解决这些挑战，我们提出了UniMo，一个基于统一点云的运动生成框架，通过将参数化骨架转换为非参数化表示来绕过拓扑差异，并通过动态采样进一步增强，该方法为活跃关节分配更多点。此外，我们还提出了UniML3D，这是一个涵盖人类和动物类别的大规模运动-语言数据集，包含145,907个运动序列和433,388个标题，规模超过现有动物数据集的102倍。我们的方法在UniML3D和三个公共基准（包括HumanML3D、KIT-ML和AnimalML3D）上实现了最先进的结果，展示了统一人类与动物运动生成的可行性和有效性。网站：https://steve-zeyu-zhang.github.io/UniMo。

View on arXiv Download PDF AI Translation

cs.CV / 12 / 2609.12343

VS-Splat: Voxel-Selective feed-forward Gaussian Splatting for end-to-end 3D object reconstruction from sparse-views

VS-Splat：基于体素选择的前馈高斯溅射用于从稀疏视图进行端到端的三维物体重建

Jeong, Yunsu, Heo, Hyuk, Kwak, Youngsang, Kwak, Jaehwa, Chun, Il Yong

Abstract

Feed-forward Gaussian splatting models have demonstrated remarkable effectiveness in reconstructing three-dimensional (3D) objects from a few two-dimensional (2D) images, even if they are unseen. As existing methods typically predict Gaussian primitives uniformly across the 3D space, most primitives are placed in non-object regions. This may hinder the representation of fine object details. This paper proposes a Voxel-Selective Gaussian Splatting model (VS-Splat), a new end-to-endfeed-forward Gaussian splatting framework that predicts many primitives only within selected voxels that are likely to belong to an object, without 3D structural supervision. To achieve this, we propose a new learnable voxel selection approach that identifies object-centric voxels only with 2D rendering supervision.Our sparse-view rendering experiments with three benchmark datasets show that proposed VS-Splat outperforms several state-of-the-art methods. We further demonstrate its effectiveness as a backbone for an existing densification method and show that anoptional extension improves its robustness to inaccurate camera pose estimates.

Chinese Translation

前馈高斯溅射模型在从少量二维（2D）图像重建三维（3D）物体方面表现出显著的有效性，即使这些图像未曾见过。由于现有方法通常在三维空间中均匀预测高斯原语，大多数原语被放置在非物体区域，这可能妨碍细节的表现。本文提出了一种体素选择高斯溅射模型（VS-Splat），这是一种新的端到端前馈高斯溅射框架，它仅在可能属于物体的选定体素内预测许多原语，而不需要三维结构监督。为此，我们提出了一种新的可学习体素选择方法，仅通过二维渲染监督来识别以物体为中心的体素。我们在三个基准数据集上的稀疏视图渲染实验表明，所提出的VS-Splat优于几种最先进的方法。我们进一步证明了其作为现有密集化方法的骨干网络的有效性，并显示出可选扩展提高了其对不准确相机姿态估计的鲁棒性。

View on arXiv Download PDF AI Translation

cs.CV / 13 / 2609.12350

EgoMaize: A First-Person Maize Instance Segmentation Benchmark under Severe Field Occlusion

EgoMaize：在严重田间遮挡下的第一人称玉米实例分割基准

Li, Jiayi, Zhang, Zihan, Yan, Erhankang, Chen, Yitian, Li, Yuze, Ding, Chengzhang, Cao, Jianxin

Abstract

Close-range first-person field images are important for mobile maize phenotyping because many plant-level traits depend on in-canopy structures that are difficult to ob serve from overhead views. However, post-seedling maize fields create a difficult in stance segmentation setting: stems, leaves, tassels, and neighboring plants are elon gated, repetitive, and strongly occluded. We introduce EgoMaize, a compact benchmark for first-person maize instance segmentation, where the task is to predict ownership consistent plant masks and plant-owned stem/tassel cues from close-range field images with severe same-class overlap. Existing visible-only labels can fragment one physi cal plant into disconnected supervision, while full-amodal labels may require unverifi able completion behind neighboring plants or field objects. EgoMaize therefore uses an evidence-closed annotation workflow for occluded maize regions and assigns unreli able maize regions to ignore rather than background. Baseline results show that pre trained query-based grouping, boundary refinement, and high-resolution crop refine ment help different aspects of the task, but no architecture solves the coupled chal lenges of fine structure recovery, same-class instance ownership, and occlusion reason ing; occlusion-level analysis further shows that performance decreases as plant visi bility becomes more limited. The dataset and code are publicly available at https: //github.com/JaaaaaaaD/EgoMaize.

Chinese Translation

近距离第一人称田间图像对于移动玉米表型分析至关重要，因为许多植物级特征依赖于从上方视角难以观察的冠层内部结构。然而，后期幼苗阶段的玉米田创造了一个困难的实例分割环境：茎、叶、雄穗和邻近植物呈现出拉长、重复且强烈遮挡的特征。我们引入了EgoMaize，这是一个紧凑的第一人称玉米实例分割基准，任务是从具有严重同类重叠的近距离田间图像中预测一致的植物掩膜和植物拥有的茎/雄穗线索。现有的仅可见标签可能将一个物理植物分割成不连贯的监督，而全模态标签可能需要在邻近植物或田间物体后面进行无法验证的补全。因此，EgoMaize采用了一种针对遮挡玉米区域的证据闭合注释工作流程，并将不可靠的玉米区域标记为忽略而非背景。基线结果表明，预训练的基于查询的分组、边界细化和高分辨率作物细化有助于任务的不同方面，但没有任何架构能够解决细结构恢复、同类实例所有权和遮挡推理的耦合挑战；遮挡级别分析进一步显示，随着植物可见性变得更加有限，性能下降。数据集和代码已公开发布在 https://github.com/JaaaaaaaD/EgoMaize。

View on arXiv Download PDF AI Translation

cs.CV / 14 / 2609.12374

Context-Aware Causal Gaze Forecasting for Human-Vehicle Interaction During In-Cabin Tracking Dropouts

基于上下文的因果注视预测在人车交互中的应用：应对车内跟踪中断

Shabani, Shabnam, Farhani, Ghazal

Abstract

Dashboard-mounted gaze trackers often lose sight of the driver's eyes during large head rotations, including shoulder checks, mirror glances, and intersection scanning. These maneuvers occur when information about the driver's visual attention is most useful. Offline gap-filling methods may reconstruct a missing interval using observations from both sides, but an online driver-monitoring system cannot rely on measurements that have not yet occurred. We therefore formulate causal gaze recovery: forecasting unavailable gaze at time t without target-tracker gaze at t or later. We introduce the Causal Context-Gated Forecaster (CCGF), which encodes a 60-frame pre-dropout history of gaze and head pose and combines it with DINOv3 scene features. A learned reliability gate controls the contribution of the history and scene representations as the dropout progresses. We evaluate two scene conditions: Live, in which the scene representation continues to update during tracker loss, and Frozen, in which the final pre-dropout representation is used throughout the missing interval. We evaluate CCGF on 2,047 eligible, naturally occurring GazeSense head\_lost events drawn from 10.5 h of naturalistic driving by ten drivers. Across all recordings, head\_lost accounts for 8.5 percent of GazeSense recording time. Synchronized gaze coordinates from a head-mounted Neon tracker provide supervision and evaluation targets but are never used as model inputs. Under leave-one-driver-out evaluation, CCGF achieves a mean per-driver median error of 175.7 px (10.5 deg) with Live scene updates, a 33 percent reduction relative to history-only causal forecasting. With Frozen scene input, the error increases to 210.8 px (12.9 deg), indicating that scene observations acquired during the dropout provide useful predictive information. We will release the dataset, evaluation protocol, and causal baselines.

Chinese Translation

安装在仪表盘上的注视追踪器在驾驶员进行大幅度头部旋转时（包括肩部检查、镜子瞥视和交叉路口扫描）常常无法捕捉到驾驶员的眼睛。这些动作发生时，关于驾驶员视觉注意力的信息最为重要。离线填补缺失数据的方法可以利用两侧的观察结果重建缺失的时间间隔，但在线驾驶员监测系统无法依赖尚未发生的测量。因此，我们提出了因果注视恢复：在时间 t 预测不可用的注视，而不依赖于时间 t 或之后的目标追踪器注视。我们引入了因果上下文门控预测器（Causal Context-Gated Forecaster, CCGF），它编码了 60 帧的预中断注视和头部姿态历史，并将其与 DINOv3 场景特征结合。一个学习的可靠性门控在中断进程中控制历史和场景表示的贡献。我们评估了两种场景条件：实时（Live），在追踪器丢失期间场景表示持续更新；以及冻结（Frozen），在整个缺失时间间隔内使用最后的预中断表示。我们在 10 名驾驶员的 10.5 小时自然驾驶中提取的 2,047 个符合条件的自然发生的 GazeSense 头部丢失事件上评估 CCGF。在所有录音中，头部丢失占 GazeSense 录音时间的 8.5%。来自头戴式 Neon 追踪器的同步注视坐标提供了监督和评估目标，但从未作为模型输入。在留一驾驶员评估下，CCGF 在实时场景更新下实现了每位驾驶员的平均中位误差为 175.7 像素（10.5 度），相较于仅使用历史的因果预测减少了 33%。使用冻结场景输入时，误差增加至 210.8 像素（12.9 度），表明在中断期间获取的场景观察提供了有用的预测信息。我们将发布数据集、评估协议和因果基线。

View on arXiv Download PDF AI Translation

cs.CV / 15 / 2609.12397

UFO: Chain-of-Evaluation for Omni-Condition Alignment in Multi-Modal Image Generation

UFO：多模态图像生成中的全条件对齐链式评估

Zhang, Danning, Lin, Yijing, Zhuang, Shuhan, Huang, Mengqi, Wu, Shaojin, Fang, Shancheng, Mao, Zhendong

Abstract

Multi-modal image generation, particularly subject-driven customization, has garnered growing attention in recent years. Despite the rapid advancement of generative models, their evaluation remains largely lagging. Existing methods, whether embedding-based or Multi-modal Large Language Model (MLLM)-based, evaluate alignment with each modal condition in isolation, which contradicts the simultaneous condition alignment objective of multi-modal image generation, leading to poor consistency with human judgments. To address this challenge, we propose UFO, the first unified framework for omni-condition alignment simultaneous evaluation. Specifically, UFO introduces a novel Atomized Chain-of-Evaluation paradigm, \emph{i.e.}, it first decomposes omni-condition alignment into a sequential chain of fine-grained, disentangled Atomic Evaluation Units (AEUs), categorizes them into distinct modality-relevance classes, and then employs general or dedicated functional calls for accurate verification of different AEU types. Experimental results demonstrate that UFO achieves the highest correlation with human evaluation preferences, delivering an average improvement of 15.25\%. Furthermore, we present UFO-Bench, a dedicated benchmark designed to holistically evaluate the performance of existing customization models under the diverse mutual interactions of textual and visual conditions.

Chinese Translation

多模态图像生成，特别是基于主题的定制，近年来受到越来越多的关注。尽管生成模型迅速发展，但其评估仍然滞后。现有方法，无论是基于嵌入的还是基于多模态大型语言模型（MLLM）的，都是孤立地评估与每个模态条件的对齐，这与多模态图像生成的同时条件对齐目标相悖，导致与人类判断的一致性较差。为了解决这一挑战，我们提出了UFO，这是第一个统一的全条件对齐的同时评估框架。具体而言，UFO引入了一种新颖的原子链式评估范式，即首先将全条件对齐分解为一系列细粒度、解耦的原子评估单元（Atomic Evaluation Units，AEUs），将其分类为不同的模态相关类别，然后采用通用或专用功能调用来准确验证不同类型的AEU。实验结果表明，UFO与人类评估偏好的相关性最高，平均提升达15.25%。此外，我们还提出了UFO-Bench，这是一个专门的基准，旨在全面评估现有定制模型在文本和视觉条件的多样相互作用下的性能。

View on arXiv Download PDF AI Translation

cs.CV / 16 / 2609.12408

GSO-Net: Visual State Machines for Hazardous Freight Transfer Compliance at Petrochemical Logistics Nodes

GSO-Net：用于石化物流节点危险货物转运合规的视觉状态机器

Xie, Yu, Xiong, Bangshu, Rao, Zhibo, Gan, Rui, Liu, Chongxuan, Ouyang, Zechu

Abstract

Hazardous-freight operations at petrochemical logistics nodes are safety-critical for intelligent transportation systems, yet existing vision benchmarks rarely address procedural compliance under realistic deployment constraints. In large infrastructure networks, cameras often operate under sparse round-robin polling, so transfer status must be inferred from incomplete observations and localized evidence. We present GSO-Net, a large-scale benchmark for visual understanding of standard operating procedures (SOPs) in petrochemical unloading scenarios. To our knowledge, GSO-Net is the first public benchmark dataset dedicated to visual SOP understanding in petrochemical hazardous-freight transfer scenarios. It contains over 50,000 independently sampled frames from 64 real expressway petrochemical logistics nodes and adopts an SOP-derived hierarchy linking 9 macroscopic procedural steps with 15 microscopic operational states. Two tasks are defined: joint detection of microscopic states and macroscopic steps as the core benchmark, and frame-level step classification as a diagnostic reference. Experiments with lightweight, transformer-based, open-vocabulary, and holistic models reveal a clear gap between object perception and transfer-stage understanding. Current models remain weak on contact-level state grounding, transient step recognition, and stage consistency, especially under sparse polling, tiny critical targets, and long-tailed operational evidence. GSO-Net provides a practical benchmark for fine-grained state perception and vision-based safety monitoring in hazardous freight transportation. The dataset is publicly available at https://github.com/yuxieHarrison/GSO-Net

Chinese Translation

石化物流节点的危险货物操作对于智能交通系统至关重要，但现有的视觉基准很少在现实部署约束下解决程序合规性的问题。在大型基础设施网络中，摄像头通常在稀疏的轮询模式下工作，因此转运状态必须通过不完整的观察和局部证据进行推断。我们提出了GSO-Net，这是一个用于石化卸货场景中标准操作程序（SOP）视觉理解的大规模基准。根据我们的了解，GSO-Net是首个专门用于石化危险货物转运场景中视觉SOP理解的公开基准数据集。该数据集包含来自64个真实高速公路石化物流节点的超过50,000帧独立采样图像，并采用基于SOP的层级结构，将9个宏观程序步骤与15个微观操作状态相连接。定义了两个任务：微观状态和宏观步骤的联合检测作为核心基准，以及帧级步骤分类作为诊断参考。使用轻量级、基于变换器的、开放词汇的和整体模型的实验显示，物体感知与转运阶段理解之间存在明显差距。当前模型在接触级状态定位、瞬态步骤识别和阶段一致性方面仍然较弱，尤其是在稀疏轮询、小型关键目标和长尾操作证据的情况下。GSO-Net为危险货物运输中的细粒度状态感知和基于视觉的安全监控提供了一个实用的基准。该数据集可在 https://github.com/yuxieHarrison/GSO-Net 上公开获取。

View on arXiv Download PDF AI Translation

cs.CV / 17 / 2609.12409

OphBiWSSD: Scaling Temporal Action Localization in Ophthalmic Surgeries with Bidirectional Weight-tied State Space Duality

OphBiWSSD：利用双向权重绑定状态空间对偶性扩展眼科手术中的时间动作定位

Liu, Yang, Ma, Qionghong, Chae, Joongwon, Luo, Lihui, Shen, Yibing, Zhuo, Yulin, Zhu, Yingting, Chang, Jiashu, Zhong, Xiaoyun, Yu, Dongmei, Lobie, Peter E., Qin, Peiwu, Yang, Chengming

Abstract

High-frequency surgical maneuvers in ophthalmology necessitate high-fidelity temporal modeling, yet characterizing long-range procedural dependencies remains computationally prohibitive for attention-based architectures. Existing models often require aggressive temporal downsampling, which compromises the detection of fine-grained action boundaries and instrument-tissue interactions. To address these scalability constraints, we present OphBiWSSD, a framework that reformulates surgical temporal action localization leveraging Bidirectional State Space Duality. By employing a weight-tied selective scan mechanism that incorporates both preceding and succeeding surgical contexts, our approach facilitates the global synthesis of non-causal temporal cues with linear complexity. This streamlined architecture is well-suited to capture the bidirectional dependencies present in ophthalmic workflows, effectively bridging the gap between local boundary precision and long-range procedural context without incurring the quadratic memory overhead of traditional Transformers. Extensive experiments on the OphNet benchmark demonstrate that OphBiWSSD achieves state-of-the-art temporal localization performance, with mean Average Precisions of 44.42% on phases and 43.08% on operations, surpassing the baselines by 6.80% and 6.66%, respectively. Empirical validation indicates that our approach ensures precise temporal localization and offers a computationally viable pathway for deploying surgical intelligence systems in clinical environments. The code is publicly available at https://github.com/yo3nglau/OphBiWSSD.

Chinese Translation

眼科手术中的高频手术操作需要高保真度的时间建模，但对于基于注意力的架构，表征长范围的程序依赖性仍然在计算上具有挑战性。现有模型通常需要进行激进的时间下采样，这会影响细粒度动作边界和器械-组织交互的检测。为了解决这些可扩展性限制，我们提出了OphBiWSSD，一个重新构建手术时间动作定位的框架，利用双向状态空间对偶性。通过采用一种权重绑定的选择性扫描机制，该机制结合了前后手术上下文，我们的方法以线性复杂度促进了非因果时间线索的全局合成。这种简化的架构非常适合捕捉眼科工作流程中的双向依赖，有效地弥合了局部边界精度与长范围程序上下文之间的差距，而不增加传统Transformer的二次内存开销。在OphNet基准上的大量实验表明，OphBiWSSD在时间定位性能上达到了最先进的水平，阶段的平均精度为44.42%，手术的平均精度为43.08%，分别超过基线6.80%和6.66%。实证验证表明，我们的方法确保了精确的时间定位，并为在临床环境中部署手术智能系统提供了可行的计算路径。代码已公开发布在 https://github.com/yo3nglau/OphBiWSSD。

View on arXiv Download PDF AI Translation

cs.CV / 18 / 2609.12410

A Multimodal Explainable Deep Learning Framework for Alzheimer's Disease Diagnosis using 3D Magnetic Resonance Imaging and Clinical Data

基于3D磁共振成像和临床数据的阿尔茨海默病诊断多模态可解释深度学习框架

Brima, Yusuf, Atemkeng, Marcellin, Namamula, Lakshmana Rao, Vacavant, Antoine

Abstract

Dementia is a major and growing global health burden, with Alzheimer's disease (AD) accounting for most cases. Timely and accurate diagnosis is central to managing this burden and increasingly depends on integrating complementary clinical and imaging information. Multimodal deep learning can combine these modalities for AD diagnosis, but how its explanations behave across modalities, fusion strategies, and cohorts remains unclear. We developed an explainable multimodal framework pairing a 3D CNN encoder for T1-weighted MRI with a feedforward network for harmonized clinical and demographic data, comparing varied model setups on three-way and pairwise diagnostic tasks using 6,479 internal records from the ADNI and 1,703 independent records from the OASIS-3. On ADNI, the tabular-only model achieved the highest three-class AUC-ROC of 0.879 and best discriminated cognitively normal (CN) versus mild cognitive impairment (MCI; 0.903), while cross-attention performed best for MCI versus AD (0.861); CN versus AD was highly discriminative overall. On OASIS-3, the vision-only model performed best (three-class AUC-ROC 0.910); CN versus MCI remained difficult, and no fusion strategy consistently outperformed single modalities across tasks and cohorts. SHAP and Integrated Gradients identified the MMSE as the dominant tabular feature in both cohorts, with global feature rankings agreeing strongly in ADNI ($\rho=0.94$) and OASIS-3 ($\rho=0.96$); CAM-based explanations, however, changed with model configuration and cohort. These findings show that multimodal performance and explanations are task, modality, fusion, and cohort-dependent: a dominant cognitive signal persisted across cohorts, but feature contributions and CAM explanations did not, underscoring the need to evaluate explainability under cohort shift rather than as a stable, intrinsic property.

Chinese Translation

痴呆症是一个主要且日益严重的全球健康负担，其中阿尔茨海默病（AD）占据了大多数病例。及时准确的诊断对于管理这一负担至关重要，并且越来越依赖于整合互补的临床和影像信息。多模态深度学习可以将这些模态结合用于AD诊断，但其解释在不同模态、融合策略和人群中的表现仍不清楚。我们开发了一个可解释的多模态框架，将用于T1加权MRI的3D CNN编码器与用于协调的临床和人口统计数据的前馈网络相结合，比较了在三类和成对诊断任务中不同模型设置的表现，使用了来自ADNI的6,479条内部记录和来自OASIS-3的1,703条独立记录。在ADNI中，仅使用表格数据的模型达到了最高的三类AUC-ROC为0.879，并且在区分认知正常（CN）与轻度认知障碍（MCI）方面表现最佳（0.903），而交叉注意力在MCI与AD的区分中表现最佳（0.861）；总体而言，CN与AD的区分能力非常强。在OASIS-3中，仅使用视觉数据的模型表现最佳（三类AUC-ROC为0.910）；CN与MCI的区分仍然困难，且没有任何融合策略在各任务和人群中始终优于单一模态。SHAP和集成梯度方法识别出MMSE是两个群体中占主导地位的表格特征，ADNI和OASIS-3的全局特征排名高度一致（$ ho=0.94$ 和 $ ho=0.96$）；然而，基于CAM的解释随着模型配置和人群的变化而变化。这些发现表明，多模态的表现和解释依赖于任务、模态、融合和人群：一个主导的认知信号在不同人群中持续存在，但特征贡献和CAM解释并不一致，强调了在群体变化下评估可解释性的必要性，而不是将其视为稳定的内在属性。

View on arXiv Download PDF AI Translation

cs.CV / 19 / 2609.12411

DERA: Detached Edge-Residual Adaptation for Prohibited item Detection

DERA：用于禁忌物品检测的分离边缘残差适应框架

Michael, Yonathan, Alansari, Mohamad, Bennamoun, Mohammed, Mahapatra, Dwarikanath, Henschel, Andreas, Werghi, Naoufel

Abstract

Prohibited-item detection in X-ray imagery remains challenging due to object superposition, weak texture, and material clutter which obscure both semantic appearance and object boundaries. We propose \textbf{DERA}, a \textbf{D}etached \textbf{E}dge-\textbf{R}esidual \textbf{A}daptation framework for prohibited item detection under X-ray imagery. DERA combines hierarchical visual features with a parallel pixel-difference edge pyramid and learns an object-specific boundary prior from training-time contours of the instance masks. The detached prior gates edge-sensitive features, which are injected into the early visual stages through residual heads. This staged design preserves the foundation detector at the start of adaptation, isolates boundary supervision from semantic feature learning, and restricts the final adaptation stage to only $14.7$K trainable parameters. Evaluated on PIDray, CLCXray, and STCray, DERA improves the baseline by \textbf{3.1}, \textbf{1.6}, and \textbf{2.4} AP points, respectively.

Chinese Translation

在X射线图像中，禁忌物品检测仍然面临挑战，主要由于物体重叠、纹理弱和材料杂乱，这些因素模糊了语义外观和物体边界。我们提出了 extbf{DERA}，一种用于X射线图像下禁忌物品检测的 extbf{D}etached extbf{E}dge- extbf{R}esidual extbf{A}daptation框架。DERA结合了分层视觉特征与并行像素差异边缘金字塔，并从实例掩膜的训练时轮廓中学习特定于物体的边界先验。分离的先验门控边缘敏感特征，这些特征通过残差头注入到早期视觉阶段。这种分阶段设计在适应开始时保留了基础检测器，将边界监督与语义特征学习隔离，并将最终适应阶段限制为仅有14.7K可训练参数。在PIDray、CLCXray和STCray上的评估表明，DERA分别提高了基线的 extbf{3.1}、 extbf{1.6}和 extbf{2.4} AP点。

View on arXiv Download PDF AI Translation

cs.CV / 20 / 2609.12417

Spectral Consistency-Guided Multiview Point Cloud Registration for Low-Overlap Scenes

基于光谱一致性引导的低重叠场景多视角点云配准

Li, Tianyu, Lin, Yanghong, Zhou, Shudong, Yang, Kui, Zhang, Jingru, Fang, Li, Yao, Wei

Abstract

Multiview point cloud registration is particularly challenging in low-overlap scenes, where reliable correspondences are limited and incorrect pairwise transformations can affect global pose estimation. In addition, registering all scan pairs is computationally expensive because many pairs provide weak geometric information. To address these problems, we propose GMPCR, a non-learning-based spectral consistency-guided framework for efficient and robust multiview point cloud registration. GMPCR builds a refined second-order compatibility structure from initial correspondences and uses its dominant spectral response to evaluate both correspondence reliability and scan-pair confidence. This allows unreliable correspondences to be filtered and informative scan pairs to be selected before relative transformation estimation, leading to a sparse pose graph and reduced pairwise registration cost. For each retained scan pair, maximal-clique-based hypothesis generation is used to estimate reliable relative transformations. The resulting pose graph is further refined by an adaptive history-aware synchronization scheme, in which the effect of residual history is adjusted according to changes in the global rotation residual. A recovery mechanism also allows down-weighted edges to regain confidence when their global consistency improves. Experiments on 3DMatch, 3DLoMatch, ScanNet, and ETH demonstrate the effectiveness of GMPCR. It achieves registration recalls of 97.2% and 89.6% on 3DMatch and 3DLoMatch, respectively, while maintaining competitive performance on ScanNet and ETH. The results show that GMPCR provides a favorable balance among registration accuracy, robustness to low overlap, and computational efficiency. The code is publicly available at https://github.com/swccj/gmpcr.

Chinese Translation

多视角点云配准在低重叠场景中尤为具有挑战性，因为可靠的对应关系有限，错误的成对变换可能影响全局姿态估计。此外，注册所有扫描对的计算开销较大，因为许多对提供的几何信息较弱。为了解决这些问题，我们提出了GMPCR，这是一种基于光谱一致性引导的非学习框架，旨在高效且稳健地进行多视角点云配准。GMPCR从初始对应关系中构建精细的二阶兼容性结构，并利用其主导光谱响应来评估对应关系的可靠性和扫描对的置信度。这使得不可靠的对应关系能够被过滤，并在相对变换估计之前选择信息量丰富的扫描对，从而形成稀疏的姿态图并降低成对配准的成本。对于每个保留的扫描对，采用基于最大团的假设生成方法来估计可靠的相对变换。生成的姿态图通过自适应历史感知同步方案进一步优化，其中残差历史的影响根据全局旋转残差的变化进行调整。恢复机制还允许权重降低的边在其全局一致性改善时重新获得信心。在3DMatch、3DLoMatch、ScanNet和ETH上的实验表明GMPCR的有效性。在3DMatch和3DLoMatch上，注册召回率分别达到了97.2%和89.6%，同时在ScanNet和ETH上保持了竞争力的性能。结果表明，GMPCR在配准精度、对低重叠的鲁棒性和计算效率之间提供了良好的平衡。代码已公开发布在 https://github.com/swccj/gmpcr。

View on arXiv Download PDF AI Translation

cs.CV / 21 / 2609.12431

An End-to-End Automated Pipeline for Controllable Crack Data Synthesis

可控裂缝数据合成的端到端自动化管道

Li, Conghui, Pu, Muxin, Lim, Chern Hong, Lin, Weiyao, Wang, Xin

Abstract

Automated crack inspection increasingly relies on deep learning, yet its reliability is limited by scarce and weakly controllable defect data. Existing generative augmentation methods often treat crack synthesis as a generic image-generation task, offering insufficient control over morphology, boundary fidelity, and scene context. This paper proposes an end-to-end automated pipeline for controllable crack data synthesis that formalizes crack geometry and inspection context into reusable computational constraints. First, procedurally sampled B\'ezier-curve skeletons are translated into realistic crack masks using a GAN, enabling scalable generation of diverse crack morphologies without manual mask design. Second, a dual-ControlNet diffusion framework disentangles appearance guidance from geometric guidance, with an edge-based branch enforcing strict boundary consistency. The framework supports both background-free synthesis and context-aware inpainting. Experiments on CRACK500 and CrackTree200 show consistent gains over existing augmentation baselines, demonstrating a scalable engineering informatics workflow for automated crack-inspection data generation.

Chinese Translation

自动化裂缝检测日益依赖深度学习，但其可靠性受到稀缺且可控性较弱的缺陷数据的限制。现有的生成增强方法通常将裂缝合成视为一种通用的图像生成任务，无法充分控制形态、边界保真度和场景上下文。本文提出了一种可控裂缝数据合成的端到端自动化管道，将裂缝几何形状和检测上下文形式化为可重用的计算约束。首先，程序化采样的贝塞尔曲线骨架通过生成对抗网络（GAN）转换为真实的裂缝掩膜，实现了多样化裂缝形态的可扩展生成，无需手动设计掩膜。其次，双重控制网络（dual-ControlNet）扩散框架将外观引导与几何引导分离，边缘基础分支强制执行严格的边界一致性。该框架支持无背景合成和上下文感知的修补。对CRACK500和CrackTree200的实验显示出相较于现有增强基线的一致性提升，展示了一种可扩展的工程信息学工作流程，用于自动化裂缝检测数据生成。

View on arXiv Download PDF AI Translation

cs.CV / 22 / 2609.12454

Bridging Vision Foundation Model Priors with CLIP for Spatial-aware Few-shot Anomaly Detection in Medical Images

将视觉基础模型先验与CLIP相结合，用于医学图像中的空间感知少样本异常检测

Miao, Juzheng, Yuan, Yuchen, Chen, Cheng, Heng, Pheng-Ann

Abstract

Vision-Language Models such as CLIP enable effective few-shot medical anomaly detection (AD) via strong image-text semantic alignment. However, their globally contrastive pretraining lacks explicit spatial supervision, limiting precise lesion localization. In contrast, Vision Foundation Models (VFMs) such as DINO learn spatially coherent patch representations via self-distillation and local-to-global consistency, better capturing fine-grained anatomical structures. Leveraging this complementarity, we propose Spatial-FAD, a spatial-aware few-shot medical AD framework that improves lesion localization by combining VFM spatial priors with CLIP semantics. Specifically, we introduce a VFM-enhanced adapter that injects a structural affinity prior derived from DINO into CLIP features. This structure-guided refinement encourages visual embeddings to better adhere to lesion boundaries while maintaining semantic alignment. To address the loss of spatial detail from patchification and the limited input resolution of CLIP, we adopt a sliding-window aggregation strategy. This generates high-resolution, spatially dense embeddings to further enhance localization granularity. Moreover, we introduce a prototype-enhanced support memory scheme to efficiently exploit the few-shot support set. This module stores compact prototypes for normal and abnormal patterns, reducing memory costs while boosting performance by fusing patch-to-prototype and image-text similarities. Extensive experiments on three benchmark datasets, including Liver CT, Retinal OCT, and Brain MRI, demonstrate that Spatial-FAD significantly outperforms state-of-the-art methods, especially in lesion segmentation. Notably, in the 4-shot scenario, our method achieves an average improvement of over 11.4% in Dice score and 1.8% in AUC. Code is available at: https://github.com/JuzhengMiao/Spatial-FAD.

Chinese Translation

视觉-语言模型如CLIP通过强大的图像-文本语义对齐，能够有效进行少样本医学异常检测（AD）。然而，它们的全局对比预训练缺乏明确的空间监督，限制了精确的病灶定位。相比之下，视觉基础模型（VFM）如DINO通过自我蒸馏和局部到全局的一致性学习空间一致的图块表示，更好地捕捉细粒度的解剖结构。利用这种互补性，我们提出了Spatial-FAD，一个空间感知的少样本医学AD框架，通过将VFM空间先验与CLIP语义相结合，改善病灶定位。具体而言，我们引入了一个VFM增强适配器，将从DINO衍生的结构亲和先验注入CLIP特征中。这种结构引导的细化鼓励视觉嵌入更好地贴合病灶边界，同时保持语义对齐。为了解决图块化带来的空间细节损失以及CLIP输入分辨率的限制，我们采用了滑动窗口聚合策略。这生成了高分辨率、空间密集的嵌入，以进一步增强定位的精细度。此外，我们引入了一个原型增强支持记忆方案，以高效利用少样本支持集。该模块存储正常和异常模式的紧凑原型，减少内存成本，同时通过融合图块到原型和图像-文本相似性来提升性能。在包括肝脏CT、视网膜OCT和脑MRI在内的三个基准数据集上的广泛实验表明，Spatial-FAD显著优于最先进的方法，尤其是在病灶分割方面。值得注意的是，在4-shot场景中，我们的方法在Dice分数上平均提高了超过11.4%，在AUC上提高了1.8%。代码可在以下链接获取：https://github.com/JuzhengMiao/Spatial-FAD。

View on arXiv Download PDF AI Translation

cs.CV / 23 / 2609.12473

Partition-Invariant Tuning for 3D Scene Understanding

用于三维场景理解的分区不变调优

Lin, Hongqiang, Wang, Tianle, Li, Shuiwang, Zhang, Dongxu, Sun, Yiding, Guo, Zihao, Yin, Dongfu

Abstract

Scene-level point cloud understanding remains challenging due to diverse geometries and spatial layouts. While pre-trained 3D point cloud foundation models (PFMs) offer strong transferability, full fine-tuning (FFT) incurs substantial computational and storage costs. Parameter-efficient fine-tuning (PEFT) provides a promising alternative, but existing PEFT methods largely focus on object-level point clouds and overlook serialization-induced partition variations in large-scale scenes. To address this issue, we propose PointPiT, a partition-invariant tuning framework for scene-level point clouds. Specifically, a Scene-aware Structural Adapter (SSA) integrates local geometric patterns with global scene context to mitigate partition-induced representation shifts. Moreover, Gradient Subspace Optimization (GSO) selects informative and partition-stable update directions, suppressing partition-dependent variations during optimization. Extensive experiments across multiple scene-level benchmarks demonstrate that PointPiT achieves competitive or even superior performance to full fine-tuning with less than 1% of backbone's parameters, while achieving consistent state-of-the-art performance among representative PEFT methods.

Chinese Translation

场景级点云理解仍然面临挑战，主要由于几何形状和空间布局的多样性。虽然预训练的三维点云基础模型（PFMs）提供了强大的迁移能力，但完全微调（FFT）会产生大量的计算和存储成本。参数高效微调（PEFT）提供了一种有前景的替代方案，但现有的PEFT方法主要集中在物体级点云上，忽视了大规模场景中序列化引起的分区变化。为了解决这个问题，我们提出了PointPiT，一种针对场景级点云的分区不变调优框架。具体而言，场景感知结构适配器（SSA）将局部几何模式与全局场景上下文相结合，以减轻分区引起的表示偏移。此外，梯度子空间优化（GSO）选择信息丰富且分区稳定的更新方向，在优化过程中抑制分区依赖的变化。在多个场景级基准测试中的广泛实验表明，PointPiT在使用不到1%的主干参数的情况下，达到了与完全微调相当甚至更优的性能，同时在代表性的PEFT方法中实现了一致的最先进表现。

View on arXiv Download PDF AI Translation

cs.CV / 24 / 2609.12491

PhysioAI: Clinical Knowledge-Guided Semantic Supervision for Skeleton-Based Physiotherapy Action Recognition

PhysioAI：基于临床知识引导的语义监督用于基于骨骼的物理治疗动作识别

Cao, Jie, Ahn, Euijoon, Hassan, Anwar, Kim, Jinman

Abstract

Skeleton-based action recognition can support automated tracking of physiotherapy exercises, particularly in remote rehabilitation settings where continuous in-person supervision is impractical. However, most existing methods are developed for large-scale daily-action benchmarks rather than rehabilitation scenarios. Public rehabilitation exercise datasets are typically small, with only subtle kinematic differences between exercise classes. For participants with motor impairments, exercise execution may also deviate from standard movement patterns in amplitude, speed, and coordination, increasing intra-class variability and making reliable recognition more difficult for skeleton-based models. We propose PhysioAI, a clinical knowledge-guided semantic supervision framework that injects structured physiotherapy knowledge into skeleton representation learning. PhysioAI combines graph-based spatiotemporal modelling of human movement with training-time semantic anchors derived from a structured Clinical Knowledge Dictionary (CKD). The CKD descriptions are encoded using a frozen Contrastive Language-Image Pre-training (CLIP) model and projected into an anchor space, where they provide class-specific semantic targets for skeleton representation learning. The resulting CKD-derived anchors are used only during skeleton-model training; inference requires only skeleton inputs. Under subject-disjoint evaluation, PhysioAI achieves $99.03\pm1.34\%$ on KiMoRe Overall, $94.64\pm7.36\%$ on the Hard-67 stress test, and $87.44\pm7.69\%$ on UI-PRMD Overall. These results exceed the strongest comparator for each endpoint by $0.27$, $2.87$, and $1.33$ percentage points (pp), respectively. These findings demonstrate that structured clinical knowledge can serve as an effective source of training-time supervision for physiotherapy action recognition.

Chinese Translation

基于骨骼的动作识别可以支持物理治疗练习的自动跟踪，特别是在面对面监督不切实际的远程康复环境中。然而，现有的大多数方法是针对大规模日常动作基准而开发的，而非康复场景。公共康复运动数据集通常较小，运动类别之间仅存在微妙的运动学差异。对于运动障碍的参与者，运动执行在幅度、速度和协调性上可能也会偏离标准运动模式，增加类内变异性，使得基于骨骼的模型更难以实现可靠识别。我们提出了PhysioAI，一种基于临床知识引导的语义监督框架，将结构化的物理治疗知识注入骨骼表示学习中。PhysioAI结合了基于图的时空建模和来自结构化临床知识字典（Clinical Knowledge Dictionary, CKD）的训练时语义锚点。CKD描述使用冻结的对比语言-图像预训练模型（Contrastive Language-Image Pre-training, CLIP）进行编码，并投影到锚点空间，在该空间中为骨骼表示学习提供类别特定的语义目标。生成的CKD衍生锚点仅在骨骼模型训练期间使用；推理仅需要骨骼输入。在受试者不重叠的评估下，PhysioAI在KiMoRe整体上达到了$99.03 ext{±}1.34 ext{ ext{%}}$，在Hard-67压力测试中达到了$94.64 ext{±}7.36 ext{ ext{%}}$，在UI-PRMD整体上达到了$87.44 ext{±}7.69 ext{ ext{%}}$。这些结果分别超过了每个端点的最强对比者$0.27$、$2.87$和$1.33$个百分点（pp）。这些发现表明，结构化的临床知识可以作为物理治疗动作识别训练时监督的有效来源。

View on arXiv Download PDF AI Translation

cs.CV / 25 / 2609.12497

RoES: Rotational Equivariant Selective-frequency Fusion for Multimodal Images

RoES：用于多模态图像的旋转等变选择性频率融合

Wang, Jiabao, Liu, Wenjian, Cai, Yaoming, Zhang, Gengyu, Zhao, Boyan, Zhang, Zijia, Ding, Yao, Liu, Xiaobo

Abstract

Infrared-visible image fusion facilitates robust multimodal perception by integrating complementary textural nuances from visible sensors with thermal signatures from infrared systems. Due to the task's inherently ill-posed nature, existing methods heavily rely on structural priors but typically enforce rotation equivariance uniformly across all features. Such a holistic approach overlooks a critical distinction where low-frequency shared structures strictly adhere to equivariant constraints while high-frequency modality-specific details require greater flexibility to preserve unique information. To bridge this gap, we propose RoES, a Rotational Equivariant Selective-frequency fusion network. Instead of employing static decomposition, we introduce a trainable rotation-enhanced updater/predictor module to dynamically decouple low- and high-frequency components. The resulting representations are then processed through a dual-branch fusion module tailored for spectral consistency. Specifically, a rotation-equivariant Mamba is employed to capture long-range structural dependencies in the low-frequency domain, while a polar spectral attention-based Dual-Fourier block refines high-frequency details under explicit low-frequency guidance. Extensive experiments demonstrate that RoES consistently achieves state-of-the-art performance in both fusion quality and downstream object detection, establishing a robust solution for multimodal fusion by reconciling frequency-selective features with equivariant constraints. The source code is available at https://github.com/BryceLosky/RoES-Fusion.

Chinese Translation

红外-可见图像融合通过将来自可见传感器的互补纹理细节与红外系统的热特征相结合，促进了稳健的多模态感知。由于这一任务本质上是病态的，现有方法严重依赖结构先验，但通常在所有特征上均匀施加旋转等变性。这种整体方法忽视了一个关键区别：低频共享结构严格遵循等变约束，而高频模态特定细节则需要更大的灵活性以保留独特信息。为了弥补这一差距，我们提出了RoES，一个旋转等变选择性频率融合网络。我们引入一个可训练的旋转增强更新/预测模块，以动态解耦低频和高频成分，而不是采用静态分解。生成的表示随后通过一个针对光谱一致性量身定制的双支路融合模块进行处理。具体而言，采用旋转等变的Mamba来捕捉低频域中的长程结构依赖关系，而基于极坐标光谱注意力的双傅里叶块则在明确的低频指导下细化高频细节。大量实验表明，RoES在融合质量和下游目标检测方面始终实现了最先进的性能，为通过调和频率选择性特征与等变约束提供了一种稳健的多模态融合解决方案。源代码可在 https://github.com/BryceLosky/RoES-Fusion 获取。

View on arXiv Download PDF AI Translation

cs.CV / 26 / 2609.12505

LettuceVisSim: A Simulator That Generates Lettuce Image Time-series for Vision-Based Reinforcement Learning

LettuceVisSim：一个生成生菜图像时间序列的模拟器，用于基于视觉的强化学习

Zhu, Ziye, Ooster, Bert van 't, Sun, Congcong, van Henten, Eldert, Boersma, Sjoerd

Abstract

Vision-based reinforcement learning holds strong potential for decision-making in controlled environment agriculture (CEA). However, its development is hindered by the scarcity of labelled crop images. To address this gap, LettuceVisSim, a lettuce growth simulator that generates labelled time series of crop images, was developed and validated. The simulator contains a process-based model (PBM) for shoot dry weight dynamics, a canopy layout algorithm for deriving canopy layout representations from shoot dry weight, and a Unity rendering engine for image generation. Five findings support the simulator. First, the PBM reproduced shoot dry weight under dynamic plant-density management with $\mathrm{R}^{2}=0.84$. Second, a piecewise cubic regression mapped shoot dry weight to potential projected area with $\mathrm{R}^{2}=0.94$. Third, the canopy layout representation was validated using 12 experimental datasets each having different dynamic environmental and spacing conditions. It reproduced the ground coverage ratio dynamics observed in measured images, achieving $\mathrm{R}^{2}=0.84$ when driven by measured shoot dry weight and $\mathrm{R}^{2}=0.40$ (0.76 excluding one outlier) when driven by PBM-simulated values. Fourth, the Unity rendering engine converted canopy layout representations into RGB and segmentation images at less than 10~ms. Fifth, a demonstration showed that a lighting-control policy can be learned and applied by observing only crop images that were generated with LettuceVisSim, providing a proof of concept of vision-based reinforcement learning in CEA using LettuceVisSim.

Chinese Translation

基于视觉的强化学习在受控环境农业（CEA）中的决策制定中具有很大的潜力。然而，其发展受到标记作物图像稀缺的制约。为了解决这一问题，开发并验证了LettuceVisSim，这是一种生成标记作物图像时间序列的生菜生长模拟器。该模拟器包含一个用于茎干干重动态的过程模型（PBM）、一个用于从茎干干重推导冠层布局表示的冠层布局算法，以及一个用于图像生成的Unity渲染引擎。五项研究结果支持该模拟器。首先，PBM在动态植物密度管理下再现了茎干干重，$ ext{R}^{2}=0.84$。其次，分段立方回归将茎干干重映射到潜在投影面积，$ ext{R}^{2}=0.94$。第三，使用12个具有不同动态环境和间距条件的实验数据集验证了冠层布局表示。它再现了在测量图像中观察到的地面覆盖率动态，当由测量的茎干干重驱动时，达到$ ext{R}^{2}=0.84$；当由PBM模拟值驱动时，达到$ ext{R}^{2}=0.40$（排除一个异常值后为0.76）。第四，Unity渲染引擎在不到10毫秒的时间内将冠层布局表示转换为RGB和分割图像。第五，演示表明，可以通过仅观察使用LettuceVisSim生成的作物图像来学习和应用照明控制策略，提供了在CEA中使用LettuceVisSim进行基于视觉的强化学习的概念验证。

View on arXiv Download PDF AI Translation

cs.CV / 27 / 2609.12509

ChitraMiti: Benchmarking Visual Grounding and Modality Reliance in Bengali Geometric Reasoning

ChitraMiti：孟加拉几何推理中的视觉定位和模态依赖性基准测试

Reza, Khan Raiyan Ibne, Maria, Sanjana Aktar, Nimi, Sumaiya Tabassum, Arefeen, Md Adnan

Abstract

Evaluation of vision-language models (VLMs) for multimodal mathematical reasoning remains limited for low-resource languages and for geometry problems that require reading a diagram and a question together. We introduce ChitraMiti-12.8k, a synthetic benchmark of 12,874 Bengali planar geometry problems paired with structured 15-attribute descriptions, and NCTB-500, a complementary set of 500 diagrams manually extracted from Bengali school textbooks. Using a three-phase protocol that separates diagram-only, diagram-plus-description, and description-only inputs, we show across five open-weight and closed-source VLMs that description-only performance is statistically indistinguishable from diagram-plus-description performance, establishing structured descriptions as a sufficient textual proxy for controlled evaluation. Despite this, models remain poor at cross-modal verification, frequently misled by a swapped spatial relation even when they answer the unmodified item correctly. We further evaluate supervised adaptation on ChitraMiti-12.8k, finding that fine-tuning improves performance on both ChitraMiti-1k and NCTB-500, although a substantial gap to the strongest zero-shot model remains. Together, ChitraMiti-12.8k, NCTB-500, and our evaluation protocol offer a standardized way to study Bengali multimodal geometry reasoning and, more broadly, whether VLMs actually check their text against what they see. Our dataset and code are publicly available on Hugging Face at https://huggingface.co/datasets/RaiyanKhaan/ChitraMiti.

Chinese Translation

对多模态数学推理的视觉语言模型（VLMs）评估在低资源语言和需要同时阅读图表与问题的几何问题上仍然有限。我们引入了ChitraMiti-12.8k，这是一个包含12,874个孟加拉平面几何问题的合成基准，配有结构化的15个属性描述，以及NCTB-500，这是一个从孟加拉学校教科书中手动提取的500个图表的补充集。通过一个三阶段协议，分别处理仅图表、图表加描述和仅描述的输入，我们在五个开放权重和闭源的VLMs上展示了，仅描述的性能在统计上与图表加描述的性能无显著差异，确立了结构化描述作为受控评估的充分文本代理。尽管如此，模型在跨模态验证方面表现不佳，常常因空间关系的交换而受到误导，即使它们正确回答了未修改的项目。我们进一步在ChitraMiti-12.8k上评估了监督适应，发现微调在ChitraMiti-1k和NCTB-500上的性能都有所提升，尽管与最强的零样本模型之间仍存在显著差距。综上所述，ChitraMiti-12.8k、NCTB-500及我们的评估协议提供了一种标准化的方式来研究孟加拉多模态几何推理，以及更广泛地探讨VLMs是否真的将其文本与所见内容进行核对。我们的数据集和代码已在Hugging Face上公开，网址为https://huggingface.co/datasets/RaiyanKhaan/ChitraMiti。

View on arXiv Download PDF AI Translation

cs.CV / 28 / 2609.12517

One Skill Does Not Fit All: Automatic Discovery and Taxonomy-Guided Routing of Frame-Selection Skills for Long-Video Question Answering

一种技能并不适用于所有：自动发现和分类引导的长视频问答中的帧选择技能路由

Hu, Jian, Cheng, Zixu, Li, Da, Li, Wei, Liu, Ziquan, Gong, Shaogang

Abstract

Long-Video Question Answering (LVQA) requires locating decisive evidence in hour-scale videos under a limited frame budget. Most training-free methods apply the same frame-selection strategy to all questions, despite substantial variation in the evidence required by different question types. Our analysis shows that the relative effectiveness of frame-selection strategies varies across semantic categories and benchmarks, motivating adaptive evidence acquisition. In this paper, we introduce AutoSkill, a source-supervised framework for automatically discovering and routing executable frame-selection skills. Starting from a small labelled source pool, LLM agents iteratively propose, implement, evaluate, and refine candidate skills. For a target benchmark, AutoSkill uses only unlabelled question and option text to induce a shared semantic taxonomy, rewrite labelled source examples into the target style, and estimate a category-to-skill mapping. Neither target videos nor target answers are used in this process. At inference time, each question is assigned one skill, which selects the frames used in a single inference of the frozen video MLLM. Across five long-video benchmark splits, AutoSkill improves Qwen2.5-VL-7B and Qwen3.5-4B by 2.4% and 1.2%, respectively, demonstrating the effectiveness of our AutoSkill.

Chinese Translation

长视频问答（LVQA）需要在有限的帧预算下，在小时级别的视频中定位决定性证据。大多数无训练方法对所有问题应用相同的帧选择策略，尽管不同问题类型所需的证据存在显著差异。我们的分析表明，帧选择策略的相对有效性在语义类别和基准测试之间有所不同，这促使了自适应证据获取。在本文中，我们介绍了AutoSkill，一个源监督框架，用于自动发现和路由可执行的帧选择技能。从一个小的标记源池开始，LLM代理迭代地提出、实施、评估和优化候选技能。对于目标基准，AutoSkill仅使用未标记的问题和选项文本来诱导共享的语义分类法，将标记的源示例重写为目标风格，并估计类别到技能的映射。在这个过程中，既不使用目标视频也不使用目标答案。在推理时，每个问题被分配一个技能，该技能选择在冻结视频MLLM的单次推理中使用的帧。在五个长视频基准拆分中，AutoSkill分别提高了Qwen2.5-VL-7B和Qwen3.5-4B的性能2.4%和1.2%，证明了我们AutoSkill的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 29 / 2609.12521

Aligned Radiometric RGB-Thermal Fusion for UAV Facade Anomaly Screening

无人机外立面异常筛查的对齐辐射RGB-热融合

Yang, Yuan, Li, Shulei, Liang, Haobo

Abstract

Unmanned aerial vehicle facade inspection can combine red, green, and blue (RGB) imagery with thermal measurements to screen surface and subsurface anomalies. However, geometric discrepancies between the sensors and thermal image rendering can obscure spatial correspondence and weak temperature contrasts. This article presents a sensor-level pipeline comprising per-sensor correction, RGB-to-thermal registration, common-support cropping, and signed local contrast encoding of 16-bit radiometric measurements. The encoding preserves the distinction between locally hotter and colder regions and supplies the fourth input channel of a compact single-stream detector. We introduce M3T, a dataset of 674 paired RGB and radiometric thermal samples from five facade-inspection projects covering eight component and anomaly categories. The median residual registration error is 3.384 pixels, and a controlled-displacement analysis characterizes how the local contrast response changes under controlled displacement. Project-grouped four-fold evaluation yields mean average precision of 0.168 over intersection-over-union thresholds from 0.5 to 0.95, using 28.50 billion floating-point operations per image. A separate single-split ablation shows improved delamination detection over RGB-only and alternative thermal inputs, although aggregate accuracy does not improve over RGB alone. Evaluation on RGBT-Tiny shows mixed performance with rendered thermal imagery. These results characterize the category-specific benefits and limitations of aligned radiometric contrast for compact facade screening.

Chinese Translation

无人机外立面检查可以将红、绿、蓝（RGB）图像与热测量相结合，以筛查表面和次表面异常。然而，传感器之间的几何差异和热图像渲染可能会模糊空间对应关系，并削弱温度对比度。本文提出了一种传感器级管道，包括每个传感器的校正、RGB到热图的配准、共同支持裁剪以及16位辐射测量的带符号局部对比度编码。该编码保留了局部较热和较冷区域之间的区别，并为紧凑型单流检测器提供了第四个输入通道。我们引入了M3T数据集，该数据集包含来自五个外立面检查项目的674对RGB和辐射热样本，涵盖八个组件和异常类别。中位数残差配准误差为3.384像素，受控位移分析表征了局部对比度响应在受控位移下的变化。按项目分组的四折评估在0.5到0.95的交并比阈值下，平均精确度为0.168，每幅图像使用28.50亿次浮点运算。单独的单分割消融实验显示，与仅使用RGB和其他热输入相比，分层检测有所改善，尽管整体准确度未超过仅使用RGB的情况。在RGBT-Tiny上的评估显示，与渲染的热图像相比，性能表现不一。这些结果表征了对齐辐射对比度在紧凑外立面筛查中的类别特定优势和局限性。

View on arXiv Download PDF AI Translation

cs.CV / 30 / 2609.12533

Earth-Agent-Pro: Towards Real-World Full-Chain Earth Observation with Agents

Earth-Agent-Pro：面向真实世界的全链条地球观测代理

Lv, Zhutao, Dang, Chenhao, Feng, Yi, Gong, Yanpei, Wang, Xiaolei, Ye, Junyan, He, Conghui, Li, Weijia

Abstract

Real-world Earth observation (EO) agents must translate high-level scientific questions into executable workflows to acquire observations, prepare data, perform domain computations, and derive conclusions from runtime evidence. Existing EO agents typically start from supplied observations, while benchmarks typically provide prepared inputs or candidate answers, leaving full-chain open-world EO execution largely untested. We present Earth-Agent-Pro, an execution-adaptive Plan-and-Execute framework using expert-authored skills to constrain planning and runtime tool use. Workflow-centered structured memory records planned steps, accepted evidence, and their dependencies, enabling repair of only the affected workflow suffix when runtime evidence invalidates a step. Separate large language model adapters use sequence-level supervised fine-tuning for planner workflow composition and node-level group relative policy optimization with locally verifiable rewards for executor tool-argument grounding. Earth-Bench-Pro instantiates 248 expert-curated task cores as 744 questions under three matched regimes. Its 248 Open-World Execution questions span RGB imagery, spectral observations, and remote sensing products, pairing high-level requests with runtime data requirements, executable trajectories, and open-ended answers grounded in execution evidence. With a shared GPT-5 backbone, Earth-Agent-Pro achieves 66.13% LLM-as-Judge accuracy, exceeding ReAct by 20.95 points in this metric and 24.44 points in Tools-In-Order. Joint adapter tuning raises Qwen3.5-9B LLM-as-Judge accuracy from 38.31% to 50.00%, an 11.69-point gain over the untuned configuration. Planning-only evaluation and execution with the reference workflow show that the adapters improve workflow composition and argument grounding, respectively. Code and datasets will be released soon.

Chinese Translation

真实世界的地球观测（EO）代理必须将高层次的科学问题转化为可执行的工作流程，以获取观测数据、准备数据、执行领域计算，并根据运行时证据得出结论。现有的EO代理通常从提供的观测数据开始，而基准测试通常提供准备好的输入或候选答案，这使得全链条开放世界EO执行在很大程度上未得到验证。我们提出了Earth-Agent-Pro，这是一种执行自适应的计划与执行框架，利用专家编写的技能来限制规划和运行时工具的使用。以工作流程为中心的结构化记忆记录了计划步骤、接受的证据及其依赖关系，使得在运行时证据使某一步骤失效时，仅修复受影响的工作流程后缀。独立的大型语言模型适配器使用序列级监督微调进行规划者工作流程的组合和节点级组相对策略优化，结合局部可验证的奖励进行执行者工具参数的基础。Earth-Bench-Pro将248个专家策划的任务核心实例化为744个问题，涵盖三种匹配的机制。其248个开放世界执行问题涵盖RGB影像、光谱观测和遥感产品，将高层次请求与运行时数据需求、可执行轨迹和基于执行证据的开放式答案相结合。在共享的GPT-5基础上，Earth-Agent-Pro在LLM-as-Judge准确率上达到了66.13%，在该指标上超越ReAct 20.95个百分点，在Tools-In-Order上超越24.44个百分点。联合适配器调优将Qwen3.5-9B LLM-as-Judge准确率从38.31%提升至50.00%，相比未调优配置提高了11.69个百分点。仅进行规划评估和使用参考工作流程的执行表明，适配器分别改善了工作流程组合和参数基础。代码和数据集将很快发布。

View on arXiv Download PDF AI Translation

cs.CV / 31 / 2609.12552

RelateAnything: Real-Time Open-Vocabulary Relation Prediction From Any Inputs

RelateAnything：来自任意输入的实时开放词汇关系预测

Neau, Maëlic

Abstract

Open-vocabulary detection accepts any class list at inference, and promptable segmentation returns regions without class names: the taxonomy has left the model and become an input. Relation prediction has not. Scene-graph models are still trained and evaluated on the 50 or 56 predicates of one annotation style, their relation head conditioned on object labels and so tied to one detector. Three obstacles explain this, none primarily modelling: no relation corpus is both free-text and verified, a label-conditioned architecture cannot accept a vocabulary it was not trained on, and the standard metric rewards agreement with the training corpus, so a larger vocabulary scores as a regression. We present RelateAnything, a 53M-parameter model taking an image and regions from any source and returning scored relations over a predicate vocabulary supplied at inference as strings. Object labels are never an input, so the region source can change without retraining, and the vocabulary is a bank of text embeddings, not a learned classifier. It runs at 20 ms/frame. Training over 19,103 predicates requires positive-unlabeled supervision and a text encoder that separates antonyms, which contrastive encoders embed at cosine 0.95. To supply the supervision we build RA-4M, 474k images and 4.3M relations over 10,102 free-text predicates, generated against numbered box markers and geometrically verified. To measure it we build OV-SGG-Bench, six axes scored across datasets that the priors standard recall rewards cannot satisfy. On three cross-dataset benchmarks and a fourth zero-shot, RelateAnything has 2.3-3.5x the mean recall of the strongest open-vocabulary method of comparable scale, margins that survive a real detector, and leads a 3B-VLM scene-graph model on both metrics at under 2% of its parameters. In-domain measurement overstates transfer gains ~5x. Model, corpus and benchmark are public.

Chinese Translation

开放词汇检测在推理时接受任何类别列表，而可提示的分割返回没有类别名称的区域：分类法已经脱离模型，成为输入。关系预测尚未如此。场景图模型仍然在一种注释风格的50或56个谓词上进行训练和评估，其关系头依赖于对象标签，因此与一个检测器紧密相连。有三个障碍解释了这一点，其中没有一个主要是建模：没有一个关系语料库既是自由文本又经过验证，标签条件架构无法接受未经过训练的词汇，标准度量奖励与训练语料库的一致性，因此更大的词汇在回归中得分。我们提出了RelateAnything，一个拥有5300万个参数的模型，接受来自任何来源的图像和区域，并在推理时返回以字符串形式提供的谓词词汇的得分关系。对象标签从未作为输入，因此区域来源可以在不重新训练的情况下变化，而词汇是文本嵌入的库，而不是学习的分类器。它的运行速度为每帧20毫秒。对19,103个谓词的训练需要正无标记监督和一个能够区分反义词的文本编码器，而对比编码器在余弦相似度0.95处嵌入。为了提供监督，我们构建了RA-4M，包含474,000张图像和4.3百万个关系，覆盖10,102个自由文本谓词，这些谓词是针对编号框标记生成并经过几何验证的。为了测量它，我们构建了OV-SGG-Bench，六个维度在数据集上得分，而先验标准召回奖励无法满足。在三个跨数据集基准和一个零样本基准上，RelateAnything的平均召回率是同规模最强开放词汇方法的2.3-3.5倍，这一差距在真实检测器中依然存在，并且在两个指标上领先于一个3B-VLM场景图模型，参数量不到其2%。领域内测量高估了转移增益约5倍。模型、语料库和基准都是公开的。

View on arXiv Download PDF AI Translation

cs.CV / 32 / 2609.12557

DRS-VPT: Directly Relocalizing in a Scan with Vision Point Transformers

DRS-VPT：使用视觉点变换器直接在扫描中重新定位

Fu, Lanke Frank Tarimo, Fallon, Maurice

Abstract

We present DRS-VPT, a feed-forward transformer architecture for foundational image-to-scan registration. Given query images and a reference 3D point cloud, the model predicts the scan pose and point map alongside the poses and point maps of each camera, all expressed in the first camera's frame. It additionally predicts a coarse-to- fine pyramid of per-point and per-pixel features for direct reprojective alignment of the scan to the first image. This formulation unifies downstream tasks such as camera-LiDAR calibration in autonomous driving and indoor camera-to-map relocalization. A single DRS-VPT model achieves state-of-the-art performance for image-to-LiDAR registration in autonomous driving, competitive indoor relocalization without training map-specific weights, and strong zero-shot transfer to unseen environments. We also show qualitatively that the model learns complex scan-to-image projection properties such as occlusion of back-facing points.

Chinese Translation

我们提出了DRS-VPT，这是一种用于基础图像与扫描配准的前馈变换器架构。给定查询图像和参考3D点云，该模型预测扫描姿态和点图，同时预测每个相机的姿态和点图，所有信息均以第一台相机的坐标系表示。它还预测了一种粗到细的金字塔结构，用于每个点和每个像素特征，以实现扫描与第一幅图像的直接重投影对齐。这种表述统一了下游任务，例如自动驾驶中的相机-LiDAR校准和室内相机到地图的重新定位。单个DRS-VPT模型在自动驾驶中的图像与LiDAR配准方面实现了最先进的性能，在无需训练特定地图权重的情况下实现了竞争力的室内重新定位，并在未见环境中表现出强大的零样本迁移能力。我们还定性地展示了该模型学习了复杂的扫描到图像投影特性，例如背向点的遮挡。

View on arXiv Download PDF AI Translation

cs.CV / 33 / 2609.12559

KAD-Net: Kinematics-Aware Decoupled Learning for Robust 3D Hand Pose Estimation from a Single Depth Image

KAD-Net：基于运动学感知的解耦学习用于从单幅深度图像中进行鲁棒的3D手势估计

Lu, Jun, Chen, Zhenming, Chen, Lin, Tan, Kanlun, Li, Xiaoling, Liu, Qiao

Abstract

Due to the complexity of hand kinematics and self-occlusion, existing 3D hand pose estimation methods based on single depth images struggle to comprehensively model the topological dependencies among hand joints. Furthermore, traditional hierarchical multitask architectures enforce a shared feature space for both 2D joint localization and depth estimation, which can induce mutual interference. To address these challenges, we propose a Kinematics-Aware Decoupled Learning Network (KAD-Net) for robust 3D hand pose estimation. Specifically, we first design a Finger Topology Constraint (FTC) module to enhance the representation of distal joints. This module utilizes three consecutive finger joints to construct a local kinematic representation to impose topological constraints, which supplements the kinematic features of the distal joints. The FTC module leverages the structural context from visible joints to assist in locating occluded distal joints, thereby improving robustness to occlusion. Additionally, we propose a task-decoupled hierarchical multitask framework. This framework separates 2D joint localization from depth estimation and incorporates a dedicated multitask learning strategy for depth regression, effectively isolating the UV and depth features to mitigate mutual interference and negative transfer. Extensive experiments demonstrate that KAD-Net outperforms existing methods on several benchmark datasets (ICVL, NYU, and MSRA), achieving state-of-the-art accuracy in 3D hand pose estimation. Potential applications of KAD-Net include human-computer interaction, virtual reality and gesture-based control systems.

Chinese Translation

由于手部运动学的复杂性和自遮挡现象，现有基于单幅深度图像的3D手势估计方法难以全面建模手指关节之间的拓扑依赖关系。此外，传统的层次化多任务架构强制为2D关节定位和深度估计共享特征空间，这可能导致相互干扰。为了解决这些挑战，我们提出了一种运动学感知解耦学习网络（KAD-Net），用于鲁棒的3D手势估计。具体而言，我们首先设计了一个指尖拓扑约束（FTC）模块，以增强远端关节的表示能力。该模块利用三个连续的手指关节构建局部运动学表示，以施加拓扑约束，从而补充远端关节的运动学特征。FTC模块利用可见关节的结构上下文来帮助定位被遮挡的远端关节，从而提高对遮挡的鲁棒性。此外，我们提出了一个任务解耦的层次化多任务框架。该框架将2D关节定位与深度估计分离，并为深度回归引入了专门的多任务学习策略，有效隔离UV和深度特征，以减轻相互干扰和负迁移。大量实验表明，KAD-Net在多个基准数据集（ICVL、NYU和MSRA）上优于现有方法，在3D手势估计中达到了最先进的准确性。KAD-Net的潜在应用包括人机交互、虚拟现实和基于手势的控制系统。

View on arXiv Download PDF AI Translation

cs.CV / 34 / 2609.12574

An Ultra-Widefield Swept-Source OCTA Dataset and a Polar-Gated Mamba Network for Retinal Vessel Segmentation

超广视场扫频源光学相干断层血管成像数据集及极化门控Mamba网络用于视网膜血管分割

Liu, Yang, Shen, Yibing, Zhao, Keming, Jiang, Cenk, Qian, Zhenghang, Du, Zhicheng, Xiong, Chen, Shao, Qidong, Lin, Zijun, Hu, Yunqi, Zhou, Jingjing, Zhang, Lian, Lobie, Peter E., Qin, Peiwu, Yang, Chengming

Abstract

Ultra-widefield (UWF) swept-source optical coherence tomography angiography (SS-OCTA) enables large-area retinal vascular imaging, yet vessel segmentation at this scale lacks dedicated public benchmarks and comprehensive evaluation for quantitative vascular analysis. We introduce WOIVES, to our knowledge the first publicly available UWF SS-OCTA vessel-segmentation dataset, comprising 206 eyes from 152 participants with a 24x20mm^2 field of view. WOIVES spans emmetropia to high myopia and provides soft probability vessel annotations. We further propose PG-Mamba, a visual state space model that enhances conventional directional scans with two complementary polar-coordinate scan orders. An auxiliary Dynamic FOV Gating module performs spatial modulation at the bottleneck. PG-Mamba outperformed seven competitive approaches on broad segmentation metrics under cross-validation. It achieved the lowest median absolute errors for vessel density, fractal dimension, and vessel length density. WOIVES is publicly available on Zenodo (DOI: 10.5281/zenodo.21904672), and the PG-Mamba code is available at https://github.com/syb1234567/PG-Mamba.

Chinese Translation

超广视场（UWF）扫频源光学相干断层血管成像（SS-OCTA）能够实现大范围的视网膜血管成像，但在这一尺度下，血管分割缺乏专门的公共基准和全面的定量血管分析评估。我们介绍了WOIVES，这是我们所知的第一个公开可用的UWF SS-OCTA血管分割数据集，包含来自152名参与者的206只眼睛，视野范围为24x20mm²。WOIVES涵盖了从正视到高度近视的范围，并提供了软概率血管注释。我们进一步提出了PG-Mamba，这是一种视觉状态空间模型，通过两种互补的极坐标扫描顺序增强了传统的方向扫描。一个辅助的动态视场门控模块在瓶颈处执行空间调制。在交叉验证下，PG-Mamba在广泛的分割指标上优于七种竞争方法。它在血管密度、分形维数和血管长度密度方面实现了最低的中位绝对误差。WOIVES已在Zenodo上公开可用（DOI: 10.5281/zenodo.21904672），PG-Mamba代码可在https://github.com/syb1234567/PG-Mamba获取。

View on arXiv Download PDF AI Translation

cs.CV / 35 / 2609.12577

SCORE: SubDistribution-aware Collaborative Knowledge Reinforcing for Cloth-Hybrid Lifelong Person Re-Identification

SCORE：针对服装混合的终身人物重识别的子分布感知协作知识强化

Xu, Kunlun, Ma, Liangyu, Li, Jiangmeng, Tong, Xin, Liu, Xiaode, Guo, Yufei, Zhou, Jiahuan

Abstract

Lifelong Person Re-Identification (LReID) aims to train a unified person retrieval model from a non-stationary data stream. Existing LReID methods mainly focus on scenarios where the clothing of each person is consistent. Recently, the Cloth-Hybrid LReID (CH-LReID) where cloth-consistent and cloth-changing data alternately occur, has emerged as a more practical and challenging scenario. Due to the conflict between clothing-relevant and clothing-irrelevant knowledge, the well-known catastrophic forgetting problem is significantly exacerbated in this task. To address this issue, we propose a SubDistribution-aware COllaborative Knowledge REinforcing (SCORE) framework, where our key idea is explicitly modeling the intra-identity diversity to continually consolidate distinct cloth-consistent and cloth-changing knowledge. Specifically, an Adaptive SubDistribution Modeling mechanism is developed, where a set of distributional subprototypes is assigned to each identity to capture the intra-identity diversity, improving the compatibility between cloth-consistent and cloth-changing knowledge. Then, a Distributional Knowledge Reinforcement scheme is introduced, where the knowledge of old distributional subprototypes is retained in the new ones by a collaborative aligning mechanism. Extensive experiments show that our SCORE achieves the state-of-the-art performance. Our code is available at https://github.com/zhoujiahuan1991/ECCV2026-SCORE

Chinese Translation

终身人物重识别（LReID）旨在从非平稳数据流中训练一个统一的人物检索模型。现有的LReID方法主要集中在每个人的服装一致的场景中。最近，服装混合终身人物重识别（CH-LReID）作为一种更实际且具有挑战性的场景，出现了服装一致和服装变化的数据交替出现。由于与服装相关和与服装无关知识之间的冲突，著名的灾难性遗忘问题在这一任务中显著加剧。为了解决这一问题，我们提出了一种子分布感知协作知识强化（SCORE）框架，我们的核心思想是明确建模同一身份内部的多样性，以持续巩固不同的服装一致和服装变化知识。具体而言，我们开发了一种自适应子分布建模机制，为每个身份分配一组分布子原型，以捕捉同一身份内部的多样性，从而提高服装一致和服装变化知识之间的兼容性。然后，引入了一种分布知识强化方案，通过协作对齐机制将旧分布子原型的知识保留到新的分布子原型中。大量实验表明，我们的SCORE达到了最先进的性能。我们的代码可在https://github.com/zhoujiahuan1991/ECCV2026-SCORE获取。

View on arXiv Download PDF AI Translation

cs.CV / 36 / 2609.12622

RA-SOD: Reliability-Aware RGB-T Salient Object Detection under Modality Degradation

RA-SOD：考虑可靠性的RGB-T显著目标检测在模态退化下的应用

Gao, Hongbo, Li, Zhengyu, Nie, Xueru, Zhu, Dihao, Zhao, Lijun, Wang, Yunke, Xu, Chang

Abstract

RGB-Thermal (RGB-T) salient object detection leverages complementary cues from visible and thermal modalities to improve robustness in challenging environments. However, in real-world scenarios, the reliability of each modality is inherently unstable: RGB images degrade under low illumination, motion blur, and noise, while thermal imagery often suffers from contrast compression and sensor artifacts. Such degradation introduces unreliable perceptual evidence that can mislead cross-modal fusion and significantly deteriorate detection performance. To address this challenge, we propose RA-SOD, a reliability-aware RGB-T salient object detection framework that explicitly models modality reliability and integrates it into feature learning and cross-modal fusion. First, we introduce a reliability-conditioned representation that adaptively compensates degraded modality features while preserving structural cues. Second, an uncertainty-guided dual-stream refinement strategy progressively corrects cross-modal representations while suppressing unreliable evidence. Finally, we propose a pixel-wise modality competition mechanism that dynamically selects modality cues according to spatial reliability for fine-grained fusion. Extensive experiments on four benchmarks (VT821, VT1000, VT5000, and VT-IMAG) demonstrate that RA-SOD achieves state-of-the-art performance and exhibits strong robustness under severe modality degradation. Code and models are available at https://github.com/zaoxienian/RA-SOD.

Chinese Translation

RGB-热成像（RGB-T）显著目标检测利用可见光和热成像模态的互补线索，以提高在挑战性环境中的鲁棒性。然而，在现实场景中，每种模态的可靠性本质上是不稳定的：在低光照、运动模糊和噪声条件下，RGB图像会退化，而热成像通常会受到对比度压缩和传感器伪影的影响。这种退化引入了不可靠的感知证据，可能误导跨模态融合，并显著降低检测性能。为了解决这一挑战，我们提出了RA-SOD，一个考虑可靠性的RGB-T显著目标检测框架，明确建模模态可靠性并将其整合到特征学习和跨模态融合中。首先，我们引入了一种可靠性条件表示，能够自适应地补偿退化的模态特征，同时保留结构线索。其次，一种不确定性引导的双流精炼策略逐步校正跨模态表示，同时抑制不可靠的证据。最后，我们提出了一种逐像素模态竞争机制，根据空间可靠性动态选择模态线索，以实现精细融合。在四个基准数据集（VT821、VT1000、VT5000和VT-IMAG）上的大量实验表明，RA-SOD实现了最先进的性能，并在严重模态退化下表现出强大的鲁棒性。代码和模型可在 https://github.com/zaoxienian/RA-SOD 获取。

View on arXiv Download PDF AI Translation

cs.CV / 37 / 2609.12668

Beyond Ambiguous Visual Cues: Studying Physiological Disruptions and Cross-Modal Inconsistencies in Deepfake Videos

超越模糊视觉线索：研究深度伪造视频中的生理干扰和跨模态不一致性

Yang, Chenxi, Ouzar, Yassine, Boubchir, Larbi

Abstract

Recent deepfake detection studies increasingly suggest remote photoplethysmography (rPPG) signals as an authenticity cue. However, existing benchmarks lack physiological ground truth, and current detectors underexplore the cross-level relationship between facial features and physiological dynamics, often relying on late fusion or rPPG features alone. In this paper, we construct high-fidelity deepfake manipulations on established real rPPG datasets (COHFACE and UBFC-rPPG) to investigate how forgeries disrupt natural physiological signals and facial behavior at the same time. Building on this analysis, we propose a bidirectional co-attention fusion detector that jointly models rPPG and facial behavior tokens. This mechanism explicitly captures the cross-level dependencies between pulse dynamics and facial motion to learn a robust, joint authenticity representation. Extensive experiments using a subject-disjoint 5-fold evaluation demonstrate the superiority of our approach. Achieving a 92.80\% AUC on constructed datasets using face swapping and 96.78\% AUC on motion transfer, our model outperforms both the rPPG-only single modality baseline and the best feature-level fusion methods. Furthermore, transfer-learning result of the fusion detector on Celeb-DF-v2 while keeping both feature extractors fixed achieves 91.20\% accuracy and 86.08\% AUC, which suggests applicability under target-domain adaptation.

Chinese Translation

近期的深度伪造检测研究越来越多地将远程光电容积脉搏信号（rPPG）作为真实性线索。然而，现有基准缺乏生理真实数据，当前的检测器对面部特征与生理动态之间的跨层关系探讨不足，通常仅依赖于后期融合或单独的rPPG特征。本文构建了高保真度的深度伪造操控，基于已建立的真实rPPG数据集（COHFACE和UBFC-rPPG），以研究伪造如何同时干扰自然生理信号和面部行为。在此分析的基础上，我们提出了一种双向共同注意力融合检测器，该检测器联合建模rPPG和面部行为标记。该机制明确捕捉脉搏动态与面部运动之间的跨层依赖关系，以学习稳健的联合真实性表示。通过使用受试者不重叠的5折评估进行广泛实验，证明了我们方法的优越性。在使用面部交换构建的数据集上实现92.80\%的AUC，在运动转移上实现96.78\\%的AUC，我们的模型优于仅使用rPPG的单模态基线和最佳特征级融合方法。此外，在保持两个特征提取器固定的情况下，融合检测器在Celeb-DF-v2上的迁移学习结果达到了91.20\\%的准确率和86.08\\%的AUC，这表明其在目标领域适应中的适用性。

View on arXiv Download PDF AI Translation

cs.CV / 38 / 2609.12678

Detecting and Explaining Fake News Short Videos with Multimodal Content and Real-World Evidence

利用多模态内容和现实证据检测与解释假新闻短视频

Luo, Yifeng, Li, Yupeng, Tang, Ming, Guo, Jianxiong, Lan, Liang

Abstract

Short-video platforms have become a primary news source for the public, which has also enabled the widespread dissemination of fake news videos. We study the task of fake news video detection and explanation (FNVDE). Existing methods face two critical limitations. First, commonly used frame selection strategies may omit veracity-relevant cues or provide insufficient temporal context for understanding news videos. Second, prior methods neglect either multimodal understanding or evidence retrieval. To address these limitations, we propose NVKE-CEI, a unified system that integrates a news video keyframes extraction method (NVKE) and an FNVDE framework leveraging both content and evidence information (CEI). NVKE selects keyframes based on chronological changes in combined visual and OCR-text similarity. CEI employs two specialized LLM-based fact checkers (content-based and evidence-based) whose outputs are fused by a lightweight judge model. Extensive experiments show that NVKE-CEI outperforms state-of-the-art baselines while generating high-quality content-grounded explanations.

Chinese Translation

短视频平台已成为公众获取新闻的主要来源，这也促进了假新闻视频的广泛传播。我们研究假新闻视频检测与解释（FNVDE）任务。现有方法面临两个关键限制。首先，常用的帧选择策略可能会遗漏与真实性相关的线索，或提供不足的时间上下文以理解新闻视频。其次，之前的方法忽视了多模态理解或证据检索。为了解决这些限制，我们提出了NVKE-CEI，一个统一的系统，整合了新闻视频关键帧提取方法（NVKE）和利用内容与证据信息的FNVDE框架（CEI）。NVKE基于视觉与OCR文本相似性的时间变化选择关键帧。CEI采用两个专门的基于大型语言模型（LLM）的事实检查器（基于内容和基于证据），其输出通过一个轻量级判断模型进行融合。大量实验表明，NVKE-CEI在生成高质量的基于内容的解释的同时，超越了最先进的基线。

View on arXiv Download PDF AI Translation

cs.CV / 39 / 2609.12682

NOVA-GS: Noise-Aware View-Consistent Gaussian Splatting for Low-Light Novel View Synthesis

NOVA-GS：噪声感知的视图一致性高斯点云合成用于低光照新视图合成

A, Shaurya Pavan, Madhuri, Vemunuri Divya, Gawande, Yash Pradeep, Mitra, Kaushik

Abstract

Reconstructing 3D scenes under real-world low-light conditions remains challenging due to severe sensor noise, low signal-to-noise ratios, and degraded photometric consistency, which destabilize geometry estimation and novel view synthesis. Existing approaches often rely on well-lit reference data for reliable Structure-from-Motion (SfM) initialization under degraded inputs or apply per-view enhancement methods that introduce cross-view inconsistencies. To address these limitations, we propose \textbf{NOVA-GS}, a unified noise-aware framework for low-light 3D Gaussian Splatting that subsumes enhancement, denoising, and geometry optimization within a single process. Our method leverages VGGT-based feed-forward estimation to obtain robust camera poses and geometry directly from degraded inputs, eliminating the need for SfM. Building on this initialization, NOVA-GS integrates three coupled components: a structure-aware enhancement module for exposure correction, a self-supervised denoising module with blind-spot masking for pseudo-supervision, and a consistency-driven Gaussian Splatting optimization enforcing cross-view geometric coherence. We further introduce a noise-guided spherical harmonic regularization to suppress view-dependent artifacts in noisy regions. Extensive experiments on diverse real-world low-light datasets demonstrate improved geometric fidelity, color consistency, and robustness without requiring paired supervision or well-lit references. https://shaurya2524.github.io/nova-gs/

Chinese Translation

在现实世界的低光照条件下重建三维场景仍然具有挑战性，因为严重的传感器噪声、低信噪比和退化的光度一致性会破坏几何估计和新视图合成。现有方法通常依赖于良好照明的参考数据，以便在退化输入下进行可靠的运动结构（Structure-from-Motion, SfM）初始化，或者应用逐视图增强方法，这会引入视图间不一致性。为了解决这些局限性，我们提出了 extbf{NOVA-GS}，一个统一的噪声感知框架，用于低光照三维高斯点云合成，将增强、去噪和几何优化整合在一个单一过程中。我们的方法利用基于VGGT的前馈估计，从退化输入中直接获取稳健的相机姿态和几何信息，消除了对SfM的需求。在此初始化基础上，NOVA-GS集成了三个耦合组件：用于曝光校正的结构感知增强模块、具有盲点掩蔽的自监督去噪模块用于伪监督，以及驱动一致性的高斯点云优化，以强制执行视图间几何一致性。我们进一步引入了一种噪声引导的球面谐波正则化，以抑制噪声区域中的视图依赖伪影。在多种现实世界低光照数据集上的大量实验表明，我们的方法在几何保真度、颜色一致性和鲁棒性方面得到了改善，而无需配对监督或良好照明的参考。

View on arXiv Download PDF AI Translation

cs.CV / 40 / 2609.12691

Semantically Aligned Gradient-Driven Context-Preserving Image Editing

语义对齐的梯度驱动上下文保留图像编辑

Chiranjeev, Chiranjeev, Dosi, Muskan, Vatsa, Mayank, Singh, Richa

Abstract

Instruction-guided image editing has a training-time blind spot. Generative editors are never required to semantically verify whether their outputs actually satisfy the instruction. Supervision stops at reconstruction and input textual-level conditioning. This produces incomplete edits, spatial spillover, and poor localization. We present IABEdit, a model-agnostic framework that embeds differentiable semantic verification into training. A frozen vision-language model extracts spatially-aware descriptors from the ground-truth edit. A trainable aligner then reproduces them from the generated output. The residual between the two becomes a gradient that teaches the generator both what to edit and where, with no inference-time VLM cost. IABEdit is compatible with diverse backbones, including U-Net (Stable Diffusion) and MMDiT (FLUX), without altering their inference pipelines. On MagicBrush, it improves structural fidelity by +3.49 DINO-I over the best diffusion baseline and +1.26 over the best overall baseline, while remaining competitive on instruction alignment. It also achieves state-of-the-art instruction adherence performance on RealEdit and EMU Edit benchmarks based on embedding-based metrics. Most consequentially, on the D-LORD surveillance benchmark, it surpasses the proprietary Gemini agent by +5.13 DINO-P under heavy occlusion, where preserving identity is hardest. This shows that gradient-aligned VLM distillation holds up under real-world-like surveillance and occlusion conditions. Human and GPT-4o evaluations confirm perceptually precise, well-localized edits.

Chinese Translation

基于指令的图像编辑存在训练时的盲点。生成编辑器从未被要求在语义上验证其输出是否真正满足指令。监督仅停留在重建和输入文本级别的条件下。这导致了不完整的编辑、空间溢出和较差的定位。我们提出了IABEdit，一个模型无关的框架，将可微分的语义验证嵌入到训练中。一个冻结的视觉-语言模型从真实编辑中提取空间感知描述符。一个可训练的对齐器随后从生成的输出中再现这些描述符。两者之间的残差成为一个梯度，教导生成器编辑什么以及在哪里，无需推理时的视觉-语言模型成本。IABEdit与多种主干网络兼容，包括U-Net（Stable Diffusion）和MMDiT（FLUX），而不改变它们的推理管道。在MagicBrush上，它在结构保真度上比最佳扩散基线提高了+3.49 DINO-I，比最佳整体基线提高了+1.26，同时在指令对齐方面保持竞争力。它还在基于嵌入的指标上，在RealEdit和EMU Edit基准测试中实现了最先进的指令遵循性能。最重要的是，在D-LORD监控基准测试中，它在重度遮挡下超越了专有的Gemini代理，提升了+5.13 DINO-P，正是在这种情况下，身份的保留最为困难。这表明，梯度对齐的视觉-语言模型蒸馏在类似真实世界的监控和遮挡条件下表现良好。人类和GPT-4o的评估确认了感知上精确、定位良好的编辑。

View on arXiv Download PDF AI Translation

cs.CV / 41 / 2609.12705

ExpertHTR: Unified Handwritten Text Recognition with Multi-Task Learning and Sparse Mixture-of-Experts

ExpertHTR：基于多任务学习和稀疏专家混合模型的统一手写文本识别

Nam, Dang Hoai, Hieu, Nguyen Duy, Hieu, Quang Huu, Duy, Vo Nguyen Le

Abstract

Handwritten text recognition resources are often small and distributed across collections that differ in language, script, document structure, and annotation format, making joint page-level training difficult. We propose ExpertHTR, a unified vision-language framework that addresses this problem through complementary supervision and conditional model capacity. Structural annotations from heterogeneous datasets are first organized through a common Page-Region-Line representation and used to construct four related training tasks for complete transcription, physical-line coverage, text localization, and localized recognition, without requiring additional manual labels. Building on a jointly trained dense model, ExpertHTR introduces a sparse Mixture-of-Experts architecture with an always-active shared branch and conditionally routed full-MLP experts. Sparsegen allows the number of active routed experts to vary with the hidden representation, while routing regularization reduces persistent concentration on a small subset of experts. Experiments on seven heterogeneous handwriting benchmarks show that complementary supervision consistently improves training with page transcription alone, while joint multi-source training provides further gains on most datasets. The proposed sparse expert model further improves the dense baseline on six of the seven sources. The final unified model also substantially outperforms the evaluated general-purpose OCR and vision-language systems on most benchmarks and achieves state-of-the-art performance on the IAM paragraph-level benchmark, while specialized HTR systems remain stronger on several challenging collections.

Chinese Translation

手写文本识别资源通常较少，并且分布在语言、书写体、文档结构和注释格式各异的集合中，这使得联合页面级训练变得困难。我们提出了ExpertHTR，一个统一的视觉-语言框架，通过互补监督和条件模型能力来解决这一问题。来自异构数据集的结构化注释首先通过共同的页面-区域-行表示进行组织，并用于构建四个相关的训练任务，包括完整转录、物理行覆盖、文本定位和局部识别，而无需额外的人工标签。在共同训练的密集模型基础上，ExpertHTR引入了一种稀疏的专家混合架构，具有始终活跃的共享分支和条件路由的全多层感知机（full-MLP）专家。Sparsegen允许活跃路由专家的数量随隐藏表示而变化，而路由正则化减少了对小部分专家的持续集中。对七个异构手写基准的实验表明，互补监督在仅依赖页面转录的训练中始终能带来改进，而联合多源训练在大多数数据集上提供了进一步的提升。所提出的稀疏专家模型在七个源中的六个上进一步改善了密集基线。最终的统一模型在大多数基准测试中也显著超越了评估的通用OCR和视觉-语言系统，并在IAM段落级基准上实现了最先进的性能，而专门的手写文本识别系统在若干具有挑战性的集合上仍然表现更强。

View on arXiv Download PDF AI Translation

cs.CV / 42 / 2609.12723

Multimodal Floorplan Encoding: Learning Dense Modality-Invariant Representations

多模态平面图编码：学习密集的模态不变表示

Anadón, Xavier, Pautrat, Rémi, Wang, Rui

Abstract

Floorplans arise in many forms, from vector CAD drawings to raster renderings and sensor-derived density maps. This heterogeneity makes it difficult to build learning systems that transfer across modalities and support geometry-centric tasks such as alignment and retrieval. We introduce the Multimodal Floorplan Encoder (MMFE), which maps diverse 2D indoor representations into a shared dense latent grid. MMFE combines a frozen DINOv3 backbone with a trainable Dense Prediction Transformer (DPT) head, and is trained with a per-cell Information Noise-Contrastive Estimation (InfoNCE) objective that aligns spatially corresponding regions across modalities while using all other cells as negatives. To improve robustness to geometric distortions, we incorporate controlled similarity transformations and enforce geometric consistency through feature-grid warping. On Structured3D, a held-out out-of-domain dataset, MMFE improves cross-modal dense matching, enables robust similarity alignment with RANSAC, and yields strong retrieval when paired with learned aggregation.

Chinese Translation

平面图以多种形式出现，从矢量CAD图纸到光栅渲染以及传感器生成的密度图。这种异质性使得构建能够跨模态转移并支持以几何为中心的任务（如对齐和检索）的学习系统变得困难。我们提出了多模态平面图编码器（Multimodal Floorplan Encoder, MMFE），该编码器将多样的二维室内表示映射到共享的密集潜在网格中。MMFE结合了一个冻结的DINOv3主干网络和一个可训练的密集预测变换器（Dense Prediction Transformer, DPT）头，并采用每个单元的信息噪声对比估计（Information Noise-Contrastive Estimation, InfoNCE）目标进行训练，该目标在跨模态对齐空间对应区域的同时，将所有其他单元作为负样本。为了提高对几何失真的鲁棒性，我们引入了受控相似性变换，并通过特征网格扭曲强制执行几何一致性。在结构化3D（Structured3D）这一保留的域外数据集上，MMFE提高了跨模态密集匹配的效果，支持与RANSAC的鲁棒相似性对齐，并在与学习的聚合相结合时实现了强大的检索性能。

View on arXiv Download PDF AI Translation

cs.CV / 43 / 2609.12731

GRACE: Adaptive Concept Erasure with Geometry-Guided Retention in Diffusion Models

GRACE：基于几何引导保留的自适应概念消除框架在扩散模型中的应用

Gong, Qinghui, Liang, Yihuai, Xie, Yuanlun, Jain, Deepak Kumar, Štruc, Vitomir, Zhou, Zhengchun

Abstract

Text-to-image (T2I) diffusion models inevitably internalize sensitive or non-compliant concepts from large-scale pretraining data, necessitating post-hoc concept erasure. However, existing erasure methods often lack explicit constraints on parameter updates, leading to over-intervention and unintended semantic drift. In addition, many methods rely on manually crafted counterfactual supervision, such as surrogate prompts, which incurs substantial data construction costs that limit scalability to new concepts. To address these limitations, we propose GRACE, a structured concept erasure framework designed to enable localized and selective intervention. Specifically, we introduce a semantically weighted sensitive subspace estimation to precisely lock intervention directions, and employ lightweight subspace-constrained adapters to prevent global semantic disturbance. To eliminate the dependency on manual prompt engineering, we design an automatically decoupled safe-anchor mechanism. To mitigate semantic drift induced by excessive intervention, we introduce an energy-driven dynamic gating mechanism that adaptively controls the timing and strength of intervention at inference. Extensive experiments demonstrate that our method achieves a superior balance between erasure effectiveness and generation fidelity. Compared with the average performance of five state-of-the-art (SOTA) concept erasure methods, our method improves the fine-grained NSFW reduction rate by $17.86\%$, while reducing the macro-averaged target CLIP Score and preservation-oriented Fr\'echet Inception Distance (FID) by $4.75\%$ and $50.58\%$, respectively, indicating stronger concept suppression with substantially improved preservation of the original model's generative utility.

Chinese Translation

文本到图像（T2I）扩散模型不可避免地会从大规模预训练数据中内化敏感或不合规的概念，因此需要进行后期概念消除。然而，现有的消除方法往往缺乏对参数更新的明确约束，导致过度干预和意外的语义漂移。此外，许多方法依赖于手动构建的反事实监督，例如替代提示，这会产生大量的数据构建成本，从而限制了对新概念的可扩展性。为了解决这些局限性，我们提出了GRACE，一个结构化的概念消除框架，旨在实现局部和选择性的干预。具体而言，我们引入了一种语义加权的敏感子空间估计，以精确锁定干预方向，并采用轻量级的子空间约束适配器，以防止全局语义干扰。为了消除对手动提示工程的依赖，我们设计了一种自动解耦的安全锚机制。为了减轻过度干预引起的语义漂移，我们引入了一种基于能量驱动的动态门控机制，能够自适应地控制推理时干预的时机和强度。大量实验表明，我们的方法在消除有效性和生成保真度之间达到了优越的平衡。与五种最先进（SOTA）概念消除方法的平均性能相比，我们的方法将细粒度的NSFW减少率提高了17.86%，同时将宏观平均目标CLIP得分和以保留为导向的Fréchet Inception Distance（FID）分别降低了4.75%和50.58%，这表明在显著提高原始模型生成效用的同时，概念抑制效果更强。

View on arXiv Download PDF AI Translation

cs.CV / 44 / 2609.12744

AquaCubeAI-Powered Monitoring Turbidity on-board {\Phi}sat-2

基于AquaCubeAI的{}sat-2卫星上浑浊度监测

Di Stasio, Pietro, Razzano, Francesca, Liparulo, Elisa, Meoni, Gabriele, Longépé, Nicolas, Tapete, Deodato, Gamba, Paolo, Schirinzi, Gilda, Ullo, Silvia Liberata

Abstract

Timely monitoring of coastal water quality is critical for environmental protection, yet conventional satellite workflows rely on downlink and ground processing, introducing latency that can limit responsiveness to rapidly evolving turbidity events. To address this limitation, we propose AquaCubeAI, a lightweight machine-learning approach for onboard estimation of coastal water turbidity from {\Phi}sat-2 multispectral imagery. By shifting inference from the ground segment to the satellite, AquaCubeAI aims to enable lower-latency, more responsive, and more operationally useful turbidity monitoring under the strict compute and bandwidth constraints of spaceborne platforms. The model is trained on simulated {\Phi}sat-2 acquisitions spatially aligned with Copernicus Marine Service (CMEMS) High-Resolution Ocean Color (HR-OC) turbidity products over selected localized coastal sites spanning four European marine macro-regions. To provide a realistic evaluation of generalization in the presence of spatial correlation, we adopt a spatial block splitting protocol that mitigates data leakage between training and evaluation subsets. The main contributions of this work are: (i) a scalable dataset generation pipeline pairing simulated {\Phi}sat-2 multispectral patches with CMEMS HR-OC turbidity labels across selected localized European coastal sites; (ii) a compact Multi-Layer Perceptron (MLP)-based turbidity regressor trained under a leakage-aware geospatial split and tailored to embedded constraints; and (iii) a reformulation for dense spatial prediction via parameter sharing, enabling turbidity mapping and simple threshold-based anomaly masks for onboard decision logic. Embedded deployment on an Intel Myriad Vision Processing Unit (VPU) further confirms the feasibility of low-power hardware and supports low-latency inference from multispectral inputs.

Chinese Translation

及时监测沿海水质对环境保护至关重要，但传统卫星工作流程依赖于下行链路和地面处理，导致延迟，从而限制了对快速变化的浑浊事件的响应能力。为了解决这一限制，我们提出了AquaCubeAI，这是一种轻量级的机器学习方法，用于从{}sat-2多光谱影像中进行沿海水浑浊度的卫星上估计。通过将推理从地面段转移到卫星，AquaCubeAI旨在在空间平台严格的计算和带宽限制下，实现更低延迟、更具响应性和更具操作实用性的浑浊度监测。该模型在与Copernicus Marine Service (CMEMS)高分辨率海洋色彩（HR-OC）浑浊度产品空间对齐的模拟{}sat-2获取数据上进行训练，覆盖四个欧洲海洋大区的选定局部沿海站点。为了在存在空间相关性的情况下提供对泛化能力的真实评估，我们采用了一种空间块分割协议，以减轻训练和评估子集之间的数据泄漏。本研究的主要贡献包括：(i) 一个可扩展的数据集生成管道，将模拟的{}sat-2多光谱图像与选定的欧洲沿海站点的CMEMS HR-OC浑浊度标签配对；(ii) 一个基于紧凑型多层感知器（MLP）的浑浊度回归模型，在考虑泄漏的地理空间分割下进行训练，并针对嵌入式约束进行了定制；(iii) 通过参数共享进行密集空间预测的重新表述，使得浑浊度映射和基于简单阈值的异常掩膜能够用于卫星上的决策逻辑。在Intel Myriad视觉处理单元（VPU）上的嵌入式部署进一步确认了低功耗硬件的可行性，并支持从多光谱输入进行低延迟推理。

View on arXiv Download PDF AI Translation

cs.CV / 45 / 2609.12761

Same Encoder, Different Winner: A Paired-View Framework for Cell Painting Encoder Evaluation

相同编码器，不同赢家：一种用于细胞绘画编码器评估的配对视图框架

Treis, Tim, Moshkov, Nikita, Haslum, Johan Fredin, Singh, Shantanu, Theis, Fabian J.

Abstract

Vision encoders for Cell Painting are typically ranked by a single evaluation, commonly replicate mean average precision (mAP). We introduce CP-BG-Bench, a paired-view evaluation framework that holds the central cell fixed across four matched views (raw crop C, segmented S, and density-augmented variants CD and SD), ablating or augmenting surrounding pixels as a controlled intervention. Instantiating the framework on three datasets (JUMP-CP, RxRx1, RxRx3-core) and three encoders (DINOv3 ViT-B/16, OpenPhenom, SubCell) under four community-standard protocols (replicate mAP, scIB batch integration, CellProfiler feature prediction, cross-batch perturbation recall), we find that the four protocols rank the same encoders systematically differently, with disagreements decomposing along three axes: cell versus background, morphology versus context, and within-study versus across-batch. The largest effect: on RxRx3-core, SubCell with segmented inputs retains 94% of crop replicate mAP but only 32% of crop R@10, so the within-study signal preserved under segmentation is largely non-transferable; density augmentation recovers 84% of the within-study C-to-S gap but only 8% of the cross-batch gap. Segmented views predict CellProfiler features as well as or better than crops on two of three datasets, inverting the replicate-mAP ranking, and the C-to-S gap varies by an order of magnitude across datasets while remaining similar across encoders, indicating that background-driven gain is set by experimental design rather than by the encoder. Single-metric ranking of Cell Painting encoders is therefore sensitive to the protocol used, and protocol disagreements are interpretable as projections onto the three axes the paired-view design exposes. We will release the paired-view datasets, reconstruction pipelines, 36 trained checkpoints, aggregated embeddings, and the full evaluation suite.

Chinese Translation

细胞绘画的视觉编码器通常通过单一评估进行排名，通常是重复均值平均精度（mAP）。我们引入了CP-BG-Bench，这是一种配对视图评估框架，固定中心细胞并在四个匹配视图（原始裁剪C、分割S以及密度增强变体CD和SD）之间进行评估，作为一种受控干预，消融或增强周围像素。在三个数据集（JUMP-CP、RxRx1、RxRx3-core）和三个编码器（DINOv3 ViT-B/16、OpenPhenom、SubCell）下，采用四种社区标准协议（重复mAP、scIB批次集成、CellProfiler特征预测、跨批次干扰召回）实例化该框架，我们发现这四种协议对相同编码器的排名系统性地不同，分歧沿着三个轴解构：细胞与背景、形态学与上下文、以及研究内与跨批次。最大的影响：在RxRx3-core上，使用分割输入的SubCell保持了94%的裁剪重复mAP，但仅有32%的裁剪R@10，因此在分割下保留的研究内信号在很大程度上是不可转移的；密度增强恢复了84%的研究内C到S的差距，但仅恢复了8%的跨批次差距。分割视图在三个数据集中的两个上预测CellProfiler特征的效果与裁剪相当或更好，颠覆了重复mAP排名，而C到S的差距在数据集之间变化一个数量级，同时在编码器之间保持相似，这表明背景驱动的增益是由实验设计而非编码器设定的。因此，细胞绘画编码器的单一指标排名对所使用的协议敏感，协议之间的分歧可以被解释为对配对视图设计所揭示的三个轴的投影。我们将发布配对视图数据集、重建管道、36个训练检查点、聚合嵌入以及完整的评估套件。

View on arXiv Download PDF AI Translation

cs.CV / 46 / 2609.12787

LG-PF: Lightweight Confidence-Guided Polarization Image Fusion

LG-PF：轻量级置信引导的偏振图像融合

Huang, Zhuangfan, Kuang, Zhenyu, Wang, Gao, Liu, Yang, Tan, Haishu, Li, Xiaosong

Abstract

Polarization image fusion combines the stable luminance and structural information of the total- intensity image S0 with the material-sensitive details of the degree of linear polarization (DoLP) image. However, the reliability of DoLP varies spatially, and indiscriminate polarization transfer may amplify unstable responses or disturb the structural appearance anchored by S0. We therefore propose LG-PF, a lightweight confidence-guided framework that formulates polarization fusion as a selective residual transfer process. A Polarization Confidence Prior estimates spatially reliable polarization responses, a Mask-guided Multi-scale Fusion module regulates their transfer across three feature scales, and a Lightweight Context-aware Bounded Correction Head stabilizes local photometric and structural transitions. Confidence guidance is also incorporated into the optimization objectives to preserve reliable polarization details while suppressing unsupported responses. We also construct MSP, a multi-scene polarization fusion dataset containing 1000 pixel-aligned image pairs from 17 indoor and outdoor scene categories. LG-PF achieves the best results across all six evaluated metrics on MSP, while subset-based evaluations on PIF and GAND show promising transferability without fine-tuning. With only 0.2936 M parameters and an inference time of 21.712 ms per image, LG-PF achieves competitive fusion quality with low computational cost. The source code, dataset, and official data splits will be made publicly available upon publication.

Chinese Translation

偏振图像融合将总强度图像 S0 的稳定亮度和结构信息与线性偏振度 (DoLP) 图像的材料敏感细节相结合。然而，DoLP 的可靠性在空间上存在差异，随意的偏振传递可能会放大不稳定的响应或干扰由 S0 固定的结构外观。因此，我们提出了 LG-PF，一种轻量级的置信引导框架，将偏振融合表述为选择性残差传递过程。偏振置信先验估计空间上可靠的偏振响应，掩膜引导的多尺度融合模块调节它们在三个特征尺度上的传递，而轻量级上下文感知有界校正头则稳定局部光度和结构过渡。置信引导也被纳入优化目标，以保留可靠的偏振细节，同时抑制不支持的响应。我们还构建了 MSP，一个包含来自 17 个室内和室外场景类别的 1000 对像素对齐图像的多场景偏振融合数据集。LG-PF 在 MSP 上的六个评估指标中均取得了最佳结果，而在 PIF 和 GAND 上的子集评估显示出良好的迁移性，无需微调。LG-PF 仅需 0.2936 M 参数和每幅图像 21.712 毫秒的推理时间，即可实现具有低计算成本的竞争性融合质量。源代码、数据集和官方数据划分将在发表时公开。

View on arXiv Download PDF AI Translation

cs.CV / 47 / 2609.12798

LGFN: Lightweight Gated RGB-Polarization Fusion with Modality-Availability Conditioning for Camouflaged Object Detection

LGFN：轻量级门控RGB-极化融合框架，支持模态可用性条件下的伪装物体检测

Huang, Zhuangfan, Li, Xiaosong, Liu, Yang, Ye, Tao, Tan, Haishu

Abstract

Camouflaged object detection (COD) is an important engineering task in intelligent optical perception, but it remains challenging when targets closely resemble their surroundings. Polarization imaging provides complementary physical cues, whereas existing methods typically assume fixed multimodal input configurations and entangle intra-polarization coordination with interaction between red-green-blue (RGB) and polarization representations. We propose LGFN, a lightweight gated RGB-polarization fusion framework supporting separately optimized RGB-only and polarization-assisted configurations. A deterministic Modality Router selects the appropriate configuration according to polarization availability. In the multimodal configuration, an availability-conditioned Modality Gate calibrates the available polarization branches; the Gated Polarization Hub coordinates learned degree of linear polarization (DoLP) and angle of polarization (AoP) representations with explicit polarization cues; and RGB-Polarization Cross Fusion introduces the coordinated representation into the RGB hierarchy through controlled residual interaction. The multimodal configuration requires neither sample-dependent statistics nor handcrafted quality descriptors during inference. On the complete 230-image PCOD_1200 test set, the RGB-only configuration achieves a mean absolute error of 0.0090, a Dice score of 0.8806, and an intersection over union of 0.8144, obtaining the best results on all six metrics among the evaluated RGB-based methods. Under a common local reevaluation protocol, the multimodal configuration outperforms PolarNet and IPNet on all six metrics. Relative to IPNet, it reduces the parameter count, floating-point operations, and latency by 53.1%, 73.6%, and 63.0%, respectively.

Chinese Translation

伪装物体检测（COD）是智能光学感知中的一项重要工程任务，但当目标与其周围环境高度相似时，仍然具有挑战性。极化成像提供了互补的物理线索，而现有方法通常假设固定的多模态输入配置，并将内极化协调与红绿蓝（RGB）和极化表示之间的交互纠缠在一起。我们提出了LGFN，一个轻量级的门控RGB-极化融合框架，支持单独优化的RGB-only和极化辅助配置。一个确定性的模态路由器根据极化的可用性选择适当的配置。在多模态配置中，一个基于可用性的模态门校准可用的极化分支；门控极化中心协调学习到的线性极化度（DoLP）和极化角（AoP）表示与显式的极化线索；RGB-极化交叉融合通过受控的残差交互将协调表示引入RGB层次。多模态配置在推理过程中既不需要样本依赖的统计信息，也不需要手工设计的质量描述符。在完整的230幅图像PCOD_1200测试集上，RGB-only配置实现了0.0090的平均绝对误差、0.8806的Dice得分和0.8144的交并比，在评估的所有RGB基础方法中获得了六项指标的最佳结果。在一个共同的局部重新评估协议下，多模态配置在所有六项指标上均优于PolarNet和IPNet。相较于IPNet，它分别减少了53.1%、73.6%和63.0%的参数数量、浮点运算和延迟。

View on arXiv Download PDF AI Translation

cs.CV / 48 / 2609.12818

Online Video Agent Harness for Long Video Understanding

在线视频代理工具用于长视频理解

Yang, Sen, Duan, Boqiang, Yang, Jing, Bo, Weihao, Liu, Jie, Tong, Boyuan, Feng, Ze, Zhang, Wenkang, Wang, Jingdong, Wu, Hua

Abstract

Long video understanding often behaves like a visual needle-in-a-haystack problem: query-relevant evidence is sparsely distributed across long temporal spans, while packing dense frames into a single VLM context incurs \textit{context rot} and high cost. Existing video agents often rely on query-agnostic offline preprocessing or ad hoc tool sets, which can miss query-specific details and waste computation. In this work, we present VideoXAgent, a purely online video-agent harness for long video understanding that starts from the given video file and user query, plans and decomposes the task, invokes specialized expert tools on demand, and aggregates multimodal evidence to produce a final answer while resolving conflicts among observations. To support this on-demand invocation, we design a suite of heterogeneous expert tools guided by a data-driven taxonomy of atomic capabilities, spanning scripts, VLMs, and domain models (e.g., detection, OCR, ASR, face recognition). The harness further enforces objective evidence prompting and budget-aware control to curb hallucination and non-termination. Across Video-MME-Long, LongVideoBench-Long, LVBench, and MINERVA, VideoXAgent is competitive with frontier LMMs and video agents under a smaller context footprint---about 50k tokens of agent context per sample, even on hour-long videos. In particular, on complex video-reasoning benchmarks such as MINERVA, it matches this level while using only about 15\% of the context of a 1,024-frame dense-packing baseline. Notably, the harness remains effective with a visually weak or even text-only orchestrator, suggesting that strong long-video understanding can emerge from progressive agentic evidence seeking rather than from packing the full video into a single context. Project page: https://go-agent-x.github.io/video_agent_harness/

Chinese Translation

长视频理解通常表现得像是在大海捞针：与查询相关的证据在长时间跨度内稀疏分布，而将密集帧打包到单一的视觉语言模型（VLM）上下文中则会导致 extit{上下文衰退}和高成本。现有的视频代理通常依赖于与查询无关的离线预处理或临时工具集，这可能会遗漏查询特定的细节并浪费计算资源。在本研究中，我们提出了VideoXAgent，这是一种纯在线视频代理工具，旨在进行长视频理解，起始于给定的视频文件和用户查询，规划和分解任务，根据需求调用专业的专家工具，并聚合多模态证据以生成最终答案，同时解决观察之间的冲突。为了支持这种按需调用，我们设计了一套异构专家工具，依据数据驱动的原子能力分类法进行指导，涵盖脚本、VLM和领域模型（例如，检测、光学字符识别（OCR）、自动语音识别（ASR）、人脸识别）。该工具进一步强化了客观证据提示和预算感知控制，以抑制幻觉和非终止现象。在Video-MME-Long、LongVideoBench-Long、LVBench和MINERVA等基准测试中，VideoXAgent在较小的上下文占用下与前沿的语言模型（LMM）和视频代理相竞争——每个样本约50k个代理上下文，即使在长达一小时的视频中也是如此。特别是在复杂的视频推理基准测试如MINERVA中，它在仅使用约15 ext{%}的1,024帧密集打包基准的上下文的情况下达到了这一水平。值得注意的是，该工具在视觉上较弱或甚至仅为文本的协调者下仍然有效，这表明强大的长视频理解可以通过渐进的代理证据寻求而不是将整个视频打包到单一上下文中产生。项目页面：https://go-agent-x.github.io/video_agent_harness/

View on arXiv Download PDF AI Translation

cs.CV / 49 / 2609.12825

SCDM: Spatial-Contextual Disentanglement Mamba via Differential Inference for Efficient Image Classification

SCDM：通过差异推理实现高效图像分类的空间-上下文解缠绕Mamba

Çelik, Mustafa Bora, Dinçer, Hayriye Aktaş, Keles, Ayse

Abstract

State Space Models (SSMs), particularly VMamba, have emerged as efficient alternatives for modeling long-range dependencies in medical image analysis. However, distinguishing subtle pathological features from visually similar anatomical backgrounds remains a significant challenge. Existing SSM architectures often learn entangled representations, lacking explicit mechanisms to separate disease-specific signals from normal anatomy. To address this limitation, we propose Spatial-Contextual Differential Mamba (SCDM), an asymmetric dual-branch architecture designed for selective representational disentanglement. SCDM introduces a Positive Branch for extracting discriminative features and a Negative Branch that actively models and suppresses normal anatomical context. This separation is achieved through a similarity-driven repulsion gate and a differential inference rule, which promote competitive feature learning without requiring additional branch labels or increasing model capacity. Evaluated on the RSNA Pneumonia dataset, SCDM achieves competitive classification performance (AUC of 0.858) while requiring significantly fewer parameters (29.4M) and FLOPs (1.44G) compared to standard VMamba and vision transformer baselines. Furthermore, activation analyses demonstrate that our differential mechanism yields highly precise localization, effectively isolating lesions by inhibiting irrelevant anatomical distractors.

Chinese Translation

状态空间模型（SSMs），特别是VMamba，已成为在医学图像分析中建模长程依赖关系的高效替代方案。然而，从视觉上相似的解剖背景中区分微妙的病理特征仍然是一个重大挑战。现有的SSM架构往往学习到纠缠的表示，缺乏明确的机制来分离特定疾病信号与正常解剖结构。为了解决这一限制，我们提出了空间-上下文差异Mamba（SCDM），这是一种不对称双分支架构，旨在实现选择性表示解缠绕。SCDM引入了一个正分支用于提取区分性特征，以及一个负分支，积极建模并抑制正常解剖背景。通过相似性驱动的排斥门和差异推理规则实现这种分离，促进竞争性特征学习，而无需额外的分支标签或增加模型容量。在RSNA肺炎数据集上的评估显示，SCDM实现了具有竞争力的分类性能（AUC为0.858），同时所需参数显著减少（29.4M）和FLOPs（1.44G），相较于标准VMamba和视觉变换器基线。此外，激活分析表明，我们的差异机制产生了高度精确的定位，有效隔离病灶，抑制无关的解剖干扰因素。

View on arXiv Download PDF AI Translation

cs.CV / 50 / 2609.12826

CoralscapesV2: Panoptic and Fine-Grained Visual Scene Understanding in Coral Reefs

CoralscapesV2：珊瑚礁中的全景与细粒度视觉场景理解

Sauder, Jonathan, Ruckli, Thomas, Strodomskytė, Gabrielė, Abdallah, Ibrahim Souleiman, Abdi, Rahma Hassan, Awaleh, Djama Goumaneh, Farah, Mohamed Houssein, Nour, Moustapha, Saad, Osama Sharhubil, Altaib, Mustafa Mohammed Khalafallah, Kteifan, Maysoon, Alsoqi, Farah, Zgool, Eyad, Al-Omari, Jafar, Gebreluel, Temesgen Gebremeskel, Abdulkerim, Zekaria Zekeria, Ghirmay, Meron, Beraki, Teklehaimanot, Tuia, Devis, Banc-Prandi, Guilhem

Abstract

In order to design conservation and restoration strategies to counter the global decline of coral reefs, ecological monitoring of reefs needs to be scaled up dramatically. Computer vision methods are increasingly used to tackle the vast amount of data: as the paradigm of data collection in reefs shifts from highly standardized and constrained survey images to unconstrained imagery on scalable platforms, it is necessary to design machine learning methods that help to get a fine-grained understanding of reefs from general-purpose reef imagery. This paper provides CoralscapesV2, an extension of the Coralscapes dataset for general-purpose visual scene understanding in reefs. CoralscapesV2 increases the dataset size, scope, label completeness and quality for semantic segmentation, and extends the number of classes from 39 to 95 fine-grained visual categories. Furthermore, CoralscapesV2 provides 65k exhaustive fish instance mask annotations, meticulously annotated to completeness by using the video, revealing that annotation of fish based on only static images is insufficient. CoralscapesV2 is the first dataset for panoptic segmentation in coral reefs, capturing a wide range of scenarios in the wild, posing a challenging benchmark for contemporary semantic segmentation and instance segmentation models. CoralscapesV2 is an important step towards general-purpose panoptic segmentation in coral reefs, which has substantial implications for scaling up coral reef monitoring, as it can be employed in a wide range of applications from benthic cover mapping from robot or handheld videos to designing methods for automated quantification and understanding of fish behavior and fish-reef interactions.

Chinese Translation

为了设计应对全球珊瑚礁衰退的保护和恢复策略，珊瑚礁的生态监测需要大幅度提升。计算机视觉方法越来越多地被用于处理大量数据：随着珊瑚礁数据收集范式从高度标准化和受限的调查图像转向可扩展平台上的无约束图像，有必要设计机器学习方法，以帮助从通用珊瑚礁图像中获得对珊瑚礁的细粒度理解。本文提供了CoralscapesV2，这是Coralscapes数据集的扩展，旨在实现珊瑚礁的通用视觉场景理解。CoralscapesV2增加了数据集的规模、范围、标签的完整性和语义分割的质量，并将类别数量从39个扩展到95个细粒度视觉类别。此外，CoralscapesV2提供了65,000个详尽的鱼类实例掩膜注释，使用视频进行细致的完整性标注，揭示仅基于静态图像的鱼类注释是不够的。CoralscapesV2是珊瑚礁全景分割的首个数据集，捕捉了野外广泛的场景，为当代语义分割和实例分割模型提供了具有挑战性的基准。CoralscapesV2是实现珊瑚礁通用全景分割的重要一步，对提升珊瑚礁监测的规模具有重要意义，因为它可以应用于从机器人或手持视频的底栖覆盖映射到设计自动化量化和理解鱼类行为及鱼-礁相互作用的方法等广泛的应用。

View on arXiv Download PDF AI Translation

cs.CV / 51 / 2609.12827

A Dual Cross-Attention Framework for Colposcopic CIN Grading and Swede Score Prediction Using a New Multi-Center Dataset

一种双重交叉注意力框架用于基于新多中心数据集的阴道镜CIN分级和Swede评分预测

Khan, Dania, Shaikh, Nuzhat Aisha, Juicy, Asfina Hassan, Kabir, Raiyun, Shahida, S M, Hasan, Taufiq

Abstract

Cervical cancer is a major global health challenge, with disease burden falling disproportionately on low- and middle-income countries (LMICs) due to a shortage of trained specialists and the subjective nature of colposcopy-based screening. To address this challenge, we propose a novel deep learning framework for the automated grading of Cervical Intraepithelial Neoplasia (CIN) and the prediction of clinical Swede scores. We also introduce the BUET Multi-Center Colposcopy Dataset, a novel, multi-center cohort designed and annotated for Swede score prediction and CIN grading. Our proposed dual-stream cross-attention architecture mimics the visual reasoning of an expert colposcopist by explicitly fusing paired multimodal cervigrams to evaluate comparative tissue responses. Furthermore, we introduce a custom composite loss function to address severe class imbalances and scoring inconsistencies across the five Swede score components. The proposed framework achieved 71.85% accuracy and an 86.23% AUC-ROC for three-class CIN grading, outperforming existing methods. For Swede score component prediction, the architecture achieved AUC-ROC values ranging from 75.7% to 88.4%, with the composite loss function yielding consistent F1-score improvements. Finally, the total predicted Swede Score, which ranges between 0 and 10, shows a Mean Absolute Error (MAE) of 1.489. The results show that the proposed method can pave the way towards developing AI-assisted colposcopy screening tools to support risk-based triage in resource-limited healthcare settings. The dataset and source code are publicly available(url: https://github.com/mHealthBuet/BUET-colposcopy)

Chinese Translation

宫颈癌是全球面临的重大健康挑战，疾病负担在低收入和中等收入国家（LMICs）中尤为严重，这主要是由于缺乏训练有素的专业人员以及基于阴道镜检查的主观性。为了解决这一挑战，我们提出了一种新颖的深度学习框架，用于自动化宫颈上皮内瘤变（CIN）的分级和临床Swede评分的预测。我们还引入了BUET多中心阴道镜数据集，这是一个专为Swede评分预测和CIN分级设计和注释的新型多中心队列。我们提出的双流交叉注意力架构通过明确融合成对的多模态宫颈图像，模拟专家阴道镜医生的视觉推理，以评估比较组织反应。此外，我们引入了一种自定义复合损失函数，以解决五个Swede评分组件之间的严重类别不平衡和评分不一致性。所提框架在三类CIN分级中达到了71.85%的准确率和86.23%的AUC-ROC，优于现有方法。在Swede评分组件预测中，该架构的AUC-ROC值范围从75.7%到88.4%，复合损失函数带来了F1-score的一致性提升。最后，预测的总Swede评分范围在0到10之间，平均绝对误差（MAE）为1.489。结果表明，所提方法可以为开发AI辅助的阴道镜筛查工具铺平道路，以支持资源有限的医疗环境中的基于风险的分诊。数据集和源代码已公开可用（网址：https://github.com/mHealthBuet/BUET-colposcopy）

View on arXiv Download PDF AI Translation

cs.CV / 52 / 2609.12830

Balancing Emotional Alignment and Semantic Consistency in Image Generation via Reinforcement Learning with Valence-Arousal Anchoring

通过带有情感-唤醒锚定的强化学习在图像生成中平衡情感一致性和语义一致性

Dang, Jisheng, Wang, Zhenxuan, Li, Bin, Lin, Ronghao, Hu, Bin, Chua, Tat-Seng

Abstract

Continuous emotion control in text-to-image generation requires a model to improve affective alignment without changing the objects, layout, or scene described by the prompt. Existing supervised emotion-injection methods often optimize feature-space proxies and may therefore exhibit emotion-semantic drift, in which stronger emotional conditioning is accompanied by unintended content changes. We address this problem with a flow-matching image-generation framework that combines continuous valence-arousal (VA) conditioning, Group Relative Policy Optimization (GRPO), and a neutral semantic anchor. The deterministic probability-flow ODE is converted into a marginal-preserving SDE, yielding non-degenerate transition densities for trajectory sampling and policy-ratio estimation. A frozen CLIP-based VA regressor supplies a terminal reward measuring the distance between the predicted and target VA coordinates, while an image generated from the same prompt under zero VA conditioning provides a feature-space reference for semantic preservation. A reduced denoising schedule is used for online RL sampling, whereas the original schedule is retained at inference. Experiments on 3,300 prompt-emotion combinations show substantially lower valence and arousal errors than the VA-conditioned baseline and an improved CLIPScore relative to EmotiCrafter, with a measurable trade-off in reference-free image quality. The results support anchor-regularized Flow-GRPO as a practical approach to balancing emotional alignment and semantic consistency in continuous-affect image synthesis.

Chinese Translation

在文本到图像生成中实现连续情感控制需要模型在不改变提示所描述的对象、布局或场景的情况下改善情感一致性。现有的监督情感注入方法通常优化特征空间代理，因此可能会出现情感-语义漂移，即更强的情感调节伴随着意外的内容变化。我们通过一种流匹配图像生成框架解决了这一问题，该框架结合了连续的情感-唤醒（VA）调节、群体相对策略优化（GRPO）和中性语义锚。确定性概率流常微分方程（ODE）被转换为边际保持的随机微分方程（SDE），从而为轨迹采样和策略比率估计产生非退化的转移密度。一个基于CLIP的冻结VA回归器提供了一个终端奖励，测量预测的VA坐标与目标VA坐标之间的距离，而在零VA调节下从相同提示生成的图像则提供了语义保持的特征空间参考。在线强化学习采样使用了减少的去噪调度，而在推理时保留了原始调度。在3,300个提示-情感组合上的实验显示，与VA调节的基线相比，情感-唤醒误差显著降低，并且相对于EmotiCrafter，CLIPScore有了改善，同时在无参考图像质量上有可测量的权衡。结果支持锚定正则化的Flow-GRPO作为在连续情感图像合成中平衡情感一致性和语义一致性的实用方法。

View on arXiv Download PDF AI Translation

cs.CV / 53 / 2609.12834

Self-supervised Pre-training Helps Retinal Disease Progression Modelling Most When Data Is Scarce

自监督预训练在数据稀缺时对视网膜疾病进展建模的帮助最大

Nwabufo, Ifeoma Veronica, Gervelmeyer, Julius, Müller, Sarah, Berens, Philipp

Abstract

Modelling how a disease progresses over time requires longitudinal imaging cohorts, which are scarce and small, whereas cross-sectional data -- one image per participant -- is abundant. Self-supervised pre-training on such data offers a way to bridge this gap, but it is unclear which strategy best supports progression modelling, or how that answer depends on the amount of labelled longitudinal data. We study this for age-related macular degeneration (AMD), pre-training encoders on the large cross-sectional NAKO cohort and predicting time to late AMD on the longitudinal AREDS dataset. We compare in-house self-supervised encoders against a general-purpose (DINOv2) and a domain-specific (RETFound) foundation model, across contrastive, masked-autoencoding, and self-distillation objectives, under frozen and fine-tuned protocols, and across labelled training sets from 100 to 32,250 examples. Which model performs best depends on how the encoder is used. When the encoder is frozen and labels are few -- the regime typical of longitudinal cohorts -- pre-trained representations reach clinically reasonable discrimination from a few hundred labelled samples, while models trained from scratch do not; this advantage fades under fine-tuning. Transfer is governed by the self-supervision objective rather than corpus scale or domain match, so that an encoder pre-trained on a modest cross-sectional cohort matches or exceeds a far larger in-domain foundation model. Together, these results offer a practical recipe for building progression models where longitudinal data is scarce: a frozen self-supervised encoder with a lightweight survival head.

Chinese Translation

建模疾病随时间进展的方式需要纵向成像队列，但这些队列稀缺且样本量小，而横断面数据——每位参与者一张图像——则非常丰富。在此类数据上进行自监督预训练提供了一种弥补这一差距的方法，但尚不清楚哪种策略最能支持进展建模，或者这一答案如何依赖于标记的纵向数据量。我们以年龄相关性黄斑变性（AMD）为研究对象，在大型横断面NAKO队列上预训练编码器，并在纵向AREDS数据集上预测晚期AMD的发生时间。我们比较了内部自监督编码器与通用（DINOv2）和特定领域（RETFound）基础模型，在对比、掩蔽自编码和自蒸馏目标下，采用冻结和微调协议，并在从100到32,250个样本的标记训练集上进行比较。哪种模型表现最佳取决于编码器的使用方式。当编码器被冻结且标签较少时——这是纵向队列的典型情况——预训练表示在几百个标记样本中达到了临床合理的区分能力，而从头训练的模型则没有；这一优势在微调下减弱。迁移受自监督目标的支配，而非语料库规模或领域匹配，因此在适度的横断面队列上预训练的编码器可以匹配或超过一个更大的领域内基础模型。综合来看，这些结果为在纵向数据稀缺的情况下构建进展模型提供了一个实用的方案：一个冻结的自监督编码器与一个轻量级的生存头。

View on arXiv Download PDF AI Translation

cs.CV / 54 / 2609.12835

HemaHier: Chain-Conditioned Ordinal Hierarchies for Lineage-Aware Bone-Marrow Cytology

HemaHier：面向谱系感知骨髓细胞学的链条件序数层次结构

Bozorgpour, Afshin, Schüffler, Peter, Jost, Edgar, Merhof, Dorit

Abstract

Bone-marrow cytology is inherently structured: each cell belongs to a hematopoietic lineage, and many cell types lie on ordered maturation trajectories. Standard flat classifiers ignore this structure, treating a mild same-lineage confusion the same as a severe cross-lineage mistake and predicting only discrete labels. We propose HemaHier, an ordinal-hierarchical prediction head for a frozen or lightly adapted cytology foundation model. Its central component is a chain-conditioned maturity score that reads a single maturity value under a per-chain query, supervised only on biologically valid healthy chains, while dysplastic and off-chain cell types remain classes but are excluded from maturity supervision. Fine and lineage predictions are coupled through a shared posterior that guarantees hierarchical consistency, and a staged objective first stabilizes recognition, then adds lineage and maturity supervision. On three bone-marrow datasets under a shared ontology, HemaHier achieves competitive recognition while reducing biologically severe errors and adding a within-lineage maturity ordering that flat classifiers lack. Code is available at https://github.com/xmindflow/HemaHier.

Chinese Translation

骨髓细胞学本质上是有结构的：每个细胞属于一个造血谱系，许多细胞类型沿着有序的成熟轨迹发展。标准的平面分类器忽视了这种结构，将轻微的同谱系混淆与严重的跨谱系错误视为相同，并仅预测离散标签。我们提出了HemaHier，一种用于冻结或轻度调整的细胞学基础模型的序数层次预测头。其核心组件是一个链条件成熟度评分，在每个链查询下读取单一成熟度值，仅在生物学上有效的健康链上进行监督，而发育异常和离链细胞类型仍然作为类别存在，但不参与成熟度监督。细胞类型和谱系预测通过共享后验相结合，确保层次一致性，而分阶段目标首先稳定识别，然后增加谱系和成熟度监督。在三个共享本体的骨髓数据集上，HemaHier实现了具有竞争力的识别，同时减少了生物学上严重的错误，并增加了平面分类器所缺乏的同谱系成熟度排序。代码可在 https://github.com/xmindflow/HemaHier 获取。

View on arXiv Download PDF AI Translation

cs.CV / 55 / 2609.12843

Pre-Trained Low-Rank Tensor Decomposition for Multi-Dimensional Image Recovery

预训练低秩张量分解用于多维图像恢复

Fu, Bing-Zhang, Han, Zhi-Long, Huang, Ting-Zhu, Zhao, Xi-Le, Meng, Deyu

Abstract

Recently, tensor decompositions are prevalent for multi-dimensional image representation, which learn the instance-specific structure of each image from scratch. However, tensor decompositions neglect the common structure across different images, leading to limited semantic modeling capability, high computational cost, and a large number of learnable parameters. To address this challenge, we suggest the first pre-trained low-rank tensor decomposition (PLTD) framework, which organically integrates the pre-trained large vision model into the classical tensor decomposition framework. Beyond the shallow and untrained deep tensor decomposition, the suggested PLTD achieves an unprecedented balance among higher recovery fidelity, fewer learnable parameters, and smaller carbon footprint. Specifically, PLTD factorizes the target tensor into a latent tensor and a learnable transform that maps the latent tensor back to the original data domain. The latent tensor consists of two indispensable and complementary terms, i.e., a fixed pre-trained latent tensor and a learnable low-rank latent tensor. The fixed pre-trained latent tensor is distilled from a pre-trained large vision model (i.e., DINOv3) to capture the common structure of the target tensor, while the learnable low-rank latent tensor characterizes the instance-specific structure of the target tensor. To examine the potential of PLTD, we develop the corresponding multi-dimensional image recovery model and theoretically justify the advantages of this framework. Additionally, we discuss the connections between PLTD and classical tensor decomposition frameworks. Extensive experiments on multi-dimensional image recovery demonstrate that PLTD consistently achieves superior performance compared with state-of-the-art methods.

Chinese Translation

近年来，张量分解在多维图像表示中广泛应用，它从零开始学习每幅图像的实例特定结构。然而，张量分解忽视了不同图像之间的共性结构，导致语义建模能力有限、计算成本高以及可学习参数数量庞大。为了解决这一挑战，我们提出了首个预训练低秩张量分解（PLTD）框架，该框架将预训练的大型视觉模型有机地融入经典的张量分解框架中。与浅层和未训练的深度张量分解相比，所建议的PLTD在更高的恢复保真度、更少的可学习参数和更小的碳足迹之间实现了前所未有的平衡。具体而言，PLTD将目标张量分解为一个潜在张量和一个可学习的变换，该变换将潜在张量映射回原始数据域。潜在张量由两个不可或缺且互补的项组成，即一个固定的预训练潜在张量和一个可学习的低秩潜在张量。固定的预训练潜在张量从预训练的大型视觉模型（即DINOv3）中提取，以捕捉目标张量的共性结构，而可学习的低秩潜在张量则表征目标张量的实例特定结构。为了检验PLTD的潜力，我们开发了相应的多维图像恢复模型，并理论上证明了该框架的优势。此外，我们讨论了PLTD与经典张量分解框架之间的联系。在多维图像恢复的广泛实验中，PLTD始终表现出优于最先进方法的性能。

View on arXiv Download PDF AI Translation

cs.CV / 56 / 2609.12850

MGAvatar: Mesh-Bound Gaussians for Head Avatar Geometry and Appearance Modeling

MGAvatar：用于头部头像几何和外观建模的网格绑定高斯模型

Shi, Lei, Peng, Sen, Deng, Zhiyang, Chen, Zhonggui, Guo, Xiaohu, Yang, Baorong, Dong, Xiao

Abstract

Accurate head modeling requires a stable yet expressive geometric representation. Existing Gaussian-based head avatars commonly rely on parametric templates (e.g., FLAME) for Gaussian initialization and deformation, but these templates lack personalized priors and struggle to represent structures such as hair and clothing. To address this issue, we propose MGAvatar, a Gaussian-mesh hybrid representation that jointly models geometry and appearance through two Gaussian-mesh binding modes. Specifically, we introduce vertex-bound Gaussians and constrain their learnable parameters, enabling progressive mesh deformation to represent complex head geometry, while a pose-dependent offset module accounts for non-rigid deformations. Once geometry is stabilized, MGAvatar switches to face-bound Gaussians for appearance modeling. To improve appearance consistency across novel poses and viewpoints, we introduce a view-conditioned neural color field that alleviates artifacts caused by independently optimized Gaussian colors. In addition, we design a Gaussian offset network to predict Gaussian offset maps in the observation space, providing greater flexibility for face-bound Gaussians to capture dynamic facial textures. Extensive experiments on multi-view and monocular videos show that MGAvatar outperforms existing methods in rendering quality, producing high-fidelity head avatars with rich texture details.

Chinese Translation

准确的头部建模需要一种稳定而富有表现力的几何表示。现有的基于高斯的头部头像通常依赖于参数化模板（例如，FLAME）进行高斯初始化和变形，但这些模板缺乏个性化的先验知识，并且难以表示诸如头发和衣物等结构。为了解决这个问题，我们提出了MGAvatar，一种高斯-网格混合表示，通过两种高斯-网格绑定模式联合建模几何和外观。具体而言，我们引入了顶点绑定高斯，并限制其可学习参数，使得逐步网格变形能够表示复杂的头部几何，同时依赖于姿态的偏移模块考虑了非刚性变形。一旦几何稳定，MGAvatar就会切换到面部绑定高斯进行外观建模。为了提高在新姿势和视角下的外观一致性，我们引入了一种视图条件神经颜色场，以缓解因独立优化高斯颜色而造成的伪影。此外，我们设计了一个高斯偏移网络，以预测观察空间中的高斯偏移图，为面部绑定高斯捕捉动态面部纹理提供更大的灵活性。在多视角和单目视频上的广泛实验表明，MGAvatar在渲染质量上优于现有方法，生成高保真度的头部头像，具有丰富的纹理细节。

View on arXiv Download PDF AI Translation

cs.CV / 57 / 2609.12860

3D CT-to-PET Translation via Latent Brownian Bridge Diffusion

基于潜在布朗桥扩散的3D CT到PET转换

Mourya, Sarita, Di Feola, Francesco, Veltri, Pierangelo, Soda, Paolo

Abstract

Computed tomography (CT) and positron emission tomography (PET) provide complementary anatomical and functional information for cancer diagnosis and treatment planning. However, the widespread use of PET is limited by high radiation exposure, elevated costs, and restricted availability. To address these limitations, deep learning-based CT-to-PET translation has emerged as a promising approach for synthesizing PET-like information directly from CT images, although accurately modeling the large cross-modal gap remains challenging. In this work, we propose a 3D CT-to-PET translation framework based on latent Brownian Bridge Diffusion (BBDM). The method consists of two stages. First, a Variational Autoencoder (VAE) is trained on paired CT-PET patches, integrating contrastive learning to improve latent alignment between anatomical and metabolic representations. Second, a BBDM is trained in the latent space to translate CT latent representations into their corresponding PET counterparts. The translated PET latents are then decoded and stitched to reconstruct the final 3D PET volume. We evaluate the proposed approach on two publicly available datasets. Quantitative results based on image fidelity and lesion-level PET-specific metrics demonstrate improved performance compared with competing methods. In particular, the proposed approach improves PET signal fidelity, better preserves clinically relevant uptake patterns, and shows improved performance in preserving small-lesion metabolic activation, paving the way for virtual imaging applications.

Chinese Translation

计算机断层扫描（CT）和正电子发射断层扫描（PET）为癌症诊断和治疗规划提供了互补的解剖和功能信息。然而，PET的广泛使用受到高辐射暴露、成本高昂和可用性受限的限制。为了解决这些限制，基于深度学习的CT到PET转换作为一种有前景的方法应运而生，可以直接从CT图像合成类似PET的信息，尽管准确建模大跨模态差距仍然具有挑战性。在本研究中，我们提出了一种基于潜在布朗桥扩散（BBDM）的3D CT到PET转换框架。该方法分为两个阶段。首先，训练一个变分自编码器（VAE），使用配对的CT-PET图块，结合对比学习以改善解剖和代谢表示之间的潜在对齐。其次，在潜在空间中训练BBDM，将CT潜在表示转换为相应的PET对应物。然后对转换后的PET潜在表示进行解码和拼接，以重建最终的3D PET体积。我们在两个公开可用的数据集上评估了所提出的方法。基于图像保真度和病灶级PET特定指标的定量结果表明，与竞争方法相比，性能有所提升。特别是，所提出的方法提高了PET信号的保真度，更好地保留了临床相关的摄取模式，并在保留小病灶代谢活性方面表现出改善，为虚拟成像应用铺平了道路。

View on arXiv Download PDF AI Translation

cs.CV / 58 / 2609.12874

VideoTok4D: A 4D-Aware Video Tokenizer for Compact World Representation

VideoTok4D：一种用于紧凑世界表示的4D感知视频标记器

Chen, Xinyi, Zhu, Hanxin, Wang, Xijun, Wang, Xingrui, Liang, Sen, Li, Xin, Chen, Zhibo

Abstract

Video tokenizers have emerged as a cornerstone of modern video modeling, underpinning progress in compression, reconstruction and generation by mapping high-dimensional visual signals into compact latent spaces. However, despite this progress, current tokenization paradigms largely remain within the 2D visual domain, treating videos as image sequences rather than observations of an underlying dynamic 3D world. Consequently, the learned tokens inherit this observation-centric bias, limiting their capacity to compactly represent real-world 4D scenes. To mitigate this issue, we propose VideoTok4D, a novel 4D-aware video tokenizer for compact world representation. Specifically, our approach comprises three key designs: 1) a spatiotemporal disentanglement strategy that factorizes videos into static and dynamic tokens for holistic world modeling; 2) a track-aware dynamic attention mechanism that aggregates trajectory-aligned cues to promote cross-view motion consistency; and 3) Co4DGen, a diffusion prior learned over the resulting VideoTok4D token space for efficient 4D scene generation. Extensive experiments have demonstrated that our proposed method achieves state-of-the-art performance while requiring up to 4 orders of magnitude less storage than dense 4D representations. Moreover, the compact token space substantially shortens diffusion sequences, enabling efficient generation.

Chinese Translation

视频标记器已成为现代视频建模的基石，通过将高维视觉信号映射到紧凑的潜在空间，推动了压缩、重建和生成的进展。然而，尽管取得了这些进展，目前的标记化范式仍主要局限于2D视觉领域，将视频视为图像序列，而非潜在动态3D世界的观察。因此，学习到的标记继承了这种以观察为中心的偏见，限制了它们紧凑表示真实世界4D场景的能力。为了解决这个问题，我们提出了VideoTok4D，一种新颖的4D感知视频标记器，用于紧凑的世界表示。具体而言，我们的方法包括三个关键设计：1）一种时空解耦策略，将视频分解为静态和动态标记，以实现整体世界建模；2）一种轨迹感知动态注意机制，聚合轨迹对齐的线索，以促进跨视图运动一致性；3）Co4DGen，一种在生成的VideoTok4D标记空间上学习的扩散先验，用于高效的4D场景生成。大量实验表明，我们提出的方法在性能上达到了最先进水平，同时所需存储量比密集的4D表示减少了多达4个数量级。此外，紧凑的标记空间显著缩短了扩散序列，从而实现高效生成。

View on arXiv Download PDF AI Translation

cs.CV / 59 / 2609.12885

Learning Sign Language Recognition under Label Noise: A Study of Noise-Robust Losses for Isolated and Continuous Settings

在标签噪声下学习手语识别：孤立和连续设置下噪声鲁棒损失的研究

Shitara, Akihisa, Ochiai, Yoichi

Abstract

In sign language recognition, the isolated (ISLR) classification loss treats a single label as ground truth, as does the frame-level auxiliary classifier over pseudo-labels we add to continuous (CSLR) methods, which lack one. Stylistic variation blurs ISLR annotation and the lack of temporal boundaries in CSLR forces pseudo-labels; both are noisy. We therefore apply symmetric and generalized cross entropy (SCE, GCE), robust alternatives to cross entropy (CE) from image classification, not to connectionist temporal classification but to the preceding single-label classifier. On ASL Citizen with injected symmetric noise on three backbones (three seeds for ST-GCN), robust losses cost at most 2.5 pt when labels are clean and beat CE by 2.9-10.0 pt in all six conditions at noise rate 0.2, one of which only after q was re-selected on dev. GCE gains more, but its optimal q does not transfer across backbones, whereas one SCE setting works in all nine conditions; both vary 2-11 times more than CE across runs, so a favorable point estimate does not establish stability. For CSLR (PHOENIX-2014) we report no gain; our frame-level targets carry a systematic assignment bias, making that study a diagnosis of a single configuration. At lambda_aux = 25 the pseudo-label CE auxiliary raises word error rate above the no-auxiliary baseline on VAC, CorrNet and SlowFastSign, and GCE/SCE improve on CE by 1.7-3.2 pt (three of six conditions return below that baseline). However, the three losses differ by more than an order of magnitude in effective gradient at a common lambda_aux: matching the initial gradient shrinks the gap to 0.4-0.9 pt, and lowering the CE weight alone already beats that baseline, so neither the degradation nor the improvement can be separated from the effect of the weight. We use only symmetric noise; multi-seed evaluation covers only ST-GCN and VAC isolated.

Chinese Translation

在手语识别中，孤立分类损失（ISLR）将单个标签视为真实标签，连续方法（CSLR）中我们添加的基于伪标签的帧级辅助分类器也是如此，而后者缺乏这一点。风格变异模糊了ISLR注释，而CSLR中缺乏时间边界则迫使使用伪标签；这两者都存在噪声。因此，我们应用对称和广义交叉熵（SCE, GCE），作为图像分类中交叉熵（CE）的鲁棒替代方案，而不是用于连接主义时间分类，而是用于前面的单标签分类器。在注入了对称噪声的ASL Citizen上，使用三种主干网络（ST-GCN的三个种子），当标签干净时，鲁棒损失最多损失2.5个点，在噪声率为0.2的六种条件下，鲁棒损失比CE提高了2.9-10.0个点，其中一种情况只有在开发集上重新选择q后才实现。GCE的增益更大，但其最佳q在不同主干网络之间无法转移，而一个SCE设置在所有九种条件下均有效；两者在不同运行中变化的幅度是CE的2-11倍，因此一个有利的点估计并不能建立稳定性。对于CSLR（PHOENIX-2014），我们报告没有增益；我们的帧级目标存在系统性分配偏差，使该研究成为单一配置的诊断。在lambda_aux = 25时，伪标签CE辅助使VAC、CorrNet和SlowFastSign的字错误率高于无辅助基线，而GCE/SCE在CE上提高了1.7-3.2个点（六种条件中有三种返回低于该基线）。然而，这三种损失在相同的lambda_aux下有效梯度的差异超过一个数量级：匹配初始梯度将差距缩小至0.4-0.9个点，仅降低CE权重就已经超过了该基线，因此降级和改进的效果无法与权重的影响分开。我们仅使用对称噪声；多种子评估仅覆盖ST-GCN和VAC孤立。

View on arXiv Download PDF AI Translation

cs.CV / 60 / 2609.12892

Beyond Accuracy: Uncertainty-Guided Boundary Refinement for Reliable Biomedical Image Segmentation

超越准确性：基于不确定性的边界细化方法用于可靠的生物医学图像分割

Kujur, Anima

Abstract

Accurate biomedical image segmentation requires not only high global overlap but also reliable delineation of clinically meaningful boundaries. In blood-smear microscopy, cytoplasm and nucleus contours provide the structural basis for downstream morphology analysis; however, deep segmentation models may remain uncertain or overconfident near ambiguous boundary regions even when achieving strong Dice scores. This work proposes a Reliability-Aware Boundary Refinement Network (RABR-Net), a two-stage framework for trustworthy image segmentation. A strong UNet++ EfficientNet-B4 base segmenter first produces initial class probabilities and logits. Predictive entropy, test-time augmentation variance, margin uncertainty, probability gradients, and soft boundary cues are then combined into a boundary-aware reliability representation. This representation guides a gated residual refiner that selectively corrects uncertain boundary pixels while preserving confident regions of the base prediction. The framework is evaluated using overlap accuracy, class-wise Dice, Boundary Dice, HD95/ASSD, calibration, risk--coverage analysis, robustness under image perturbations, qualitative correction maps, and paired statistical testing. On the held-out test set, the proposed method improves Dice from 0.9602 to 0.9614, Boundary Dice from 0.3448 to 0.3611, and HD95 from 3.0354 to 2.8274 compared with the cached base prediction. Statistical analysis confirms significant improvements in Dice, Boundary Dice, and HD95. Qualitative results show that the learned gate concentrates around uncertain cytoplasm and nucleus boundaries, and correction maps confirm localized boundary refinement. Although calibration does not automatically improve after refinement, the proposed framework provides an interpretable and reliability-focused strategy for boundary-sensitive biomedical image segmentation.

Chinese Translation

准确的生物医学图像分割不仅需要高全局重叠度，还需要可靠地勾勒出临床意义的边界。在血涂片显微镜下，细胞质和细胞核的轮廓为后续的形态分析提供了结构基础；然而，即使在获得强大的 Dice 分数时，深度分割模型在模糊边界区域附近仍可能存在不确定性或过度自信。本研究提出了一种可靠性意识边界细化网络（Reliability-Aware Boundary Refinement Network, RABR-Net），这是一个用于可信图像分割的两阶段框架。强大的 UNet++ EfficientNet-B4 基础分割器首先生成初始类别概率和 logits。然后，将预测熵、测试时增强方差、边际不确定性、概率梯度和软边界线索结合成一种边界感知的可靠性表示。该表示指导一个门控残差细化器，选择性地修正不确定的边界像素，同时保留基础预测的自信区域。该框架通过重叠准确性、类别 Dice、边界 Dice、HD95/ASSD、校准、风险-覆盖分析、图像扰动下的鲁棒性、定性修正图和配对统计测试进行评估。在保留的测试集上，所提方法将 Dice 从 0.9602 提高到 0.9614，将边界 Dice 从 0.3448 提高到 0.3611，将 HD95 从 3.0354 降低到 2.8274，相较于缓存的基础预测。统计分析确认了 Dice、边界 Dice 和 HD95 的显著改善。定性结果表明，学习到的门控集中在不确定的细胞质和细胞核边界附近，修正图确认了局部边界细化。尽管细化后校准并未自动改善，但所提框架为边界敏感的生物医学图像分割提供了一种可解释且以可靠性为中心的策略。

View on arXiv Download PDF AI Translation

cs.CV / 61 / 2609.12898

UniPart: Towards Zero-shot Language-Grounded 3D Part Segmentation for Embodied Interaction

UniPart：迈向零-shot语言基础的3D部件分割以实现具身交互

Yu, Xinqiang, qi, Zekun, He, Jiawei, Zhang, Wenyao, Chen, Xuchuan, Yao, Guaocai, Yi, Li, Zhang, Zhaoxiang, Wang, He

Abstract

Fine-grained robotic manipulation depends on understanding parts, not only whole objects. Existing 3D foundation models tend to be either generalized but object-aware, or part-aware but limited to closed-set taxonomies, which weakens zero-shot transfer. We study text-conditioned 3D part segmentation, where a free-form phrase selects a functional part on point cloud. We introduce UniPart, a feed-forward cross-modal 3D Transformer that conditions CLIP text embedding. To scale supervision, we build LangPart-1M with 160K+ Objaverse assets and 8M text to part pairs using multi-view consistent part generation. We further manually label a high-quality subset, LangPart-4K, for fine-tuning and evaluation. UniPart achieves strong zero-shot results on open-vocabulary part benchmarks and transfers to language-conditioned part grasping in real world.

Chinese Translation

细粒度的机器人操作依赖于对部件的理解，而不仅仅是整体物体。现有的3D基础模型往往要么是通用但具对象意识，要么是具部件意识但仅限于封闭集分类，这削弱了零-shot迁移。我们研究文本条件的3D部件分割，其中自由形式的短语选择点云上的功能部件。我们提出了UniPart，一种前馈跨模态3D Transformer，能够对CLIP文本嵌入进行条件处理。为了扩大监督范围，我们构建了LangPart-1M，包含超过16万的Objaverse资产和800万对文本与部件的配对，采用多视图一致的部件生成方法。我们进一步手动标注了一个高质量子集LangPart-4K，用于微调和评估。UniPart在开放词汇的部件基准测试中取得了强大的零-shot结果，并成功迁移到现实世界中的语言条件部件抓取任务。

View on arXiv Download PDF AI Translation

cs.CV / 62 / 2609.12902

Parallel Training Using a CNN-DNN Architecture for Accelerated Development of Diagnostic Models

基于CNN-DNN架构的并行训练以加速诊断模型的开发

Weber-Hamacher, Janine, Jaiswal, Astha, Fervers, Philipp, Móré, Dorotya, Giannakis, Athanasios, Fischbach, Ricarda, Bucher, Andreas Michael, Shahzad, Rahil, Kottlors, Jonathan, Persigehl, Thorsten, Klawonn, Axel

Abstract

Artificial intelligence has shown promise in assisting radiologists in imaging-based diagnosis across a wide range of diseases. Efficient training of large deep learning models is essential to cope with extremely large data sets or dynamically growing disease data, like in a pandemic like situation. In this retrospective study, we collected 300 CT scans from COVID-19 and non-COVID-19 pneumonia patients from three different centers in Germany. We investigated a hybrid CNN-DNN network model based on image decomposition and localization that naturally supports parallel and efficient training of deep learning models. In total, 156 models with three different architectures were trained to capture features at different levels resulting in 12 patient-level COVID-19 diagnosis models. Diagnostic performance as well as time saving were measured. The highest accuracy was obtained from DenseNet121 and 3D CNN models with a parallel CNN-DNN approach, resulting in $88.78\%$ training, $76.67\%$ validation and $76.03\%$ test accuracy for the DenseNet121 with $4\times4\times1$ subdomains and $87.72\%$ training, $76.82\%$ validation and $74.86\%$ test accuracy, respectively, for the 3D CNN with $4\times4\times1$ subdomains. The strongest reduction in parallel training time by a factor of $31$ was observed for the 3D CNN model and $4\times4\times2$ subdomains. Our parallel training approach improves efficiency as well as performance enabling rapid model development, among others crucial for pandemic preparedness.

Chinese Translation

人工智能在协助放射科医生进行基于影像的多种疾病诊断方面展现出了良好的前景。高效训练大型深度学习模型对于应对极大数据集或动态增长的疾病数据（如疫情情况下）至关重要。在这项回顾性研究中，我们从德国三个不同中心收集了300例COVID-19及非COVID-19肺炎患者的CT扫描图像。我们研究了一种基于图像分解和定位的混合CNN-DNN网络模型，该模型自然支持深度学习模型的并行和高效训练。总共训练了156个具有三种不同架构的模型，以捕捉不同层次的特征，最终得到了12个患者级别的COVID-19诊断模型。我们测量了诊断性能和时间节省。DenseNet121和3D CNN模型采用并行CNN-DNN方法获得了最高的准确率，其中DenseNet121在$4 imes4 imes1$子域下的训练准确率为$88.78\%$，验证准确率为$76.67\\%$，测试准确率为$76.03\\%$；3D CNN在$4 imes4 imes1$子域下的训练准确率为$87.72\\%$，验证准确率为$76.82\\%$，测试准确率为$74.86\\%$。在$4 imes4 imes2$子域下，3D CNN模型的并行训练时间减少了$31$倍。我们的并行训练方法提高了效率和性能，促进了快速模型开发，这对于疫情准备等方面至关重要。

View on arXiv Download PDF AI Translation

cs.CV / 63 / 2609.12917

PA-CDM: Position-Aware Character Detection Matching for Evaluating Handwritten Mathematical Expression Recognition

PA-CDM：用于评估手写数学表达式识别的基于位置的字符检测匹配

Luo, Shiliang

Abstract

Handwritten mathematical expression recognition (HMER) is conventionally scored by exact-match rates and string-similarity metrics that are blind to where an error occurs: two predictions with identical token-error counts receive identical scores whether they misplace a subscript or swap the operands of a fraction. Render-based character detection matching (CDM) aligns glyphs robustly but remains position-blind---on controlled fraction-operand swaps it scores 0.8595 where position-aware scoring yields 0.6253. Tree-edit metrics exhibit a complementary blind spot: rewrites outside the parser's normalization coverage are penalized as structural errors (0.8552 where render-based metrics score 1.0). We propose PA-CDM, a position-aware metric that couples character detection matching with position-forest encoding and divergence-level weighting; StructPerturb v2.0, a frozen benchmark of 1,340 controlled perturbation pairs across 15 type--intensity cells; and a cross-metric consistency protocol combining a sensitivity matrix, a human study, and LLM-judge calibration. In a six-annotator study, PA-CDM attains the highest correlation with human judgments among seven automatic metrics (Spearman rho=0.9535, n=990). A frontier LLM judge correlates slightly higher (rho=0.9613) but is costly, nondeterministic, and API-dependent; PA-CDM approaches it at zero marginal cost with deterministic, diagnosable behavior.

Chinese Translation

手写数学表达式识别（HMER）通常通过精确匹配率和字符串相似性度量进行评分，这些度量对错误发生的位置视而不见：两个预测即使在下标错误或分数操作数交换的情况下具有相同的标记错误计数，仍会获得相同的分数。基于渲染的字符检测匹配（CDM）能够稳健地对齐字形，但仍然对位置盲目——在控制的分数操作数交换中，其得分为0.8595，而基于位置的评分则为0.6253。树编辑度量显示出一个互补的盲点：在解析器的标准化覆盖范围之外的重写被惩罚为结构错误（得分为0.8552，而基于渲染的度量得分为1.0）。我们提出了PA-CDM，一种基于位置的度量，将字符检测匹配与位置森林编码和发散级别加权相结合；StructPerturb v2.0，一个包含15个类型-强度单元的1,340对控制扰动的冻结基准；以及一个结合灵敏度矩阵、人类研究和LLM评判校准的跨度量一致性协议。在一项六位注释者的研究中，PA-CDM在七个自动度量中与人类判断的相关性最高（Spearman rho=0.9535，n=990）。一个前沿的LLM评判者的相关性略高（rho=0.9613），但成本高、非确定性且依赖于API；而PA-CDM以零边际成本接近该评判者，具有确定性和可诊断的行为。

View on arXiv Download PDF AI Translation

cs.CV / 64 / 2609.12920

Input Resolution Matters: Real-Time Object Detection Latency

输入分辨率的重要性：实时目标检测延迟

Zhang, Qingyang, Machida, Fumio, Carnevali, Laura

Abstract

We model total latency as the convolution of preprocessing, inference, and postprocessing distributions under a simplifying independence approximation, with selected stage parameters expressed as functions of source-image resolution. Under this assumption, the probability density of the total latency is the convolution of the stage-wise densities, and its cumulative distribution function (CDF) provides the distribution of end-to-end detection time. Each stage is modeled by a parametric distribution (e.g., Exponential, Erlang, Normal, Gamma), with parameters expressed as functions of the source-image resolution. Experiments with YOLOv11n on NVIDIA Jetson Orin NX using COCO2017 images across multiple resolutions assess the proposed models against fixed-parameter baselines using Kolmogorov Smirnov, Anderson Darling, and Cram\'er von Mises statistics. The results indicate that resolution-aware parameterization can improve distributional approximation in the measured setting, particularly for the more flexible Normal and Gamma models, while the quality of fit remains distribution dependent. Our contribution is a theoretically grounded and lightweight formulation for studying resolution-dependent latency distributions in a measured object detection pipeline.

Chinese Translation

我们将总延迟建模为预处理、推理和后处理分布的卷积，基于简化的独立性近似，并将选定阶段的参数表示为源图像分辨率的函数。在此假设下，总延迟的概率密度为各阶段密度的卷积，其累积分布函数（CDF）提供了端到端检测时间的分布。每个阶段都通过参数分布（例如，指数分布、埃尔朗分布、正态分布、伽马分布）建模，参数表示为源图像分辨率的函数。我们在NVIDIA Jetson Orin NX上使用COCO2017图像进行YOLOv11n的实验，评估所提模型与固定参数基线的表现，采用Kolmogorov Smirnov、Anderson Darling和Cramér von Mises统计量。结果表明，考虑分辨率的参数化可以改善测量环境中的分布近似，特别是对于更灵活的正态分布和伽马分布模型，而拟合质量仍然依赖于分布。我们的贡献是一个理论基础扎实且轻量化的公式，用于研究测量的目标检测流程中的分辨率依赖延迟分布。

View on arXiv Download PDF AI Translation

cs.CV / 65 / 2609.12929

DementiaCare-Bench: A Modality-Validated Video Benchmark

痴呆护理基准：一种经过模态验证的视频基准

Sheikholeslami, Afrouz, Qi, Yuankai, Zhang, Xuyun, Zhou, Luping, Beheshti, Amin, Sheng, Quan Z., Yang, Ming-Hsuan

Abstract

Dementia affects an estimated 57 million people worldwide, and for most families the hardest part of care is not memory loss but the behavioral and psychological symptoms of dementia (BPSD): agitation, wandering, resistance to care, sundowning. Understanding these symptoms requires more than recognizing the behavior itself; it also requires knowing what happened beforehand. The same behavior may call for a different response depending on its trigger. Video-language models (VLMs) could potentially support caregivers, yet no existing benchmark evaluates this capability. To fill this gap, we present DementiaCare-Bench: 56 professionally produced caregiver training videos segmented into 94 clips across nine BPSD categories, with 2023 questions generated by a multi-agent pipeline that grounds every clinical claim in a verbatim transcript span. Each question is then probed under four visual conditions and labelled by the least it requires, so its visual demand is measured rather than assumed. Measurement contradicts intent: we wrote 77.7% of the questions to require ordered frames, and 34.8% do. Across 12 current VLMs the pattern is uniform. The best reach 85% overall, but that average is carried by questions a language model can answer from clinical knowledge alone; accuracy falls by 17 points on average on questions that require the ordered clip, and a leading open model scores at chance on judging whether a caregiver's response was appropriate. A lightweight LoRA fine-tune, DemCare-VLM, moves video dependence from -3.3 to +4.5 points, so what the benchmark exposes can be repaired and not only measured.

Chinese Translation

痴呆症在全球影响约5700万人，对于大多数家庭来说，护理中最困难的部分不是记忆丧失，而是痴呆症的行为和心理症状（BPSD）：烦躁、徘徊、抵抗护理、日落现象。理解这些症状不仅需要识别行为本身，还需要了解其背后的原因。同样的行为可能因触发因素的不同而需要不同的反应。视频语言模型（VLMs）有可能支持护理人员，但现有的基准并未评估这一能力。为填补这一空白，我们提出了痴呆护理基准（DementiaCare-Bench）：56个专业制作的护理人员培训视频，分为94个片段，涵盖九个BPSD类别，并通过多代理管道生成2023个问题，每个临床主张都基于逐字转录的范围。每个问题在四种视觉条件下进行探测，并根据所需的最低条件进行标记，以便测量其视觉需求而非假设。测量与意图相悖：我们编写了77.7%的问题需要有序帧，而实际上只有34.8%符合这一要求。在12个当前的VLM中，这一模式是一致的。最佳模型的整体准确率达到85%，但这一平均值是由语言模型仅凭临床知识能够回答的问题所支撑；在需要有序片段的问题上，准确率平均下降17个百分点，而一个领先的开放模型在判断护理人员的反应是否适当时的得分接近随机水平。一个轻量级的LoRA微调模型DemCare-VLM将视频依赖度从-3.3提升至+4.5分，因此基准所揭示的问题可以被修复，而不仅仅是被测量。

View on arXiv Download PDF AI Translation

cs.CV / 66 / 2609.12953

Fast and Faithful: Principled Conditional Flow Matching for Inverse Problems

快速而可靠：用于逆问题的原则性条件流匹配

Shoushtari, Shirin, Chandler, Edward P., Shi, Xiao, Kamilov, Ulugbek S.

Abstract

Flow matching approaches to imaging inverse problems commonly incorporate measurements in two ways. Conditioning-based approaches supply measurement-derived information as a network input, often through concatenation, while inference-guided approaches combine an unconditional velocity field with a separate data-consistency update. In these common formulations, the forward model is not explicitly enforced within the learned conditional velocity field. We propose a principled parametrization of the measurement-conditional velocity field to solve inverse problems. Under linear interpolation, we express the conditional velocity $v(x_t,t,y)$ in terms of the posterior mean $E[x_1 | x_t,y]$, and characterize that mean as the unique minimizer of a variational objective whose data-consistency term is explicit. We further prove that the velocity field defines a probability flow from the source distribution to the measurement-conditioned posterior. Splitting the variational objective yields a conditional velocity parameterization with operator-dependent data-consistency updates, which we train end-to-end under the flow-matching objective, with no additional guidance at inference. Our method achieves state-of-the-art PSNR with $50\times$ fewer function evaluations than the strongest flow baseline. Varying the sampling steps provides test-time control over the distortion-perception trade-off without retraining.

Chinese Translation

流匹配方法在成像逆问题中通常以两种方式结合测量数据。基于条件的方法将测量导出的信息作为网络输入，通常通过拼接的方式，而推理引导的方法则将无条件速度场与单独的数据一致性更新相结合。在这些常见的公式中，前向模型并未在学习的条件速度场中明确强制执行。我们提出了一种原则性的测量条件速度场参数化方法来解决逆问题。在线性插值下，我们将条件速度 $v(x_t,t,y)$ 表达为后验均值 $E[x_1 | x_t,y]$ 的函数，并将该均值表征为一个变分目标的唯一最小化解，其数据一致性项是显式的。我们进一步证明，速度场定义了从源分布到测量条件后验的概率流。将变分目标分解后，得到了一种具有操作符依赖的数据一致性更新的条件速度参数化，我们在流匹配目标下进行端到端训练，在推理时无需额外指导。我们的方法在函数评估次数比最强流基线少 $50 imes$ 的情况下实现了最先进的 PSNR。改变采样步数可以在测试时控制失真感知的权衡，而无需重新训练。

View on arXiv Download PDF AI Translation

cs.CV / 67 / 2609.12965

Generative Retrieval for Unsupervised Text-Based Person Search

用于无监督文本基础人像搜索的生成检索

Ye, Mang, Ji, Yucheng, Bai, Yang, Cao, Min, Chai, Siyuan, Du, Bo, Zhang, Min

Abstract

Text-based person search (TBPS) aims to retrieve images of a target person from a large image gallery based on a given natural language description. Most existing methods rely on supervised learning with manually annotated image-text pairs. In this paper, we explore unsupervised TBPS, with only unlabeled images. We propose GTR+, a two-stage generation-then-retrieval framework. In the generation stage, we introduce a tiered description generation framework designed to produce fine-grained and stylistically diverse textual descriptions through a three-tier sequential process. The base tier leverages an automated question-and-answer mechanism to generate basic visual attribute descriptions; the intermediate tier enhances fine-grained detail using an inter-sample contrastive mechanism; the advanced tier further enriches textual diversity via a stylized expansion mechanism. In the retrieval stage, to mitigate the impact of noisy pseudo texts, we develop an adaptive confidence-weighted retrieval learning framework. We model image-text pairs as clean or noisy using a Gaussian Mixture Model, calibrated by real-time image-text similarity and static text generation probability from the prior stage, yielding adaptive sample weights during training. Beyond that, we also contribute LargeFine-Person, a large-scale TBPS dataset with high-quality, fine-grained, and diverse textual annotations, enabling a practical and generalizable TBPS pre-training benchmark under unsupervised setting. Experiments on multiple TBPS benchmarks demonstrate the effectiveness and generalization of both GTR+ and LargeFine-Person. Code is available at: https://github.com/Flame-Chasers/GTR.

Chinese Translation

基于文本的人像搜索（TBPS）旨在根据给定的自然语言描述从大型图像库中检索目标人物的图像。现有的大多数方法依赖于带有手动标注的图像-文本对的监督学习。本文探讨了仅使用未标记图像的无监督TBPS。我们提出了GTR+，一种两阶段的生成-检索框架。在生成阶段，我们引入了一种分层描述生成框架，旨在通过三层顺序过程生成细粒度和风格多样的文本描述。基础层利用自动问答机制生成基本的视觉属性描述；中间层通过样本间对比机制增强细粒度细节；高级层则通过风格化扩展机制进一步丰富文本多样性。在检索阶段，为了减轻噪声伪文本的影响，我们开发了自适应置信加权检索学习框架。我们使用高斯混合模型将图像-文本对建模为干净或噪声，依据来自前一阶段的实时图像-文本相似性和静态文本生成概率进行校准，从而在训练过程中产生自适应样本权重。除此之外，我们还贡献了LargeFine-Person，一个具有高质量、细粒度和多样化文本注释的大规模TBPS数据集，使得在无监督设置下能够实现实用且可推广的TBPS预训练基准。在多个TBPS基准上的实验表明，GTR+和LargeFine-Person的有效性和泛化能力。代码可在以下链接获取：https://github.com/Flame-Chasers/GTR。

View on arXiv Download PDF AI Translation

cs.CV / 68 / 2609.12993

Investigating Temporal Motion Features for Pose-to-Text Indian Sign Language Translation

研究姿态到文本的印度手语翻译中的时间运动特征

Dhamecha, Manav, Chandaliya, Praveen Kumar, Mishra, Pruthwik

Abstract

We investigate the effect of pretrained T5 model scale and explicit motion features on pose-to-text Indian Sign Language Translation (SLT) for the WSLP 2026 Shared Task. Pose sequences are projected into the embedding space of T5 through a lightweight pose encoder, with the complete model fine-tuned to generate English text. The shared task data used for this work consists of a test set with 5,334 examples and a validation set with 5,257 examples. We compare T5-small, T5-base, and T5-large, and additionally introduce a motion-augmented variant, T5-small + Motion, that adds explicit frame-to-frame pose differences to the input representation. T5-small achieves the best BLEU and ROUGE scores among the spatial-only models, while T5-large obtains the highest chrF score. Augmenting T5-small with motion features yields the largest single improvement observed in our study, substantially improving BLEU over the spatial-only baseline and making it the strongest model overall on this metric. Our submitted system ranked 5th on the official WSLP 2026 SLT testing leaderboard. The source code and trained models are publicly available on GitHub and HuggingFace.

Chinese Translation

我们研究了预训练的 T5 模型规模和显式运动特征对姿态到文本的印度手语翻译（SLT）在 WSLP 2026 共享任务中的影响。通过轻量级姿态编码器将姿态序列投影到 T5 的嵌入空间中，完整模型经过微调以生成英文文本。本研究使用的共享任务数据包括一个包含 5,334 个示例的测试集和一个包含 5,257 个示例的验证集。我们比较了 T5-small、T5-base 和 T5-large，并额外引入了一种运动增强变体 T5-small + Motion，该变体将显式的帧间姿态差异添加到输入表示中。在仅使用空间特征的模型中，T5-small 在 BLEU 和 ROUGE 分数上表现最佳，而 T5-large 获得了最高的 chrF 分数。将运动特征与 T5-small 结合使用在我们的研究中观察到的单一最大改进，显著提高了 BLEU 分数，相较于仅使用空间特征的基线，使其成为该指标上整体最强的模型。我们提交的系统在官方 WSLP 2026 SLT 测试排行榜中排名第五。源代码和训练模型已在 GitHub 和 HuggingFace 上公开。

View on arXiv Download PDF AI Translation

cs.CV / 69 / 2609.12997

SV-Cine: Diagnosis-Conditioned Segmentation of Single Ventricle Physiology via Generative Data Augmentation

SV-Cine：基于诊断条件的单心室生理分割通过生成数据增强

Cunge, Lila, Liu, Yuehong, Xu, Hang, Coudert, Thomas, Renella, Pierangelo, Finn, J Paul, Hsu, William, Nguyen, Kim-Lien

Abstract

Single Ventricle Physiology (SVP) is a rare subtype of congenital heart disease characterized by the presence of a single functional cardiac ventricle with atypical anatomic configurations that challenge conventional image segmentation approaches. The scarcity of clinical data and the morphological diversity across SVP subtypes make the development of robust segmentation methods particularly difficult. To address these limitations, we propose a cardiac MRI segmentation framework focused on ventricular chambers and myocardium segmentation tailored for SVP. First, we introduce a data augmentation pipeline that generates synthetic 3D cardiac meshes using SDF4CHD and corresponding synthetic cardiac MRI through generative modeling. Second, we introduce SV-Cine, a diagnosis-conditioned adaptation of the foundation model CineMA that incorporates patient-level diagnostic information through Feature-wise Linear Modulation layers, enabling diagnosis-aware feature adaptation during segmentation. We evaluated the framework on an internal cohort with varying SVP subtypes. SV-Cine achieved median Dice scores of 0.89 (IQR: 0.80--0.91) for the left ventricle and 0.72 (IQR: 0.54--0.84) for the right ventricle, outperforming the strongest baseline, nnU-Net, by 0.39 Dice points on right ventricle segmentation. It also yields a median ejection fraction error of 5.55 percentage points (IQR: 3.41--7.69) for the dominant ventricle. Compared with the internal cohort, LV and myocardium segmentation performance was lower for the external cohort; whereas RV Dice scores were comparable for both cohorts. Our findings suggest that a pretrained foundation model can be adapted for highly specialized downstream tasks through usage of diagnosis priors while leveraging anatomic knowledge learned from large-scale MRI datasets during pretraining.

Chinese Translation

单心室生理（SVP）是一种罕见的先天性心脏病亚型，其特征是存在一个功能性心室，并具有非典型的解剖结构，这对传统的图像分割方法构成了挑战。临床数据的稀缺性以及SVP亚型之间的形态多样性使得开发稳健的分割方法尤其困难。为了解决这些限制，我们提出了一种心脏MRI分割框架，专注于针对SVP的心室腔和心肌的分割。首先，我们引入了一种数据增强管道，该管道使用SDF4CHD生成合成的3D心脏网格，并通过生成建模生成相应的合成心脏MRI。其次，我们引入了SV-Cine，这是一种基于诊断条件的基础模型CineMA的适应，利用特征线性调制层（Feature-wise Linear Modulation）整合患者级别的诊断信息，使得在分割过程中能够进行基于诊断的特征适应。我们在一个具有不同SVP亚型的内部队列上评估了该框架。SV-Cine在左心室的中位Dice分数为0.89（四分位数范围：0.80--0.91），右心室为0.72（四分位数范围：0.54--0.84），在右心室分割上超越了最强基线nnU-Net，提升了0.39 Dice点。它还为主导心室提供了中位射血分数误差为5.55个百分点（四分位数范围：3.41--7.69）。与内部队列相比，外部队列的左心室和心肌分割性能较低，而右心室的Dice分数在两个队列中相当。我们的研究结果表明，预训练的基础模型可以通过利用诊断先验并结合从大规模MRI数据集中学习到的解剖知识，在高度专业化的下游任务中进行适应。

View on arXiv Download PDF AI Translation

cs.CV / 70 / 2609.13006

Physics-Aware Video Generation via Agentic Planning and Graph-Guided Optimization

基于物理知识的视频生成：代理规划与图引导优化

Nguyen, Minh-Loi, Le, Xuan-Vu, Do, Thanh-Toan, Nguyen, Tam V., Tran, Minh-Triet, Le, Trung-Nghia

Abstract

Video diffusion models (VDMs) have demonstrated remarkable capabilities in synthesizing high-fidelity, photorealistic video content. However, they fundamentally lack an intrinsic understanding of physical laws and frequently produce visually appealing but causally illogical sequences characterized by structural hallucinations and physically implausible dynamics. Injecting physical awareness via training-free test-time optimization is a promising alternative, yet existing methods rely on global gradient updates and rigid scheduling heuristics that inadvertently corrupt passive backgrounds and fail to model complex dynamic state changes. To address this, we propose PhysPlan, a novel training-free guidance framework that shifts the paradigm from stochastic visual interpolation to agentic physics simulation. First, a VLM operates as an iterative cognitive simulator, decomposing multimodal inputs into a Chain-of-Visual-Thought to create a multimodal representation of kinematic trajectories and 3D depth geometries. Second, these signals drives an object-centric test-time optimization. Unlike prior training-free methods that rely on global gradients and rigid scheduling heuristics, PhysPlan introduces Object-Centric Gradient Routing to isolate kinematic modifications and completely lock the passive environment. Furthermore, our Kinetic Intensity Profiling dynamically parameterizes framework hyperparameters to accommodate the varying severity of physical deformations. Extensive evaluations on the PhyGenBench and Physics-IQ benchmarks demonstrate that PhysPlan significantly outperforms both foundational and controllable VDM baselines, offering a promising approach for improving the physical understanding of video generation.

Chinese Translation

视频扩散模型（VDMs）在合成高保真、逼真的视频内容方面展现了显著的能力。然而，它们在本质上缺乏对物理法则的内在理解，常常生成视觉上吸引人但因果上不合逻辑的序列，这些序列表现出结构性幻觉和物理上不合理的动态。通过无训练的测试时优化注入物理意识是一种有前景的替代方案，但现有方法依赖于全局梯度更新和僵化的调度启发式，这无意中破坏了被动背景，并未能有效建模复杂的动态状态变化。为了解决这一问题，我们提出了PhysPlan，一种新颖的无训练指导框架，旨在将范式从随机视觉插值转变为代理物理模拟。首先，VLM作为一个迭代认知模拟器，将多模态输入分解为视觉思维链，以创建运动轨迹和三维深度几何的多模态表示。其次，这些信号驱动以对象为中心的测试时优化。与以往依赖全局梯度和僵化调度启发式的无训练方法不同，PhysPlan引入了对象中心梯度路由，以隔离运动学修改并完全锁定被动环境。此外，我们的动能强度分析动态参数化框架超参数，以适应物理变形的不同严重程度。在PhyGenBench和Physics-IQ基准上的广泛评估表明，PhysPlan显著优于基础和可控VDM基线，为改善视频生成的物理理解提供了一种有前景的方法。

View on arXiv Download PDF AI Translation

cs.CV / 71 / 2609.13012

Pixel Decodability Is Not a Compression Signal: Causally Evaluating Importance Proxies for Visual KV-Cache Eviction

像素可解码性不是压缩信号：因果评估视觉KV缓存驱逐的重要性代理

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

Abstract

Vision-language models retain a substantial amount of pixel-decodable visual content in their visual key-value cache. We show, in our setting, that this retention is task-inert: across our preregistered tests, how much a unit retains never positively tracks whether the computation that answers the question causally relies on it. We measure retention with a learned pixel-inversion decoder and causal use with single-super-patch KV ablation, the teacher-forced drop in gold-answer log-probability, and relate the two within images under a preregistered, sign-calibrated, held-out design. Retention is decoupled from attention and, in a well-powered null, from causal utilization. Utilization is not inert to every proxy: attention weakly but significantly tracks it, the only signal we find that does and the design's positive control. We characterize pixel-decodable retention as an informational axis of the visual KV cache, orthogonal to the functional one. How much task-inert content a cache holds differs by architecture in our model pair: the encoder-free model retains 2.7 times more than the encoder-based one. The engineering consequence is a controlled negative result. At super-patch granularity, deconfounded pixel-decodable retention ranks KV eviction no better than random; at token granularity it acquires only a weak inverse-importance signal at larger budgets, dominated at every budget by attention magnitude. In our setting, pixel-decodable reconstructability is not a competitive KV-compression signal at any granularity we test.

Chinese Translation

视觉语言模型在其视觉键值缓存中保留了大量像素可解码的视觉内容。我们在我们的研究中表明，这种保留是任务惰性的：在我们预注册的测试中，一个单元保留的内容量与回答问题的计算是否因果依赖于它之间没有正相关。我们通过学习的像素反演解码器来测量保留，并通过单超补丁KV消融、教师强制的黄金答案对数概率下降来测量因果使用，并在预注册的、符号校准的、保留设计中将两者关联。保留与注意力解耦，并且在一个有力的无效结果中，与因果利用无关。利用并非对每个代理都是惰性的：注意力弱但显著地与之相关，这是我们发现的唯一信号，也是设计的正控制。我们将像素可解码的保留特征化为视觉KV缓存的信息轴，与功能轴正交。在我们的模型对中，缓存中持有的任务惰性内容量因架构而异：无编码器模型的保留量是基于编码器模型的2.7倍。工程上的结果是一个受控的负结果。在超补丁粒度下，去混淆的像素可解码保留在KV驱逐中的排名不比随机更好；在标记粒度下，它在更大的预算下仅获得一个弱的逆重要性信号，在每个预算中都被注意力幅度主导。在我们的研究中，像素可解码的重构能力在我们测试的任何粒度上都不是一个竞争性的KV压缩信号。

View on arXiv Download PDF AI Translation

cs.CV / 72 / 2609.13013

TileNet: Tile-Based CNN-SVM Architecture for Autonomous Unmanned Aerial Systems Inspection of Flat Roofs

TileNet：基于瓦片的卷积神经网络-支持向量机架构用于自主无人机系统对平屋顶的检查

Dunthorne, Samuel, Hashim, Hashim A.

Abstract

Flat roofs are among the most influential components of the building envelope, governing both structural performance and thermal efficiency, and thereby contributing directly to household energy consumption, carbon emissions, and long-term environmental sustainability. Timely detection of roof defects is essential for reducing heating and cooling losses, preventing moisture-driven degradation such as mold growth, and supporting national climate-change mitigation goals. This paper presents a real-time, Unmanned Aerial System (UAS)-based deep learning framework that autonomously detects defects using live imagery captured during dual-altitude aerial passes. The multi-resolution flight strategy is designed to aid the identification of both small, fine-scale defects and larger structural issues, enabling more comprehensive assessments. To meet the strict computational and power constraints of embedded UAS hardware, the proposed framework integrates a tile-based architecture with a lightweight Convolution Neural Network-Support Vector Machine (CNN-SVM) classifier designed for low-latency onboard inference. The final model-comprising five convolutional layers and four dense layers, the last a linear SVM head, achieved a mean test accuracy of $94.4\%$ ($95\%$ confidence interval $\pm0.4\%$ over three seeds) on a photo-level split ($43,383$ training, $3,869$ validation, and $2,540$ test tiled and augmented images), outperforming GoogLeNet ($89.2\%$) and AlexNet ($79.8\%$). Experimental evaluations using real UAS imagery collected by onsite visits with DJI Matrice 350 RTK drone demonstrate that the system supports rapid, repeatable, and safe roof inspections while reducing human risk, lowering operational costs, and enabling more sustainable building maintenance.

Chinese Translation

平屋顶是建筑围护结构中最重要的组成部分之一，影响着结构性能和热效率，从而直接影响家庭能源消耗、碳排放和长期环境可持续性。及时检测屋顶缺陷对于减少供暖和制冷损失、预防由湿气引起的劣化（如霉菌生长）以及支持国家气候变化减缓目标至关重要。本文提出了一种基于无人机系统（UAS）的实时深度学习框架，能够利用在双高度航拍过程中捕获的实时图像自主检测缺陷。多分辨率飞行策略旨在帮助识别小型细微缺陷和较大结构问题，从而实现更全面的评估。为了满足嵌入式UAS硬件的严格计算和功率限制，所提出的框架集成了一种基于瓦片的架构，并采用了为低延迟机载推理设计的轻量级卷积神经网络-支持向量机（CNN-SVM）分类器。最终模型由五个卷积层和四个全连接层组成，最后一个为线性SVM头，在照片级别分割上（43,383个训练图像，3,869个验证图像和2,540个测试图像，均为瓦片化和增强处理）实现了94.4%的平均测试准确率（95%置信区间±0.4%，基于三次随机种子），超越了GoogLeNet（89.2%）和AlexNet（79.8%）。使用DJI Matrice 350 RTK无人机在现场访问中收集的真实UAS图像进行的实验评估表明，该系统支持快速、可重复和安全的屋顶检查，同时降低了人类风险、降低了运营成本，并促进了更可持续的建筑维护。

View on arXiv Download PDF AI Translation

cs.CV / 73 / 2609.13024

Label-Guided Knowledge Distillation for 3D-CNNs in Action Recognition

基于标签引导的知识蒸馏在动作识别中的3D-CNN应用

Shi, Yanjiang, Zhao, Peng, Qi, Nan, Wang, Guiqin

Abstract

As a key model compression technique, knowledge distillation aims to transfer knowledge from a high-capacity teacher model to a lightweight student model for enhancing the latter's performance. In this work, we reviewed the feature knowledge distillation for 3D-CNNs and observed that most feature distillation methods in video analysis are simple adaptations of those used in image analysis, often neglecting the differences of video features in the temporal dimension. To address this issue, we proposed Label-Guided Knowledge Distillation (LGKD) to guide the distillation of student model features using ground truth labels. Our method entails two components: sample-wise distillation and class-wise distillation, enabling the student model to learn feature representation of the teacher model at two levels. Sample-wise distillation utilizes label information and the teacher's probability distribution to guide the learning of features that significantly impact temporal accuracy while mitigating noise. Meanwhile, class-wise feature distillation employs a prototype network to further capture the relational knowledge among samples within the same category, enhancing the student's ability to learn higher-dimensional semantic information and improving model generalization. To demonstrate the effectiveness and superiority of our method, we conducted comprehensive experiments on two benchmark action recognition datasets, UCF101 and HMDB51, achieving competitive results.

Chinese Translation

作为一种关键的模型压缩技术，知识蒸馏旨在将知识从高容量的教师模型转移到轻量级的学生模型，以提升后者的性能。在本研究中，我们回顾了3D-CNN的特征知识蒸馏，并观察到大多数视频分析中的特征蒸馏方法只是对图像分析中方法的简单改编，往往忽视了视频特征在时间维度上的差异。为了解决这一问题，我们提出了基于标签引导的知识蒸馏（Label-Guided Knowledge Distillation, LGKD），通过真实标签引导学生模型特征的蒸馏。我们的方法包含两个组成部分：样本级蒸馏和类别级蒸馏，使学生模型能够在两个层面上学习教师模型的特征表示。样本级蒸馏利用标签信息和教师的概率分布来指导学习对时间准确性影响显著的特征，同时减小噪声。与此同时，类别级特征蒸馏采用原型网络进一步捕捉同一类别样本之间的关系知识，增强学生学习高维语义信息的能力，提高模型的泛化能力。为了验证我们方法的有效性和优越性，我们在两个基准动作识别数据集UCF101和HMDB51上进行了全面的实验，取得了具有竞争力的结果。

View on arXiv Download PDF AI Translation

cs.CV / 74 / 2609.13043

Unified CT and MRI Pancreas Segmentation for Label-Efficient Cross-Modality Subregion Transfer

统一的CT和MRI胰腺分割用于标签高效的跨模态子区域转移

Hong, Ziliang, Pan, Hongyi, Aktas, Halil Ertugrul, Bejar, Andrea, Keles, Elif, Miller, Frank H., Wallace, Michael B., Keswani, Rajesh N., Durak, Gorkem, Bagci, Ulas

Abstract

Robust medical image segmentation across imaging modalities is challenging because of large differences in appearance and intensity distributions. Models trained on a single modality often show substantial performance drops when applied to unseen domains. In this work, we develop a unified 3D pancreas segmentation framework that applies domain-adversarial learning to 4,604 heterogeneous CT and MRI scans to learn anatomical representations. A shared nnU-Net encoder-decoder is trained for whole-pancreas segmentation, with a latent domain discriminator encouraging CT-MRI feature alignment. The learned encoder is subsequently transferred to pancreatic head-body-tail segmentation using limited MRI-only subregion annotations. An average Dice score of 87.31% on the in-distribution test set and Dice scores ranging from 84.20% to 88.09% across external OOD datasets were achieved in whole pancreas segmentation. Dice scores of 80.53% on MRI and 83.05% on CT were achieved for downstream subregion segmentation, without using CT subregion annotations. These results demonstrate that a unified anatomical representation can support both cross-modality pancreas segmentation and label-efficient downstream transfer.

Chinese Translation

在不同成像模态下进行稳健的医学图像分割具有挑战性，因为外观和强度分布存在较大差异。基于单一模态训练的模型在应用于未见领域时，通常会出现显著的性能下降。在本研究中，我们开发了一个统一的3D胰腺分割框架，该框架应用领域对抗学习于4604个异构的CT和MRI扫描，以学习解剖表示。一个共享的nnU-Net编码器-解码器被训练用于整体胰腺分割，同时潜在领域鉴别器促进CT-MRI特征对齐。学习到的编码器随后被转移到胰腺头-体-尾的分割，使用有限的仅MRI子区域注释。在整体胰腺分割中，在分布内测试集上获得了87.31%的平均Dice分数，而在外部OOD数据集上的Dice分数范围为84.20%到88.09%。在下游子区域分割中，MRI的Dice分数为80.53%，CT的Dice分数为83.05%，且未使用CT子区域注释。这些结果表明，统一的解剖表示可以支持跨模态胰腺分割和标签高效的下游转移。

View on arXiv Download PDF AI Translation

人工智能 (Artificial Intelligence)

cs.AI / 1 / 2609.11977

Occamy-1.0: Open Pareto-frontier 35B Intelligence for Co-work

Occamy-1.0：开放的帕累托前沿35B智能协作模型

Chen, Wenhui, Cheng, Shiwen, Dong, Hao, Duan, Chenda, Feng, Ruixiang, Guan, Zhong, Guo, Boqiang, Han, Xueyuan, Hao, Haojie, Huang, Liangmeng, Huang, Zhelong, Kong, Xinke, Li, Hongyu, Li, Jiazheng, Li, Junbo, Li, Qingchuan, Lian, Yukun, Liu, Chang, Liu, Tianyu, Liu, Zicheng, Ouyang, Shuyi, Pan, Yijun, Shi, Kunyu, Tang, Xiaojun, Wang, Bingquan, Wang, Kesu, Wang, Yuchen, Wei, Sibo, Xie, Sicong, Xing, Xiaoying, Xu, Yi, Xu, Zhijun, Xue, Hongwei, Zeng, Qingcheng, Zhang, Di, Zhang, Guannan, Zhang, Haochen, Zhang, Tianlong, Zhao, Tianyu, Zhao, Tianyu, Zheng, Yanjun, Zhu, Jialong, Zou, Zijian

Abstract

Co-work agents execute complex workflows that combine information gathering, tool use, coding, and file manipulation across many model invocations. Because cost and latency accumulate over the full episode, their practical value depends not only on peak capability but also on how efficiently that capability is delivered. Yet many steps in everyday work emphasize state tracking, coordination, recovery, and follow-through rather than frontier-scale reasoning. We present Occamy-1.0, a cost-efficient co-work model obtained by further training the post-trained Qwen3.6-35B-A3B checkpoint. We construct execution-grounded data and environments, capture replayable long-horizon trajectories across multiple harnesses, and use staged post-training to develop and consolidate complementary execution capabilities. Across a broad suite of co-work benchmarks, Occamy-1.0 is consistently among the strongest comparably sized models and remains competitive with substantially larger frontier systems on several tasks. Under our stated evaluation and pricing protocol, its aggregate performance across four representative benchmarks places it at the low-cost knee of the observed cost--performance Pareto frontier. Supporting evaluations in tool calling, coding, and instruction following further show that this specialization preserves broad agentic capability. We release the model weights and a subset of the training data to support research on practical co-work agents and agentic post-training.

Chinese Translation

协作代理执行复杂的工作流程，这些流程结合了信息收集、工具使用、编码和文件操作，涉及多个模型调用。由于成本和延迟在整个过程中累积，它们的实际价值不仅取决于峰值能力，还取决于这种能力的高效交付。然而，日常工作中的许多步骤强调状态跟踪、协调、恢复和后续行动，而不是前沿规模的推理。我们提出了Occamy-1.0，这是一个通过进一步训练后训练的Qwen3.6-35B-A3B检查点获得的成本效益高的协作模型。我们构建了基于执行的数据和环境，捕捉多个环境中的可重放长时间轨迹，并使用分阶段后训练来开发和巩固互补的执行能力。在一系列广泛的协作基准测试中，Occamy-1.0始终位于同类模型中最强的行列，并在多个任务上与显著更大的前沿系统保持竞争力。在我们所述的评估和定价协议下，它在四个代表性基准测试中的综合表现使其处于观察到的成本-性能帕累托前沿的低成本拐点。支持在工具调用、编码和指令跟随方面的评估进一步表明，这种专业化保持了广泛的代理能力。我们发布了模型权重和部分训练数据，以支持对实用协作代理和代理后训练的研究。

View on arXiv Download PDF AI Translation

cs.AI / 2 / 2609.11987

Harness or Model? Isolating the Harness Effect in Agentic Coding with a Contamination-Controlled Private Suite

利用还是建模？在受污染控制的私人套件中隔离代理编码的利用效应

Arjmandi, Mohsen

Abstract

An agentic coding system couples a language model to a harness: the tools, prompts and control flow that turn a chat model into an autonomous software engineer. Vendors ship harnesses tuned to their own models, and practitioners assume the vendor-native pairing solves more tasks. We measure that assumption with paired same-model contrasts on a private, contamination-controlled suite of 256 repository and post-cutoff contest tasks. The same 80 tasks ran under claude-agent-sdk and under deepagents on claude-opus-4-8, and under the openai-codex SDK and deepagents on gpt-5.5, with gemini-3.5-flash and deepseek-v3.2 as side cells. 792 of 800 planned runs were graded by an isolated oracle. Neither contrast resolves an average advantage for either harness: -1.25 pp for Opus 4.8 (48.8% vs 50.0%, task-bootstrap 95% CI [-10.0, +7.5]) and +1.25 pp for GPT-5.5 (55.6% vs 54.4%, CI [-4.4, +6.9]). The Opus average combines opposite strata: the native harness trails by 9.0 pp on the 61 repository tasks and leads by 23.7 pp on the 19 contest tasks (label-permutation p = 0.003). The partition was chosen after seeing the data and needs a designed replication. Correctness and completion also separate: 22 of 81 runs cancelled at the wall-clock ceiling had produced a passing patch. Re-priced from raw per-turn usage at frozen list prices, the neutral harness cost 1.3 to 1.6 times as much per solved task on Opus 4.8 and 1.2 times on GPT-5.5. These are observed-usage estimates. On the Anthropic account 58 runs left no usage record, and allocating that spend to either cell would move the Opus ratio between 0.7 and 2.3, so the billed ordering is unresolved. This revision corrects an August 2026 manuscript whose cost figures rested on a usage-semantics defect in our own telemetry (Section 5.1). We release the orchestrator, grading oracle, reanalysis code and derived aggregates. The tasks stay private.

Chinese Translation

代理编码系统将语言模型与一个工具包相结合：这些工具、提示和控制流程将聊天模型转变为自主软件工程师。供应商提供调优至其自身模型的工具包，实践者假设供应商原生配对能够解决更多任务。我们通过在一个包含256个代码库和后截止竞赛任务的私人、受污染控制的套件上进行同模型对比来测量这一假设。相同的80个任务在claude-agent-sdk和claude-opus-4-8下运行，以及在openai-codex SDK和gpt-5.5下运行，gemini-3.5-flash和deepseek-v3.2作为辅助单元。800个计划运行中有792个由一个独立的预言者评分。两个对比都没有显示出任何工具包的平均优势：Opus 4.8的结果为-1.25个百分点（48.8%对50.0%，任务引导95%置信区间[-10.0, +7.5]），而GPT-5.5的结果为+1.25个百分点（55.6%对54.4%，置信区间[-4.4, +6.9]）。Opus的平均值结合了相反的层次：在61个代码库任务中，原生工具包落后9.0个百分点，而在19个竞赛任务中领先23.7个百分点（标签置换p = 0.003）。该分区是在查看数据后选择的，需要设计的复制实验。正确性和完成度也有所不同：在墙钟时间上限取消的81次运行中，有22次产生了通过的补丁。根据冻结列表价格的每轮原始使用情况重新定价，中立工具包在Opus 4.8上的每个解决任务成本为1.3到1.6倍，而在GPT-5.5上为1.2倍。这些是观察到的使用估计。在Anthropic账户中，58次运行没有留下使用记录，将该支出分配给任一单元将使Opus比例在0.7到2.3之间波动，因此账单顺序尚未解决。本修订版纠正了2026年8月的手稿，其成本数据基于我们自身遥测中的使用语义缺陷（第5.1节）。我们发布了调度器、评分预言者、重新分析代码和派生汇总。任务仍然保持私密。

View on arXiv Download PDF AI Translation

cs.AI / 3 / 2609.12035

Reading the Whole Heart: Latent-Attention Masked Autoencoders for Multimodal Cardiac Representation Learning

全面理解心脏：用于多模态心脏表征学习的潜在注意力掩蔽自编码器

Agostini, Andrea, Böhi, Simon, Vandenhirtz, Moritz, Ruiperez-Campillo, Samuel, Krähenmann, Max, Mühlstedt, Silke, Cannistraci, Irene, Elsen, Ece Özkan, Vogt, Julia E., Sutter, Thomas M.

Abstract

Cardiovascular diagnosis rests on integrating complementary modalities, like ECG, echocardiography, chest radiographs, and clinical variables, each capturing distinct but correlated aspects of cardiac physiology. Yet most medical foundation models remain modality-specific, combining modalities only for finetuning or post-training. This discards the cross-modal evidence clinicians naturally integrate and ignores the structure within each modality. We introduce Latent-Attention Masked Autoencoders (LAMAE), a multimodal, structure-aware masked autoencoder that jointly learns patient-level representations during self-supervised pretraining. Rather than fusing modalities post hoc, LAMAE exchanges information directly in the latent space through a shared latent-attention module operating over a study-view-entity hierarchy, enabling aggregation of variable observations and graceful handling of missing modalities. Pretrained on over 1.2 million MIMIC-IV hospital stays, LAMAE outperforms modality-specific pretraining and strong contrastive and vision-language baselines across multimodal hospital-stay tasks, such as in-hospital mortality, ICD-10 and DRG coding, and length of stay, while remaining competitive on unimodal tasks. These gains persist even when only a single modality is available at test time, showing that modeling both intra- and inter-modal structure yields more robust, transferable representations.

Chinese Translation

心血管诊断依赖于整合互补的模态，如心电图（ECG）、超声心动图、胸部X光片和临床变量，每种模态捕捉心脏生理的不同但相关的方面。然而，大多数医学基础模型仍然是特定于模态的，仅在微调或训练后结合模态。这忽视了临床医生自然整合的跨模态证据，并忽略了每种模态内部的结构。我们提出了潜在注意力掩蔽自编码器（Latent-Attention Masked Autoencoders，LAMAE），这是一种多模态、结构感知的掩蔽自编码器，在自监督预训练期间共同学习患者级别的表征。与其在后期融合模态，LAMAE通过一个在研究-视图-实体层次结构上操作的共享潜在注意力模块直接在潜在空间中交换信息，从而实现可变观察的聚合，并优雅地处理缺失模态。在超过120万次MIMIC-IV住院记录上进行预训练后，LAMAE在多模态住院任务（如院内死亡率、ICD-10和DRG编码以及住院时长）上超越了特定模态的预训练和强对比及视觉-语言基准，同时在单模态任务上仍具竞争力。这些提升在测试时仅有单一模态可用的情况下依然存在，表明建模内部和跨模态结构能够产生更稳健、可迁移的表征。

View on arXiv Download PDF AI Translation

cs.AI / 4 / 2609.12101

Competence-Gated Pooling of Language Models and Priors for Event Forecasting

基于能力门控的语言模型与先验信息融合用于事件预测

Tiwari, Aditi, Bandaru, Aashrith, Ji, Heng

Abstract

In hybrid forecasting, a language model is often one of several available signals. A system may already have a market, crowd, or statistical forecast and must decide whether the model adds useful information or should be ignored. The relevant target is therefore not standalone model accuracy, but relative competence, defined as the model's marginal value beyond the available external forecast. Under Brier loss, we characterize when model disagreement can improve an external forecast and derive the gain from using domain-specific rather than global pooling weights. We then introduce a competence gate that estimates domain-level source weights from resolved outcomes, shrinks uncertain estimates toward a global weight, and recalibrates the pooled forecast. Across 2,357 resolved binary questions and five language models, the gate improves the main external baseline from 0.0771 to 0.0732 Brier and significantly outperforms global forecast combinations. The gain remains significant under leakage controls and against a leakage-safe time-series prior on the pooled structured set, with separate evidence on FRED. In contrast, the gate gives no significant improvement on the official ForecastBench market subset, where it largely defers to the market. Across four Qwen models, verbal confidence does not reliably identify when the model outperforms the external forecast, while outcome-estimated competence supports better abstention decisions. These results provide a practical approach for selective model use based on measured marginal value.

Chinese Translation

在混合预测中，语言模型通常是多种可用信号之一。系统可能已经拥有市场、群体或统计预测，必须决定该模型是否提供了有用信息或应被忽略。因此，相关的目标不是模型的独立准确性，而是相对能力，即模型在现有外部预测基础上的边际价值。在Brier损失下，我们刻画了模型分歧何时能够提升外部预测，并推导出使用领域特定而非全局融合权重所带来的增益。随后，我们引入了能力门控机制，该机制通过已解决的结果估计领域级别的来源权重，将不确定估计收缩至全局权重，并对融合预测进行重新校准。在2357个已解决的二元问题和五个语言模型上，该门控机制将主要外部基线的Brier分数从0.0771提升至0.0732，且显著优于全局预测组合。该增益在泄漏控制下以及针对融合结构化集合上的泄漏安全时间序列先验中依然显著，并在FRED数据集上获得了独立验证。相比之下，该门控机制在官方ForecastBench市场子集上未带来显著提升，主要依赖于市场预测。在四个Qwen模型中，口头置信度无法可靠识别模型何时优于外部预测，而基于结果估计的能力则支持更优的弃权决策。这些结果为基于测量边际价值的选择性模型使用提供了实用方法。

View on arXiv Download PDF AI Translation

cs.AI / 5 / 2609.12105

Language Is an Insufficient Substrate for Quantitative Reasoning, and Consequential Domains Need Large Quantitative Models

语言不足以作为定量推理的基础，后果性领域需要大型定量模型

Vandeventer, Reuben, Imrem, David, Wild, David J.

Abstract

The prevailing assumption in applied machine learning is that progress on consequential quantitative decisions such as pricing risk, allocating capital, triaging patients, or containing a network intrusion will follow from progress in large language models (LLMs). A language model is trained on a representation of the world that was produced by human description; description is a lossy encoding of the quantitative record, and the loss is irreversible: no downstream model, at any scale, can recover from a description what the description did not encode. We formalize this as a property of the representation on which a model is trained rather than of the model capacity, and we identify three further properties that consequential settings demand of a model and that a language substrate cannot supply by construction: reproducibility, lineage from every output back to the source records that produced. it, and calibrated uncertainty. We argue that these properties define a distinct model class, which we call the Large Quantitative Model (LQM).

Chinese Translation

在应用机器学习中，普遍的假设是，关于定量决策（例如定价风险、资本分配、患者分诊或网络入侵控制）的进展将依赖于大型语言模型（LLMs）的进展。语言模型是基于人类描述所产生的世界表现进行训练的；描述是定量记录的有损编码，而这种损失是不可逆的：任何规模的下游模型都无法从描述中恢复出描述未编码的内容。我们将这一点形式化为模型训练所依赖的表现的属性，而不是模型能力的属性，并识别出后果性环境对模型的三个进一步要求，而语言基础由于其构造无法提供：可重复性、从每个输出追溯到生成它的源记录的来源，以及经过校准的不确定性。我们认为这些属性定义了一个独特的模型类别，我们称之为大型定量模型（LQM）。

View on arXiv Download PDF AI Translation

cs.AI / 6 / 2609.12115

DU-NO: A Parameter-Efficient Double U-Shaped Neural Operator for Phase-Resolving Wave Modeling

DU-NO：一种参数高效的双U形神经算子用于相位分辨波浪建模

Noguera, Enrique Hernandez, Ferdaus, Md Meftahul, Cooper, Nathan, Ioup, Elias, Abdelguerfi, Mahdi

Abstract

Phase-resolving wave models such as FUNWAVE-TVD are the accuracy standard for nearshore dynamics, resolving the shoaling, refraction, and breaking of individual waves, but their cost rules them out for the ensembles, uncertainty quantification, and real-time warning that operational forecasting demands. Neural operators promise solver-level accuracy at a fraction of that cost, yet on wave-dominated fields the accurate ones are large: hybrid spectral-convolutional operators such as U-FNO (the strongest baseline in our study after DU-NO) buy their fidelity with tens of millions of parameters. We introduce DU-NO (Double U-shaped Neural Operator), a multiscale U-shaped spectral operator that attaches lightweight convolutional U-Net branches only at its two shallowest encoder and decoder levels. The placement follows a sampling argument: high-wavenumber content exists only on fine grids, so the local, full-band pathways go where that content lives, while the coarse, band-limited levels stay purely spectral. A depth-decaying mode schedule holds the model to 3.64M parameters, an order of magnitude below U-FNO. On our publicly released FUNWAVE-TVD benchmark, DU-NO attains the best autoregressive rollout error of six identically trained architectures, improving on U-FNO by 14.9% with 10.8x fewer parameters, and a frequency-band analysis shows the gain holds across all bands, including the high-wavenumber band where truncated-spectral operators collapse. Parameter-matched controls confirm the gain is architectural: rescaled to the same 3.6M budget, the best baseline still trails DU-NO by 28.6%. The advantage carries beyond nearshore waves: DU-NO matches the strongest baselines on 2D Navier-Stokes and wins clearly on PDEBench shallow-water rollouts. Code, trained models, and evaluation artifacts are available at https://anonymous.4open.science/r/duno-code-5A7B/.

Chinese Translation

相位分辨波浪模型如FUNWAVE-TVD是近岸动力学的准确性标准，能够解析单个波浪的淹没、折射和破碎，但其成本使其无法满足操作性预报所需的集合、误差量化和实时预警。神经算子承诺以较低的成本实现求解器级别的准确性，但在波主导的场景中，准确的模型通常较大：混合谱卷积算子如U-FNO（在我们的研究中DU-NO之后的最强基线）以数千万个参数换取其精度。我们提出DU-NO（双U形神经算子），这是一种多尺度U形谱算子，仅在其两个最浅的编码器和解码器层附加轻量级卷积U-Net分支。该布局遵循采样论点：高波数内容仅存在于细网格上，因此局部的全频带路径走向内容所在的位置，而粗糙的带限层则保持纯谱特性。深度衰减模式调度将模型参数限制在3.64M，数量级低于U-FNO。在我们公开发布的FUNWAVE-TVD基准测试中，DU-NO在六个相同训练架构中获得了最佳自回归展开误差，相比U-FNO提高了14.9%，且参数减少了10.8倍，频带分析显示该增益在所有频带中均有效，包括截断谱算子崩溃的高波数频带。参数匹配的对照实验确认该增益是架构性的：在相同的3.6M预算下，最佳基线仍落后于DU-NO 28.6%。这一优势超越了近岸波浪：DU-NO在2D Navier-Stokes上与最强基线相匹配，并在PDEBench浅水展开中明显胜出。代码、训练模型和评估文档可在 https://anonymous.4open.science/r/duno-code-5A7B/ 获取。

View on arXiv Download PDF AI Translation

cs.AI / 7 / 2609.12116

When Successful Knowledge Graph Edits Displace Correct Answers: Rank-Level Locality beyond Parameter Support

当成功的知识图谱编辑取代正确答案时：超越参数支持的排名级局部性

Lai, Yi-Cheng, Wang, Jerry, Hsu, Hsin-Ling, Chi, Li-Chu, Teng, Ya-Wen, Huang, Hen-Hsen

Abstract

Editing a knowledge graph embedding (KGE) model to promote a desired answer can displace correct answers from the returned list. Locality tests based only on facts that reuse the edited parameter can miss this ranking effect. We introduce a common rank-displacement audit at three scopes: facts supported by the edited parameter, other correct answers to the target query, and correct answers across queries with the same relation. We also derive dimensional and geometric conditions for an update to improve the target while exactly preserving selected scores. On FB15k-237 with DistMult and ComplEx, direct promotion always moves the target into the top ten, but does so without damage in only 23.0--23.2\% of edits. Strict preservation causes no measured damage, yet succeeds in only 1.3--1.4\%. Support-regularized entity editing gives the highest joint success, 36.3--37.7\%, while rank-truncated preservation reaches 32.8--34.7\% and reduces the mean number of displaced answers from about 14 to 1.2. Experiments across dimensions, scorers, ranking conventions, and a learned editor show that locality depends on both the protected scope and the editing mechanism. KGE editing should therefore report correction success together with the incidence and severity of rank displacement.

Chinese Translation

编辑知识图谱嵌入（KGE）模型以促进期望答案可能会将正确答案从返回列表中排除。仅基于重用编辑参数的事实进行的局部性测试可能会忽略这种排名效应。我们在三个范围内引入了一个共同的排名位移审计：由编辑参数支持的事实、目标查询的其他正确答案，以及在具有相同关系的查询中正确答案的情况。我们还推导了更新的维度和几何条件，以在精确保留选定分数的同时改善目标。在使用 DistMult 和 ComplEx 的 FB15k-237 数据集上，直接促进始终将目标移动到前十名，但仅在 23.0% 到 23.2% 的编辑中没有造成损害。严格保留没有造成任何测量损害，但成功率仅为 1.3% 到 1.4%。支持正则化的实体编辑实现了最高的联合成功率，达到 36.3% 到 37.7%，而排名截断保留达到了 32.8% 到 34.7%，并将位移答案的平均数量从约 14 降低到 1.2。跨维度、评分器、排名惯例和学习编辑器的实验表明，局部性依赖于保护范围和编辑机制。因此，KGE 编辑应报告修正成功率以及排名位移的发生率和严重性。

View on arXiv Download PDF AI Translation

cs.AI / 8 / 2609.12139

Mined from Scientific Literature: Process Schemas for Atomic Layer Deposition and Etching in Materials Science

从科学文献中提取：材料科学中原子层沉积和刻蚀的过程模式

Sadruddin, Sameer, Poupaki, Eleni, Watkins, Alex, Karasulu, Bora, Mackus, Adriaan J. M., Kessels, Erwin, Auer, Sören, D'Souza, Jennifer

Abstract

Atomic layer deposition (ALD) and atomic layer etching (ALE) are reported heterogeneously across experimental and simulation literature in materials science, hindering comparison and machine-actionable reuse. We present four domain-expert-reviewed JSON Schemas for ALD and ALE experimental and simulation processes. Curated with schema-miner and grounded in QUDT using schema-miner pro, the schemas structure materials, process conditions, configurations , and measured or predicted results. We compare their scope, structure, and semantic grounding, and demonstrate their use for schema-guided literature extraction and publication of structured records through ORKG templates.

Chinese Translation

原子层沉积（ALD）和原子层刻蚀（ALE）在材料科学的实验和模拟文献中呈现出异质性，阻碍了比较和机器可操作的重用。我们提出了四个经过领域专家审查的 ALD 和 ALE 实验及模拟过程的 JSON 模式。这些模式通过 schema-miner 进行整理，并基于 QUDT 使用 schema-miner pro 构建，结构化了材料、过程条件、配置以及测量或预测的结果。我们比较了它们的范围、结构和语义基础，并展示了它们在基于模式的文献提取和通过 ORKG 模板发布结构化记录中的应用。

View on arXiv Download PDF AI Translation

cs.AI / 9 / 2609.12162

Can LLMs in Draft-Verify-Revise Pipelines Resolve Deictic Ambiguity?

草拟-验证-修订流程中的大型语言模型能否解决指示性歧义？

Ekekezie, Obinna I.

Abstract

Draft-verify-revise is a common LLM orchestration pattern for scaling inference-time compute. One LLM drafts, a second critiques the draft and provides feedback, and a third uses that feedback to revise the draft into the final output. As context cascades between stages, LLMs at different stages can resolve a context-dependent expression such as "previous" differently. When that happens, the expression undergoes a deictic shift, a change in what it refers to. This phenomenon was studied with a synthetic dataset of 10 base examples, each rendered in three conditions. Holding the shared components constant, the conditions varied whether the draft stage LLM (the assistant) or the verify stage LLM (the grader) resolved the expression correctly, and how much independent reasoning the revise stage LLM (the meta-evaluator) needed to determine which reading was correct. Six models from three providers were tested across 21 reasoning effort configurations using e-values for sequential testing, in a primary experiment and an ablation experiment that removed error classification labels from the grader's feedback. A separate LLM analyzed the meta-evaluator's stated rationale for each wrong verdict. Balanced accuracy (the unweighted mean of sensitivity and specificity) ranged from 0.156, below chance, to near-perfect. GPT-5.2 rose from 0.156 without reasoning to 0.942 at its highest reasoning effort level, while Gemini 3 Pro stayed above 0.94 at every level. Gemini 3 Pro at low reasoning effort outscored GPT-5.2 at xhigh reasoning effort for roughly 5% of the cost per trial. When the meta-evaluator erred, it tended to rely on surface cues rather than operational reasoning. Context engineers implementing draft-verify-revise pipelines should be wary of deictic shifts and make the intended referent explicit at each stage.

Chinese Translation

草拟-验证-修订是一种常见的大型语言模型（LLM）编排模式，用于扩展推理时的计算能力。一个LLM进行草拟，第二个对草稿进行批评并提供反馈，第三个利用这些反馈将草稿修订为最终输出。随着上下文在各个阶段之间的传递，不同阶段的LLM可能会对上下文依赖的表达（如“之前”）做出不同的解释。当这种情况发生时，该表达经历了指示性转变，即其所指内容的变化。该现象通过一个包含10个基本示例的合成数据集进行了研究，每个示例在三种条件下呈现。在保持共享组件不变的情况下，这些条件变化了草拟阶段的LLM（助手）或验证阶段的LLM（评分者）是否正确解析了该表达，以及修订阶段的LLM（元评估者）需要多少独立推理来确定哪个解释是正确的。测试了来自三个提供商的六个模型，在21种推理努力配置下使用e值进行顺序测试，包括一个主要实验和一个消融实验，该实验移除了评分者反馈中的错误分类标签。一个独立的LLM分析了元评估者对每个错误裁决的陈述理由。平衡准确率（敏感性和特异性的无权重平均值）范围从0.156（低于随机猜测）到接近完美。GPT-5.2在没有推理时的准确率为0.156，在其最高推理努力水平时上升至0.942，而Gemini 3 Pro在每个水平上均保持在0.94以上。Gemini 3 Pro在低推理努力下的表现超过了GPT-5.2在高推理努力下的表现，成本仅为每次试验约5%。当元评估者出错时，往往依赖表面线索而非操作性推理。实施草拟-验证-修订流程的上下文工程师应对指示性转变保持警惕，并在每个阶段明确意图指代的对象。

View on arXiv Download PDF AI Translation

cs.AI / 10 / 2609.12165

GLARE: Generative Learning via Adversarial Reward Estimation For Social Dynamics Forecasting

GLARE：通过对抗奖励估计进行生成学习以预测社会动态

Huang, Tenghao, Tan, Zhaoxuan, Chen, Muhao, May, Jonathan, Wan, Mengting, Yang, Longqi, Zhou, Pei, Chen, Sihao

Abstract

Meeting continuation requires tracking the agenda, speaker roles, participant intentions, and disagreement across long multi-party discussions. We introduce the Meeting Dynamic Forecasting Benchmark (MDFB), constructed from 2,207 real-world meetings and 24,794 future-facing queries. Given a transcript prefix and an active question, a model generates a plausible multi-turn continuation in one call. We evaluate utility---progress toward the question---and human-likeness---plausible conversational flow and role consistency---without requiring exact reproduction of the observed future. We further present GLARE, an adaptation of adversarial imitation learning to conditional language generation. A discriminator ranks the observed continuation above samples from the current actor, and its score supplies a KL-regularized policy reward; retraining on current-policy negatives allows the reward landscape to evolve with the actor. GLARE attains average human-evaluated win rates of 0.66 on utility and 0.70 on human-likeness, outperforming SFT and SPIN while remaining below the observed human continuation. We also demonstrate MDFB as a social reasoning arena for comparing general-purpose models, including closed-source systems, through reference-assisted judgments. Together, these studies illustrate the benchmark's use for both task-specific learning and output-based evaluation of meeting behavior.

Chinese Translation

会议的延续需要跟踪议程、发言者角色、参与者意图以及在长时间多方讨论中的分歧。我们介绍了会议动态预测基准（Meeting Dynamic Forecasting Benchmark, MDFB），该基准由2207个真实会议和24794个面向未来的查询构成。在给定的转录前缀和一个活跃问题下，模型能够在一次调用中生成一个合理的多轮延续。我们评估了效用——朝向问题的进展——和人类相似性——合理的对话流和角色一致性——而不要求完全重现观察到的未来。我们进一步提出了GLARE，这是一种将对抗模仿学习适应于条件语言生成的方法。一个判别器将观察到的延续排名高于当前参与者的样本，其得分提供了一个KL正则化的策略奖励；在当前策略的负样本上进行再训练使得奖励空间能够随着参与者的变化而演变。GLARE在效用上的平均人类评估胜率为0.66，在人类相似性上为0.70，超越了SFT和SPIN，但仍低于观察到的人类延续。我们还展示了MDFB作为一个社会推理平台，用于通过参考辅助判断比较通用模型，包括闭源系统。这些研究共同展示了该基准在任务特定学习和基于输出的会议行为评估中的应用。

View on arXiv Download PDF AI Translation

cs.AI / 11 / 2609.12171

WinSyn: An Automated Pipeline for Realistic Enterprise Question-Answering Evaluation

WinSyn：一种用于真实企业问答评估的自动化流程

Varhade, Amey, Sutradhar, Ananya, Krishnaswamy, Ravishankar, Goyal, Navin

Abstract

Enterprise settings provide a challenging environment for question-answering agents, which often rely on Retrieval-Augmented Generation, Deep Research (DR), and related techniques. Much of this challenge comes from the complexity of enterprise data: information is often spread across evolving and potentially conflict- ing emails, chat messages, documents, and other artifacts. Existing benchmarks typically have limited real-world complexity, short-form responses, and unnatural queries, so they often fail to capture the challenges of enterprise settings. In this work, we introduce an automated pipeline for generating synthetic datasets of emails reflecting realistic workplace scenarios, along with long- and short-form questions and gold answers grounded in the data. Our method simulates long-running enterprise projects spanning several months and involving up to 25 interacting employees across multiple roles. The data emphasizes ambiguity, distributed information, and naturally occurring queries. To validate the pipeline, we evaluate few standard agentic baselines on our datasets using the latest frontier models. We find that aggregate scores averaged over all queries remain below 80% for each dataset, indicating significant room for improvement. These findings suggest that more work remains to be done for enterprise deployment and underscore the importance of realistic, high-complexity evaluation data for developing stronger real-world enterprise DR systems.

Chinese Translation

企业环境为问答代理提供了一个具有挑战性的环境，这些代理通常依赖于检索增强生成（Retrieval-Augmented Generation）、深度研究（Deep Research, DR）及相关技术。这个挑战的主要原因在于企业数据的复杂性：信息通常分散在不断演变且可能存在冲突的电子邮件、聊天消息、文档及其他工件中。现有的基准测试通常具有有限的现实世界复杂性、短小的响应和不自然的查询，因此往往无法捕捉到企业环境中的挑战。在本研究中，我们引入了一种自动化流程，用于生成反映真实工作场景的电子邮件合成数据集，包含长短问题及基于数据的标准答案。我们的方法模拟了跨越数月的长期企业项目，涉及多达25名在多个角色中互动的员工。数据强调了模糊性、分布式信息和自然发生的查询。为了验证该流程，我们在我们的数据集上使用最新的前沿模型评估了几个标准的代理基线。我们发现，所有查询的平均聚合得分在每个数据集上均低于80%，这表明仍有显著的改进空间。这些发现表明，企业部署仍需更多工作，并强调了为开发更强大的现实世界企业DR系统而需要真实、高复杂度评估数据的重要性。

View on arXiv Download PDF AI Translation

cs.AI / 12 / 2609.12247

Soft Symbol Grounding for Prototypical Concepts

原型概念的软符号基础

Galván-López, Marcos, Upreti, Nijesh, Calvo, Hiram, Aguilar-Ibáñez, Carlos, Belle, Vaishak

Abstract

Neuro-symbolic models are usually trained with supervision only on final labels, leaving the intermediate concepts unobserved. Since many concept assignments are consistent with a given label, training can predict labels correctly while recovering the wrong concepts, a failure known as a reasoning shortcut. Prototypical networks reduce shortcuts by anchoring each concept to a few labeled examples, but existing methods still couple perception and reasoning through a hand-crafted, task-specific differentiable loss that must be redesigned for every task. We introduce \textbf{Soft-PNet}, which removes this loss: it reframes concept grounding as a Metropolis walk over a precomputed cache of feasible symbolic solutions, guided by a prototype distribution built from a single labeled anchor per concept, and trains against one KL objective between the prototype-weighted cache and the network's concept predictions. The objective is identical across tasks and remains applicable when the solution space cannot be enumerated. On \texttt{MNIST-EvenOdd}, Visual Sudoku, and \texttt{Kand-Logic} under scarce supervision, Soft-PNet matches loss-engineered prototypical networks at the concept and label levels and recovers concepts that soft-grounding baselines miss, with no loss engineering and lower training time.

Chinese Translation

神经符号模型通常仅在最终标签上进行监督训练，导致中间概念未被观察。由于许多概念分配与给定标签一致，训练可以正确预测标签，同时恢复错误的概念，这种失败被称为推理捷径。原型网络通过将每个概念锚定到少量标记示例来减少捷径，但现有方法仍通过手工设计的特定任务可微损失将感知与推理耦合，这需要为每个任务重新设计。我们引入了 extbf{Soft-PNet}，它消除了这种损失：它将概念基础重新构建为在预计算的可行符号解缓存上的Metropolis游走，由从每个概念的单个标记锚点构建的原型分布引导，并针对原型加权缓存与网络概念预测之间的一个KL目标进行训练。该目标在各个任务中是相同的，并且在解空间无法枚举时仍然适用。在稀缺监督下的 exttt{MNIST-EvenOdd}、视觉数独和 exttt{Kand-Logic} 上，Soft-PNet 在概念和标签层面上与损失工程优化的原型网络相匹配，并恢复了软基础基线遗漏的概念，且无需损失工程，训练时间更短。

View on arXiv Download PDF AI Translation

cs.AI / 13 / 2609.12265

GTA: Graph Theory Agent and Benchmark for Algorithmic Graph Reasoning with LLMs

GTA：图论智能体及其在大规模语言模型算法图推理中的基准测试

Xu, Zixiang, Wang, Yanbo, Wang, Chenxi, Gao, Lang, Song, Zirui, Huang, Yue, Chen, Zhaorun, Zhang, Xiangliang, Chen, Xiuying

Abstract

Large Language Models (LLMs) are increasingly asked to reason over structured data such as graphs, yet how reliably they can carry out multi-step graph algorithms in language remains unclear. Existing evaluations tend to use simple tasks on small graphs, to score code generation rather than reasoning over the graph itself, or to fix a single input format. We introduce Graph Theory Bench (GT Bench), a benchmark covering 24 classical graph problems in 44 task-structure settings, with over 100,000 examples across four representations: natural language, structured language, adjacency list, and adjacency matrix. Evaluating eight LLMs on GT Bench shows that accuracy is strongly tied to the input representation, that the best representation shifts with graph density, size, and topology as well as with the model, and that this sensitivity persists, attenuated, in the strongest reasoning models. Building on these observations, we propose the Graph Theory Agent (GTA), which pairs a preference-trained representation selector with plan-and-decompose scaffolding around a frozen executor LLM. GTA lifts Phi-4 from 53.5% to 69.1% on the benchmark's easy split and from 33.0% to 41.5% on its hard split, outperforming eight prompting and agent baselines, and transfers without retraining to GraCoRe and NLGraph. Code for benchmark generation and evaluation: https://github.com/xzx34/GTA. The project homepage is available at https://xzx34.github.io/gta/.

Chinese Translation

大型语言模型（LLMs）越来越多地被要求对结构化数据（如图）进行推理，但它们在语言中执行多步骤图算法的可靠性仍不明确。现有评估往往使用小图上的简单任务，以评估代码生成而非对图本身的推理，或者固定单一输入格式。我们引入了图论基准（Graph Theory Bench，GT Bench），这是一个涵盖44个任务结构设置下24个经典图问题的基准，包含超过100,000个示例，涵盖四种表示形式：自然语言、结构化语言、邻接表和邻接矩阵。在GT Bench上评估的八个LLMs显示，准确性与输入表示形式密切相关，最佳表示形式随着图的密度、大小和拓扑结构以及模型的不同而变化，这种敏感性在最强的推理模型中仍然存在，尽管有所减弱。基于这些观察，我们提出了图论智能体（Graph Theory Agent，GTA），它将偏好训练的表示选择器与围绕冻结执行器LLM的计划与分解框架相结合。GTA在基准的简单分割中将Phi-4的准确率从53.5%提升至69.1%，在困难分割中从33.0%提升至41.5%，超越了八个提示和智能体基线，并且在不重新训练的情况下迁移到GraCoRe和NLGraph。基准生成和评估的代码可在：https://github.com/xzx34/GTA获取。项目主页可访问：https://xzx34.github.io/gta/。

View on arXiv Download PDF AI Translation

cs.AI / 14 / 2609.12267

Learning Symbolic Constraint Representations from Examples: A Neuro-Symbolic Approach

从示例中学习符号约束表示：一种神经符号方法

Belmecheri, Nassim, Gotlieb, Arnaud, Lazaar, Nadjib, Spieker, Helge

Abstract

Learning user-defined concepts as constraint networks has been extensively studied in the constraint acquisition (CA) literature. However, existing approaches typically rely on intensive interactions with a human oracle, making the learning process costly in terms of time and number of queries. In this paper, we propose a neuro-symbolic framework for automatic CA that significantly reduces user involvement by introducing neural Oracle Transformer models which learn to emulate user responses and to generalize conceptual knowledge. Trained on previously available examples, the learned oracle interacts with a dedicated CA engine, FastCA, which systematically refines the oracle's responses into a sound, consistent, and interpretable constraint network. This neuro-symbolic interaction enables the recovery of structured symbolic models from data without prior domain knowledge. Our results demonstrate that this neuro-symbolic interplay effectively aligns data-driven pattern recognition with symbolic reasoning, offering a robust approach to automating model construction in combinatorial domains.

Chinese Translation

用户定义概念作为约束网络的学习在约束获取（CA）文献中得到了广泛研究。然而，现有的方法通常依赖于与人类oracle的密集交互，这使得学习过程在时间和查询次数上都成本高昂。本文提出了一种神经符号框架用于自动化的CA，通过引入神经Oracle Transformer模型显著减少用户参与，模型学习模仿用户响应并推广概念知识。在先前可用示例的基础上训练的学习oracle与专门的CA引擎FastCA进行交互，系统地将oracle的响应精炼为一个可靠、一致且可解释的约束网络。这种神经符号交互使得从数据中恢复结构化符号模型成为可能，而无需先前的领域知识。我们的结果表明，这种神经符号的相互作用有效地将数据驱动的模式识别与符号推理对齐，为组合领域中的模型构建自动化提供了一种稳健的方法。

View on arXiv Download PDF AI Translation

cs.AI / 15 / 2609.12286

T-GADE: Thermodynamical Generative-AI-Driven Evolution of LLM Artifacts

T-GADE：热力学生成式人工智能驱动的大型语言模型工件演化

Ogawa, Kyoko, Mori, Naoki

Abstract

Integrating evolutionary computation and large language models (LLMs) requires control of population diversity as well as generative capability. Among LLM outputs, those with explicit structure, such as a description paired with code, are structured artifacts; we use artifact for short. We propose T-GADE, which evolves these artifacts by extending thermodynamical genetic algorithms through LLM-based genetic operators and artifact-level diversity evaluation. A common free-energy objective supports generational and steady-state updates, with Fermi-type occupancy excluding repeated genotypes and Bose-type occupancy permitting them. We establish exact one-member removal and conditions for recovering the zero-temperature survival rule of Evolution of Heuristics (EoH). On the online bin-packing task studied in the EoH paper, excess measures relative bin-count overhead above a volume lower bound. Training excess uses search instances; transfer excess uses instances with another bin capacity. Generational Bose-type T-GADE at $T=0.003$ reduced median training excess by approximately 29%, from 1.152% to 0.815%, over 20 runs per configuration (two-sided Mann-Whitney $p=0.042$, Cliff's $\delta=0.378$). Validation selection among its two highest-ranked final candidates reached the same median transfer excess as EoH, 0.496%. These results demonstrate the utility of thermodynamical selection and validation-based use of retained artifacts.

Chinese Translation

将进化计算与大型语言模型（LLMs）相结合，需要控制种群多样性以及生成能力。在LLM的输出中，具有明确结构的输出，例如与代码配对的描述，被称为结构化工件；我们简称为工件。我们提出了T-GADE，通过扩展热力学遗传算法，结合基于LLM的遗传操作符和工件级多样性评估，来演化这些工件。一个共同的自由能目标支持代际和稳态更新，其中费米型占据排除重复基因型，而玻色型占据则允许它们。我们建立了精确的单成员移除及恢复零温度生存规则的条件，适用于启发式进化（EoH）。在EoH论文研究的在线装箱任务中，超额度量相对于体积下限的箱子计数开销。训练超额使用搜索实例；转移超额使用具有不同箱子容量的实例。在$T=0.003$的代际玻色型T-GADE中，经过20次每个配置的运行，减少了中位数训练超额约29%，从1.152%降至0.815%（双侧Mann-Whitney $p=0.042$，Cliff's $ heta=0.378$）。在其两个最高排名的最终候选者中进行验证选择，达到了与EoH相同的中位数转移超额0.496%。这些结果展示了热力学选择和基于验证的保留工件使用的实用性。

View on arXiv Download PDF AI Translation

cs.AI / 16 / 2609.12287

Robust Prototypical Networks for Few-Shot Sensor Fault Diagnosis

用于少样本传感器故障诊断的鲁棒原型网络

Belay, Mohammed Ayalew, Haghipour, Amirshayan, Rossi, Pierluigi Salvo

Abstract

Industrial fault diagnosis often operates with only a handful of labeled fault examples, making few-shot learning attractive for sensor monitoring. Standard prototypical networks are simple and effective; however, their class prototypes may become unstable in the very-low-shot regime because each decision relies on a small support set. We propose \emph{Multi-Episode Prototypical Networks} (MEPN), which aggregate prototypes from multiple disjoint support episodes and use their mean as the final class representative, reducing prototype variance without changing the encoder architecture. We evaluate MEPN on the DeFACTO sensor dataset using five-way fault classification with synthetic bias, drift, spike, and noise faults injected into real industrial measurements. Over 100 independent runs, MEPN reaches \textbf{\SensorOneShotGcpn\%} in the per-episode one-shot setting ($K\!=\!1$ shot, aggregated over $N_{\text{agg}}\!=\!10$ support episodes), substantially above single-episode baselines. Under an equal 10-sample support budget, MEPN and ProtoNet at $K\!=\!10$ are statistically indistinguishable, confirming prototype accumulation as the mechanism rather than superior fixed-budget learning.

Chinese Translation

工业故障诊断通常仅依赖少量标记的故障示例，这使得少样本学习在传感器监测中变得具有吸引力。标准原型网络简单且有效；然而，在极少样本的情况下，其类别原型可能变得不稳定，因为每个决策依赖于一个小的支持集。我们提出了 extit{多回合原型网络}（Multi-Episode Prototypical Networks, MEPN），该网络从多个不相交的支持回合中聚合原型，并使用它们的均值作为最终类别代表，从而在不改变编码器架构的情况下减少原型方差。我们在DeFACTO传感器数据集上评估MEPN，使用五类故障分类，注入合成偏差、漂移、尖峰和噪声故障到真实工业测量中。在超过100次独立运行中，MEPN在每回合一-shot设置下（$K eq1$，在$N_{ ext{agg}}=10$个支持回合上聚合）达到了 extbf{ ext{SensorOneShotGcpn ext{ extbackslash%}}}，显著高于单回合基线。在相同的10样本支持预算下，MEPN和ProtoNet在$K=10$时在统计上不可区分，确认了原型积累作为机制，而不是优越的固定预算学习。

View on arXiv Download PDF AI Translation

cs.AI / 17 / 2609.12304

Hybrid Physics-AI Framework of Body Center of Mass Dynamics from Wrist-Worn Sensors

基于腕部传感器的身体质心动态混合物理-人工智能框架

Que, Shuhao, Breschi, Valentina, Wang, Ying

Abstract

Wrist-worn IMU has been widely used for daily-life health monitoring. Yet, it does not fully represent whole-body dynamics, for which the body center of mass (COM) is considered the physiological reference standard. Therefore, this work proposes a simplified kinematic model (KM), which is designed to map the wrist IMU to the COM acceleration. It is built upon several reductive assumptions that enable the solvability of the dynamic equations based on wrist IMU measurements alone. This work further proposes three types of hybrid AI modeling methods, namely human kinematic model-based neural network (HKM-NN) models, to leverage the power of both grey-box and black-box modeling. The HKM-NN methods include serial learning (ser-) and two approaches of simultaneous learning (sim1- and sim2-). The proposed models are trained and tested using our dataset, which includes wrist IMU measurements and ground-truth COM measurements from 10 healthy volunteers during six gait activities and sit-to-stand (SS) transitional movement. The results demonstrate the feasibility of estimating COM acceleration from wrist IMU measurements. Our KM model yields satisfactory results, with an error ranging from 6.7% to 12.5% for gait activities and 5.6% for the SS. In comparison with the KM model, our HKM-NN models significantly enhance the performance, achieving 5.3% to 9.3% errors for gait activities, and the best error of 3.9% for the SS. In addition, the HKM-NN models demonstrate distinct robustness characteristics under noisy test conditions, with sim1-/sim2- generally maintaining greater robustness under Gaussian perturbations, while the KM model exhibits comparatively strong robustness under salt-and-pepper noise. These findings highlight the importance of combining biomechanical structure with data-driven learning for wearable sensing applications operating under imperfect and noisy measurement conditions.

Chinese Translation

腕部佩戴的惯性测量单元（IMU）已广泛应用于日常健康监测。然而，它并不能完全代表全身动态，其中身体质心（COM）被视为生理参考标准。因此，本研究提出了一种简化的运动学模型（KM），旨在将腕部IMU映射到COM加速度。该模型基于若干简化假设，能够仅基于腕部IMU测量解算动态方程。本研究进一步提出了三种混合人工智能建模方法，即基于人体运动学模型的神经网络（HKM-NN）模型，以利用灰箱和黑箱建模的优势。HKM-NN方法包括串行学习（ser-）和两种同时学习的方法（sim1-和sim2-）。所提出的模型使用我们的数据集进行训练和测试，该数据集包括10名健康志愿者在六种步态活动和坐立（SS）过渡运动中的腕部IMU测量和真实的COM测量。结果表明，从腕部IMU测量中估计COM加速度是可行的。我们的KM模型产生了令人满意的结果，步态活动的误差范围为6.7%至12.5%，而SS的误差为5.6%。与KM模型相比，我们的HKM-NN模型显著提高了性能，步态活动的误差达到5.3%至9.3%，而SS的最佳误差为3.9%。此外，HKM-NN模型在噪声测试条件下表现出明显的鲁棒性特征，其中sim1-/sim2-在高斯扰动下通常保持更大的鲁棒性，而KM模型在椒盐噪声下表现出相对较强的鲁棒性。这些发现强调了在不完美和噪声测量条件下，将生物力学结构与数据驱动学习相结合的重要性，以用于可穿戴传感应用。

View on arXiv Download PDF AI Translation

cs.AI / 18 / 2609.12313

Do Influence-Derived Data Perturbations Enable Machine Unlearning? A Controlled Study of Three Plausible Roles

影响导向的数据扰动是否能够实现机器遗忘？三种可能角色的控制研究

Wu, Chenkai, Kambimbi, Chrispine, Zeng, Qinyang, Yan, Jun

Abstract

We evaluate Deep Perturbation Learning (DPL), which perturbs training images and labels along influence-derived directions, in three roles in which prior work has positioned it for machine unlearning: a direct deletion signal (the strongest claim), a utility-preserving regularizer, and a warm start for adversarial unlearning. Evidence for the weaker roles has been used to support the stronger one, so we test each role separately under a matched protocol with exact-seed retraining baselines. An audit of the public implementation identifies two correctness issues: image directions are computed on augmented, normalized tensors but applied to raw images, and the label perturbation falls below float32 resolution, leaving labels unchanged. After correcting the image-perturbation pipeline, DPL fails the direct-deletion criterion on CIFAR-10/ResNet-18 in all three paired seeds. Its utility effects are inconsistent in sign across seeds, and once direction-computation time is counted it underperforms simple warm-start baselines. A one-seed Tiny ImageNet check likewise does not favor DPL as a regularizer or warm start; preprocessing inconsistencies in the released code make the direct comparison there inconclusive. These results cover random instance deletion only and do not rule out influence-based methods in other deletion regimes. We release a role-matched evaluation protocol and an audit checklist for perturbation-based deletion claims.

Chinese Translation

我们评估了深度扰动学习（Deep Perturbation Learning, DPL），该方法沿着影响导向的方向对训练图像和标签进行扰动，并在先前研究中将其定位于机器遗忘的三种角色：直接删除信号（最强的主张）、保持效用的正则化器，以及对抗性遗忘的热启动。对于较弱角色的证据被用来支持较强的主张，因此我们在匹配协议下分别测试每个角色，并使用精确种子重训练基线。对公共实现的审计发现了两个正确性问题：图像方向是在增强的、标准化的张量上计算的，但应用于原始图像；标签扰动低于 float32 分辨率，导致标签保持不变。在修正图像扰动流程后，DPL 在 CIFAR-10/ResNet-18 的所有三个配对种子上均未能满足直接删除标准。其效用效果在种子之间的符号不一致，并且一旦计算方向的时间，其表现不及简单的热启动基线。对单一种子的 Tiny ImageNet 检查同样未能支持 DPL 作为正则化器或热启动；发布代码中的预处理不一致使得直接比较结果不确定。这些结果仅涵盖随机实例删除，并未排除在其他删除机制中基于影响的方法。我们发布了一个角色匹配的评估协议和一个针对基于扰动的删除主张的审计清单。

View on arXiv Download PDF AI Translation

cs.AI / 19 / 2609.12320

AIM: A Privacy-Aware Interoperable Memory Framework for Multi-Agent Multi-User LLM Systems

AIM：一种隐私意识的可互操作多智能体多用户大语言模型系统内存框架

Johnson, Zachary, Kumankumah, Nigel Boachie, Chatterjee, Somya, Sathyamurthi, Tejas, Chen, Min, Li, Xinyi Alice, Wang, Xiao, Gelchie, Emily Morgan, Lin, Jessica, Hasan, Sadid A., Vesal, Sulaiman

Abstract

Traditional large language models (LLMs) are scoped to individual user sessions, limiting their knowledge to a single conversation and preventing them from learning user preferences that evolve over time. Existing agentic memory systems address this limitation but generally operate at the individual-user level, restricting the public knowledge that could be shared across users to improve downstream responses. We introduce AIM (Agentic Interoperable Memory), a unified, privacy-aware memory framework that enables multi-agent, multi-user LLM systems to persistently manage private and shared memory. AIM dynamically classifies information as private, scoped to one user and inaccessible to others, or public, accessible to all users. It enforces index-level access controls so that private memories are retrievable only by their owner, protecting sensitive data while allowing beneficial shared knowledge to improve coordination and consistency. We also introduce MUMBench (Multi-User Memory Benchmark), a dataset of multi-user interactions containing private and shareable information across four domains. To our knowledge, MUMBench is the first public dataset designed to evaluate multiple memory operations, including retrieval, creation, update, and deletion, in a multi-user environment. Across three independent runs on MUMBench, AIM achieves 96.0% visibility classification accuracy, 58.8% strict operation accuracy, and 70.5% state-aware operation accuracy.

Chinese Translation

传统的大语言模型（LLMs）仅限于单个用户会话，限制了它们的知识范围仅在一次对话中，无法学习随时间演变的用户偏好。现有的智能记忆系统解决了这一限制，但通常在个体用户层面操作，限制了可以在用户之间共享的公共知识，从而改善下游响应。我们提出了AIM（Agentic Interoperable Memory），一个统一的、隐私意识的内存框架，使多智能体、多用户的LLM系统能够持久地管理私有和共享内存。AIM动态地将信息分类为私有信息，仅限于一个用户且对其他用户不可访问，或公共信息，所有用户均可访问。它实施索引级访问控制，以确保私有记忆仅能被其所有者检索，从而保护敏感数据，同时允许有益的共享知识改善协调性和一致性。我们还引入了MUMBench（Multi-User Memory Benchmark），这是一个包含四个领域内私有和可共享信息的多用户交互数据集。据我们所知，MUMBench是第一个旨在评估多用户环境中多个内存操作（包括检索、创建、更新和删除）的公共数据集。在对MUMBench进行的三次独立运行中，AIM实现了96.0%的可见性分类准确率、58.8%的严格操作准确率和70.5%的状态感知操作准确率。

View on arXiv Download PDF AI Translation

cs.AI / 20 / 2609.12322

Affective Agent: On-Device Personalized Intervention Reasoning for Wearable Systems

情感代理：可穿戴系统的设备内个性化干预推理

Mun, Reina, Wan, Zishen, Reddi, Vijay Janapa

Abstract

Affective computing has advanced wearable state inference, but on-device reasoning about whether, when, and how to intervene remains challenging. We present Affective Agent, a three-layer reference architecture for personalized intervention reasoning under uncertainty on wearable-class hardware. It combines a compact sub-billion-parameter language model with physiological evidence, context, and user history to decide whether, when, and how to intervene, without cloud dependency or per-user retraining. The architecture is organized into three interacting layers (perception, personalization, and reasoning), adapting to individual users through host-managed structured memory evolution rather than per-user weight updates. We instantiate Affective Agent in indoor environmental quality control and evaluate it on held-out, simulator-generated longitudinal scenarios spanning physiological variation, context, signal quality, and intervention history. Results show that memory-driven personalization and two-pass structured reasoning improve intervention decisions within this synthetic evaluation. By moving the decision layer on-device, this work demonstrates a path from wearable state inference toward closed-loop, personalized intervention on wearable-class hardware.

Chinese Translation

情感计算已经推动了可穿戴状态推断的发展，但在设备上推理何时、如何进行干预仍然具有挑战性。我们提出了情感代理（Affective Agent），这是一种针对可穿戴硬件的不确定性下个性化干预推理的三层参考架构。它结合了一个紧凑的亚十亿参数语言模型与生理证据、上下文和用户历史，以决定是否、何时以及如何进行干预，而无需依赖云计算或对每个用户进行重新训练。该架构分为三个相互作用的层次（感知、个性化和推理），通过主机管理的结构化记忆演变适应个体用户，而不是通过每个用户的权重更新。我们在室内环境质量控制中实例化了情感代理，并在保留的、模拟生成的纵向场景中进行了评估，这些场景涵盖了生理变化、上下文、信号质量和干预历史。结果表明，基于记忆的个性化和双通道结构化推理在这一合成评估中改善了干预决策。通过将决策层移至设备上，这项工作展示了从可穿戴状态推断到在可穿戴硬件上实现闭环个性化干预的路径。

View on arXiv Download PDF AI Translation

cs.AI / 21 / 2609.12327

LoRA-RC: Reservoir Computing with Low-Rank Adaptation

LoRA-RC：低秩适应的水库计算

Wan, Wenbin

Abstract

Reservoir computing (RC) trains only a linear readout over a fixed recurrent layer, making it fast and data-efficient for online prediction. However, a static reservoir degrades under system drift, readout-only adaptation is then insufficient, and unconstrained reservoir adaptation can destroy the echo-state and incremental stability properties that make RC reliable. This paper proposes LoRA-RC, which adapts the recurrent matrix through a low-rank correction driven by streaming prediction errors. The base reservoir and adaptation bases are fixed offline; a small core matrix is adapted online, projected onto a spectral-norm ball, and low-pass filtered at each step. The projection guarantees that every applied recurrent matrix remains within a certified contraction set, and an incremental input-to-state stability bound is established for the reservoir along each online adaptation path, with path-independent rate and gain. On a Lorenz system with an abrupt parameter drift, LoRA-RC cuts post-drift prediction error by 56% versus a fixed RC and 51% versus readout-only adaptation; ablations over 20 seeds show that removing the projection inflates this error by more than a factor of 40.

Chinese Translation

水库计算（RC）仅在固定的递归层上训练线性读出，使其在在线预测中快速且数据高效。然而，静态水库在系统漂移下会退化，仅依赖读出适应是不够的，而不受限制的水库适应可能破坏回声状态和增量稳定性特性，这些特性使得水库计算可靠。本文提出了LoRA-RC，通过基于流预测误差的低秩修正来适应递归矩阵。基本水库和适应基在离线时固定；一个小的核心矩阵在线适应，投影到谱范数球上，并在每一步进行低通滤波。该投影保证每个应用的递归矩阵保持在一个经过认证的收缩集内，并为水库沿每条在线适应路径建立了增量输入到状态的稳定性界限，具有路径独立的速率和增益。在一个具有突发参数漂移的洛伦兹系统中，LoRA-RC将漂移后的预测误差相比于固定的水库减少了56%，相比于仅读出的适应减少了51%；对20个种子的消融实验表明，去除投影会使该误差增加超过40倍。

View on arXiv Download PDF AI Translation

cs.AI / 22 / 2609.12373

Toward Robust Personalized Alignment for LLMs: Mitigating Persona Drift in Multi-Turn Dialogue

朝着强健的个性化对齐：缓解多轮对话中的个性漂移

Zhang, Youyuan, Li, Siyuan, Liu, Fangming, Li, Jing

Abstract

Persona drift remains a central challenge for personalized language models, as user profiles evolve over long interactions rather than remain permanently fixed. Models must therefore revise persistent persona states when preferences genuinely change, while avoiding updates driven by transient, ambiguous, or unresolved observations. We propose CORE, which separates turn-local evidence from persistent persona-state revision and selectively updates grounded user preferences through uncertainty-aware belief revision. We also introduce PERSIST, a held-out post-anchor benchmark for persona-state robustness under sequential interaction stress, covering ambiguity, conflict, and controlled social influence. Across ALOE, PersonaChat, and PERSIST, CORE improves personalized alignment and robustness, with complementary gains in normalized closed-slot state fidelity. Human evaluation and mechanistic controls further support explicit update control beyond stronger generation or persistent memory alone.

Chinese Translation

个性漂移仍然是个性化语言模型面临的一个核心挑战，因为用户档案在长时间交互中会演变，而不是永久固定。因此，当用户偏好真正发生变化时，模型必须修订持久的个性状态，同时避免因瞬时、模糊或未解决的观察而驱动的更新。我们提出了CORE，它将轮次局部证据与持久个性状态修订分开，并通过不确定性感知的信念修订选择性地更新基于用户偏好的信息。我们还引入了PERSIST，这是一个用于评估个性状态在顺序交互压力下的稳健性的后锚定基准，涵盖模糊性、冲突和控制社会影响。在ALOE、PersonaChat和PERSIST数据集上，CORE提高了个性化对齐和稳健性，并在归一化的封闭槽状态保真度上获得了互补的提升。人类评估和机制控制进一步支持了超越更强生成或持久记忆的显式更新控制。

View on arXiv Download PDF AI Translation

cs.AI / 23 / 2609.12394

BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI Agents

BlueLM-GUI技术报告：一个以真实设备为中心的自我改进移动GUI代理的飞轮

Ye, Tong, Han, Kunyang, Wang, Guozhi, Luo, Longqiang, Ding, Zhifeng, Zhang, Yongxiang, Shen, Xiaolei, Zhang, Yuxuan, Zhang, Zhuping, Xu, Tao, Pan, Yue, Zhao, Yucheng, Hu, Yupei, Ouyang, Yuanjiang, Shen, Danfeng, Lin, Runqi, Cai, Hongda, Wang, Zhaoxiong, Yan, Mengjia, Zhong, Yingjie, Zhou, Chen, Zhang, Zeyu, Zhu, Xuwen, Shi, Penggang, Luo, Mingcheng, Wu, Ziyang, Jin, Min, Shen, Mingfu, Xu, Zairong, Zhang, Fan, Wang, Hao, Liu, Liang, Xie, Zhulin, Yao, Lijun, Liang, Xiao, Wen, Liangmin, Feng, Liqiang, Wu, Feilong, Hu, Min, Chen, Min, Xiong, Guanjing, Ruan, Xiaohu, Chen, Xiaoxin

Abstract

Mobile GUI agents are shifting from multi-module frameworks to native models trained end-to-end, yet industrial deployment faces three persistent gaps. Sandbox training produces a distribution mismatch with production environments; expensive real-device failures remain underutilized; and fixed benchmarks saturate, losing the power to guide iteration. We present BlueLM-GUI, a 35B-A3B mobile GUI agent built as a real-device-centric flywheel that closes these gaps through three principles. Every Sample Matters: a dual-track pipeline with Heterogeneous Triple-System Consensus evaluation and an Error Correction \& Derivation Module salvages every trajectory into usable supervision. Every Rollout Is Real: a three-stage recipe---continual pre-training, supervised fine-tuning, and agentic reinforcement learning on hundreds of real phones---grounds every rollout in real production environments, so the capability the model learns transfers directly to deployment. Every Query Evolves: a quota-driven benchmark methodology with three orthogonal axes enables precise attribution and allows the benchmark to be systematically upgraded as the model improves. BlueLM-GUI achieves 87.4 on MobileGUI-VBench, surpassing the best closed-source model by 5.1 points, and 84.9 on AndroidWorld, the best result among open-source models and competitive with closed-source models. These results demonstrate that grounding model training and iterative improvement in both real devices and the three Every principles yields strong, robust, and transferable mobile GUI capability.

Chinese Translation

移动GUI代理正从多模块框架转向端到端训练的原生模型，但工业部署面临三个持续存在的差距。沙盒训练导致与生产环境的分布不匹配；昂贵的真实设备故障仍未得到充分利用；固定基准测试饱和，失去了指导迭代的能力。我们提出了BlueLM-GUI，一个基于真实设备中心的飞轮，旨在通过三个原则来弥合这些差距。每个样本都重要：一个双轨管道结合异构三系统共识评估和错误修正与推导模块，将每个轨迹转化为可用的监督信号。每次推出都是真实的：一个三阶段的配方——持续预训练、监督微调和在数百部真实手机上的代理强化学习——将每次推出扎根于真实的生产环境中，使模型学习的能力能够直接转移到部署中。每个查询都在演变：一个基于配额的基准方法，具有三个正交轴，使得精确归因成为可能，并允许基准随着模型的改进而系统性地升级。BlueLM-GUI在MobileGUI-VBench上取得了87.4的成绩，超越了最佳闭源模型5.1分，在AndroidWorld上取得了84.9的成绩，是开源模型中的最佳结果，并与闭源模型具有竞争力。这些结果表明，将模型训练和迭代改进扎根于真实设备以及三个“每个”原则，能够产生强大、稳健且可转移的移动GUI能力。

View on arXiv Download PDF AI Translation

cs.AI / 24 / 2609.12395

Is Gaussian Splatting Becoming Neural Again? A Taxonomy and Controlled Study of Learned Parameterization

高斯点云是否再次变得神经化？学习参数化的分类与控制研究

Wang, YuanHang, Cao, Xin, Zhang, Yi

Abstract

Three-dimensional Gaussian Splatting (3DGS) combines explicit primitives with efficient rasterization, yet recent systems increasingly use neural networks to generate or share Gaussian parameters. We characterize this trend along five axes: attribute decoding, spatial sharing, view-conditioned decoding, topology generation, and amortized inference. An analysis of 19 representative methods shows that these choices address different limitations and cannot be reduced to a binary neural label. We also isolate three forms of neural parameterization in a controlled mip-NeRF 360 study. Sharing appearance and opacity improves reconstruction quality, while decoding geometric structure offers no further gain. The evidence favors selective neuralization: shared functions help when they capture reusable correlations without sacrificing the local geometric freedom of explicit splats.

Chinese Translation

三维高斯点云（3DGS）结合了显式原语与高效光栅化，然而最近的系统越来越多地使用神经网络来生成或共享高斯参数。我们从五个方面对这一趋势进行了描述：属性解码、空间共享、视图条件解码、拓扑生成和摊销推理。对19种代表性方法的分析表明，这些选择解决了不同的局限性，无法简单归结为二元的神经标签。我们还在一个控制的 mip-NeRF 360 研究中隔离了三种形式的神经参数化。共享外观和不透明度提高了重建质量，而解码几何结构并未带来进一步的提升。证据支持选择性神经化：共享函数在捕捉可重用的相关性时有助于提升性能，而不牺牲显式点云的局部几何自由度。

View on arXiv Download PDF AI Translation

cs.AI / 25 / 2609.12398

Niching Agents in The Core

核心中的细分智能体

Parker, Gary B., O'Connor, Jim, Asaro, John

Abstract

The Core is a unique competitive co-evolution algorithm that allows agents to evolve autonomous control without utilizing a traditional fitness function. The agents evolve via local interactions through tournament selection, crossover, and mutation, producing offspring by evolving better controllers. Previous works have shown The Core's ability to evolve agents capable of combat and navigation in the Xpilot video game. This research expands upon that premise by niching agents to specific subsets of the original environment The Core was tested in. Our results demonstrate the niched agents capacity for success over agents niched to the entire system and agents niched to different sub-environments.

Chinese Translation

核心是一种独特的竞争性共进化算法，它允许智能体在不使用传统适应度函数的情况下进化自主控制。智能体通过局部交互、锦标赛选择、交叉和变异进行进化，产生后代并进化出更好的控制器。先前的研究已展示了核心在Xpilot视频游戏中进化出能够进行战斗和导航的智能体的能力。本研究在此基础上扩展，通过将智能体细分到核心测试的原始环境的特定子集。我们的结果表明，细分智能体在成功率上优于细分到整个系统的智能体和细分到不同子环境的智能体。

View on arXiv Download PDF AI Translation

cs.AI / 26 / 2609.12399

OneLA: Scaling Linear-Attention Decoding to Large Beams in Generative Recommendation

OneLA：将线性注意力解码扩展至生成推荐中的大束

Yang, Xiangrui, Peng, Cheng, Zhao, Yunfeng, Zeng, Liang, Hu, Ao, Yang, Jiawei, Wang, Shengzhe, Lv, Jingshan, Liang, Xiao, Yang, Chen, Liu, Jiaqiang, Qiu, Yiming

Abstract

Generative recommendation (GR) relies on large-beam decoding to generate hundreds of candidate items, creating a new scaling challenge for recurrent linear attention. Existing linear attention serving systems either materialize a full recurrent state for every beam or repeatedly replay shared history, incurring substantial memory and traffic overhead. To address this, we present OneLA, a linear-attention decoding framework that exploits the shared prompt and short divergent suffixes of GR workloads. Specifically, OneLA represents all beam states using a single shared prompt-derived state and compact, append-only records of their divergent transitions. Using this representation, OneLA computes only the state information required at each decoding step, without reconstructing a full recurrent state for every beam. Furthermore, OneLA uses a lightweight ancestry index to track the transition records that make up each beam's history, allowing beams to be updated without moving or copying existing records. A fused GPU kernel further reuses the shared state across beams. Our analysis shows that OneLA achieves 1.54-2.46x end-to-end decode speedups while substantially reducing recurrent-state memory use and data movement.

Chinese Translation

生成推荐（Generative Recommendation, GR）依赖于大束解码生成数百个候选项，这为递归线性注意力带来了新的扩展挑战。现有的线性注意力服务系统要么为每个束生成完整的递归状态，要么重复回放共享历史，导致显著的内存和流量开销。为了解决这个问题，我们提出了OneLA，一个线性注意力解码框架，利用GR工作负载的共享提示和短的发散后缀。具体而言，OneLA使用单个共享提示派生状态和紧凑的仅附加记录来表示所有束状态及其发散转变。通过这种表示，OneLA仅计算每个解码步骤所需的状态信息，而无需为每个束重建完整的递归状态。此外，OneLA使用轻量级的祖先索引来跟踪构成每个束历史的转变记录，从而允许在不移动或复制现有记录的情况下更新束。融合的GPU内核进一步在束之间重用共享状态。我们的分析表明，OneLA实现了1.54-2.46倍的端到端解码加速，同时显著减少了递归状态的内存使用和数据移动。

View on arXiv Download PDF AI Translation

cs.AI / 27 / 2609.12400

Decentralized Evolution of Hexapod Gaits with Independent Leg Controllers

具有独立腿部控制器的六足步态的分散进化

Parker, Gary B., Asaro, John, O'Connor, Jim

Abstract

This paper presents a novel approach to hexapod locomotion by evolving each leg's gait independently through a decentralized evolutionary algorithm. Using the Webots simulator and the Mantis hexapod robot, we optimize individual leg controllers without centralized coordination, allowing emergent behaviors to drive the development of efficient, coordinated locomotion. Our decentralized method is benchmarked against cooperative coevolution, demonstrating improved efficacy in generating stable and adaptive gaits while showing interesting emergent coordination. By enabling independent evolution of leg controllers, this method reduces the complexity of gait optimization and highlights the potential of decentralized strategies for scalable and adaptive robotic systems.

Chinese Translation

本文提出了一种新颖的六足运动方法，通过分散进化算法独立进化每条腿的步态。利用Webots仿真器和Mantis六足机器人，我们在没有集中协调的情况下优化各个腿部控制器，使得自发行为推动高效、协调运动的发展。我们的分散方法与合作共进化进行了基准测试，显示出在生成稳定和自适应步态方面的有效性提升，同时展现出有趣的自发协调。通过实现腿部控制器的独立进化，该方法降低了步态优化的复杂性，并突显了分散策略在可扩展和自适应机器人系统中的潜力。

View on arXiv Download PDF AI Translation

cs.AI / 28 / 2609.12403

Beyond ID Embeddings: Process-Grounded Language Modeling for Cognitive Diagnosis

超越身份嵌入：基于过程的语言建模用于认知诊断

Liu, Minghang, Wang, Yuanzhuo, Qiu, Qiang, Shen, Huawei, Cheng, Xueqi

Abstract

Cognitive Diagnosis Models (CDMs) play a pivotal role in personalized online learning. Traditional CDMs rely on discrete, ID-based embeddings to represent students, exercises, and concepts. This paradigm diverges from the nature of learner cognition, where knowledge is not stored and retrieved as isolated symbols. As a result, CDMs suffer from semantic limitations when new exercises or concepts appear. In this paper, we propose a Process-aware Language Cognitive Diagnosis (PLCD) framework that uses language-derived structures as cognitive priors and response records to calibrate student posterior states. PLCD leverages large language models (LLMs) to construct concept schemas and cognitive process graphs, and uses target-conditioned semantic memory to retrieve historical responses that are relevant to each target exercise. A process-grounded Language-to-Cognition Mapper with DA-MoE experts and process-level contrastive learning then maps the textual evidence into a unified cognitive space. Experimental results show that PLCD not only outperforms traditional baselines in predicting student performance but also exhibits strong cognitive transfer capabilities. These results connect the computational power of LLMs with the psychometric goal of measuring latent knowledge states, suggesting that structured language priors calibrated by response records can improve cold-start robustness and cognitive grounding.

Chinese Translation

认知诊断模型（CDMs）在个性化在线学习中发挥着关键作用。传统的CDMs依赖于离散的、基于身份的嵌入来表示学生、练习和概念。这一范式与学习者认知的本质相悖，因为知识并不是以孤立符号的形式存储和检索的。因此，当出现新的练习或概念时，CDMs会遭遇语义限制。在本文中，我们提出了一种过程感知的语言认知诊断（PLCD）框架，该框架利用语言衍生的结构作为认知先验和响应记录，以校准学生的后验状态。PLCD利用大型语言模型（LLMs）构建概念架构和认知过程图，并使用目标条件的语义记忆来检索与每个目标练习相关的历史响应。然后，采用具有DA-MoE专家和过程级对比学习的过程基础语言-认知映射器将文本证据映射到统一的认知空间。实验结果表明，PLCD不仅在预测学生表现方面优于传统基线，而且还表现出强大的认知迁移能力。这些结果将LLMs的计算能力与测量潜在知识状态的心理测量目标联系起来，表明通过响应记录校准的结构化语言先验可以提高冷启动的鲁棒性和认知基础。

View on arXiv Download PDF AI Translation

cs.AI / 29 / 2609.12404

VRL-Bench: Benchmarking agents on computer control tasks under finite trial budgets

VRL-Bench：在有限试验预算下对计算机控制任务中的智能体进行基准测试

Bai, Yu, Miao, Yukai, Wang, Dawei, Chen, Li, Ren, Yanyu, Shi, Yuqian, Li, Dan, Xiong, Ying, Tan, Chengqiu, Zhou, Run, Li, Li

Abstract

Learning from trial and error is a promising way to improve language agents on complex tasks such as computer control. Reflexion introduced verbal reinforcement learning, which turns failed trials into text that guides later attempts without updating model parameters. We introduce VRL-Bench, a harness for fair evaluation of trial-and-error learning under finite trial budgets. Across three models on MiniWoB and WebShop, we evaluate updates from several prominent verbal-memory methods spanning Reflexion and later work: each improves observed success over memory-free retry in some settings but reduces it in others. Replay experiments show that using reflection can reduce success rates, revealing a trade-off between exploiting experience and continued exploration. We propose VEX$^2$, a verbal exploration--exploitation scheduler that uses a language model to jointly select policies and allocate the remaining trial budget. VEX$^2$ is the only evaluated update to achieve positive observed success-rate gains over retry in all six settings.

Chinese Translation

通过试错学习是提高语言智能体在复杂任务（如计算机控制）中的表现的一种有前景的方法。Reflexion 引入了语言强化学习，它将失败的试验转化为文本，以指导后续尝试，而无需更新模型参数。我们介绍了 VRL-Bench，这是一个在有限试验预算下公平评估试错学习的工具。在 MiniWoB 和 WebShop 上的三个模型中，我们评估了来自多个显著语言记忆方法的更新，这些方法涵盖了 Reflexion 及其后续工作：在某些设置中，每种方法都提高了相较于无记忆重试的成功率，但在其他设置中则降低了成功率。重放实验表明，使用反思可能会降低成功率，揭示了利用经验与持续探索之间的权衡。我们提出了 VEX$^2$，一种语言探索-利用调度器，它利用语言模型共同选择策略并分配剩余的试验预算。VEX$^2$ 是唯一在所有六个设置中实现相较于重试的正向观察成功率提升的评估更新。

View on arXiv Download PDF AI Translation

cs.AI / 30 / 2609.12413

SoK: Rethinking Jailbreaking in the Era of Agentic AI: Attacks, Defenses, and Practical Consideration

系统化知识：在自主智能时代重新思考越狱问题：攻击、防御与实践考虑

Mia, Md Jueal, Wu, Yanzhao, Uluagac, Selcuk, Amini, M. Hadi

Abstract

Large language models (LLMs) are rapidly evolving from conversational assistants into agentic AI systems that reason, plan, invoke tools, maintain persistent memory, communicate with other agents, and execute multi-step tasks. At the same time, modern models exhibit substantially stronger native safety alignment than earlier generations on which many jailbreak attacks and defenses were originally studied. This shift raises a fundamental question: \textit{which established jailbreak-security findings remain valid in the era of modern LLMs and agentic AI?} We address this question through a Systematization of Knowledge (SoK) that reframes jailbreak security around the full agentic execution pipeline. We develop unified taxonomies of attacks and defenses spanning user interaction, planning and reasoning, memory, tool use, and inter-agent communication, and introduce a security--utility--efficiency evaluation framework that separates native harmful-prompt safety, adversarial jailbreak robustness, and agent-level security outcomes. We further conduct a controlled empirical study of representative attacks and defenses within a common agentic framework. Our results reveal three important gaps. First, strong native alignment does not imply robustness to adversarial jailbreaks. Second, defense effectiveness is highly model-, attack-, and component-dependent and can come at substantial cost in over-refusal, utility, and latency. Third, low final-response attack success can mask severe intermediate compromise: planning, memory, and tool interactions may remain unsafe even when the final response is successfully filtered. These findings motivate a shift from response-centric jailbreak defense toward cross-layer, execution-aware security that protects agent state, component transitions, and external actions while preserving practical utility and efficiency.

Chinese Translation

大型语言模型（LLMs）正迅速从对话助手演变为能够推理、规划、调用工具、维持持久记忆、与其他智能体沟通并执行多步骤任务的自主智能系统。同时，现代模型展现出比早期世代更强的本地安全对齐性，而许多越狱攻击和防御最初是在这些早期模型上进行研究的。这一转变提出了一个根本性的问题： extit{在现代LLMs和自主智能的时代，哪些既有的越狱安全发现仍然有效？} 我们通过系统化知识（SoK）来解决这个问题，重新框定越狱安全，围绕完整的自主执行流程展开。我们开发了一个统一的攻击和防御分类法，涵盖用户交互、规划与推理、记忆、工具使用和智能体间通信，并引入一个安全-效用-效率评估框架，区分本地有害提示安全性、对抗性越狱鲁棒性和智能体级安全结果。我们还在一个共同的自主框架内进行了代表性攻击和防御的控制实证研究。我们的结果揭示了三个重要的差距。首先，强大的本地对齐性并不意味着对抗性越狱的鲁棒性。其次，防御的有效性高度依赖于模型、攻击和组件，并可能在过度拒绝、效用和延迟方面带来显著成本。第三，低最终响应攻击成功率可能掩盖严重的中间妥协：即使最终响应成功过滤，规划、记忆和工具交互仍可能不安全。这些发现促使我们从以响应为中心的越狱防御转向跨层、执行感知的安全性，保护智能体状态、组件转换和外部行为，同时保持实用效用和效率。

View on arXiv Download PDF AI Translation

cs.AI / 31 / 2609.12422

Hierarchical Belief Modeling for Zero-Shot Opponent Adaptation in Partially Observable Multi-Agent Navigation

用于部分可观测多智能体导航的零-shot对手适应的层次信念建模

Shih, Kowei, Cheng, Lu, Wang, Zeyu, Xu, Yeyun, Tong, Kejian

Abstract

Lux AI Season 3 requires agents to act under partial observability, randomized episode level dynamics, and a best of five match structure that rewards both tactical execution and fast adaptation. We present HORIZON, a hierarchical agent that combines symmetry aware spatial perception, dual memory belief tracking, relic centric graph attention, information gain driven exploration, and an opponent conditioned policy mixture. HORIZON separates short horizon control from cross match meta reasoning, while auxiliary belief and world model objectives stabilize learning. Trained with PPO in a large scale JAX simulator, the resulting agent explicitly infers hidden game parameters and opponent style. Experiments show consistent gains in match win rate, episode win rate, adaptation gain, and league rating over strong recurrent and feed forward baselines.

Chinese Translation

Lux AI 第三季要求智能体在部分可观测性、随机化的回合动态以及奖励战术执行和快速适应的五局三胜赛制下进行行动。我们提出了 HORIZON，一种层次化智能体，结合了对称感知的空间感知、双重记忆信念跟踪、以遗物为中心的图注意力、基于信息增益的探索以及对手条件的策略混合。HORIZON 将短期控制与跨局的元推理分离，同时辅助信念和世界模型目标稳定学习。在大规模 JAX 模拟器中使用 PPO 进行训练，所得到的智能体能够明确推断隐藏的游戏参数和对手风格。实验表明，与强大的递归和前馈基线相比，比赛胜率、回合胜率、适应增益和联赛评分均有持续提升。

View on arXiv Download PDF AI Translation

cs.AI / 32 / 2609.12436

LifeFuse-Mem: Lifecycle-Aware State Fusion Against Temporary Overwriting for Long-Term Memory

LifeFuse-Mem：针对临时覆盖的生命周期感知状态融合用于长期记忆

Zhao, Hanyu, Feng, Yuqian, Song, Zhenyu, Cheng, Yuanchao, Jiao, Yance, Pan, Tengfei, Du, Li

Abstract

Long-running LLM agents require memory mechanisms that maintain coherent internal states across interactions. We study a lifecycle-labeled memory setting in which write episodes provide lifecycle metadata during training, and phase-aware readout is used during evaluation. This setting reflects the need to distinguish information that should remain influential across future interactions from information that should affect only the current context. A mismatch between these lifecycles can cause temporary information to overwrite durable knowledge, leading to behavioral drift in persistent agents. Within this setting, we introduce \textbf{LifeFuse-Mem}, a lifecycle-aware neural memory framework that separates information according to its temporal commitment. LifeFuse-Mem uses dedicated memory components and lifecycle-aware updates to allow stable and transient knowledge to evolve locally without converting temporary context into durable state. On the controlled anti-overwrite benchmark, LifeFuse-Mem improves acquisition-controlled retention and reduces temporary overwrite; on two public long-memory benchmarks, it remains broadly competitive. These results suggest that explicit lifecycle signals can help diagnose and mitigate overwrite in compact online memory.

Chinese Translation

长期运行的LLM代理需要能够在交互中维持一致内部状态的记忆机制。我们研究了一种带有生命周期标签的记忆设置，在该设置中，写入事件在训练期间提供生命周期元数据，而在评估期间使用阶段感知的读取方式。该设置反映了区分应在未来交互中持续影响的信息与仅应影响当前上下文的信息的必要性。这些生命周期之间的不匹配可能导致临时信息覆盖持久知识，从而导致持久代理的行为漂移。在此设置中，我们引入了 extbf{LifeFuse-Mem}，一种生命周期感知的神经记忆框架，根据信息的时间承诺对其进行分类。LifeFuse-Mem使用专用的记忆组件和生命周期感知的更新，使得稳定和瞬态知识能够在局部演变，而无需将临时上下文转换为持久状态。在受控的反覆盖基准测试中，LifeFuse-Mem提高了获取控制的保留率，并减少了临时覆盖；在两个公共的长期记忆基准测试中，它仍然保持广泛的竞争力。这些结果表明，明确的生命周期信号可以帮助诊断和减轻紧凑在线记忆中的覆盖问题。

View on arXiv Download PDF AI Translation

cs.AI / 33 / 2609.12446

Do LLMs Trust the Accuser or the Accusation? Measuring Belief Shifts in Werewolf

大型语言模型信任控诉者还是控诉内容？在《狼人》游戏中测量信念转变

Yang, Yu-Yu, Wu, Ti-Rong, Guei, Hung, Chen, Hsing-Yu, Wu, I-Chen

Abstract

Social-deduction games such as Werewolf are increasingly used to evaluate LLM agents, but existing evaluations often rely on final game outcomes. We propose a belief-shift evaluation benchmark in Werewolf for analyzing communication skills through belief updating. Using LLM-played games, we annotate suspicion and accusation messages and measure how an observing village-side model's beliefs change after each message. We evaluate 40 open-weight LLM configurations on 1,224 annotated messages. Our results show that larger models better distinguish true wolves from villagers based on game history, but accusations still strongly influence their beliefs. Models become more suspicious of the accused target and less suspicious of the accuser, especially when the accuser is trusted, even if the accuser is wolf-aligned. Larger models better resist accusations from accusers they already distrust. Overall, our findings suggest that current open-weight LLMs up to 120B parameters still struggle to integrate accusation content with source trust in strategic communication. Our benchmark and code are available at https://rlg.iis.sinica.edu.tw/papers/werewolf-accusation-benchmark.

Chinese Translation

社交推理游戏如《狼人》越来越多地被用来评估大型语言模型（LLM）代理，但现有评估往往依赖于最终的游戏结果。我们提出了一种在《狼人》游戏中用于分析通过信念更新进行沟通技能的信念转变评估基准。通过LLM参与的游戏，我们对怀疑和控诉信息进行了注释，并测量观察村庄方模型在每条信息后的信念变化。我们对40种开放权重的LLM配置进行了评估，共分析了1,224条注释信息。我们的结果表明，较大的模型能够更好地区分真实的狼人和村民，基于游戏历史进行判断，但控诉仍然对其信念产生强烈影响。模型对被控诉目标的怀疑程度增加，而对控诉者的怀疑程度降低，尤其是在控诉者被信任的情况下，即使控诉者与狼人阵营一致。较大的模型更能抵抗来自他们已经不信任的控诉者的控诉。总体而言，我们的研究结果表明，目前开放权重的LLM（参数最多可达120B）在战略沟通中仍然难以将控诉内容与来源信任整合。我们的基准和代码可在 https://rlg.iis.sinica.edu.tw/papers/werewolf-accusation-benchmark 获取。

View on arXiv Download PDF AI Translation

cs.AI / 34 / 2609.12459

EvoRS: On-Policy Self-Evolution of Reward Systems for Open-Ended Reinforcement Learning

EvoRS：开放式强化学习中奖励系统的在线自我演化

Li, Weiyuan, Chen, Aili, Wang, Xintao, Zhang, Yikai, Dong, Qingqing, Xu, Jinghan, Hou, Hongru, Zhao, Wenxuan, Lang, Chengkun, Gao, Jun, Guo, Yuanli, Guo, Hongcheng, Xiao, Yanghua, Yang, Deqing

Abstract

Open-ended reinforcement learning often relies on rubric-based rewards for tasks without directly verifiable answers. Yet the policy and reward system form a dynamic feedback loop: as the policy optimizes the current reward, an initially useful reward system may become unreliable due to reward hacking or reduced response discriminability. The reward system should therefore evolve rather than remain fixed during training. Existing dynamic-rubric methods adapt evaluation criteria, but reward failures can also arise from scoring mechanisms or signal composition. We introduce EvoRS, a self-evolving RL framework that evolves the reward system from on-policy experience, representing it as an executable Reward-DAG. Specifically, an agentic designer updates this system from on-policy rollouts and reward traces to maintain train-time reliability. Across writing and roleplay, EvoRS achieves the best quality under all three judges, outperforming the policy by $2.107$ and $4.767$ points, respectively, while reducing reward hacking and coverage failures and preserving reward informativeness. Ablations confirm that a comprehensive fixed reward system cannot remain reliable in open-ended tasks and must evolve throughout training.

Chinese Translation

开放式强化学习通常依赖于基于评分标准的奖励来处理没有直接可验证答案的任务。然而，策略和奖励系统形成了一个动态反馈循环：随着策略优化当前奖励，最初有效的奖励系统可能因奖励黑客行为或响应可辨别性降低而变得不可靠。因此，奖励系统在训练过程中应当演化，而非保持固定。现有的动态评分标准方法适应评估标准，但奖励失败也可能源于评分机制或信号组合。我们提出了EvoRS，一个自我演化的强化学习框架，它从在线策略经验中演化奖励系统，并将其表示为可执行的奖励有向无环图（Reward-DAG）。具体而言，代理设计者从在线策略的回滚和奖励轨迹中更新该系统，以保持训练期间的可靠性。在写作和角色扮演任务中，EvoRS在所有三位评审中都达到了最佳质量，分别超越了策略2.107和4.767分，同时减少了奖励黑客行为和覆盖失败，并保持了奖励的信息性。消融实验确认，在开放式任务中，全面固定的奖励系统无法保持可靠性，必须在训练过程中不断演化。

View on arXiv Download PDF AI Translation

cs.AI / 35 / 2609.12464

Beyond Vector Similarity: Hierarchical Context-Aware Graph RAG vs Standard RAG in Enterprise Code Migration

超越向量相似性：层次上下文感知图 RAG 与标准 RAG 在企业代码迁移中的比较

Jaiswal, Nilesh, Agrawal, Aniket, Shukla, Arjit, Malhotra, Divya, Garg, Saurabh, Puri, Suchit, Bhaumik, Suddhasatwa

Abstract

As enterprises modernize legacy monolithic systems to microservices, Large Language Models (LLMs) are heavily utilized for automated code translation. However, traditional vector-based Retrieval-Augmented Generation (Standard RAG) struggles to capture topological relationships. It fetches isolated chunks that sever inheritance chains, leading to high compilation failure rates. This paper introduces a Hierarchical Context-Resident Graph (HCRG) methodology to resolve these limitations. Our pipeline uses tree-sitter for Abstract Syntax Tree (AST) extraction, maps architectural edges into a Google Cloud Spanner Property Graph, and serializes this structure into a Gemini Context Cache for topological, parent-first code translation. We shift evaluation from naive text-overlap to a custom 7-metric Software Engineering framework. Traditional metrics like CodeBLEU (which scored 91% for both methods) effectively masked Standard RAG's structural failures behind syntactically plausible but broken code. Empirically, Graph RAG decisively mitigates dependency loss: API hallucination rates dropped from 56.4% to 16.2%, Dependency Resolution Quality improved from 34.8% to 65.9%, and Parent-Child Consistency rose from 26.7% to 45.5%. However, Graph RAG introduces specific trade-offs. The dense global context causes defensive over-engineering by the LLM, reducing Cyclomatic Complexity Consistency from 71.6% to 46.7%, and slightly degrades Docstring Preservation (67.0% to 61.0%). Ultimately, while trading code complexity for reduced hallucinations, Graph RAG provides a substantially more viable, architecturally sound path for automated enterprise codebase modernization.

Chinese Translation

随着企业将传统的单体系统现代化为微服务，大型语言模型（LLMs）被广泛应用于自动化代码翻译。然而，传统的基于向量的检索增强生成（标准 RAG）在捕捉拓扑关系方面存在困难。它提取的孤立代码块切断了继承链，导致高编译失败率。本文提出了一种层次上下文驻留图（HCRG）方法来解决这些局限性。我们的流程使用 tree-sitter 进行抽象语法树（AST）提取，将架构边映射到 Google Cloud Spanner 属性图，并将该结构序列化为 Gemini 上下文缓存，以实现拓扑优先的代码翻译。我们将评估标准从简单的文本重叠转向自定义的 7 指标软件工程框架。传统指标如 CodeBLEU（两种方法均得分 91%）有效掩盖了标准 RAG 在结构上的失败，导致生成语法上合理但逻辑上破损的代码。实证结果表明，图 RAG 显著减轻了依赖损失：API 幻觉率从 56.4% 降低至 16.2%，依赖解析质量从 34.8% 提升至 65.9%，父子一致性从 26.7% 提高至 45.5%。然而，图 RAG 也引入了特定的权衡。密集的全局上下文导致 LLM 进行防御性过度工程，使圈复杂度一致性从 71.6% 降低至 46.7%，并略微降低了文档字符串保留率（从 67.0% 降至 61.0%）。最终，尽管在减少幻觉的同时牺牲了一定的代码复杂性，图 RAG 提供了一条更为可行、架构上更为合理的自动化企业代码库现代化路径。

View on arXiv Download PDF AI Translation

cs.AI / 36 / 2609.12472

TripPattern: A Pattern-based Text Watermarking Method for Large Language Models

TripPattern：一种基于模式的大语言模型文本水印方法

Moon, Sangjun, Choi, Dasom, Kwon, Jingun, Kamigaito, Hidetaka, Watanabe, Taro, Okumura, Manabu

Abstract

Text watermarking techniques have gained significant attention for identifying machine-generated text and mitigating risks from large language models (LLMs). Existing methods typically divide an LLM's vocabulary into green and red tokens, but encouraging generation toward green tokens can reduce text quality and naturalness. To address this, we propose TripPattern, a watermarking framework that formulates text watermarking as a pattern-based matching task using three vocabulary partitions. TripPattern divides the vocabulary into one neutral group and two pattern groups. During generation, the model alternates token selection between the two pattern groups to embed detectable patterns, while neutral tokens are selected independently to improve flexibility and preserve naturalness. For detection, TripPattern uses pattern-based statistical tests that provide interpretable p-values by measuring how often adjacent tokens alternate between the pattern groups. Theoretical analysis and empirical evaluations on four multilingual datasets show that TripPattern maintains LLM generation quality while achieving robust watermark detectability.

Chinese Translation

文本水印技术在识别机器生成文本和降低大语言模型（LLMs）风险方面引起了广泛关注。现有方法通常将LLM的词汇分为绿色和红色标记，但鼓励生成绿色标记可能会降低文本质量和自然性。为了解决这一问题，我们提出了TripPattern，一种将文本水印化形式化为基于模式的匹配任务的水印框架，使用三种词汇分区。TripPattern将词汇分为一个中性组和两个模式组。在生成过程中，模型在两个模式组之间交替选择标记，以嵌入可检测的模式，同时独立选择中性标记以提高灵活性并保持自然性。在检测方面，TripPattern使用基于模式的统计测试，通过测量相邻标记在模式组之间交替的频率，提供可解释的p值。理论分析和在四个多语言数据集上的实证评估表明，TripPattern在保持LLM生成质量的同时，实现了强大的水印可检测性。

View on arXiv Download PDF AI Translation

cs.AI / 37 / 2609.12482

When Does AI Augment Work? A Workflow-Level Framework for Human-Agent Collaboration

人工智能何时增强工作？人机协作的工作流程级框架

AI Collaboration, Wu, Jiaying, Ziems, Caleb, Chan, Raymond, Chen, Nancy F., Chua, Corlyss, Chung, Gerard, Hahn, Jungpil, Lee, Wee Sun, Liu, Zhengyuan, Ng, Jamie, Ong, Desmond C., Ong, Jeryl, Soon, Da Ren, Song, Tianqi, Tan, Zhi-Xuan, Tao, Sixing, Yang, Emily, Yang, Yajing, Yin, Stella Xin, Kan, Min-Yen, Yang, Diyi

Abstract

We aim to characterise the value of artificial intelligence in the workplace. Current studies largely measure this value in terms of the current automation capabilities and public adoption of AI. However, such metrics ignore the greater impacts of human--agent collaboration in transforming the nature of work. To account for this, we must expand the scope of our analysis beyond atomised tasks of today, and instead focus on how AI can augment entire workflows of the future. To ground this analysis, we establish a precise definition of AI augmentation comprising six conditions, spanning durable net value, meaningful human control, accountability and recovery, and long-term human development through learning, career pathways, and job purpose. We elaborate on these conditions and apply the framework in a case study of AI-mediated social surveys. We conclude by outlining how organisations, researchers, and government leaders can use this framework to make sense of the future of work.

Chinese Translation

我们旨在描述人工智能在工作场所的价值。目前的研究主要通过当前的自动化能力和公众对人工智能的采纳程度来衡量这一价值。然而，这些指标忽视了人机协作在改变工作性质方面的更大影响。为了考虑这一点，我们必须将分析的范围扩展到当今的细化任务之外，而是关注人工智能如何增强未来的整个工作流程。为了使这一分析更为扎实，我们建立了一个精确的人工智能增强定义，包括六个条件，涵盖持久的净价值、有意义的人类控制、问责制与恢复能力，以及通过学习、职业发展和工作目的实现的长期人类发展。我们详细阐述了这些条件，并在一个人工智能介导的社会调查案例研究中应用了该框架。最后，我们概述了组织、研究人员和政府领导者如何利用这一框架来理解工作的未来。

View on arXiv Download PDF AI Translation

cs.AI / 38 / 2609.12489

Confidence-Gated Transductive Test Generation for Code Reranking

基于置信度的传导测试生成用于代码重排序

Lee, Sungjae, Yoon, Youngsik, Song, Seockbean, Wang, Siwei, Chen, Wei, Ok, Jungseul

Abstract

Test case synthesis is crucial for evaluating and ranking programs generated by large language models (LLMs). However, constructing high-quality test cases remains challenging because reliable expected outputs are often difficult to obtain. We propose Confidence-Gated Transductive Test Generation (CoTT), which first uses an efficient inductive procedure and invokes transductive generation only when inductive confidence is low. This adaptive design improves output reliability while allocating extra computation only when needed. On code reranking benchmarks, CoTT outperforms prior baselines across the reported metrics while reducing cost relative to applying transductive generation to every input. These results show that confidence-based allocation of test-time computation provides a favorable efficiency-effectiveness trade-off with a single efficient LLM.

Chinese Translation

测试用例合成对于评估和排名由大型语言模型（LLMs）生成的程序至关重要。然而，构建高质量的测试用例仍然具有挑战性，因为可靠的预期输出往往难以获得。我们提出了基于置信度的传导测试生成（CoTT），该方法首先使用高效的归纳过程，仅在归纳置信度较低时调用传导生成。这种自适应设计提高了输出的可靠性，同时仅在需要时分配额外的计算资源。在代码重排序基准测试中，CoTT在报告的各项指标上均优于先前的基线，同时相较于对每个输入应用传导生成，降低了成本。这些结果表明，基于置信度的测试时间计算分配提供了单个高效LLM的良好效率与效果的权衡。

View on arXiv Download PDF AI Translation

cs.AI / 39 / 2609.12495

Information Specialization and Constrained Synthesis in Multi-Agent LLM Forecasting: A Prospective Live-Study of the 2026 FIFA World Cup

信息专业化与多智能体大语言模型预测中的受限综合：2026年国际足联世界杯的前瞻性实时研究

Varghese, Julian, Bickmann, Lucas, Sandmann, Sarah

Abstract

Large language models are being organized into multi-agent systems with specialized roles, but whether such specialization produces distinct forecasts and whether subsequent synthesis improves utility remains unclear. In this study, we carried out a live, prospective evaluation over the final 56 matches of the information-dense 2026 FIFA World Cup, keeping a frontier foundation model constant while assigning two primary forecasting agents contrasting specialist roles: a quantitative specialist focusing on structured performance statistics and a news specialist focusing on current injuries, tactics and information from press conferences. Their forecasts were then reviewed by a separate critic before being combined by a meta-agent, resulting in a sequential four-agent model. Forecasts from the betting market served as an external benchmark. The news specialist obtained the highest mean probability-weighted Top-3 utility and matched the betting market in Top-3 exact-score hits. Nevertheless, the two specialist forecasters agreed on at least two of the three scorelines in 50 out of 56 matches, and the meta-agent never generated more than one scoreline outside the specialists' forecast set. These findings show that rapidly changing, unstructured information can provide a valuable forecasting signal alongside structured statistics, whereas adding critic and meta-agent stages does not necessarily create complementary information or improve on the strongest specialist.

Chinese Translation

大型语言模型正被组织成具有专业化角色的多智能体系统，但这种专业化是否产生了不同的预测，以及随后的综合是否提高了效用仍不清楚。在本研究中，我们对信息密集的2026年国际足联世界杯最后56场比赛进行了实时前瞻性评估，保持一个前沿基础模型不变，同时为两个主要预测智能体分配了对比鲜明的专业角色：一个定量专家专注于结构化的表现统计，另一个新闻专家则关注当前的伤病、战术和新闻发布会的信息。随后，他们的预测由一个独立的评论员进行审查，然后由一个元智能体进行组合，形成一个顺序的四智能体模型。博彩市场的预测作为外部基准。新闻专家获得了最高的平均概率加权前3效用，并在前3准确比分上与博彩市场相匹配。然而，这两位专业预测者在56场比赛中的50场比赛中至少在三种比分中达成一致，且元智能体从未生成超出专家预测集的比分。这些发现表明，快速变化的非结构化信息可以与结构化统计数据一起提供有价值的预测信号，而增加评论员和元智能体阶段并不一定创造互补信息或改善最强专业者的预测。

View on arXiv Download PDF AI Translation

cs.AI / 40 / 2609.12578

From Collaboration to Capability: Internalizing Routed LLM Experts into Compact Reasoners

从协作到能力：将路由的 LLM 专家内化为紧凑推理器

Nie, Frank, Wang, Shuyao, Liu, Ethan B.

Abstract

A compact controller can coordinate stronger experts by selecting whom to consult, formulating requests, and integrating their responses. We study whether learning from both the controller's decisions and the experts' reasoning and code improves its generation after expert removal. We introduce \textsc{Rivet} for \emph{collaboration internalization}: expert-augmented reinforcement learning applies a shared outcome signal to controller decisions and returned expert spans, and verified trajectory internalization consolidates complete successful interactions through format-aware supervised training. The deployed controller generates reasoning, code, and interaction structure with local Python execution and no external LLM. Across seven competition-mathematics benchmarks, RIVET-1.7B and RIVET-4B achieve average accuracies of $28.25\%$ and $44.16\%$; Stage~II improves RIVET-4B's accuracy after expert removal by $6.49$ points, and GPQA-Diamond results provide evidence of generalization to scientific reasoning. Ablations show gains from ordinary trajectory supervision and additional format weighting, supporting the effectiveness of training on the content and structure of verified collaborations.

Chinese Translation

紧凑控制器可以通过选择咨询对象、制定请求并整合其响应来协调更强的专家。我们研究了在去除专家后，学习控制器的决策以及专家的推理和代码是否能提高其生成效果。我们引入了 extsc{Rivet} 进行 extit{协作内化}：专家增强的强化学习将共享结果信号应用于控制器决策和返回的专家跨度，而验证的轨迹内化通过格式感知的监督训练巩固了完整的成功交互。部署的控制器生成推理、代码和交互结构，能够在本地 Python 执行且不依赖外部 LLM。在七个竞争数学基准测试中，RIVET-1.7B 和 RIVET-4B 的平均准确率分别为 $28.25\%$ 和 $44.16\\%$；第二阶段在去除专家后将 RIVET-4B 的准确率提高了 $6.49$ 个百分点，GPQA-Diamond 结果提供了对科学推理的泛化证据。消融实验表明，普通轨迹监督和额外格式加权带来了收益，支持在验证的协作内容和结构上进行训练的有效性。

View on arXiv Download PDF AI Translation

cs.AI / 41 / 2609.12586

Reproducing and Evaluating the Generalizability of Subliminal Learning in Open-Weight Models

再现与评估开放权重模型中潜意识学习的普适性

van der Weijden, Daan, Brack, Nathan, Santamaria, Selene Baez

Abstract

In this reproduction paper we investigate subliminal learning, a consequence of distillation where teacher models transmit behavioral preference traits through semantically unrelated data. The original paper explores two types of traits (animal preferences and misalignment), three data modalities (number sequences, code, and chain of thought), and several model families. We reproduce their experiments and extend the setup along three axes: new preference categories (actors and politicians), a new task (chess move generation), and an additional open-weight model (Ministral8B). We also run a controlled ablation on the numbers task's answer-space size (1-, 2-, and 3-digit sequences). We focus on open-weight models with accessible checkpoints on HuggingFace, since the original paper's GPT-4.x fine-tuning is no longer available. Our reproduction supports the original paper's claims, but our extensions show they are not universal as transmission strength varies across traits and tasks, and one model shows almost no effect at all.

Chinese Translation

在这篇再现论文中，我们研究了潜意识学习，这是一种蒸馏的结果，其中教师模型通过语义上无关的数据传递行为偏好特征。原始论文探讨了两种特征（动物偏好和不一致性）、三种数据模态（数字序列、代码和思维链）以及几种模型家族。我们再现了他们的实验，并在三个方面扩展了设置：新的偏好类别（演员和政治家）、一个新任务（国际象棋走法生成）以及一个额外的开放权重模型（Ministral8B）。我们还对数字任务的答案空间大小进行了控制消融实验（1位、2位和3位数字序列）。我们专注于在 HuggingFace 上可访问的检查点的开放权重模型，因为原始论文中的 GPT-4.x 微调已不再可用。我们的再现支持了原始论文的主张，但我们的扩展显示这些主张并非普适，因为传递强度在不同特征和任务之间有所不同，并且有一个模型几乎没有任何效果。

View on arXiv Download PDF AI Translation

cs.AI / 42 / 2609.12606

Beyond Generation and Accuracy: Diagnosing and Enhancing Visual Chain-of-Thought for Geometry Problem Solving

超越生成与准确性：诊断与增强几何问题解决中的视觉思维链

Dong, Zhitong, Pan, Jicai, Gao, Yingguo, Ding, Jingting, Chen, Hao, Gu, Jinjie

Abstract

While multimodal reasoning has advanced rapidly, solving complex geometry problems critically hinges on active visual assistance, such as constructing auxiliary lines, spurring the rise of Visual Chain-of-Thought (VCoT). However, existing evaluations typically assess visual generation quality and final answer accuracy in isolation, failing to examine whether intermediate visual aids are geometrically valid, effectively utilized in subsequent reasoning, or causally responsible for task success. To bridge this gap, we introduce GeoVAD-Bench, a diagnostic benchmark that pairs a fine-grained five-dimensional trajectory diagnosis covering perception, auxiliary quality, utilization, deductive reasoning, and final correctness with controlled No-Aux, Auto-Aux, and GT-Aux intervention settings to systematically isolate intermediate error modes, the causal gains of visual aids, and the resulting autonomy gap. Our findings reveal that while high-quality auxiliary aids offer substantial theoretical gains for geometric problem solving, autonomous generation is frequently hampered by compounding errors across geometric perception, faithful visual manipulation, visual-state grounding, and deductive reasoning. Guided by these diagnostic insights, we establish a specialized data construction pipeline encompassing geometric perception, diagram editing, and interleaved visual-textual reasoning trajectories, and develop a progressive SFT and multimodal RL training framework. The resulting model, GeoWeave-8B, outperforms the base model by +25.3% in final geometric accuracy and achieves a +30.4% gain in process average across the four intermediate diagnostic dimensions.

Chinese Translation

尽管多模态推理迅速发展，解决复杂的几何问题仍然严重依赖于主动的视觉辅助，例如构建辅助线，这促使了视觉思维链（Visual Chain-of-Thought, VCoT）的兴起。然而，现有评估通常孤立地评估视觉生成质量和最终答案的准确性，未能检验中间视觉辅助是否在几何上有效、是否在后续推理中得到有效利用，或是否对任务成功具有因果责任。为了解决这一问题，我们引入了GeoVAD-Bench，这是一个诊断基准，结合了细粒度的五维轨迹诊断，涵盖感知、辅助质量、利用、推理和最终正确性，并与受控的无辅助（No-Aux）、自动辅助（Auto-Aux）和真实辅助（GT-Aux）干预设置相结合，以系统性地隔离中间错误模式、视觉辅助的因果增益以及由此产生的自主性差距。我们的研究发现，尽管高质量的辅助工具为几何问题解决提供了显著的理论增益，但自主生成常常受到几何感知、忠实视觉操作、视觉状态基础和推理中的累积错误的阻碍。在这些诊断见解的指导下，我们建立了一个专门的数据构建管道，涵盖几何感知、图示编辑和交错的视觉-文本推理轨迹，并开发了一个渐进式的SFT和多模态强化学习训练框架。最终模型GeoWeave-8B在最终几何准确性上比基础模型提高了25.3%，并在四个中间诊断维度上实现了30.4%的过程平均增益。

View on arXiv Download PDF AI Translation

cs.AI / 43 / 2609.12623

SteerDuplex: Steerable Duplex Speech Dialogue Models

SteerDuplex：可引导的双工语音对话模型

Tyagi, Utkarsh, Selvakumar, Ramaneswaran, Gosai, Advait, Kumar, Sonal, Barhate, Nikhil, Sagar, Isabell, Li, Steven, Bavare, Miheer, Quigley, Daniel, Carrillo, Fabiola Tapia, E, Jose M Patron, Gutiérrez, Diego Macías, Song, Paul, Duraiswami, Ramani, Manocha, Dinesh, He, Yunzhong

Abstract

Full-duplex spoken dialogue models support low-latency turn taking, interruption handling, and backchanneling, yet a key capability remains underexplored: steerability, the ability to reliably shift conversational behavior along attributes such as tone, persona, speaking rate, and voice style in response to user instructions. We introduce a taxonomy of text- and audio-based steerability that identifies substantial gaps in current full-duplex models. To address this gap, we introduce SteerDuplex, a Moshi-based full-duplex speech model fine-tuned on natural conversations and synthetic dialogues targeting instruction following, vocal delivery, reasoning, and duplex interaction. We further apply two-stage reinforcement learning (RL) with hybrid rewards, combining verifiable interaction checks and judge-based semantic feedback to improve timing and response continuity. To evaluate full-duplex spoken steerability, we introduce SteerBench, a benchmark with 390 spoken prompts and 1,067 human-authored binary audio and text rubrics spanning tone, persona, style/accent, and speed/length. On SteerBench, supervised training improves audio-steering average pass rate by 44.5 percentage points over the strongest evaluated open baseline. On Audio MultiChallenge, task average pass rate improves by 7 points over its strongest evaluated open baseline. RL further raises source-clean interruption response from 72.5% to 82.5% and reduces synthetic pause barge-in from 26.5% to 9%. Steering and aggregate task scores remain comparable or higher, while reward probes reveal reward hacking through incomplete responses. Our model and benchmark support systematic research on spoken steerability, with reward analysis showing why timing gains must be evaluated alongside response completeness.

Chinese Translation

全双工语音对话模型支持低延迟的轮流发言、打断处理和反馈交流，但一个关键能力仍未得到充分探索：可引导性，即能够可靠地根据用户指令在语调、角色、说话速度和声音风格等属性上调整对话行为。我们提出了一种基于文本和音频的可引导性分类法，识别当前全双工模型中的重大缺口。为了解决这一缺口，我们引入了SteerDuplex，这是一种基于Moshi的全双工语音模型，经过自然对话和合成对话的微调，旨在实现指令跟随、语音表达、推理和双工交互。我们进一步应用了两阶段强化学习（RL）与混合奖励，结合可验证的交互检查和基于评审的语义反馈，以改善时机和响应连贯性。为了评估全双工语音的可引导性，我们引入了SteerBench，这是一个包含390个语音提示和1,067个由人类创作的二元音频和文本评分标准的基准，涵盖语调、角色、风格/口音和速度/长度。在SteerBench上，监督训练使音频引导的平均通过率比最强评估的开放基线提高了44.5个百分点。在Audio MultiChallenge上，任务的平均通过率比其最强评估的开放基线提高了7个百分点。RL进一步将源清晰的打断响应从72.5%提高到82.5%，并将合成暂停插入从26.5%降低到9%。引导和综合任务得分保持相当或更高，而奖励探测显示通过不完整响应进行奖励操控。我们的模型和基准支持对语音可引导性的系统研究，奖励分析显示时机收益必须与响应完整性一起评估。

View on arXiv Download PDF AI Translation

cs.AI / 44 / 2609.12684

Generative AI Use Cases In Real Estate Marketing: Adoption and Constraints in Germany

生成性人工智能在房地产营销中的应用案例：德国的采用与限制

Kolominsky-Rabas, Victor, Müller, Leopold, Perpina, Felicia, Kühl, Niklas

Abstract

Generative artificial intelligence (GenAI) is changing how work is organized and performed. Real estate marketing is a prime example of this, yet evidence of GenAI in real estate agents' day-to-day practice remains scarce. In this work, we report on our insights from a German-based empirical study with eleven semi-structured interviews. GenAI is already utilized across different activities, with marketing communication being the most prominent. Concrete use cases are emergent and unevenly adopted, with writing expos\'e texts being the only widely established one. Interaction is predominantly human-in-the-loop: GenAI drafts, structures, and retrieves, while real estate agents curate, verify, and decide. Constraints stem less from model capability than from integration with listings and documents, data availability, and compliance in sensitive tasks. The study contributes a grounded map of existing and potential use cases and identifies tentative practical implications for adoption.

Chinese Translation

生成性人工智能（GenAI）正在改变工作组织和执行的方式。房地产营销是这一变化的典型例子，但在房地产经纪人的日常实践中，关于GenAI的证据仍然稀缺。在本研究中，我们报告了基于德国的实证研究的见解，该研究通过十一场半结构化访谈进行。GenAI已经在不同活动中得到应用，其中营销沟通是最为显著的。具体的应用案例正在出现，但采用情况不均，撰写房产说明文本是唯一广泛建立的应用。交互主要是人机协作：GenAI进行草拟、结构化和检索，而房地产经纪人则负责策划、验证和决策。限制因素主要不是来自模型能力，而是与房源和文件的整合、数据可用性以及在敏感任务中的合规性相关。本研究提供了现有和潜在应用案例的基础地图，并识别出对采用的初步实践影响。

View on arXiv Download PDF AI Translation

cs.AI / 45 / 2609.12686

Residual Vector-based Reconstruction as Long-Context Recall Regardless of Context Window Size

基于残差向量的重建方法用于长上下文回忆，无论上下文窗口大小如何

Ryu, MyungHoon, Piao, XinYu, Kim, Jong-Kook

Abstract

Large language models (LLMs) process long contexts, including long documents and lengthy conversations, but face token-level memory usage that increases proportionally to input length. Although model optimization and lossy prompt compression are widely used, these methods still fail to solve the long-context recall problem beyond pretrained and size-constrained context windows. This paper proposes a long-context recall method that maintains near-constant GPU memory usage as context length increases, without additional training. The main idea is to reconstruct facts using parameter activations in the LLM's feed-forward layers, which store residual vectors representing facts from the source document. Utilizing residual vectors allows the LLM to deterministically reconstruct query relevant facts without referencing the original document, preserving high fidelity and reducing memory usage without fine-tuning weights. Experimental results show that the proposed method enables answering single-fact questions in two-million-token story contexts where previous methods fail.

Chinese Translation

大型语言模型（LLMs）处理长上下文，包括长文档和冗长对话，但面临着与输入长度成正比的令牌级内存使用问题。尽管模型优化和有损提示压缩被广泛应用，但这些方法仍未能解决超出预训练和大小受限上下文窗口的长上下文回忆问题。本文提出了一种长上下文回忆方法，该方法在上下文长度增加时保持近乎恒定的GPU内存使用，而无需额外训练。其主要思想是利用LLM的前馈层中的参数激活重建事实，这些激活存储表示源文档中事实的残差向量。利用残差向量使得LLM能够确定性地重建与查询相关的事实，而无需参考原始文档，从而保持高保真度并减少内存使用，而无需微调权重。实验结果表明，所提出的方法能够在两百万令牌的故事上下文中回答单一事实问题，而之前的方法则无法做到。

View on arXiv Download PDF AI Translation

cs.AI / 46 / 2609.12694

I Am AdMan: A Pipeline for Automatic Generation of Personalized Advertising Imagery

我是广告生成器：个性化广告图像自动生成的管道

Kolominsky-Rabas, Victor, Müller, Leopold, Budcke, Claudius, Kühl, Niklas

Abstract

Personalized marketing can increase customer engagement, satisfaction, and conversion. While existing personalization approaches have become effective at matching the right product to the right customer, the visual representation of advertisements remains generic and only weakly tailored to the individual. Prior research shows that generative artificial intelligence can improve the creation of personalized advertisements, particularly for text, and that image generation models can support scalable advertisement production. However, little research has examined how detailed customer information can be systematically translated into fully AI-generated, personalized advertising imagery at scale on a technical level. To address this gap, we propose AdMan, a multi-agent pipeline that transforms customer data into personas, generates personalized advertisement images conditioned on product reference images, and applies an LLM-based judge agent for automated quality control. We implement the pipeline with two different model configurations and evaluate it across four products, using six celebrity personas for qualitative inspection, and 100 real customer profiles, producing 1745 advertisements. The evaluation combines a qualitative expert focus group and a quantitative artifact-rate assessment. The results show that the pipeline can generate photorealistic and personalized advertisements. At the same time, performance varies substantially by product complexity and model configuration. Our findings extend the literature on AI-based personalized advertising by demonstrating the feasibility and current limitations of fully automated image generation for advertising.

Chinese Translation

个性化营销可以提高客户参与度、满意度和转化率。虽然现有的个性化方法在将合适的产品与合适的客户匹配方面已变得有效，但广告的视觉表现仍然较为通用，仅对个体的定制程度较低。先前的研究表明，生成性人工智能可以改善个性化广告的创作，特别是在文本方面，并且图像生成模型可以支持可扩展的广告生产。然而，关于如何在技术层面上系统地将详细的客户信息转化为完全由人工智能生成的个性化广告图像的研究仍然较少。为了解决这一空白，我们提出了AdMan，一个多智能体管道，它将客户数据转化为用户画像，基于产品参考图像生成个性化广告图像，并应用基于大型语言模型（LLM）的评估代理进行自动化质量控制。我们使用两种不同的模型配置实现了该管道，并在四种产品上进行了评估，使用六个名人用户画像进行定性检查，并使用100个真实客户档案生成了1745个广告。评估结合了定性专家焦点小组和定量工件率评估。结果表明，该管道能够生成逼真且个性化的广告。同时，性能因产品复杂性和模型配置的不同而有显著差异。我们的研究结果扩展了基于人工智能的个性化广告文献，展示了完全自动化图像生成在广告中的可行性及当前局限性。

View on arXiv Download PDF AI Translation

cs.AI / 47 / 2609.12697

Enabling and Understanding Personalization in AI-Generated Advertising Imagery

实现与理解AI生成广告图像中的个性化

Kolominsky-Rabas, Victor, Müller, Leopold, Budcke, Claudius, Germelmann, Claas Christian, Kühl, Niklas

Abstract

Personalized marketing traditionally matches static products to customers, while dynamic creative optimization focuses mainly on AI-driven text personalization or basic product image modifications. We address this gap by developing and implementing an AI-based framework that generates personalized advertising imagery directly from customer data. We evaluate this framework in a two-stage within-subject study with N=100 participants across four products and three levels of personalization, varied by the amount and specificity of customer data used. Participants rated each image on attitude toward the advertisement, attitude toward the product, and purchase intention. Results show that participants perceive differences across personalization levels and evaluate AI-generated advertising imagery most positively at a moderate level of personalization. High personalization increases perceived personalization, which is positively associated with all three outcome measures, but also increases perceived creepiness, which is negatively associated with the outcomes and dominates the total effect.

Chinese Translation

个性化营销传统上将静态产品与客户匹配，而动态创意优化主要集中在基于AI的文本个性化或基本产品图像修改上。我们通过开发和实施一个基于AI的框架，直接从客户数据生成个性化广告图像，来填补这一空白。我们在一项包含N=100名参与者的两阶段内被试研究中评估了该框架，研究涉及四种产品和三种个性化水平，个性化水平的变化取决于使用的客户数据的数量和具体性。参与者对每个图像在广告态度、产品态度和购买意图方面进行了评分。结果显示，参与者感知到不同的个性化水平，并在中等个性化水平下对AI生成的广告图像给予了最积极的评价。高个性化水平增加了感知个性化，这与所有三个结果指标呈正相关，但同时也增加了感知的不适感，这与结果呈负相关，并主导了总体效果。

View on arXiv Download PDF AI Translation

cs.AI / 48 / 2609.12704

Implicit Personality Representations in Humans and LLMs

人类与大型语言模型中的隐性人格表征

Geng, Yilin, Abend, Omri, Hovy, Eduard, Frermann, Lea

Abstract

A century of psychology has found that the trait words people use to describe one another vary, but the relational structure among those traits, which ones go together and which oppose, is strikingly consistent across raters and cultures. We test whether the LLM (Qwen 2.5-7B-Instruct) reproduces this structure in its internal trait representations. From millions of crowd-sourced personality ratings of fictional characters, we build a human implicit-personality matrix over hundreds of traits; from contrastive model activations, we build a matching matrix over the same traits. The two relational structures align strongly (Mantel r = 0.77), and the agreement holds trait by trait as well as in aggregate. Two dominant axes of the model's trait representations recover the social and intellectual dimensions long known to organize human personality impressions, social warmth and intellectual competence. On held-out dialogue, projecting model activations onto these directions yields personality profiles that agree with human ratings. This work establishes a framework that enables comprehensive, human-grounded comparison between internal model trait geometry and the shared structure of human personality impressions.

Chinese Translation

一个世纪的心理学研究发现，人们用来描述彼此的特质词汇各不相同，但这些特质之间的关系结构，即哪些特质相互关联，哪些特质相对立，在不同的评估者和文化中却出奇一致。我们测试了大型语言模型（Qwen 2.5-7B-Instruct）是否在其内部特质表征中再现了这一结构。通过对虚构角色的数百万个众包人格评分，我们构建了一个涵盖数百种特质的人类隐性人格矩阵；通过对比模型激活，我们构建了一个相同特质的匹配矩阵。这两个关系结构高度一致（Mantel r = 0.77），并且在逐个特质和整体上均保持一致。模型特质表征的两个主导轴恢复了长期以来被认为组织人类人格印象的社会和智力维度，即社会温暖和智力能力。在保留的对话中，将模型激活投影到这些方向上，得到的人格轮廓与人类评分一致。这项工作建立了一个框架，使得内部模型特质几何与人类人格印象的共享结构之间能够进行全面且以人类为基础的比较。

View on arXiv Download PDF AI Translation

cs.AI / 49 / 2609.12718

When Rubrics Fail: Hallucinations Reveal Blind Spots in Medical AI Evaluation

当评分标准失效时：幻觉揭示医疗人工智能评估中的盲点

Farrow, Griffin, Li, Lily Sijia, Johnson, Jack, Wang, Tingyan, Torr, Philip, Bolton, William, Fehr, Fabio J.

Abstract

Hallucinations can undermine clinician trust in LLMs, making it important that evaluation methods capture clinically relevant errors. Rubric-based evaluation has become the leading approach for assessing LLMs in medicine, but it is unclear whether rubric scores reflect such errors. We first study this in a controlled setting using MedHallu, finding that more specific rubrics better distinguish correct from hallucinated responses. To test this systematically, we develop a taxonomy of medical hallucination types and a clinician-validated error-injection pipeline that creates matched correct and error-injected responses. Across HealthBench, HealthBench Professional, and LiveMedBench, our clinically relevant hallucinations are missed by rubrics, often leaving scores unchanged. We find that rubrics are most effective when explicitly checking facts, and are less effective for additional or unexpected errors they do not anticipate. A preliminary retrieval-based factuality check recovers some of the rubric-blind errors, suggesting a complementary approach. These findings reveal systematic blind spots in current medical evaluation of LLMs and suggest that rubric scores alone are insufficient to establish clinical reliability, potentially undermining clinician trust and confidence in clinical deployment.

Chinese Translation

幻觉可能削弱临床医生对大型语言模型（LLMs）的信任，因此评估方法捕捉临床相关错误显得尤为重要。基于评分标准的评估已成为评估医疗领域LLMs的主要方法，但尚不清楚评分标准的得分是否反映了这些错误。我们首先在受控环境中使用MedHallu进行研究，发现更具体的评分标准能够更好地区分正确响应与幻觉响应。为了系统性地测试这一点，我们开发了一种医疗幻觉类型的分类法和一个经过临床医生验证的错误注入管道，该管道生成匹配的正确响应和注入错误的响应。在HealthBench、HealthBench Professional和LiveMedBench中，我们的临床相关幻觉被评分标准遗漏，通常导致得分不变。我们发现，当明确检查事实时，评分标准最为有效，而对于它们未预见的额外或意外错误则效果较差。初步的基于检索的事实检查能够恢复一些评分标准盲区的错误，暗示了一种互补的方法。这些发现揭示了当前医疗领域LLMs评估中的系统性盲点，并表明仅依靠评分标准得分不足以建立临床可靠性，可能会削弱临床医生的信任和信心，从而影响临床应用。

View on arXiv Download PDF AI Translation

cs.AI / 50 / 2609.12742

Skill Issue: Lessons from Optimizing Repository SKILLs for Coding Agents

技能问题：从优化代码代理的仓库技能中获得的经验教训

Kozyrev, Mykhailo, Kozyrev, Andrei, Podkopaev, Anton

Abstract

Coding agents increasingly read repository knowledge from SKILLs --- plain \texttt{.md} files versioned alongside the code. Recent work synthesizes these files automatically, by optimizing the document against a benchmark. A bare repository comes with no benchmark, and the synthetic tasks prior work builds are small enough that a capable agent saturates them with no document at all. We mine harder tasks --- merged pull requests of the repository, reverted at a single frozen base commit; and score a candidate document by whether the same agent does better with it than without it. On three Kotlin repositories, the documents GEPA finds raise this score by $4.9$pp on average, and the ones SkillOpt finds leave it where it started, $0.1$pp above the seed. The GEPA gain matches what prior work reports with the same optimizer, and at the dataset size a single repository supplies it cannot be separated from the agent's run-to-run variance; settling that would take more tasks than one repository's history yields. The documents themselves read better than the score: a maintainer of one repository found in them knowledge one only gets by working in the project.

Chinese Translation

代码代理越来越多地从 SKILLs 中读取仓库知识——这些是与代码一起版本控制的普通 exttt{.md} 文件。最近的研究通过针对基准优化文档，自动合成这些文件。一个空的仓库没有基准，而之前工作的合成任务小到足以让一个有能力的代理在没有任何文档的情况下就能完成它们。我们挖掘更难的任务——合并的拉取请求，这些请求在一个固定的基础提交上被还原；并通过同一个代理在使用该文档时的表现是否优于不使用该文档来评估候选文档。在三个 Kotlin 仓库中，GEPA 找到的文档平均提高了 $4.9$ 个百分点，而 SkillOpt 找到的文档则保持在起始水平，仅高出 $0.1$ 个百分点。GEPA 的增益与之前工作使用相同优化器所报告的结果相符，而在单个仓库提供的数据集大小下，这一增益无法与代理的运行间方差分开；要解决这一点需要比单个仓库历史所提供的更多任务。这些文档本身的可读性超出了得分：一个仓库的维护者在其中发现了只有在项目中工作才能获得的知识。

View on arXiv Download PDF AI Translation

cs.AI / 51 / 2609.12746

What Drives Recovery in Agentic Text-to-Cypher? LAST-CQ: An LLM Agent Self-Refinement Framework

驱动代理文本到密码的恢复因素？LAST-CQ：一种LLM代理自我精炼框架

Prokopiou, Ioannis, Aidinis, Athanasios, Kyrmpatsos, Panagiotis-Christos, Vikatos, Pantelis

Abstract

Agentic pipelines for structured-query generation are rapidly expanding, but it is unclear which part of the loop produces the gain. We use LAST-CQ -- a five-agent, training-free, execution-grounded Text-to-Cypher framework -- as an instrumented testbed, running three counterfactuals over 2,471 live-database queries and six backbones spanning three vendor scale tiers. Removing correction is worth between 3.1% aggregate execution-BLEU against the single-pass system and 12.3% against a no-refinement counterfactual (up to 80.7% for the weakest backbone). Replacing schema-grounded, LLM-synthesised feedback with raw database error strings costs almost nothing (20.9% vs. 19.9% naive exact match; <0.2% end-to-end; equivalent within $\pm 0.075$ set-F1 by two one-sided tests). Spending the same call budget on parallel sampling degrades quality by 10-11%. What works is detecting failure and routing it to a retry, not the feedback sophistication or number of samples. LAST-CQ itself recovers 91.7% of queries that fail under single-pass generation, while a query that succeeds first time still costs exactly one LLM call. We also show that n-gram overlap on serialised results is not a bound in either direction: it over-scores against set equivalence on 65.9% of results while under-scoring against judged semantics. Finally, we calibrate our LLM judge against blind human labels and find it optimistic by 9 points.

Chinese Translation

结构化查询生成的代理管道正在迅速扩展，但尚不清楚循环中的哪一部分产生了增益。我们使用LAST-CQ——一个五代理、无训练、基于执行的文本到密码框架——作为一个仪器化测试平台，在2471个实时数据库查询和跨越三个供应商规模层级的六个基础架构上运行了三个反事实实验。去除修正的收益在单遍系统中为3.1%的整体执行-BLEU，而在无精炼反事实中为12.3%（对于最弱的基础架构可达80.7%）。用原始数据库错误字符串替代基于模式的、LLM合成的反馈几乎没有成本（20.9%对比19.9%的天真精确匹配；<0.2%的端到端；在两个单侧测试中，等效于$ ext{±}0.075$的集合-F1）。在并行采样上花费相同的调用预算会使质量下降10-11%。有效的做法是检测失败并将其路由到重试，而不是反馈的复杂性或样本数量。LAST-CQ本身能够恢复91.7%在单遍生成下失败的查询，而第一次成功的查询仍然只消耗一个LLM调用。我们还表明，序列化结果上的n-gram重叠在任一方向上都不是界限：在65.9%的结果中，它对集合等价的评分过高，而对判断语义的评分则过低。最后，我们将我们的LLM评估者与盲人类标签进行校准，发现其乐观程度高出9分。

View on arXiv Download PDF AI Translation

cs.AI / 52 / 2609.12747

Assisted Spatial Cognition Through Vision-Language Models

通过视觉-语言模型辅助空间认知

Riaz, H., Fernandez, J. B., Mills, I., Hickey, D., Cleary, F., Ali, M. I.

Abstract

Multimodal AI, powered by Large Language Models (LLMs) and Vision-Language Models (VLMs), is transforming assistive technologies by enabling simultaneous processing of visual and textual data. This advancement holds significant promise for over 43 million visually impaired and neuro-divergent individuals worldwide who face persistent challenges in navigating indoor and outdoor environments due to limited spatial awareness and insufficient environmental cues. Existing navigation aids often lack comprehensive 3D scene understanding, relying on constrained route-based strategies that hinder user autonomy. In this paper, we introduce a novel end-to-end framework that integrates LLMs, VLMs and digital twin technologies to deliver a spatially cognitive navigation support for visually impaired and neuro-divergent users. Our system captures video input via standard mobile phone cameras, and employs SLAM3R to generate dense 3D point clouds from monocular RGB sequences in real-time. Our custom post-processing algorithm ensures accurate point cloud alignment across multiple viewpoints without requiring predefined reference points. This enhances the capabilities of SpatialLM to produce structured 3D representations, including architectural elements and oriented object bounding boxes. The enriched spatial data is then processed by a locally deployed LLM, which interprets 3D contexts to generate detailed scene descriptions and precise distance measurements between users and surrounding objects. We evaluated our approach across diverse video scenarios featuring various perspectives, looped walking views and captured in multiple environments. The evaluation results demonstrate consistent accuracy in 3D scene interpretation and object localisation, underscoring the potential of our system as a transformative assistive navigation solution that combines advanced visual perception with spatial reasoning

Chinese Translation

多模态人工智能（AI），由大型语言模型（LLMs）和视觉-语言模型（VLMs）驱动，正在通过实现视觉和文本数据的同时处理来变革辅助技术。这一进展对全球超过4300万名视力障碍和神经多样性个体具有重要意义，他们在室内和室外环境中由于空间意识有限和环境线索不足而面临持续的导航挑战。现有的导航辅助工具通常缺乏全面的3D场景理解，依赖于受限的基于路径的策略，限制了用户的自主性。在本文中，我们介绍了一种新颖的端到端框架，该框架集成了LLMs、VLMs和数字双胞胎技术，为视力障碍和神经多样性用户提供空间认知导航支持。我们的系统通过标准手机摄像头捕获视频输入，并采用SLAM3R实时生成单目RGB序列的稠密3D点云。我们的自定义后处理算法确保在多个视角下准确对齐点云，而无需预定义参考点。这增强了SpatialLM生成结构化3D表示的能力，包括建筑元素和定向物体边界框。随后，丰富的空间数据由本地部署的LLM处理，该模型解释3D上下文以生成详细的场景描述和用户与周围物体之间的精确距离测量。我们在多种视频场景中评估了我们的方法，这些场景具有不同的视角、循环行走视图，并在多个环境中捕获。评估结果表明，在3D场景解释和物体定位方面具有一致的准确性，突显了我们的系统作为一种变革性辅助导航解决方案的潜力，结合了先进的视觉感知与空间推理。

View on arXiv Download PDF AI Translation

cs.AI / 53 / 2609.12749

SCQ: Stabilizing Conservative Q-Learning with Sigmoid-Bounded Entropy

SCQ：使用 sigmoid 有界熵稳定保守 Q 学习

Wu, Xiefeng, Zhang, Shu, Chu, Zhaojie, Hu, Mingyu

Abstract

Offline-to-online reinforcement learning reduces interaction cost for real-world robot learning but suffers from persistent value estimation instability. Existing methods address this through pessimistic regularization, lower-bound calibration, and architectural normalization, but an overlooked source of instability lies in the entropy formulation: the standard log-entropy term can become negative, destabilizing policy updates. We introduce SCQ (Sigmoid-Bounded Conservative Q-Learning), which replaces this term with a sigmoid-bounded formulation that stays strictly positive. SCQ retains conservative Q regularization and return-based lower-bound calibration, stabilizing policy optimization without sacrificing exploration. We evaluate SCQ on D4RL (Minari) benchmarks under both single-demonstration and standard dataset settings, as well as on simulation and real-world visual tasks. SCQ matches or exceeds baseline performance while exhibiting more stable training dynamics across state-based and visual benchmarks, and transfers to four real-robot platforms including manipulation, wheeled, quadruped, and humanoid systems. A direct clipping intervention that removes negative log-probability contributions, together with gradient-matched positive-score controls, indicates that positivity rather than a particular score shape alone drives much of the improvement. Project website: https://scq-rl.github.io.

Chinese Translation

离线到在线的强化学习减少了现实世界机器人学习的交互成本，但面临持续的价值估计不稳定性。现有方法通过悲观正则化、下界校准和架构归一化来解决这个问题，但一个被忽视的不稳定来源在于熵的公式：标准的对数熵项可能变为负值，从而使策略更新不稳定。我们提出了 SCQ（Sigmoid-Bounded Conservative Q-Learning），它用一个严格为正的 sigmoid 有界公式替代了这个项。SCQ 保留了保守 Q 正则化和基于回报的下界校准，稳定了策略优化而不牺牲探索。我们在 D4RL（Minari）基准上评估了 SCQ，包括单一演示和标准数据集设置，以及在模拟和现实世界的视觉任务上。SCQ 的性能与基线相匹配或超过，同时在基于状态和视觉基准上展现出更稳定的训练动态，并且可以迁移到四种真实机器人平台，包括操控、轮式、四足和类人系统。直接的剪切干预去除了负的对数概率贡献，加上与梯度匹配的正分数控制，表明正性而非特定的分数形状单独驱动了大部分改进。项目网站：https://scq-rl.github.io。

View on arXiv Download PDF AI Translation

cs.AI / 54 / 2609.12769

Unified Agentic Video Editing Across Levels of Complexity and Creativity

跨复杂性和创造性水平的统一自主视频编辑

Nath, Surabhi S., Ferres, Kim, Petrović, Milan, Schulz, Lion

Abstract

Editing is a core component of video production, requiring creative planning and decisions under multiple constraints. Here, we report methods for agentic tooling for automated video editing across three tasks varying in editorial goal, complexity and creativity, namely scene previews, video summaries and cinematic trailers. We evaluate the outputs and discuss implications for automation and agency.

Chinese Translation

编辑是视频制作的核心组成部分，需要在多重约束下进行创造性的规划和决策。在此，我们报告了针对三项任务的自主工具方法，这些任务在编辑目标、复杂性和创造性上各不相同，分别为场景预览、视频摘要和电影预告片。我们评估了输出结果，并讨论了自动化和自主性的影响。

View on arXiv Download PDF AI Translation

cs.AI / 55 / 2609.12771

MPT: Missing Prototype Tracking via Barycentric Reconstruction in Vehicular Federated Learning

MPT：通过重心重建在车辆联邦学习中的缺失原型跟踪

Jang, Hanju, Han, Gyeongmin, Lee, Sungmin, Lee, Kichang, Lee, Chunghan, Ko, JeongGil

Abstract

Cross-vehicle federated learning enables vehicles to collaboratively improve perception models while keeping locally collected driving data private. However, vehicle participation is transient, and a vehicle may depart before training converges while permanently taking its local data. When this departing vehicle holds most samples of a target class, the class becomes rare in the remaining FL network, and its recognition can silently degrade as the shared backbone continues to evolve. Recovering the class is difficult since the few remaining samples provide a noisy prototype estimate, while FL privacy constraints prevent centralized access to raw data or per-sample features. This paper presents MPT, a cross-vehicle FL framework that maintains rare-class recognition by reconstructing its prototype at every round from privacy-preserving class-level statistics. MPT combines a barycentric decomposition that tracks drift shared with remaining-class prototypes, a covariance-based residual prediction that estimates out-of-span drift, and an adaptive calibration that weighs the remaining rare-class samples according to their reliability. We evaluate MPT on three vehicle classification tasks and four backbones against representative calibration and drift-compensation baselines. MPT outperforms all baselines in rare class F1, reaching 0.516 on the nuImages dataset with only 1\% of rare-class samples remaining, without raw data, per-sample features, or retraining.

Chinese Translation

跨车辆联邦学习使车辆能够在保持本地收集的驾驶数据私密的同时，共同改善感知模型。然而，车辆的参与是暂时的，车辆可能在训练收敛之前就离开，同时永久带走其本地数据。当离开的车辆持有目标类别的大部分样本时，该类别在剩余的联邦学习网络中变得稀有，并且其识别可能会随着共享骨干网络的持续演变而悄然下降。恢复该类别是困难的，因为剩余的少量样本提供了噪声原型估计，而联邦学习的隐私约束则阻止了对原始数据或每个样本特征的集中访问。本文提出了MPT，一个跨车辆的联邦学习框架，通过从隐私保护的类别级统计数据中在每一轮重建其原型来维持稀有类别的识别。MPT结合了跟踪与剩余类别原型共享的漂移的重心分解、基于协方差的残差预测以估计超出范围的漂移，以及根据剩余稀有类别样本的可靠性进行加权的自适应校准。我们在三个车辆分类任务和四个骨干网络上评估了MPT，并与代表性的校准和漂移补偿基线进行了比较。MPT在稀有类别F1上超越了所有基线，在nuImages数据集上仅剩1%的稀有类别样本时达到了0.516的成绩，且无需原始数据、每个样本特征或重新训练。

View on arXiv Download PDF AI Translation

cs.AI / 56 / 2609.12801

Interpreting the predictions of neural network classification based on a Taylor Coefficient Analysis (TCA)

基于泰勒系数分析（TCA）解释神经网络分类的预测

Klute, Markus, Monsch, Artur, Sowa, Lars, Wolf, Roger

Abstract

We introduce a rigid and comprehensive taxonomy and paradigm for characterizing the influence of the input feature space $X$ on the predictions $\hat{y}$ of a neural network (NN) used for event classification, based on a Taylor expansion of $\hat{y}$ in $X$. The complete process of introspection we refer to as Taylor Coefficient Analysis (TCA). Based on two simplistic example tasks, which can be easily understood and bencmarked, we illustrate the power of the TCA when it comes to revealing, what properties of $X$ have led to what value of $\hat{y}$, of a given NN model, building up intuition for the method. A more complex application is meant to represent $X$ of a typical classification task at a CERN LHC experiment. Based on this application, we play through the different levels of introspection that the TCA offers and discuss a number of practical aspects for a TCA application typical for the analysis of CERN LHC data. We conclude with a study to support the assumption that those properties of $X$ most relevant for tasks of the complexity typical for a CERN LHC experiment, are usually caught by a TCA up to the second order.

Chinese Translation

我们介绍了一种严格且全面的分类法和范式，用于表征输入特征空间 $X$ 对用于事件分类的神经网络（NN）预测 $ ilde{y}$ 的影响，该方法基于对 $ ilde{y}$ 在 $X$ 中的泰勒展开。我们将这一完整的自省过程称为泰勒系数分析（TCA）。基于两个简单的示例任务，这些任务易于理解和基准测试，我们展示了 TCA 在揭示 $X$ 的哪些属性导致了给定 NN 模型的 $ ilde{y}$ 值方面的强大能力，从而为该方法建立直觉。一个更复杂的应用旨在表示 CERN LHC 实验中典型分类任务的 $X$。基于这一应用，我们逐步演示 TCA 提供的不同自省层次，并讨论一些适用于 CERN LHC 数据分析的 TCA 应用的实际方面。最后，我们通过一项研究支持这样的假设：对于典型的 CERN LHC 实验复杂性任务，$X$ 中最相关的属性通常被 TCA 捕捉到，直到二阶。

View on arXiv Download PDF AI Translation

cs.AI / 57 / 2609.12808

K-Bench: A Benchmark for LLM Unlearning in Agentic Deployments

K-Bench：一种用于代理部署中大语言模型遗忘的基准测试

Yu, Guangsheng, Jiang, Yanna, Wang, Qin, Ma, Baihe, Wang, Xu

Abstract

Unlearning benchmarks such as TOFU and MUSE certify forgetting by reading the model's final answer, where a model that refuses to answer already counts as having forgotten. We show that this model-level certificate does not transfer once the model is deployed as an agent. We introduce K-Bench, a benchmark that scores LLM unlearning under agentic deployment. K-Bench inspects all six channels a ReAct agent exposes, including its chain-of-thought (CoT), tool calls and tool observations, and elicited summary. A query counts as leaked if the secret appears in any of them. Each experiment places the secret in exactly one of the agent's three sources (the weights, the prompt, or the retrieval store). The K-Score is computed separately for each source and credits forgetting only when the agent remains usable. Clearing the answer channel does not make the secret unrecoverable. On structured retrieval, the secret stays verbatim in the tool-observation channel and the aggregate leak rate is unchanged. When the secret lives in the prompt or the retrieval store, TOFU and MUSE report no leakage, while the deployed agent still leaks it on 22--86\% of queries. When the secret is in the weights, none of the twenty evaluated published methods demonstrably removes it, and only an input-corruption intervention reaches selective forgetting under the evaluated observer. The top-ranked method changes across base models. A refusal-tuning method resists the evaluated extraction without verified knowledge removal.

Chinese Translation

遗忘基准如 TOFU 和 MUSE 通过读取模型的最终答案来认证遗忘，其中拒绝回答的模型已经被视为遗忘。我们展示了这种模型级别的认证在模型被部署为代理后并不适用。我们引入了 K-Bench，这是一种在代理部署下评估大语言模型（LLM）遗忘的基准测试。K-Bench 检查 ReAct 代理所暴露的六个渠道，包括其思维链（CoT）、工具调用和工具观察，以及引发的摘要。如果秘密出现在其中任何一个渠道，则查询被视为泄露。每个实验将秘密放置在代理的三个来源中的一个（权重、提示或检索存储）中。K-Score 是针对每个来源单独计算的，仅在代理仍然可用时才计入遗忘。清除答案渠道并不意味着秘密不可恢复。在结构化检索中，秘密在工具观察渠道中保持逐字不变，整体泄露率不变。当秘密存在于提示或检索存储中时，TOFU 和 MUSE 报告没有泄露，而部署的代理在 22% 到 86% 的查询中仍然泄露它。当秘密在权重中时，评估的二十种已发布方法均未能明显移除它，只有输入干扰干预在评估的观察者下达到了选择性遗忘。不同基础模型的排名最高的方法各不相同。一种拒绝调优方法抵御了评估的提取，而没有经过验证的知识移除。

View on arXiv Download PDF AI Translation

cs.AI / 58 / 2609.12822

Scaling Clinical Judgment to Evaluate Medical AI

扩展临床判断以评估医疗人工智能

Buckley, Thomas A., Kanjee, Zahir, Brodeur, Peter G., Crowe, Byron, Pettinato, Anthony M., Shah, Aashna P., Haimovich, Adrian D., McCoy, Liam G., Restrepo, Daniel, Goh, Ethan, Chen, Jonathan H., Zwaan, Laura, Goodman, Katherine E., Morgan, Daniel J., Abdulnour, Raja-Elie E., Rodman, Adam, Manrai, Arjun K.

Abstract

Blinded physician evaluation has been considered by many to be the gold standard for assessing clinical reasoning in large language models (LLMs). This is difficult to scale; thus, prior studies typically rely on small physician panels, often from a single institution or specialty, which both limits the scientific questions investigated and makes it unclear whether findings would be reproduced with a different set of evaluators. To more rigorously and scalably study clinical reasoning in AI models, here we introduce PrecepTron, an LLM fine-tuned for physician-level evaluation of open-ended responses. PrecepTron was trained using low-rank adaptation (LoRA) of a 32-billion-parameter model on a small number of physician examples. We also release GRAND-ROUNDS, a new large-scale physician-annotated benchmark of 9,217 scores by 11 physicians across seven studies. We show that frontier LLMs in typical "LLM-as-a-judge" approaches often disagree with physicians and with each other, but fine-tuning PrecepTron on a small number of cases enables physician-level consistent scoring across tasks. We use PrecepTron to reproduce headline findings from five influential studies assessing LLMs for clinical care in JAMA, Science, and Nature Medicine without new human grading. Using PrecepTron, we then pose new questions about how LLMs reason in medicine that would have been infeasible with human grading alone, including measuring the diagnostic accuracy of frontier LLMs when clinical cases are provided piecemeal, even token by token. Together, PrecepTron and GRAND-ROUNDS provide a foundation for reproducible, large-scale study of how LLMs reason in medicine. All code, data, and labels are made freely available for researchers.

Chinese Translation

盲评医生的评估被许多人视为评估大型语言模型（LLMs）临床推理的金标准。然而，这种方法难以扩展；因此，先前的研究通常依赖于来自单一机构或专业的小型医生小组，这既限制了研究的科学问题，也使得不清楚不同评估者是否会重现相同的发现。为了更严格且可扩展地研究人工智能模型中的临床推理，我们在此介绍了PrecepTron，这是一种针对开放式响应的医生级评估进行微调的大型语言模型。PrecepTron使用低秩适应（LoRA）对一个320亿参数的模型进行了训练，基于少量医生示例。我们还发布了GRAND-ROUNDS，这是一个新的大规模医生标注基准，包含来自七项研究的11名医生的9,217个评分。我们展示了在典型的“LLM作为评审者”方法中，前沿LLMs常常与医生及彼此之间存在分歧，但在少量案例上微调PrecepTron能够实现跨任务的一致评分。我们利用PrecepTron重现了五项影响力研究的主要发现，这些研究评估了LLMs在JAMA、Science和Nature Medicine中的临床护理应用，而无需新的人工评分。使用PrecepTron，我们提出了关于LLMs在医学中推理的新问题，这些问题在仅依赖人工评分的情况下是不可行的，包括在临床案例逐步提供时，甚至逐个标记时，测量前沿LLMs的诊断准确性。总之，PrecepTron和GRAND-ROUNDS为可重复的大规模研究LLMs在医学中推理的方式提供了基础。所有代码、数据和标签均免费提供给研究人员。

View on arXiv Download PDF AI Translation

cs.AI / 59 / 2609.12851

MedRoundsQA: A Persona and Difficulty Aware Evaluation for Multi-Turn Medical Consultations

MedRoundsQA：一种关注角色和难度的多轮医疗咨询评估

Mohamed, Youssef, Heakl, Ahmed, Cui, Qinrong, Liang, Junhong, Ali, Rafiq, Babillie, Bdour, Dunbayeva, Nazira, Gao, Lang, Hussein, Omar, Nada, Ahmed, Mohamed, Ahmed Mohamed Magdy, Liu, Jinghui, Khan, Salman, Razzak, Imran, Wang, Yuxia, Chen, Xiuying

Abstract

Medical benchmarks are dominated by single-turn, multiple-choice clinical cases that poorly reflect real consultations. Practically, clinicians elicit evidence interactively and patient communication varies widely. We introduce MedRoundsQA, a multi-turn diagnostic benchmark derived from 1,387 board-exam cases across 17 specialties. Each case is converted into a structured 24-slot clinical record, and then instantiated as controlled doctor-patient dual-agent dialogues under varying patient personas, with the underlying clinical content held fixed. We further classify cases by difficulty using model-based uncertainty to enable easy-to-hard analysis. Evaluations of fifteen LLM doctor agents show that (i) moving from a single-turn diagnosis on the standardized records to multi-turn consultations causes large degradations of roughly 13-39 points; (ii) more turns reliably improves question relevance, but diagnostic accuracy exhibits diminishing returns and typically plateaus after 6-12 turns; and (iii) patient persona differences can shift diagnosis accuracy by about 7-8 points (lowest to highest education), highlighting equity risks that single-turn benchmarks miss.

Chinese Translation

医学基准测试通常以单轮、多选的临床案例为主，这些案例无法真实反映实际咨询情况。在实际操作中，临床医生通过互动方式获取证据，而患者的沟通方式差异很大。我们引入了MedRoundsQA，这是一个基于来自17个专业的1,387个执业考试案例的多轮诊断基准。每个案例被转换为一个结构化的24槽临床记录，然后在不同的患者角色下实例化为控制的医患双代理对话，同时保持潜在的临床内容不变。我们进一步通过基于模型的不确定性对案例进行难度分类，以便进行从易到难的分析。对十五个大型语言模型（LLM）医生代理的评估表明：（i）从标准化记录的单轮诊断转向多轮咨询导致大约13-39分的显著下降；（ii）更多的轮次可靠地提高了问题的相关性，但诊断准确性呈现递减收益，通常在6-12轮之后趋于平稳；（iii）患者角色差异可以使诊断准确性变化约7-8分（最低到最高教育水平），突显了单轮基准测试所忽视的公平风险。

View on arXiv Download PDF AI Translation

cs.AI / 60 / 2609.12897

Tracing and Coordinating Cross-Layer Influence for Multimodal Model Merging

追踪与协调跨层影响以实现多模态模型合并

Zhou, Pengyang, Tu, Xiaobin, Liu, Zhengxi, Xue, Rongkun, Li, Haochen, Liu, Miancan, Chen, Ziyuan, Wang, Yinggui, Ren, Jinkui, Zhang, Xiantao

Abstract

Multimodal model merging aims to consolidate task experts into a single model that retains their complementary capabilities. Most unimodal model merging methods combine expert updates within individual layers, and multimodal approaches largely follow this design. However, an expert update changes the representations passed to subsequent layers, allowing its influence to propagate across depth and affect how visual and textual information interact. When visual and language updates are combined, later updates act on inputs already modified by earlier ones, coupling their effects. This poses two challenges: (1) how to characterize the multimodal influence of individual expert updates across depth, and (2) how to jointly combine expert updates based on their multimodal influence. To address these challenges, we propose TAC-Merge for tracing and coordinating cross-layer influence in multimodal model merging. It contains two modules, i.e., multimodal influence mapping (MIM) and coupled merge control (CMC). MIM constructs graphs of update effects and uses Ricci curvature together with expert predictions to define a shared fusion objective. CMC models interactions among coefficient adjustments and jointly optimizes regional weights to synthesize one shared model. Experiments across diverse multimodal tasks demonstrate the effectiveness of TAC-Merge in consolidating complementary expert capabilities and supporting generalization to unseen tasks.

Chinese Translation

多模态模型合并旨在将任务专家整合为一个保留其互补能力的单一模型。大多数单模态模型合并方法在各个层内结合专家更新，而多模态方法在很大程度上遵循这一设计。然而，专家更新会改变传递给后续层的表示，从而使其影响在深度上传播，并影响视觉与文本信息的交互。当视觉和语言更新结合时，后续更新作用于已经被早期更新修改的输入，耦合了它们的效果。这带来了两个挑战：（1）如何表征单个专家更新在深度上的多模态影响，以及（2）如何基于其多模态影响共同结合专家更新。为了解决这些挑战，我们提出了TAC-Merge，用于追踪和协调多模态模型合并中的跨层影响。它包含两个模块，即多模态影响映射（MIM）和耦合合并控制（CMC）。MIM构建更新效果的图，并结合Ricci曲率与专家预测来定义一个共享的融合目标。CMC建模系数调整之间的交互，并共同优化区域权重以合成一个共享模型。针对多种多模态任务的实验表明，TAC-Merge在整合互补专家能力和支持对未见任务的泛化方面具有有效性。

View on arXiv Download PDF AI Translation

cs.AI / 61 / 2609.12949

EduFair-Bench: Evaluating Pedagogical Fairness of LLM Tutors Across Student Demographics

EduFair-Bench：评估大型语言模型辅导员在学生人口统计特征中的教学公平性

Zhao, Jiaxu, Radmehr, Bahar, Fawzi, Fares, Nazaretsky, Tanya, Käser, Tanja

Abstract

Large language models (LLMs) are increasingly deployed as tutors, but it is unclear whether they support all students equally well. We introduce \textbf{EduFair-Bench}, a benchmark for auditing the pedagogical fairness of LLM tutors---whether tutoring quality varies systematically with student demographics. EduFair-Bench pairs a multi-domain question bank (mathematics, physics, chemistry) with a controlled simulation in which a fixed LLM student interacts with each tutor across nine demographic levels spanning four dimensions: gender, immigration background, first language, and socioeconomic status (SES). Tutoring quality is scored on five turn-level pedagogical metrics and four conversation-level dimensions, using an LLM judge validated against three-annotator consensus on 180 tutor turns. Bias is measured via paired Wilcoxon signed-rank tests and bootstrap effect-size confidence intervals. Two ablations (demographic cues conveyed through names; conflicting demographic information between tutor and student) disentangle tutor-driven from student-driven bias. Across five tutors, we find that model capability and demographic fairness are largely orthogonal: the smallest model is the most consistent while the four more capable tutors all exhibit wide demographic gaps with no clear capability-to-fairness ordering, pedagogy-specific RL training redistributes rather than removes bias, and language- and immigration-related cues produce larger gaps than gender- and SES-related cues.

Chinese Translation

大型语言模型（LLMs）越来越多地被用作辅导员，但尚不清楚它们是否对所有学生提供同等的支持。我们提出了 extbf{EduFair-Bench}，这是一个用于审计LLM辅导员教学公平性的基准——即辅导质量是否随着学生人口统计特征的变化而系统性地变化。EduFair-Bench结合了一个多领域的问题库（数学、物理、化学）和一个受控模拟，其中一个固定的LLM学生与每个辅导员在九个涵盖四个维度的人口统计水平上进行互动：性别、移民背景、母语和社会经济地位（SES）。辅导质量通过五个轮次级教学指标和四个对话级维度进行评分，使用经过验证的LLM评审与180个辅导轮次的三位注释者共识进行对比。通过配对Wilcoxon符号秩检验和自助法效应大小置信区间来测量偏见。两个消融实验（通过姓名传达的人口统计线索；辅导员与学生之间的冲突人口统计信息）将辅导员驱动的偏见与学生驱动的偏见区分开来。在五个辅导员中，我们发现模型能力与人口统计公平性在很大程度上是正交的：最小的模型是一致性最高的，而四个更强大的辅导员都表现出较大的人口统计差距，没有明确的能力与公平性排序，特定于教学法的强化学习训练重新分配而不是消除偏见，而与语言和移民相关的线索产生的差距大于与性别和SES相关的线索。

View on arXiv Download PDF AI Translation

cs.AI / 62 / 2609.13009

How Good Are Frontier Models at Physics? Expert Re-Grading Reveals Broken Evaluations and Near-Saturation of Leading Benchmarks

前沿模型在物理学中的表现如何？专家重新评估揭示了评估的缺陷和领先基准的近饱和状态

Ansari, Ali, Sun, Haoran, Liu, Andy Zeyi, Jabbour, Mark, Ding, Yongshan, Girvin, Steven, He, Yu, Ismail-Beigi, Sohrab, Kubica, Aleksander, Miller, Owen D., O'Hern, Corey, Ozolins, Vidvuds, Poland, David, Stone, A. Douglas, Bosch, Frank C. van den, Wright, Logan, Akbari, Navid, Antu, Santanu, Cai, Kangle, Calabrese-Day, Andrew, Wuttig, Mateo Cárdenes, Cheng, Meng, Chiang, Barry T., Ghorashi, Ali, Gu, Shouzhen, Huang, Haoyang, Kang, Zhibo, Kienesberger, Lukas, Liu, Hantian, Lomba, Charles, Lu, Zhongling, Ma, Wenchao, McIntosh, Rohin E., McKinney, Evan, Rojkov, Ivan, Sun, Xulei, Tokayer, Yarone Meir, Umasankar, Naveen Balaji, Varma, Mira, Wang, Leda, Wang, Qimin, Wang, Tyler, Wei, Haoyu, Yang, Jinming, Zhao, Jinchen, Zhao, Sherlock Tingrui, Zheng, Qinyuan, Zou, Jay S., Baker, Lucas, Cohan, Arman, Sous, John

Abstract

Low reported scores on leading physics benchmarks, including those featured in the Artificial Analysis Intelligence Index (2026), suggest that frontier language models still struggle with advanced physics, a demanding test of their scientific reasoning and quantitative problem-solving abilities. Yet this impression does not always align with domain experts' experiences using these models in their work. We revisit these reported findings by evaluating frontier models on six widely used physics benchmarks and auditing them with experts, focusing on text-only problems with verifiable final answers. For each subfield of physics, faculty and graduate researchers with relevant expertise carefully review problem statements, reference solutions, and model responses to distinguish genuine model errors from grader errors, incorrect reference solutions, and ambiguous or underspecified questions. Most audited cases initially evaluated as incorrect reflect these benchmarking issues rather than errors in the models' physics reasoning. We then ask experts to address these benchmarking issues by correcting erroneous reference solutions and repairing or excluding flawed questions. We find that GPT-5.6-Sol's measured mean@4 rises from 47.3% to 78.7% on HLE-Physics and from 61.0% to 87.2% on CMT-Benchmark, while its corrected pass@4 reaches 94.4% on the 54 retained CritPt challenges. Corrected scores are computed on the retained evaluation subsets following expert review. Scores on the audited subsets of UGPhysics, PRISM-Physics, and PHYBench also rise substantially after correction. These findings suggest that current benchmarks substantially understate frontier models' ability to solve well-posed physics problems. Near-saturation on these closed-ended tasks highlights the need for more demanding, expert-validated evaluations.

Chinese Translation

在包括《人工分析智能指数》（2026）中列出的领先物理基准上，报告的低分数表明，前沿语言模型在高级物理方面仍然存在困难，这对它们的科学推理和定量问题解决能力提出了严峻的考验。然而，这一印象并不总是与领域专家在工作中使用这些模型的经验相符。我们通过对六个广泛使用的物理基准对前沿模型进行评估，并邀请专家进行审计，重新审视这些报告的发现，重点关注具有可验证最终答案的文本问题。对于每个物理子领域，具有相关专业知识的教师和研究生仔细审查问题陈述、参考解决方案和模型响应，以区分真正的模型错误与评分者错误、错误的参考解决方案以及模糊或不明确的问题。大多数最初评估为不正确的审计案例反映了这些基准测试的问题，而不是模型的物理推理错误。随后，我们请专家通过纠正错误的参考解决方案和修复或排除有缺陷的问题来解决这些基准测试问题。我们发现，GPT-5.6-Sol在HLE-Physics上的测得mean@4从47.3%上升到78.7%，在CMT-Benchmark上的测得mean@4从61.0%上升到87.2%，而其在保留的54个CritPt挑战中的corrected pass@4达到了94.4%。经过专家审查后，纠正的分数是在保留的评估子集上计算的。UGPhysics、PRISM-Physics和PHYBench的审计子集在纠正后分数也显著上升。这些发现表明，当前的基准测试显著低估了前沿模型解决良好设定的物理问题的能力。在这些封闭式任务上的近饱和状态突显了对更具挑战性、经过专家验证的评估的需求。

View on arXiv Download PDF AI Translation

cs.AI / 63 / 2609.13047

Diffusion Models and Concept Formation

扩散模型与概念形成

Wang, Zekun, Singaravadivelan, Karthik, MacLellan, Christopher J.

Abstract

Humans organize knowledge into a taxonomy of concepts with nested levels of abstraction and a \emph{basic level} at which people recognize and name objects with the least cognitive effort. Cobweb is a classic cognitive account of this ability, an incremental learner that builds a probabilistic concept hierarchy by maximizing category utility. We argue that diffusion models, although designed for image synthesis, implicitly perform the same computation. The noisy marginals of a diffusion model are Gaussian smoothings of the data distribution, and the modes of these marginals form a hierarchy that corresponds to a Cobweb tree of probabilistic prototypes in four respects. Both are hierarchical density models, both are hierarchical-Bayesian models with Gaussian prototypes, both treat categorization as score-following that reduces uncertainty, and in both a basic level emerges. We locate this basic level for a diffusion model at an intermediate noise level, where recent analyses show that the reverse process commits to the class identity of a sample. The two models differ mainly in how they represent and learn the taxonomy. Cobweb learns a discrete tree incrementally, whereas a diffusion model encodes a continuous, interpolable hierarchy in a single learned score field fit to the data distribution. We test the correspondence on MNIST and Fashion-MNIST by recovering the diffusion hierarchy through mode-finding and comparing the basic levels of the two models. This reframes diffusion as a cognitive model of concept formation and offers Cobweb a continuous, scalable instantiation.

Chinese Translation

人类将知识组织成一个具有嵌套抽象层次的概念分类法，并在一个 extit{基本层次}上以最小的认知努力识别和命名对象。Cobweb是这一能力的经典认知模型，它通过最大化类别效用来构建一个概率概念层次结构。我们认为，尽管扩散模型是为图像合成而设计的，但它们隐含地执行相同的计算。扩散模型的噪声边际是数据分布的高斯平滑，而这些边际的模态形成一个层次结构，在四个方面对应于Cobweb树的概率原型。两者都是层次密度模型，都是具有高斯原型的层次贝叶斯模型，都将分类视为减少不确定性的评分跟随，并且在两者中都出现了基本层次。我们将扩散模型的基本层次定位在一个中间噪声水平，在该水平下，最近的分析表明反向过程承诺于样本的类别身份。这两种模型的主要区别在于它们如何表示和学习分类法。Cobweb以增量方式学习一个离散树，而扩散模型则在一个适合数据分布的单一学习评分场中编码一个连续的、可插值的层次结构。我们通过模式发现恢复扩散层次，并比较两种模型的基本层次，在MNIST和Fashion-MNIST上测试这种对应关系。这将扩散重新框架为概念形成的认知模型，并为Cobweb提供了一个连续的、可扩展的实例化。

View on arXiv Download PDF AI Translation

cs.AI / 64 / 2609.13062

Anchoring Clinical Events in Time: UID-Preserving Multimodal Reconstruction and Source-Grounded Adjudication

将临床事件锚定于时间：UID保留的多模态重建与源基础裁定

Kumar, Sayantan, Grimaldi, Nicolas, Cummins, Jack, Weiss, Jeremy C.

Abstract

Clinical timelines support treatment-window analysis and leakage-free modeling, but discharge summaries often obscure chronology and structured EHR tables describe only part of the patient course. We present a UID-preserving framework that links each narrative event occurrence to its source span and retains that identity through text-only estimation, structured-evidence retrieval, timestamped source-row grounding, and joint revision. We also present GAVEL, an LLM judge that compares two UID-aligned timelines against the narrative and structured record, to augment prior matching and temporal assessments. Across six open-weight models and 40 mixed-critical-care summaries, the GLM 5.2 multimodal revision, as compared to its text-only variant, improved temporal agreement without reducing event recovery and performed competitively with clinician annotations, while other model revisions showed smaller gains and lower overall performance. Ablations showed that UIDs primarily preserve event retention, whereas source-row linkage supports temporal placement. Blinded human review upheld most GAVEL findings, and controlled adjudication favored multimodal over text-only GLM 5.2 but did not for DeepSeek V3.2. In developing the UID and judge pipeline, we are able to demonstrate 43\% increased event recovery, a framework competitive with clinician annotations, and a system with occurrence-level provenance for both reconstruction and evaluation.

Chinese Translation

临床时间线支持治疗窗口分析和无泄漏建模，但出院总结往往模糊时间顺序，而结构化电子健康记录（EHR）表格仅描述患者过程的一部分。我们提出了一种UID保留框架，将每个叙述事件的发生与其源跨度链接，并通过仅文本估计、结构证据检索、带时间戳的源行基础和联合修订来保持该身份。我们还提出了GAVEL，一个大型语言模型（LLM）裁判，它比较两个UID对齐的时间线与叙述和结构记录，以增强先前的匹配和时间评估。在六个开放权重模型和40个混合重症监护总结中，与其仅文本变体相比，GLM 5.2多模态修订提高了时间一致性，而没有降低事件恢复，并且在与临床医生注释的比较中表现出竞争力，而其他模型修订则显示出较小的增益和较低的整体性能。消融实验表明，UID主要保留事件的保留，而源行链接支持时间定位。盲审的人类评审支持大多数GAVEL的发现，控制裁定偏向于多模态而非仅文本的GLM 5.2，但对DeepSeek V3.2则不然。在开发UID和裁判管道的过程中，我们能够展示43%的事件恢复增加，这是一个与临床医生注释竞争的框架，并且是一个具有事件级来源的系统，适用于重建和评估。

View on arXiv Download PDF AI Translation

cs.AI / 65 / 2609.13073

Autonomous Research for Open-Ended Problems: A Case Study on Telecom Ticket Retrieval

针对开放性问题的自主研究：电信工单检索的案例研究

Min, Junghyun, Uzunalioglu, Huseyin, Trabelsi, Mohamed

Abstract

Recent breakthroughs in LLM-based systems and their abilities in problem solving and coding have allowed progress in the AI for Science paradigm, potentially replacing human roles in machine learning (ML) research. However, while several frameworks of fully autonomous end-to-end ML research have been proposed, successful implementations of them are often limited to problems with narrow search spaces, like language modeling or biomedical ML benchmarks. In this paper, we explore how autonomous research can be adapted to solve open-ended, industry-grade ML problems, by considering a case study: telecom ticket retrieval, an open-ended task with degrees of freedom in representation, architecture, and training data generation. We discover that autonomous research for open-ended problems with commercial and open-source agents shows both promise and limitations: while autonomous research can excel in narrow hyperparameter optimization, it lacks human-like intuition and creativity and requires operational overhead. Even with minimal human supervision, autonomous research can reach $90\%$ of state-of-the-art performance (0.34 vs. 0.38 Recall@1) in a much shorter time period (10 weeks vs. 10 months of human work) at a modest cost (up to \$200 per Cursor campaign). Our empirical evidence recommends that human researchers and autonomous research frameworks work together for best results in ML research.

Chinese Translation

近期基于大语言模型（LLM）系统的突破及其在问题解决和编码方面的能力，推动了科学领域的人工智能（AI）范式的发展，可能会取代人类在机器学习（ML）研究中的角色。然而，尽管已经提出了多个完全自主的端到端机器学习研究框架，但其成功实施往往仅限于具有狭窄搜索空间的问题，如语言建模或生物医学机器学习基准。在本文中，我们探讨了如何将自主研究适应于解决开放性、行业级的机器学习问题，并以电信工单检索为案例研究，这是一项在表示、架构和训练数据生成上具有自由度的开放性任务。我们发现，针对开放性问题的自主研究在商业和开源代理的应用中显示出潜力和局限性：虽然自主研究在狭窄的超参数优化中表现出色，但缺乏人类般的直觉和创造力，并且需要额外的操作开销。即使在最小的人类监督下，自主研究也能在更短的时间内（10周对比人类工作10个月）达到90%的最先进性能（0.34对比0.38 Recall@1），且成本适中（每个Cursor活动最高可达200美元）。我们的实证证据建议，人类研究人员与自主研究框架应协同工作，以在机器学习研究中获得最佳结果。

View on arXiv Download PDF AI Translation

cs.AI / 66 / 2609.13082

Embodied-BenchForge: A Closed-Loop Agentic Workflow for Embodied Benchmark Construction

具身基准锻造：一种闭环代理工作流程用于具身基准构建

Jiang, Baoyang, Zhang, Fengchun, Wang, Leyuan, Li, Haotian, Wang, Yida, Ji, Zhe, Lai, Jinshan, Ren, Xi, Li, Danyang, Yang, Zheng, Hu, Jianwei, Ma, Qiang

Abstract

Agentic systems offer a promising way to automate embodied benchmark construction, but existing approaches typically cover isolated stages or remain specialized to predefined environments and task families. More importantly, multi-step construction produces dependent intermediate artifacts that are often passed downstream without artifact-specific verification, allowing local defects to propagate into the final benchmark. We present Embodied-BenchForge, an agentic framework that transforms user-specified evaluation intents into complete embodied benchmark artifacts. It formulates construction as Closed-Loop Benchmark Synthesis, integrating forward artifact synthesis with backward verification and repair. Skill-Orchestrated Artifact Synthesis composes typed and reusable skills into executable workflows, while an artifact dependency graph records intermediate outputs and their dependencies. Requirement-Guided Verification and Repair applies artifact-specific contracts throughout construction and uses provenance to trigger local re-execution or upstream rollback when verification fails. Embodied-BenchForge constructs six benchmarks covering diverse embodied scenarios in the Offline EQA Track, together with one interactive benchmark containing 220 executable tasks in the Interactive Embodied Track. Evaluations of representative MLLMs and embodied agents show that the benchmarks distinguish model capabilities in both observation-based understanding and closed-loop execution. Quality assessment and ablations validate benchmark quality and the effectiveness of verification and repair, while repair and skill-reuse analyses demonstrate efficient localized recovery and cross-benchmark reusability.

Chinese Translation

代理系统提供了一种自动化具身基准构建的有前景的方法，但现有方法通常仅覆盖孤立阶段或专注于预定义的环境和任务家族。更重要的是，多步骤构建产生的中间工件相互依赖，通常在没有特定工件验证的情况下传递到下游，导致局部缺陷传播到最终基准。我们提出了具身基准锻造（Embodied-BenchForge），这是一个代理框架，将用户指定的评估意图转化为完整的具身基准工件。它将构建形式化为闭环基准合成（Closed-Loop Benchmark Synthesis），将前向工件合成与后向验证和修复相结合。技能协调工件合成（Skill-Orchestrated Artifact Synthesis）将类型化和可重用的技能组合成可执行的工作流程，而工件依赖图记录中间输出及其依赖关系。需求引导的验证与修复（Requirement-Guided Verification and Repair）在整个构建过程中应用特定于工件的契约，并利用来源信息在验证失败时触发局部重新执行或上游回滚。具身基准锻造构建了六个涵盖离线EQA轨道中多样化具身场景的基准，以及一个包含220个可执行任务的交互式基准，位于交互式具身轨道中。对代表性MLLMs和具身代理的评估表明，这些基准能够区分模型在基于观察的理解和闭环执行中的能力。质量评估和消融实验验证了基准质量及验证与修复的有效性，而修复和技能重用分析则展示了高效的局部恢复和跨基准的可重用性。

View on arXiv Download PDF AI Translation

cs.AI / 67 / 2609.13118

CMA-OT: Hierarchical Expert Supervision for Dance-to-Music Generation

CMA-OT：用于舞蹈与音乐生成的层次专家监督

Wang, Jinting, Li, Chenxing, Yu, Dong, Liu, Li

Abstract

Dance-to-music (D2M) generation aims to synthesize music that is rhythmically and stylistically aligned with dance videos. A key challenge arises from the semantic mismatch between sparse dance cues, such as rhythm and style, and the dense information required for music composition, including structure, instrumentation, and expressive dynamics. Existing methods typically rely on these sparse cues and supervise only the final audio output, resulting in poorly learned music representations and generated music with limited musicality and structural coherence. To address these issues, we propose Curriculum-guided Multi-scale representation Alignment with scale-aware Optimal Transport (CMA-OT), a novel paradigm that leverages an external music expert to provide hierarchical supervision for the generator's latent features, bridging the semantic gap and enhancing representation learning. To effectively incorporate hierarchical supervision, we introduce a curriculum-guided multi-scale learning strategy that progressively transfers musical knowledge from the expert to the music generator, enabling stable and effective representation learning. Moreover, to accommodate the semantic and structural variations across different expert scales and achieve fine-grained alignment under temporal mismatch, we propose a scale-aware optimal transport alignment mechanism, which models soft correspondences between hierarchical expert representations and the generator's latent features. Extensive experiments on two datasets demonstrate that CMA-OT achieves state-of-the-art performance in rhythmic synchronization, perceptual quality, and overall music generation.

Chinese Translation

舞蹈与音乐（D2M）生成旨在合成与舞蹈视频在节奏和风格上协调一致的音乐。一个关键挑战在于稀疏舞蹈线索（如节奏和风格）与音乐创作所需的密集信息（包括结构、乐器编制和表现动态）之间的语义不匹配。现有方法通常依赖于这些稀疏线索，并仅监督最终的音频输出，导致音乐表示学习不佳，生成的音乐在音乐性和结构一致性方面有限。为了解决这些问题，我们提出了一种基于课程指导的多尺度表示对齐方法，结合了规模感知的最优传输（CMA-OT），这是一种新颖的范式，利用外部音乐专家为生成器的潜在特征提供层次监督，弥合语义差距并增强表示学习。为了有效地整合层次监督，我们引入了一种课程指导的多尺度学习策略，逐步将音乐知识从专家转移到音乐生成器，从而实现稳定和有效的表示学习。此外，为了适应不同专家尺度之间的语义和结构变化，并在时间不匹配的情况下实现细粒度对齐，我们提出了一种规模感知的最优传输对齐机制，该机制建模层次专家表示与生成器潜在特征之间的软对应关系。在两个数据集上的大量实验表明，CMA-OT在节奏同步、感知质量和整体音乐生成方面达到了最先进的性能。

View on arXiv Download PDF AI Translation

cs.AI / 68 / 2609.13125

A Hybrid LSTM-XGBoost Framework for Multi-Horizon Stock Return Prediction Across Diversified Equity Portfolios

一种混合LSTM-XGBoost框架用于多时间跨度的股票收益预测跨多元化股票投资组合

Mostafa, Seif ElDein, Ahmed, Yahia, Datwish, Farah, Solayman, Marwa

Abstract

Accurate prediction of equity returns remains a major challenge in computational finance due to the non-stationary, nonlinear, and low signal-to-noise ratio nature of financial time series. This paper proposes a hybrid two-stage architecture that combines a long short-term memory (LSTM) network with an XGBoost gradient-boosted regressor for multi-horizon stock return prediction across a diversified panel of 14 U.S. equities spanning six industry sectors. The LSTM component, comprising two stacked layers with 64 hidden units, processes 60-day sliding windows of five sequential market features to produce 64-dimensional temporal embeddings that encode learned sequential market dynamics. These embeddings are concatenated with 14 hand-crafted technical indicators to form a 78-dimensional hybrid feature vector, which is subsequently passed to an XGBoost regressor tuned via 3-fold cross-validation grid search. The framework is trained on a multi-stock pooled corpus using strict chronological splits and per-stock MinMaxScaling to prevent look-ahead bias, and evaluated across four prediction horizons of 30, 90, 252, and 365 trading days. Experimental results demonstrate that the hybrid model achieves a test RMSE of 0.0949 on the 30-day horizon, roughly one-third that of the standalone LSTM baseline, while marginally matching or surpassing the XGBoost-Only baseline across the majority of stocks. Directional accuracy rises with horizon length, reaching 97.6% at 365 days; we show, however, that this largely tracks the high base rate of positive long-horizon returns in the sample, and we therefore benchmark directional accuracy against a naive always-positive predictor and treat the above-base-rate gap at short horizons as the more informative signal. A composite investment scoring framework derived from multi-horizon predictions is further proposed to support portfolio ranking and decision support.

Chinese Translation

由于金融时间序列的非平稳性、非线性和低信噪比特性，股票收益的准确预测仍然是计算金融领域的一大挑战。本文提出了一种混合的两阶段架构，将长短期记忆（LSTM）网络与XGBoost梯度提升回归器结合，用于对涵盖六个行业部门的14只美国股票进行多时间跨度的股票收益预测。LSTM组件由两个堆叠的层和64个隐藏单元组成，处理60天滑动窗口的五个连续市场特征，生成64维的时间嵌入，编码学习到的序列市场动态。这些嵌入与14个手工制作的技术指标连接形成一个78维的混合特征向量，随后传递给通过3折交叉验证网格搜索调优的XGBoost回归器。该框架在一个多股票汇总语料库上进行训练，采用严格的时间顺序分割和每只股票的MinMaxScaling，以防止前瞻性偏差，并在30、90、252和365个交易日的四个预测时间跨度上进行评估。实验结果表明，该混合模型在30天时间跨度上实现了0.0949的测试均方根误差（RMSE），大约是独立LSTM基线的三分之一，同时在大多数股票中略微匹配或超越了仅XGBoost基线。方向准确率随着时间跨度的增加而上升，在365天时达到97.6%；然而，我们显示这在很大程度上与样本中长期正收益的高基率相关，因此我们将方向准确率与一个简单的始终正预测器进行基准比较，并将短时间跨度的高于基率差距视为更具信息量的信号。此外，基于多时间跨度预测的复合投资评分框架被进一步提出，以支持投资组合排名和决策支持。

View on arXiv Download PDF AI Translation

cs.AI / 69 / 2609.13134

Rethinking Heterogeneous System Disaggregation for Subquadratic Attention

重新思考异构系统的去聚合以实现亚二次注意力

Tschand, Arya, Fu, Yaosheng, Mailthody, Vikram Sharma, Oswald, Nicolai, Tsai, Po-An, Zhao, Ritchie, Villa, Oreste, Reddi, Vijay Janapa, Sankaralingam, Karu

Abstract

Frontier language models are more aggressively using subquadratic attention to reduce the memory footprint and compute requirements during inference while still delivering frontier accuracy. While existing systems make dense attention-centric disaggregated serving decisions, we show that disaggregating inference around the unique arithmetic intensity and memory footprint of subquadratic attention LLMs can achieve significant throughput and energy efficiency gains on emerging DRAM-based and SRAM-only heterogeneous systems. We introduce SQD (SubQuadratic Disaggregation), a fine-grained heterogeneous disaggregation scheme that splits decode by quadratic and subquadratic attention rather than by operator type, and that applies across subquadratic attention variants. For sparse attention LLMs, we disaggregate decode into top-k selection, which must index through the full KV, and top-k attention plus FFN, which have static memory footprints. For linear and sliding-window attention LLMs, we disaggregate decode into dense attention layers and subquadratic attention layers plus FFN. In an adjusted 8xB200 heterogeneous system proxy, we observe average tokens/J improvements of 53% on GLM 5.2, 31% on Nemotron 3 Ultra, and 56% on Gemma 4 31B over the strongest GPU-only baselines. In an analytical model of a Rubin plus LPX system with fixed power budgets, we observe 1.2x to 1.5x tighter achievable latencies and up to 3.6x higher throughput over the best baseline of attention-FFN disaggregation. Our experiments also reveal architectural insights on chip and interconnect provisioning for next-generation heterogeneous systems serving subquadratic attention.

Chinese Translation

前沿语言模型更积极地使用亚二次注意力，以减少推理过程中的内存占用和计算需求，同时仍能提供前沿的准确性。尽管现有系统做出以密集注意力为中心的去聚合服务决策，我们展示了围绕亚二次注意力大语言模型（LLMs）的独特算术强度和内存占用进行去聚合推理，可以在新兴的基于DRAM和仅基于SRAM的异构系统上实现显著的吞吐量和能效提升。我们引入了SQD（SubQuadratic Disaggregation），一种细粒度的异构去聚合方案，它根据二次和亚二次注意力而不是操作类型来拆分解码，并适用于亚二次注意力的各种变体。对于稀疏注意力的LLMs，我们将解码去聚合为top-k选择，这必须遍历完整的键值（KV），以及top-k注意力加前馈网络（FFN），这具有静态内存占用。对于线性和滑动窗口注意力的LLMs，我们将解码去聚合为密集注意力层和亚二次注意力层加FFN。在一个调整后的8xB200异构系统代理中，我们观察到GLM 5.2的平均tokens/J提升了53%，Nemotron 3 Ultra提升了31%，Gemma 4 31B提升了56%，相较于最强的仅GPU基线。在一个具有固定功率预算的Rubin加LPX系统的分析模型中，我们观察到可实现的延迟收敛性提高了1.2倍到1.5倍，吞吐量提高了最高3.6倍，相较于最佳的注意力-FFN去聚合基线。我们的实验还揭示了下一代异构系统在提供亚二次注意力服务时的芯片和互连配置的架构洞察。

View on arXiv Download PDF AI Translation

计算语言学 (Computation and Language)

cs.CL / 1 / 2609.11955

R2VC: Modular Fact-Checking with Retrieval, Verification, and Confidence Calibration

R2VC：具有检索、验证和置信度校准的模块化事实核查

Dixit, Dhruv, Pandey, Paritosh

Abstract

Large language models are increasingly used for automated fact checking, but end-to-end prompting often entangles evidence retrieval, reasoning, and uncertainty estimation, making failures difficult to diagnose and confidence difficult to trust. We present R2VC, a modular retrieve, reason, verify, calibrate architecture for evidence-grounded fact checking with citations and abstention. R2VC combines hybrid sparse+dense retrieval over Wikipedia, a supervised fine-tuned and DPO-aligned generator that produces diverse structured verdict candidates, an external NLI cross-encoder for evidence-based candidate selection, and a lightweight sequence-level calibrator for confidence estimation and selective abstention. On FEVER, an 8B backbone with R2VC achieves 13.74% higher accuracy than baseline. Ablation studies show that verifier-based candidate selection and confidence calibration are the largest contributors to performance. Removing candidate selection drops FEVER accuracy to 76.24%, while removing calibration nearly doubles the Brier score to 0.161. A manual analysis of 250 errors further shows that retrieval failures, especially wrong-entity evidence, remain the dominant bottleneck. Together, these results show that modular fact-checking pipelines can substantially improve both predictive accuracy and confidence reliability in open-domain verification.

Chinese Translation

大型语言模型越来越多地用于自动化事实核查，但端到端提示往往将证据检索、推理和不确定性估计纠缠在一起，使得故障难以诊断，置信度难以信任。我们提出了R2VC，一种模块化的检索、推理、验证和校准架构，用于基于证据的事实核查，支持引用和选择性放弃。R2VC结合了对维基百科的混合稀疏+密集检索、经过监督微调和与DPO对齐的生成器，能够生成多样化的结构化裁决候选、一个用于基于证据的候选选择的外部NLI交叉编码器，以及一个轻量级的序列级校准器，用于置信度估计和选择性放弃。在FEVER数据集上，使用R2VC的8B主干模型的准确率比基线提高了13.74%。消融研究表明，基于验证器的候选选择和置信度校准是性能的最大贡献者。移除候选选择使FEVER的准确率降至76.24%，而移除校准几乎使Brier分数翻倍至0.161。对250个错误的手动分析进一步显示，检索失败，特别是错误实体证据，仍然是主要瓶颈。综合来看，这些结果表明，模块化的事实核查流程可以显著提高开放域验证中的预测准确性和置信度可靠性。

View on arXiv Download PDF AI Translation

cs.CL / 2 / 2609.12085

What Counts as a Mistake? Annotating Recitation Events in Quran Memorization Transcripts

什么算作错误？在古兰经背诵记录中标注背诵事件

Mdfaa, Mohamad Al, Askarbekuly, Nursultan, Helaly, Ahmed, Sandouk, Ubai, Mazzara, Manuel

Abstract

Checking Quran recitation from an ASR transcript requires distinguishing unresolved mistakes from repetitions, repairs, opening formulas and accepted spelling differences. We report a completed human annotation of 100 production recording cases: 348 scored units and 162 localized events across ten combined labels. An executable evaluator scores labels and word positions together. A plain diff reaches label-aware F1 0.525 and localization F1 0.826; adapted production cleaner/alignment components reach 0.518 and 0.786, with exact-span F1 0.505 for both. Correcting the adapter's word coordinates recovers all five annotated repetition events, showing why annotation interfaces must be checked before interpreting baseline failures. In a preliminary pilot, eight single 20-minute runs across three coding agents and eight models span label-aware F1 0.143 to 0.892: seven land far above every baseline, and one collapses below the naive diff from a missing normalization step. Across the six, 970 of 972 gold-event instances draw an overlapping prediction, so what remains is not detection but convention: span extent, and the labels whose boundary is stipulated by adjudication rather than visible in the text. Seven of 162 events defeat all six same-day runs, five of them one orthographic rule, and the strongest run still misses the same ones. No run annotated before building, so the pilot measures the algorithm half of the task only.

Chinese Translation

检查来自自动语音识别（ASR）转录的古兰经背诵需要区分未解决的错误、重复、修正、开场公式和接受的拼写差异。我们报告了对100个生产录音案例的人工标注完成情况：348个评分单元和162个跨十个组合标签的局部事件。一个可执行的评估器将标签和单词位置一起评分。普通差异（plain diff）达到标签感知F1值0.525和定位F1值0.826；改进的生产清理/对齐组件分别达到0.518和0.786，确切跨度F1值均为0.505。纠正适配器的单词坐标恢复了所有五个标注的重复事件，显示了为什么在解释基线失败之前必须检查标注接口。在一项初步试点中，三个编码代理和八个模型的八次单次20分钟运行的标签感知F1值范围为0.143到0.892：七个运行远高于每个基线，而一个因缺少归一化步骤而低于简单差异（naive diff）。在六个运行中，970个金事件实例中的972个产生了重叠预测，因此剩下的不是检测而是约定：跨度范围，以及其边界由裁决而非文本中可见的标签。162个事件中的七个击败了所有六个同日运行，其中五个是一个正字法规则，而最强的运行仍然错过了相同的事件。没有运行在构建之前进行标注，因此该试点仅测量了算法任务的一半。

View on arXiv Download PDF AI Translation

cs.CL / 3 / 2609.12107

Extracting Dataset Mentions in Forced Displacement and FCV Documents: A Weakly Supervised Framework with LLM-Based Label Refinement

在强迫迁移和脆弱、冲突与暴力（FCV）文献中提取数据集提及：一种基于弱监督的框架与大型语言模型（LLM）标签精炼

Macalaba, Rafael, Solatorio, Aivin V., Brock, Patrick Michael, Dupriez, Olivier

Abstract

Development and humanitarian organizations produce and support surveys, administrative registries, and other data resources to inform research, policy, and operations, yet systematically identifying where these datasets are referenced remains difficult. Such references are dispersed across research papers, project documents, humanitarian reports, and other unstructured text, limiting both the ability to trace data use and to identify potential gaps in data availability or dissemination. We present a weakly supervised framework for adapting dataset extraction to forced displacement and Fragile, Conflict, and Violence (FCV) documents without first constructing a large manually labeled training corpus. A lightweight model trained on general research literature generates candidate dataset mentions from unlabeled domain documents, which a frontier large language model (LLM) reviews in context, validating or rejecting candidates and correcting their extraction boundaries. The resulting annotations are supplemented with targeted synthetic and contrastive examples and used to fine-tune the lightweight model for large-scale extraction. We evaluate the resulting model on an independent gold-standard benchmark of 1,706 text passages spanning research, humanitarian, and operational documents. Across the full benchmark, the model achieves 74.1\% precision and 70.5\% recall at the mention level; among passages containing dataset references, precision reaches 89.5\%. At the passage level, the model achieves 88.2\% accuracy and 88.6\% specificity in distinguishing passages with dataset references from those without them. These results demonstrate a practical approach for constructing domain-specific supervision when labeled data are limited, and provide a technical foundation for larger-scale analysis of data use and potential gaps in the displacement data landscape.

Chinese Translation

发展和人道主义组织制作并支持调查、行政登记和其他数据资源，以为研究、政策和操作提供信息，但系统性识别这些数据集的引用仍然困难。这些引用分散在研究论文、项目文件、人道主义报告和其他非结构化文本中，限制了追踪数据使用的能力以及识别数据可用性或传播潜在缺口的能力。我们提出了一种弱监督框架，用于将数据集提取适应于强迫迁移和脆弱、冲突与暴力（FCV）文献，而无需首先构建一个大型手动标注的训练语料库。一个在一般研究文献上训练的轻量模型从未标记的领域文档中生成候选数据集提及，前沿的大型语言模型（LLM）在上下文中审查这些候选项，验证或拒绝候选项并纠正其提取边界。生成的注释通过有针对性的合成和对比示例进行补充，并用于微调轻量模型以进行大规模提取。我们在一个独立的金标准基准上评估了生成的模型，该基准包含1,706个文本段落，涵盖研究、人道主义和操作文档。在整个基准中，该模型在提及级别上实现了74.1%的精确率和70.5%的召回率；在包含数据集引用的段落中，精确率达到89.5%。在段落级别上，该模型在区分包含数据集引用的段落与不包含的段落时，准确率达到88.2%，特异性达到88.6%。这些结果展示了一种在标注数据有限时构建领域特定监督的实用方法，并为更大规模的数据使用分析和识别迁移数据领域潜在缺口提供了技术基础。

View on arXiv Download PDF AI Translation

cs.CL / 4 / 2609.12111

The Cost of Compression: A Rate-Distortion Limit on Factual Hallucination

压缩的代价：对事实幻觉的速率-失真限制

Wang, Xi, Xu, Shijia, Guo, Rongfeng

Abstract

Factual hallucination in closed-book question answering is often treated as a coverage problem: a model fails because the relevant fact is absent from its internal memory. This view misses a second source of error. Even when a fact has been observed, finite memory may force it to be stored only approximately. We study this effect through a simple coverage--compression model of factual recall. We consider an unstructured question-answering task with $N$ possible queries and $K$ possible answers. A learner observes $M$ training facts, compresses them into at most $B$ bits, and answers uniformly drawn test queries without retrieval. For a uniformly random ground-truth mapping, we prove $\mathcal{E} \geq \frac{M}{N}\delta^\star\!\left(\frac{B}{M}\right) + \left(1-\frac{M}{N}\right)\left(1-\frac{1}{K}\right)$, where $\delta^\star(r)$ is the inverse rate-distortion function of a uniform $K$-ary source under zero-one loss. The two terms separate compression distortion on observed facts from missing coverage on unobserved facts. The bound gives a compact way to reason about selective memory, forced compression, structure, retrieval, abstention, and long-context organization. We study the predicted signatures with theory-implied simulations and controlled fact-injection probes in modern language models that vary fact load and effective trainable memory. The result is not a complete theory of hallucination, but an information-theoretic account of a separable failure mode: lossy recall of observed facts under finite memory.

Chinese Translation

在闭卷问答中，事实幻觉通常被视为覆盖问题：模型失败是因为相关事实在其内部记忆中缺失。这种观点忽视了第二种错误来源。即便一个事实已经被观察到，有限的记忆可能迫使其仅以近似方式存储。我们通过一个简单的覆盖-压缩模型研究这一效应，该模型用于事实回忆。我们考虑一个无结构的问答任务，其中有 $N$ 个可能的查询和 $K$ 个可能的答案。学习者观察到 $M$ 个训练事实，将其压缩为最多 $B$ 位，并在不进行检索的情况下回答均匀抽取的测试查询。对于均匀随机的真实映射，我们证明 $ extmath{E} geq rac{M}{N} extmath{oldsymbol{eta}}^ extmath{*} extmath{igg(}rac{B}{M} extmath{igg)} + igg(1-rac{M}{N}igg)igg(1-rac{1}{K}igg)$，其中 $ extmath{oldsymbol{eta}}^ extmath{*}(r)$ 是在零一损失下均匀 $K$ 进制源的逆速率-失真函数。这两个项将观察到的事实的压缩失真与未观察到的事实的缺失覆盖分开。该界限提供了一种紧凑的方式来推理选择性记忆、强制压缩、结构、检索、弃权和长上下文组织。我们通过理论推导的模拟和在现代语言模型中控制事实注入探针的预测特征进行研究，这些模型在事实负载和有效可训练记忆方面有所不同。结果并不是对幻觉的完整理论，而是对一种可分离失败模式的信息论解释：在有限记忆下对观察到的事实的有损回忆。

View on arXiv Download PDF AI Translation

cs.CL / 5 / 2609.12122

Quantifying Consonant Contributions to Word Intelligibility via Acoustic Masking

通过声学掩蔽量化辅音对词语可懂性的贡献

Yeo, Eunjung, Choi, Kwanghee, Kothadia, Krupaben, Berisha, Visar, Liss, Julie M., Mortensen, David R., Harwath, David

Abstract

Consonants contribute unequally to whether a word is understood. Given the limited time available for therapy, ranking consonants by contribution to intelligibility helps prioritize intervention targets in motor speech disorders. However, measuring this contribution relies on perceptual studies that are difficult to scale. This paper presents a scalable method that measures consonant contribution using acoustic masking. We silence one consonant at a time in an isolated word and test whether an automatic speech recognition (ASR) model still recognizes the word. We define a consonant's contribution score as the proportion of its masked instances for which the word becomes misrecognized, which we refer to as the mask-induced misrecognition rate (MMR). We validate MMR against two linguistic factors previously reported to correlate with consonant contribution, namely phoneme frequency and functional load. We apply this analysis across four languages, English, Spanish, German, and Czech, using three ASR architectures, MMS (encoder-only), Whisper (encoder-decoder), and Qwen3-ASR (LLM-based). Using partial Spearman correlations, we find that phoneme frequency correlates negatively with MMR while functional load correlates positively. In other words, more frequent consonants are less disruptive when masked, whereas consonants carrying more lexical contrast are more disruptive. Further cross-language analysis shows that consonant rankings are not consistent, indicating that consonant contribution is language-dependent.

Chinese Translation

辅音对一个词是否被理解的贡献是不均等的。鉴于治疗时间有限，根据辅音对可懂性的贡献进行排名有助于优先确定运动语言障碍的干预目标。然而，测量这种贡献依赖于难以扩展的感知研究。本文提出了一种可扩展的方法，通过声学掩蔽来测量辅音的贡献。我们在一个孤立的词中一次性静音一个辅音，并测试自动语音识别（ASR）模型是否仍然能识别该词。我们将辅音的贡献分数定义为其被掩蔽实例中导致词语被错误识别的比例，我们称之为掩蔽引起的错误识别率（MMR）。我们将MMR与两个先前报告与辅音贡献相关的语言因素进行验证，即音素频率和功能负载。我们在英语、西班牙语、德语和捷克语四种语言中应用这一分析，使用三种ASR架构，分别为MMS（仅编码器）、Whisper（编码器-解码器）和Qwen3-ASR（基于LLM）。通过部分斯皮尔曼相关分析，我们发现音素频率与MMR呈负相关，而功能负载与MMR呈正相关。换句话说，更频繁的辅音在被掩蔽时干扰较小，而承载更多词汇对比的辅音则干扰较大。进一步的跨语言分析表明，辅音排名并不一致，表明辅音的贡献是语言依赖的。

View on arXiv Download PDF AI Translation

cs.CL / 6 / 2609.12127

Local Edits, Global Ripples: Replay-Informed Policy Adaptation for Workflow Synthesis

局部编辑，全球涟漪：基于重放的信息的工作流合成策略适应

Mao, Manqing, Wang, Hong, Koelle, Samson, Yuan, Jie, Wang, Zhuoer, Feng, James, Lin, Yanjun, Edmiston, Daniel, Kuang, Nikki Lijing, Sheng, Zhecheng, Niu, Wei

Abstract

Prompt-policy editing offers a practical way to improve agents that synthesize executable workflows without updating the underlying model. However, persistent prompt editing has two coupled properties. First, edit locality does not imply effect locality: an edit confined to one policy segment can ripple through downstream execution, altering behavior beyond the edited segment. Second, edit effects are composition-sensitive: edits that work in isolation can interfere after composition, causing one or both to lose their benefit or become harmful. Persistent adaptation must therefore support two distinct decisions: identifying where the policy should change from execution feedback, and determining whether the resulting edit remains safe to persist after composition. To address these challenges, we introduce RIPPLE (Replay-Informed Persistent Policy Localization and Editing), which separates where an edit is made from whether it remains safe after composition. It diagnoses failed trajectories, maps each actionable failure to a predefined policy segment, and restricts the correction to that part of the policy. RIPPLE then evaluates candidates against the same iteration-start policy to compare their isolated gains, before replaying promising edits after previously accepted updates to expose downstream effects and interactions. Only edits that remain safe under composition are retained. We evaluate RIPPLE on Flow-HO, a synthetic held-out benchmark for executable workflow synthesis. RIPPLE improves validation success by up to 23.1% and yields positive gains on two additional frozen language-model backbones, while maintaining edit efficiency and low execution cost. Targeted interaction analysis further demonstrates both properties: a segment-local tool-use edit changes downstream resource resolution and validation, while an edit beneficial in isolation becomes harmful after composition.

Chinese Translation

提示策略编辑提供了一种实用的方法来改进合成可执行工作流的智能体，而无需更新底层模型。然而，持续的提示编辑具有两个相互关联的特性。首先，编辑的局部性并不意味着效果的局部性：限制在一个策略片段内的编辑可以在下游执行中产生涟漪，改变超出编辑片段的行为。其次，编辑效果对组合敏感：在孤立状态下有效的编辑在组合后可能会相互干扰，导致其中一个或两个失去其效益或变得有害。因此，持续适应必须支持两个不同的决策：识别策略应从执行反馈中改变的地方，以及确定结果编辑在组合后是否仍然安全。为了解决这些挑战，我们引入了RIPPLE（基于重放的信息的持久策略定位与编辑），它将编辑的位置与组合后是否保持安全分开。它诊断失败轨迹，将每个可操作的失败映射到预定义的策略片段，并将修正限制在该策略的部分。然后，RIPPLE评估候选者与相同迭代起始策略的比较，以比较它们的孤立增益，随后在先前接受的更新后重放有前景的编辑，以揭示下游效果和交互。只有在组合下仍然安全的编辑才会被保留。我们在Flow-HO上评估RIPPLE，这是一个用于可执行工作流合成的合成保留基准。RIPPLE将验证成功率提高了多达23.1%，并在两个额外的冻结语言模型骨干上产生了积极的增益，同时保持编辑效率和低执行成本。针对性的交互分析进一步证明了这两个特性：一个局部工具使用编辑改变了下游资源解析和验证，而一个在孤立状态下有益的编辑在组合后变得有害。

View on arXiv Download PDF AI Translation

cs.CL / 7 / 2609.12132

Population-level measures of perceived food access reveal barriers beyond geographic proximity

感知食品获取的群体水平测量揭示了超越地理接近性的障碍

Groton, Teresa, rachunok, Benjamin

Abstract

Food access is multidimensional, but population-level measurement still relies heavily on geography because perceived dimensions of access are difficult to measure at scale. Here, we use 25,125 Google Maps reviews from 49 grocery stores in Raleigh, North Carolina, to measure five dimensions of food access: availability, accessibility, affordability, accommodation, and acceptability. We identify review topics with unsupervised topic modeling and assign them to access dimensions using zero-shot classification, with 85.4% agreement against manual coding. The resulting store-level measures capture distinct aspects of food access and reveal barriers that geographic proximity alone does not capture. Comparisons between nearby stores in the same chain further show that identical store policies can be perceived very differently across locations, consistent with food access reflecting the fit between residents and their food environment. Perceived food access also follows systematic socioeconomic and demographic patterns that broadly parallel, but do not replicate, those observed for geographic access. These results show that online grocery reviews can provide a scalable complement to geographic measures of food access.

Chinese Translation

食品获取是多维度的，但群体水平的测量仍然在很大程度上依赖于地理因素，因为感知的获取维度在大规模测量时较为困难。在此，我们使用来自北卡罗来纳州罗利市49家杂货店的25,125条谷歌地图评论，测量食品获取的五个维度：可获得性、可达性、可负担性、适应性和可接受性。我们通过无监督主题建模识别评论主题，并使用零样本分类将其分配到获取维度，手动编码的协议一致性达到85.4%。最终得到的商店级别测量捕捉了食品获取的不同方面，并揭示了仅靠地理接近性无法捕捉的障碍。同一连锁店附近商店之间的比较进一步表明，相同的商店政策在不同地点的感知可能截然不同，这与食品获取反映居民与其食品环境之间的适配性一致。感知的食品获取还遵循系统的社会经济和人口统计模式，这些模式大体上与地理获取观察到的模式相似，但并不完全相同。这些结果表明，在线杂货评论可以为食品获取的地理测量提供可扩展的补充。

View on arXiv Download PDF AI Translation

cs.CL / 8 / 2609.12191

GAUGE: When Not to Trust LLM-as-a-Judge in User-Simulated Evaluation of Task-Oriented Agents

GAUGE：何时不应信任 LLM 作为任务导向代理的用户模拟评估中的评判者

Bodhwani, Umesh, Tran, Thanh, Wei, Kai

Abstract

Comparing and selecting task-oriented LLM agents increasingly relies on a low-cost offline evaluation gate: persona-driven LLM user-simulators converse with each candidate, an LLM-as-a-judge scores the transcripts, and the higher-scoring agent is promoted. We introduce GAUGE, a reusable offline protocol that measures whether this gate's ranking matches a grounded verifiable reward across 25 agents from six providers on the $\tau^2$-bench and SimulatorArena benchmarks, separating two kinds of evaluation validity that release practices conflate: ranking validity and construct validity. First, a satisfaction-success gap: satisfaction carries essentially no information about task success, as conversations rated satisfied by our blind panel are decorrelated from actual success, with 57.5% of them failing the customer's task, a pattern consistent across five rater populations, both benchmarks, and every subjective dimension we rated. Second, while the gate's ranking is robust across the broad capability span, it loses resolution among the near-equal strong agents: this decision-disagreement rate jumps from $<$1% on wide-reward pairs to 31% on close pairs. The gate is thus human-validated yet mis-anchored. As a remedy, we propose a calibrate-then-trust cadence in which a judge-free completion bit is a zero-cost tripwire for truncation regressions.

Chinese Translation

比较和选择任务导向的 LLM 代理越来越依赖于一种低成本的离线评估机制：以角色驱动的 LLM 用户模拟器与每个候选者进行对话，LLM 作为评判者对对话记录进行评分，得分较高的代理被提升。我们引入了 GAUGE，这是一种可重用的离线协议，用于测量该评估机制的排名是否与在 $ au^2$-bench 和 SimulatorArena 基准上对来自六个提供商的 25 个代理的可验证奖励相一致，区分了实践中混淆的两种评估有效性：排名有效性和构念有效性。首先，存在满意度与成功之间的差距：满意度几乎不包含关于任务成功的信息，因为我们盲评小组评定为满意的对话与实际成功无关，其中 57.5% 的对话未能完成客户的任务，这一模式在五个评分人群、两个基准和我们评定的每个主观维度中是一致的。其次，尽管该评估机制的排名在广泛的能力范围内是稳健的，但在近乎相等的强代理之间却失去了分辨率：在宽奖励对上，决策不一致率低于 1%，而在接近的奖励对上则跃升至 31%。因此，该评估机制经过人工验证但却存在锚定错误。作为补救措施，我们提出了一种先校准后信任的节奏，其中无评判的完成位是截断回归的零成本触发器。

View on arXiv Download PDF AI Translation

cs.CL / 9 / 2609.12230

Repair Before Reinforce: Context-Augmented Knowledge Graph Reasoning for Multi-Hop Question Answering

修复优先于增强：用于多跳问答的上下文增强知识图谱推理

Fonseka, Tharaka D., Jha, Niraj K.

Abstract

Question-answering often requires reasoning across multiple connected facts rather than retrieving a single isolated relation. Knowledge graphs (KGs) provide a structured way to represent such facts, but training large language models (LLMs) only on isolated KG head-relation-tail triples may limit their ability to learn the surrounding context needed for multi-hop reasoning. In this work, we propose a context-augmented training framework for multi-hop question-answering. Although generally applicable, we validate the framework in the context of disease-specific KGs, extracted using a reliable KG extraction framework called GraphMERT, for Gastroparesis and Diabetes. For each primary KG triple, we attach supporting triples extracted from the same source text chunk to form a context graph (CG). This creates two supervision settings: KG-grounded supervision, which uses only the target KG triple or path, and CG-grounded supervision, which uses the target KG triple or path together with supporting context triples. We train the Qwen3-14B model using supervised fine-tuning (SFT) under both settings, producing KGModel and CGModel variants. To strengthen the lower-hop factual foundation of the models, we introduce an LLM-judged, history-aware adaptive repair pipeline that identifies unresolved one-hop failures, continually fine-tunes on targeted repair examples, and removes or quarantines problematic noisy triples. This repair stage enables the models to reach 100% accuracy on the cleaned retained one-hop validation sets. Finally, we employ reinforcement learning (RL) using lower-hop question-answer items and evaluate generalization on harder 3-hop, 4-hop, and 5-hop tasks. Across both diseases, context-augmented supervision consistently improves multi-hop performance over KG-only supervision. RL initialized from repaired SFT checkpoints yields larger and more stable gains.

Chinese Translation

问答通常需要跨多个相关事实进行推理，而不是仅仅检索单一孤立的关系。知识图谱（KGs）提供了一种结构化的方式来表示这些事实，但仅在孤立的KG头-关系-尾三元组上训练大型语言模型（LLMs）可能会限制它们学习多跳推理所需的周围上下文的能力。在本研究中，我们提出了一种用于多跳问答的上下文增强训练框架。尽管该框架具有普遍适用性，但我们在特定疾病知识图谱的背景下进行了验证，这些知识图谱是通过一个可靠的KG提取框架GraphMERT提取的，针对胃轻瘫和糖尿病。对于每个主要KG三元组，我们从相同的源文本块中附加支持三元组，以形成上下文图（CG）。这创建了两种监督设置：KG基础监督，仅使用目标KG三元组或路径，以及CG基础监督，使用目标KG三元组或路径以及支持上下文三元组。我们在这两种设置下使用监督微调（SFT）训练Qwen3-14B模型，生成KGModel和CGModel变体。为了增强模型的低跳事实基础，我们引入了一个由LLM评估的、历史感知的自适应修复管道，该管道识别未解决的一跳失败，持续在目标修复示例上进行微调，并去除或隔离有问题的噪声三元组。该修复阶段使模型在清理后的保留一跳验证集上达到100%的准确率。最后，我们使用低跳问答项进行强化学习（RL），并在更困难的3跳、4跳和5跳任务上评估泛化能力。在这两种疾病中，上下文增强监督始终优于仅使用KG的监督，显著提高了多跳性能。从修复后的SFT检查点初始化的RL产生了更大且更稳定的增益。

View on arXiv Download PDF AI Translation

cs.CL / 10 / 2609.12243

Chopthin-Consensus Power Sampling: A Diversity-Preserving Approach to LLM Decoding

Chopthin-共识功率采样：一种保留多样性的LLM解码方法

Ahmadi, Minoo, Azizi, Seyedarmin, Potraghloo, Erfan Baghaei, Kamal, Mehdi, Pedram, Massoud

Abstract

Inference-time power sampling via Sequential Monte Carlo (SMC) can substantially improve large language model (LLM) reasoning without requiring post-training. However, many existing SMC approaches rely on equal-weight resampling, which can aggressively prune low-weight trajectories, discarding potentially correct reasoning paths and degrading the genealogical diversity of the search space. To address this, we introduce Chopthin-Consensus Power Sampling (CCPS). Our method applies the Chopthin resampler to LLM decoding: rather than equalizing weights and forcing unnecessary particle duplication, it enforces an upper bound on the ratio between the largest and smallest weights and carries the unequal weights forward. This targeted intervention preserves a richer set of distinct reasoning paths, keeps the weighted SMC approximation unchanged in conditional expectation, and guarantees a lower bound on the post-resampling effective sample size (ESS). To fully exploit this enriched population, we employ a semantic-majority selection mechanism that merges token-identical final trajectories, clusters semantically equivalent answers, and returns the answer supported by the largest number of distinct trajectories. Evaluating across three open-weight models and five reasoning benchmarks, we show that Chopthin increases oracle coverage in 13 of 15 settings. Combined with semantic-majority selection, CCPS matches or exceeds the final-answer accuracy of the Power-SMC baseline in 14 of 15 settings, delivering absolute gains of up to 10.6 percentage points. These findings demonstrate that diversity-preserving resampling and diversity-aware selection are complementary mechanisms for training-free LLM reasoning. Code is available at github.com/MinooAhmadii/chopthin-consensus-power-sampling.

Chinese Translation

通过序列蒙特卡洛（SMC）进行推理时的功率采样可以显著改善大型语言模型（LLM）的推理能力，而无需进行后期训练。然而，许多现有的SMC方法依赖于等权重重采样，这可能会过度修剪低权重轨迹，丢弃潜在的正确推理路径，并降低搜索空间的谱系多样性。为了解决这个问题，我们提出了Chopthin-共识功率采样（CCPS）。我们的方法将Chopthin重采样器应用于LLM解码：它不是通过平衡权重和强制不必要的粒子重复，而是对最大和最小权重之间的比率施加上限，并将不等权重向前传递。这种有针对性的干预保留了一组更丰富的独特推理路径，保持了加权SMC近似在条件期望下不变，并保证了重采样后有效样本大小（ESS）的下限。为了充分利用这一丰富的群体，我们采用了一种语义多数选择机制，该机制合并了标记相同的最终轨迹，聚类语义等价的答案，并返回由最多不同轨迹支持的答案。在三个开放权重模型和五个推理基准上进行评估，我们显示Chopthin在15个设置中的13个中增加了oracle覆盖率。结合语义多数选择，CCPS在15个设置中的14个中匹配或超过了功率-SMC基线的最终答案准确性，绝对增益高达10.6个百分点。这些发现表明，保留多样性的重采样和关注多样性的选择是无训练LLM推理的互补机制。代码可在github.com/MinooAhmadii/chopthin-consensus-power-sampling获取。

View on arXiv Download PDF AI Translation

cs.CL / 11 / 2609.12254

Automated Detection and Structuring of Social Tipping Point Evidence in Climate related Documents: A Modular AI Framework

气候相关文献中社会临界点证据的自动检测与结构化：一个模块化的人工智能框架

Perera, Kavindu, Abaeiani, Mohammad, Gilman, Ekaterina, Loven, Lauri, Oussalah, Mourad, Kanellos, Tassos, Gobbo, Beatrice, Adami, Dante, Ferriani, Nicolò, Romero, Maximiliano, Rossel, Pierre, Bonazountas, Marc, Deligianni, Christina, Xyderis, Nikos, Bogucki, Artur, Argyriou, Lampros, Balasubramanian, Prasasthy

Abstract

The climate literature has grown faster than review teams can read it. That gap matters most for a concept like the environmental social tipping point, the threshold at which a small change triggers rapid, self-reinforcing change in a social system. Evidence of this kind of shift is usually contained in one or two paragraphs within a longer document. As a result, existing text mining tools-which categorize entire documents by topic or highlight isolated claims-leave an expanding set of important evidence without any systematic method for discovery or organization. This paper presents an open and modular transformer-based framework that detects and structures social tipping point evidence at the passage level. The framework joins five components into a single deployable workflow: a DistilBERT boundary splitter for segmentation, an iteratively augmented RoBERTa classifier for detection, a Mistral 7B model that rewrites each detected passage for clarity, a LLaMA 3.2 3B model that rates the passage against five published social tipping point criteria, and a Milvus vector store for semantic retrieval. The system is wrapped in a Streamlit interface backed by MinIO object storage. Evaluated on a 163-passage benchmark labelled by GPT-4.1 and a 51-passage set reviewed by experts, the splitter surpassed three competing methods on a nine-metric composite score (6.137). The tuned RoBERTa model achieved 71.4 percent accuracy with a Cohen's kappa of 0.337 on the full benchmark, and 87.5 percent accuracy with a kappa of 0.742 on passages with labels, outperforming both a climate-focused model and untuned language models.

Chinese Translation

气候文献的增长速度超过了评审团队的阅读能力。这一差距对于环境社会临界点这一概念尤为重要，该概念指的是小变化触发社会系统中快速、自我增强变化的阈值。这类转变的证据通常包含在较长文档中的一两个段落内。因此，现有的文本挖掘工具——这些工具通过主题对整个文档进行分类或突出孤立的主张——未能为发现或组织这一日益扩大的重要证据提供系统的方法。本文提出了一个开放的、模块化的基于变换器的框架，能够在段落级别检测和结构化社会临界点证据。该框架将五个组件整合为一个可部署的工作流程：用于分段的DistilBERT边界分割器、用于检测的迭代增强RoBERTa分类器、用于清晰重写每个检测段落的Mistral 7B模型、根据五个已发布的社会临界点标准对段落进行评分的LLaMA 3.2 3B模型，以及用于语义检索的Milvus向量存储。该系统通过Streamlit界面包裹，并由MinIO对象存储支持。在由GPT-4.1标记的163段基准和由专家审查的51段数据集上进行评估，分割器在九项指标的综合得分（6.137）上超越了三种竞争方法。调优后的RoBERTa模型在完整基准上实现了71.4%的准确率，Cohen's kappa为0.337，而在带标签的段落上则达到了87.5%的准确率，kappa为0.742，优于气候专注模型和未调优的语言模型。

View on arXiv Download PDF AI Translation

cs.CL / 12 / 2609.12260

HypoKG: Evidence-Disciplined Biomedical Hypothesis Generation Beyond Endpoint Knowledge

HypoKG：超越终点知识的证据驱动生物医学假设生成

Okonkwo, Dominic, Okunoye, Adetayo, Arpinar, Ismailcem Budak

Abstract

Large language models (LLMs) can generate biomedical hypotheses, but it remains unclear whether they truly reason from scientific evidence or simply produce convincing-sounding ideas. To study this, we combine three major biological databases: the Kyoto Encyclopedia of Genes and Genomes (KEGG), Rhea, and UniProt, into a unified biochemical knowledge graph and construct a benchmark of 550 paths connecting enzyme sources to rare disease endpoints, yielding 13,200 hypotheses from six LLMs under four conditions varying the biological information each model receives: source enzyme only, full biological path, or source and disease endpoint only. Hypotheses are scored using an expert-derived five-criterion rubric on a 1-5 scale per criterion. We find that models given both the source and disease endpoint often produce the highest-scoring hypotheses, showing that LLMs can generate compelling ideas from minimal information. However, these hypotheses are less grounded in the evidence. In contrast, models given the full biological path generate hypotheses more consistent with known mechanistic relationships. We call this evidence-disciplined reasoning. To confirm this effect, we shuffled intermediate path steps while keeping endpoints fixed. Evidence grounding dropped significantly (delta = -0.793, p < 0.001), confirming models genuinely used path structure during reasoning. Our findings show that knowledge graphs support hypothesis generation in two ways: they identify biological endpoint pairs absent from the literature, and their mechanistic paths guide how LLMs reason between them.

Chinese Translation

大型语言模型（LLMs）能够生成生物医学假设，但尚不清楚它们是否真正基于科学证据进行推理，还是仅仅产生听起来令人信服的想法。为此，我们将三个主要生物数据库：京都基因组百科全书（KEGG）、Rhea 和 UniProt，结合成一个统一的生化知识图谱，并构建了一个包含550条连接酶源与罕见疾病终点的路径基准，从六个LLMs在四种条件下生成了13,200个假设，这些条件变化了每个模型接收的生物信息：仅源酶、完整生物路径或仅源和疾病终点。假设使用专家制定的五项标准评分标准进行评分，每项标准的评分范围为1-5。我们发现，给定源和疾病终点的模型通常生成得分最高的假设，表明LLMs能够从最少的信息中生成引人注目的想法。然而，这些假设在证据基础上较弱。相比之下，给定完整生物路径的模型生成的假设与已知的机制关系更为一致。我们称之为证据驱动推理。为了确认这一效应，我们在保持终点固定的情况下打乱了中间路径步骤。证据基础显著下降（delta = -0.793，p < 0.001），确认模型在推理过程中确实使用了路径结构。我们的研究结果表明，知识图谱在假设生成中支持两种方式：它们识别文献中缺失的生物终点对，并且它们的机制路径指导LLMs在这些终点之间的推理。

View on arXiv Download PDF AI Translation

cs.CL / 13 / 2609.12268

EAR: Entity-Aware Partitioning Approach for Retrieval-Augmented Generation Development

EAR：面向实体的检索增强生成开发的分区方法

Bora, Cenab Batu, Alatlı, Oylum, Bora, Sebnem, Dikenelli, Oguz

Abstract

Retrieval-augmented generation (RAG) can improve knowledge-intensive question answering, but the first design choice is easy to overlook: how should the source corpus be partitioned into retrievable units? Fixed-size chunks often return long passages whose relation to the question is only implicit. We introduce EAR, an Entity-Aware Partitioning approach for multiple-choice question answering (MCQA). EAR extracts normalized surface anchors from the question, answer options, and corpus; retrieves local windows around matching corpus anchors; and can attach a larger parent passage through an extractive summary. We evaluate EAR on a cleaned Massive Multitask Language Understanding (MMLU)-style subset of 153 questions selected by an automatic corpus-support heuristic and using decontaminated public textbook text. Across same-protocol top-k = 3 and top-k = 8 sweeps with Mistral, Gemma, and DeepSeek, EAR entity-window reduces retrieved words by 37.5-40.2% relative to chunks. Observed accuracy changes are +5.2, +1.3, and -3.9 points at top-k = 3, and +5.9, -3.3, and -4.6 points at top-k = 8; none of the entity-window differences is statistically significant. The scoped contribution is methodological: EAR provides a compact and inspectable retrieval unit, while its rule-based anchor extractor remains domain-specific and requires separate validation before transfer.

Chinese Translation

检索增强生成（RAG）可以改善知识密集型问答，但第一个设计选择往往容易被忽视：源语料库应如何划分为可检索单元？固定大小的块通常返回与问题关系仅为隐含的长段落。我们提出了EAR，一种面向实体的多项选择问答（MCQA）分区方法。EAR从问题、答案选项和语料库中提取标准化的表面锚点；在匹配的语料锚点周围检索局部窗口；并且可以通过提取摘要附加一个更大的父段落。我们在一个经过清理的153个问题的MMLU（大规模多任务语言理解）风格子集上评估EAR，该子集由自动语料支持启发式选择，并使用去污染的公共教科书文本。在与Mistral、Gemma和DeepSeek的相同协议下，top-k = 3和top-k = 8的实验中，EAR实体窗口相较于块减少了37.5-40.2%的检索词汇。观察到的准确率变化在top-k = 3时为+5.2、+1.3和-3.9点，在top-k = 8时为+5.9、-3.3和-4.6点；没有任何实体窗口的差异在统计上显著。其贡献在于方法论：EAR提供了一个紧凑且可检查的检索单元，而其基于规则的锚点提取器仍然是领域特定的，并且在转移之前需要单独验证。

View on arXiv Download PDF AI Translation

cs.CL / 14 / 2609.12303

Breaking the Token Ceiling: Distilling Smaller, Stronger Byte Models

打破令牌上限：提炼更小、更强的字节模型

Marathe, Kalyani, Pagnoni, Artidoro, Limisiewicz, Tomasz, Li, Margaret, Lewis, Mike, Zettlemoyer, Luke, Iyer, Srinivasan

Abstract

Small models are made more capable through distillation from a larger one that shares their tokenization scheme. However, do distilled byte and token models behave similarly in terms of scaling trends as compute and data increases? To enable this comparison, we introduce two variants to efficiently convert token logits to Byte Logits: 1) approximate: Marginalize-It, and 2) exact: End-Of-Token. We then present the first large scale study of overtraining decoder-only dense transformer models varying two dimensions simultaneously: the tokenization scheme (Tokens, Bytes, Bytes w/ eot) and the training objective (Distillation vs. Cross-Entropy), sweeping layer-parameter-matched models with roughly 1 billion parameters up to 1 trillion bytes of data. Across eight benchmarks spanning three categories: Multiple Choice QA, Language Generation, and Machine Translation, we find that Token-1B models outperform byte models (End-Of-Token-1B and Bytes-1B) in the low-FLOP regime but eventually plateau; byte models start worse yet surpass Token-1B models with more compute, reaching a higher downstream task performance ceiling. Extrapolating the average top-1 error vs. validation BPB scaling laws predicts that, asymptotically, distilled End-Of-Token-1B outperforms distilled Token-1B by up to 4%. They are also far more data efficient, matching the performance of distilled Token-1B using only one-sixth of the training data. Moreover, by operating over a small vocabulary of 256 bytes instead of on the order of 100K tokens, they circumvent the need for top-k truncation during logit dumping, while also reducing logit storage costs to roughly one-fifth. Finally, our downstream performance scaling laws predict that our distilled End-Of-Token-1B models asymptotically surpass the Llama 3.2-1B, Gemma-3-1B-pt, and Gemma 2B models on averaged downstream tasks by up to 6.5%, 8.1%, and 2.1%, respectively.

Chinese Translation

通过从一个共享其令牌化方案的大模型中进行蒸馏，小模型的能力得以提升。然而，在计算和数据增加的情况下，蒸馏的字节模型和令牌模型在扩展趋势上是否表现相似？为了进行这种比较，我们引入了两种变体以有效地将令牌 logits 转换为字节 logits：1) 近似方法：Marginalize-It，2) 精确方法：End-Of-Token。然后，我们首次进行了大规模研究，探讨了仅解码器的稠密变换器模型在两个维度上同时变化的过拟合情况：令牌化方案（Tokens、Bytes、Bytes w/ eot）和训练目标（蒸馏 vs. 交叉熵），对大约 10 亿参数的层参数匹配模型进行测试，数据量高达 1 万亿字节。在涵盖三个类别的八个基准测试中：多项选择问答、语言生成和机器翻译，我们发现 Token-1B 模型在低 FLOP 范围内优于字节模型（End-Of-Token-1B 和 Bytes-1B），但最终趋于平稳；字节模型起初表现较差，但随着计算量的增加，超越了 Token-1B 模型，达到了更高的下游任务性能上限。推测平均 top-1 错误率与验证 BPB 扩展规律预测，蒸馏的 End-Of-Token-1B 在渐近情况下优于蒸馏的 Token-1B，提升幅度可达 4%。它们的数据效率也远高于后者，仅使用六分之一的训练数据便可匹配蒸馏 Token-1B 的性能。此外，通过在 256 字节的小词汇上操作，而不是在大约 10 万个令牌上操作，它们避免了在 logits 转储过程中需要进行 top-k 截断，同时将 logits 存储成本降低至大约五分之一。最后，我们的下游性能扩展规律预测，蒸馏的 End-Of-Token-1B 模型在渐近情况下在平均下游任务上超越 Llama 3.2-1B、Gemma-3-1B-pt 和 Gemma 2B 模型，提升幅度分别可达 6.5%、8.1% 和 2.1%。

View on arXiv Download PDF AI Translation

cs.CL / 15 / 2609.12310

ESTS at WMT26: Routing-Informed Expert Pruning for Model Compression

WMT26的ESTS：基于路由信息的专家剪枝用于模型压缩

Martin, Liu O., Bandarkar, Lucas, Peng, Nanyun

Abstract

We describe six submissions under the team name ESTS to the unconstrained WMT26 Model Compression Shared Task for English--Simplified Chinese and English--Egyptian Arabic. We submit three compression operating points per translation direction, all derived from GPT-OSS-20B. We use task-specific routing mass to rank experts and cross-lingual routing divergence to allocate retained capacity across layers, then physically remove low-importance experts. The resulting specialists are recovery-tuned on GPT-5.1-generated synthetic translation data and further compressed by applying MXFP4 quantization to the retained expert projection weights. We additionally implement a robust inference system for the instruction-conditioned WMT26 setting, including category inference, output validation, retries, segmented fallback, and source-owned JSON reconstruction. Across our six submissions, parameter counts range from 4.186B to 7.770B and packed artifact sizes from 4.55 to 6.33~GiB. Internal xCOMET-XL evaluation using GPT-5.1 pseudo-references provides an internal comparison across the submitted compression operating points.

Chinese Translation

我们描述了以团队名称ESTS提交的六个方案，参与无约束的WMT26模型压缩共享任务，涉及英语-简体中文和英语-埃及阿拉伯语的翻译方向。我们为每个翻译方向提交了三个压缩操作点，均源自GPT-OSS-20B。我们使用任务特定的路由质量来对专家进行排名，并利用跨语言路由差异在各层之间分配保留的容量，然后物理移除低重要性的专家。最终的专家在GPT-5.1生成的合成翻译数据上进行了恢复调优，并通过对保留的专家投影权重应用MXFP4量化进一步压缩。我们还为指令条件下的WMT26设置实现了一个稳健的推理系统，包括类别推理、输出验证、重试、分段回退和源拥有的JSON重构。在我们的六个提交中，参数数量范围从4.186B到7.770B，打包工件大小从4.55到6.33 GiB。使用GPT-5.1伪参考的内部xCOMET-XL评估提供了对提交的压缩操作点的内部比较。

View on arXiv Download PDF AI Translation

cs.CL / 16 / 2609.12341

I Am No One: Style-Aware Paraphrasing for Text Anonymization

我无名：基于风格的文本匿名化释义

Khan, Ahmed Sohair, He, Estrid, Wachowicz, Monica, Naghizade, Elham

Abstract

Authorship attribution models can re-identify users from seemingly anonymized text by exploiting stable stylistic fingerprints, even after explicit identifiers are removed, posing a growing privacy risk for text publishing and analytics. This risk extends to speech-derived text such as ASR transcripts of meetings and call-center conversations, where stylometric leakage can persist even after acoustic anonymization. Differential privacy-based anonymization often severely degrades text quality and utility. We propose a style-aware, prompt-driven anonymization approach that uses pretrained large language models to construct compact stylistic profiles from minimal samples and rewrite text to suppress identifiable style markers while preserving meaning. Across blog and review datasets, our approach reduces authorship attribution F1 by 60-70% while maintaining content quality and readability, substantially outperforming DP-based and non-DP baselines.

Chinese Translation

作者归属模型可以通过利用稳定的风格指纹重新识别用户，即使在显式标识符被移除之后，这对文本发布和分析构成了日益严重的隐私风险。这种风险还扩展到语音生成的文本，例如会议和呼叫中心对话的自动语音识别（ASR）转录，其中风格特征泄露即使在声学匿名化后仍然可能存在。基于差分隐私的匿名化往往会严重降低文本质量和实用性。我们提出了一种基于风格的、以提示驱动的匿名化方法，该方法利用预训练的大型语言模型从最小样本构建紧凑的风格特征，并重写文本以抑制可识别的风格标记，同时保留其含义。在博客和评论数据集上，我们的方法将作者归属的F1值降低了60-70%，同时保持内容质量和可读性，显著优于基于差分隐私和非差分隐私的基线。

View on arXiv Download PDF AI Translation

cs.CL / 17 / 2609.12353

SynthSentry: Detecting Synthetic Data Contamination in Language Model Training Data

SynthSentry：检测语言模型训练数据中的合成数据污染

Myakala, Praveen Kumar, Namburi, Ravichandra, Jayaramu, Sowmya Keragodu, Thomas, Sooraj George

Abstract

Large language models trained recursively on their own or other models' outputs undergo model collapse, in which distributional tails and factual accuracy deteriorate while fluency survives. Prior work diagnoses collapse after training; the actionable problem is screening a corpus of unknown provenance before training. We introduce SynthSentry, a corpus-level, model-agnostic contamination signal requiring no access to the generating model, no generation history, and no synthetic labels. The score is a distributional divergence over three statistics: lexical diversity collapse, n-gram tail truncation, and perplexity variance across reference models. We evaluate on corpora contaminated by small open-weight generators and an instruction-tuned open-weight model under a leave-one-generator-out protocol. A domain-stratified study measures false positives on naturally repetitive human text (legal, clinical, source code). The score ranks corpora by severity with little loss when whole generator families are held out. Per-domain calibration holds near its nominal false-positive budget once covariance shrinkage and a bootstrap threshold replace a naive quantile, which runs four times over budget. A downstream fine-tuning check showed no contamination-driven accuracy deficit at our scale, so whether pruning recovers one remains open; the same run shows over-pruning risk once pruning exceeds the true contamination fraction. We frame screening as a data-curation defense rather than a post-hoc diagnosis and release the scoring toolkit. All results are small-scale; scope is English-language, batch-mode corpus screening. Contamination sources are single-generation or hand-authored rather than recursively generated, so results speak to synthetic contamination generally and not to recursion depth.

Chinese Translation

大型语言模型在自身或其他模型输出上递归训练时会出现模型崩溃现象，此时分布尾部和事实准确性会恶化，而流畅性得以保留。之前的研究在训练后诊断崩溃；可操作的问题是在训练前筛选来源不明的语料库。我们提出了SynthSentry，这是一种语料库级别、模型无关的污染信号，不需要访问生成模型、生成历史或合成标签。该评分是基于三项统计数据的分布差异：词汇多样性崩溃、n-gram尾部截断和参考模型间的困惑度方差。我们在受到小型开放权重生成器和经过指令调优的开放权重模型污染的语料库上进行评估，采用逐个生成器排除的协议。领域分层研究测量了自然重复人类文本（法律、临床、源代码）上的假阳性。该评分根据严重程度对语料库进行排名，当整个生成器家族被排除时，损失很小。每个领域的校准在协方差收缩和引导阈值替代简单分位数后，保持在其名义假阳性预算附近，而简单分位数的预算超出四倍。下游微调检查表明，在我们的规模下没有因污染导致的准确性缺失，因此是否修剪能恢复准确性仍然是一个开放问题；同一轮实验显示，一旦修剪超过真实污染比例，就存在过度修剪的风险。我们将筛选框架视为数据策展防御，而不是事后诊断，并发布了评分工具包。所有结果均为小规模；范围为英语语言的批量模式语料库筛选。污染来源为单次生成或手动撰写，而非递归生成，因此结果普遍适用于合成污染，而非递归深度。

View on arXiv Download PDF AI Translation

cs.CL / 18 / 2609.12354

CueMem: Cue-Guided Context Reconstruction for Long-Term Conversational Memory

CueMem：基于线索引导的长期对话记忆上下文重建

Wang, Changjian, Li, Rongzhen, Guan, Weili, Shi, Shuming, Lu, Quan, Jiang, Ning

Abstract

Long-term conversational agents must answer user queries by recalling information from extended dialogue histories, yet directly using the full history is costly and often unreliable, while compressed memory units may lose fine-grained evidence needed for question answering. Motivated by the reconstructive view of autobiographical memory, we propose CueMem, a cue-guided framework that treats extracted memory records as retrieval cues rather than self-contained evidence and reconstructs query-relevant dialogue context from their source turns. During memory construction, CueMem extracts fine-grained memory cues from dialogue turns and links each cue to its source turn. At query time, it retrieves query-relevant cues, maps them to source-turn anchors, and expands from these anchors over a turn graph that captures temporal proximity and semantic relatedness, reconstructing a compact evidence context from the original dialogue for LLM answer generation. Experiments on LoCoMo and LongMemEval show that CueMem consistently outperforms representative long-term memory baselines. Further analyses show that graph-based context reconstruction helps recover supporting dialogue evidence while reducing query-time input tokens and latency compared with the full-history LLM setting. These results highlight retrieval cues as an effective alternative to self-contained memory evidence for long-term conversational question answering.

Chinese Translation

长期对话代理必须通过回忆扩展对话历史中的信息来回答用户查询，但直接使用完整历史代价高昂且常常不可靠，而压缩的记忆单元可能会丢失回答问题所需的细粒度证据。受到自传记忆重建视角的启发，我们提出了CueMem，一个将提取的记忆记录视为检索线索而非自包含证据的框架，并从其源轮次中重建与查询相关的对话上下文。在记忆构建过程中，CueMem从对话轮次中提取细粒度记忆线索，并将每个线索与其源轮次关联。在查询时，它检索与查询相关的线索，将其映射到源轮次锚点，并从这些锚点扩展到捕捉时间接近性和语义相关性的轮次图，从原始对话中重建紧凑的证据上下文以生成LLM（大语言模型）答案。在LoCoMo和LongMemEval上的实验表明，CueMem始终优于代表性的长期记忆基线。进一步的分析表明，与完整历史LLM设置相比，基于图的上下文重建有助于恢复支持对话证据，同时减少查询时的输入令牌和延迟。这些结果突显了检索线索作为长期对话问答中自包含记忆证据的有效替代方案。

View on arXiv Download PDF AI Translation

cs.CL / 19 / 2609.12366

ORQA: An Occupation-Realistic Question and Answer Framework for LLM Professional Knowledge

ORQA：一个针对大型语言模型职业知识的职业现实问答框架

Krishnan, Shreyas, Chang, Serina, Nagaraj, Abhishek

Abstract

We present ORQA, a method for testing occupation-level knowledge in large language models. Prior methods either map abstract LLM skills to occupations via task definitions or utilize expert knowledge which is difficult to obtain at scale and expensive. ORQA complements both of these methods by connecting O*NET occupations to trusted occupation-specific websites (such as regulatory agencies, licensing bodies, professional organizations, and government publications) and converting these into source-traceable question-answer pairs. A combination of an automated pipeline and human review produces a set of high quality questions about occupations. The question set created via our method covers 116 occupations from all 21 major groups in the SOC, with 480 questions sourced from 187 different websites. Each question is designed to probe a real-world skill question that is relevant to the occupation in question. We test 15 state-of-the-art frontier and open-weight models via this method. Claude Opus 4.6, GPT-5.4 and Claude Sonnet 4.6 all perform the best at approximately 58-62% while smaller open-weight models achieve approximately 33-41% performance. Performance varies significantly across occupations. Healthcare-related occupations achieve the highest performance (78%) while Office and Administrative Support achieve approximately 40%. Performance on individual occupations (e.g. Sheet Metal Workers and Fish and Game Wardens) is essentially zero. We also find that open-ended questions and weighting by wage bill do not significantly affect the ranking of models on this benchmark. We believe that leveraging existing trusted occupation-specific information to test LLM knowledge in professional domains may be a scalable and useful method for evaluating occupation-level AI performance in the future. Results and data are available at orqabench.org.

Chinese Translation

我们提出了ORQA，一种用于测试大型语言模型职业水平知识的方法。以往的方法要么通过任务定义将抽象的LLM技能映射到职业上，要么利用难以大规模获取且成本高昂的专家知识。ORQA通过将O*NET职业与可信的职业特定网站（如监管机构、许可机构、专业组织和政府出版物）连接起来，并将其转化为可追溯来源的问题-答案对，来补充这两种方法。通过自动化流程与人工审核相结合，生成了一组高质量的职业相关问题。我们的方法创建的问题集涵盖了SOC中所有21个主要组别的116个职业，共有480个问题来源于187个不同的网站。每个问题旨在探讨与特定职业相关的真实技能问题。我们通过该方法测试了15个最先进的前沿模型和开放权重模型。Claude Opus 4.6、GPT-5.4和Claude Sonnet 4.6的表现最佳，约为58-62%，而较小的开放权重模型的表现约为33-41%。不同职业的表现差异显著。与医疗相关的职业表现最高（78%），而办公室和行政支持职业的表现约为40%。个别职业（如钣金工和鱼类及游戏管理员）的表现几乎为零。我们还发现，开放式问题和按工资总额加权对模型在该基准上的排名没有显著影响。我们相信，利用现有的可信职业特定信息来测试LLM在专业领域的知识，可能是一种可扩展且有用的方法，用于评估未来职业级AI的表现。结果和数据可在orqabench.org获取。

View on arXiv Download PDF AI Translation

cs.CL / 20 / 2609.12382

Representation-based Masked Diffusion Model

基于表示的掩码扩散模型

Hu, Yangrong, Huang, Ding, Zhou, Xueyu, Huang, Jian

Abstract

Masked Diffusion Models (MDMs) have emerged as a compelling paradigm for language modeling, offering the capability for efficient parallel text generation. However, existing parallel sampling methods typically update multiple masked tokens independently and ignore the complex mutual dependencies among the masked tokens. This independent updating mechanism lacks global coordination and might lead to incoherent outputs. To address this limitation, we propose Representation-based Masked Diffusion Model (RMDM), a framework that leverages the text representation to explicitly encode global semantics and help to parallel update tokens more precisely. Specifically, we first encode text into a continuous semantic space using a pretrained encoder and learn an invertible transformation that normalizes the representation distribution to a Gaussian prior, facilitating efficient sampling during generation. Conditioned on this latent semantic representation, we train a masked diffusion model to learn the conditional text distribution, where the representation serves as global semantic guidance to coordinate parallel token updates and faithfully approximate the target distribution. Empirical results demonstrate that RMDM significantly improves generation quality, particularly in aggressive few-step sampling regimes.

Chinese Translation

掩码扩散模型（Masked Diffusion Models, MDMs）作为一种有力的语言建模范式，具备高效并行文本生成的能力。然而，现有的并行采样方法通常独立更新多个掩码标记，忽视了掩码标记之间复杂的相互依赖关系。这种独立更新机制缺乏全局协调，可能导致生成结果不连贯。为了解决该限制，我们提出了基于表示的掩码扩散模型（Representation-based Masked Diffusion Model, RMDM），该框架利用文本表示显式编码全局语义，帮助更精确地并行更新标记。具体而言，我们首先使用预训练编码器将文本编码到连续语义空间，并学习一个可逆变换，将表示分布归一化为高斯先验，从而促进生成过程中的高效采样。在该潜在语义表示的条件下，我们训练掩码扩散模型以学习条件文本分布，其中该表示作为全局语义指导，协调并行标记更新，忠实地逼近目标分布。实证结果表明，RMDM显著提升了生成质量，尤其在激进的少步采样策略下表现突出。

View on arXiv Download PDF AI Translation

cs.CL / 21 / 2609.12448

GraphProfiler: Source-Linked Sensitive Attribute Inference via Personal Knowledge Graphs

图谱分析器：通过个人知识图谱进行源链接的敏感属性推断

Khan, Ahmed Sohair, He, Estrid, Ma, Chenglong, Wachowicz, Monica, Naghizade, Elham

Abstract

Sensitive attributes such as age, income, and occupation can be inferred from user-generated content by aggregating indirect cues across many ordinary posts. LLM-based profilers can perform this aggregation automatically and with high accuracy, which makes large-scale personal attribute inference a major privacy threat. Existing LLM-based profilers, however, offer limited insight into which specific posts, concepts, and relationships made an inference possible, which is key to targeted privacy mitigation, i.e., redacting or rewriting only the few posts that actually leak an attribute, rather than perturbing entire histories. We introduce GraphProfiler, an auditable LLM-based profiler that represents each user's post history as a source-linked personal knowledge graph where nodes and edges trace back to the originating post and resolves attribute predictions to cited graph records and source texts. GraphProfiler reaches 86.7% attack success rate on the eight-attribute SynthPAI benchmark, within two points of strong text-only baselines, and 84.6% on PANDORA, while citing supporting evidence for over 98% of predictions. Our controlled ablation experiments provide evidence that the cited posts contribute to attack success, as removing them reduces the attack success rate substantially more than removing an equal number of random posts.

Chinese Translation

敏感属性如年龄、收入和职业可以通过聚合许多普通帖子中的间接线索，从用户生成的内容中推断出来。基于大型语言模型（LLM）的分析工具能够自动且高精度地执行这种聚合，这使得大规模个人属性推断成为一个主要的隐私威胁。然而，现有的基于LLM的分析工具对哪些特定帖子、概念和关系使得推断成为可能提供的洞见有限，这对于有针对性的隐私缓解至关重要，即仅对实际泄露属性的少数帖子进行编辑或重写，而不是扰动整个历史记录。我们引入了图谱分析器（GraphProfiler），这是一个可审计的基于LLM的分析工具，它将每个用户的帖子历史表示为一个源链接的个人知识图谱，其中节点和边追溯到原始帖子，并将属性预测解析为引用的图谱记录和源文本。图谱分析器在八个属性的SynthPAI基准测试中达到了86.7%的攻击成功率，距离强大的仅文本基线仅差两分，在PANDORA上达到了84.6%，同时为超过98%的预测引用了支持证据。我们的控制消融实验提供了证据，表明引用的帖子对攻击成功率的贡献，因为去除这些帖子会显著降低攻击成功率，远超去除相同数量的随机帖子。

View on arXiv Download PDF AI Translation

cs.CL / 22 / 2609.12469

Not All Speech Is Intent: Adaptive Self-Correcting Inference Layer for Post-ASR False Wake-Up

并非所有语音都是意图：自适应自我纠正推理层用于后ASR错误唤醒

Saraswat, Preeti, Neelagiri, Divya, Yadav, Anil

Abstract

False wake-up activations remain a persistent challenge in conversational AI. Speech phonetically similar to a device's wake word can produce a syntactically valid and semantically coherent ASR transcript that the assistant incorrectly executes. Most existing systems make a single intent decision in isolation, without a mechanism to learn from recurring errors over time or adapt to individual users through personalized learning. We introduce the Feedback-Driven Adaptive Self-Correcting Inference Layer (ASCIL), a complementary post-ASR correction framework that re-evaluates wake-up intent before response generation by fusing acoustic embeddings, linguistic cues, device context, and patterns from past misclassifications. ASCIL interprets implicit signals, including hesitation, disengagement, and silence, and explicit signals, including cancellation and repetition, as automatically inferred, noisy behavioral indicators of potential misclassification. These signals drive online pattern updates without manual annotation, whereas the intentional/unintentional reference labels used for offline evaluation are human-annotated. It generalizes from prior errors, applies corrective adjustments at inference time, and continuously updates in parallel with natural-language execution. Evaluated on a proprietary dataset of 3,667 interactions with human-annotated intentional/unintentional reference labels spanning 14 acoustic and contextual conditions, ASCIL achieves 54.27% relative error reduction on a session-disjoint subset constructed from baseline failures, and up to 24.39% relative error reduction at threshold 0.90 on the issue-tagged evaluation slice. These gains are achieved while improving intentional acceptance rates, with a median added latency below 60 ms in the reported benchmark.

Chinese Translation

错误唤醒激活在对话式人工智能中仍然是一个持续的挑战。与设备唤醒词在语音上相似的语音可能会生成一个语法上有效且语义上连贯的ASR转录，导致助手错误地执行。大多数现有系统在孤立的情况下做出单一意图决策，没有机制从重复错误中学习或通过个性化学习适应个别用户。我们提出了反馈驱动的自适应自我纠正推理层（ASCIL），这是一个补充的后ASR纠正框架，通过融合声学嵌入、语言线索、设备上下文和过去错误分类的模式，在生成响应之前重新评估唤醒意图。ASCIL将隐含信号（包括犹豫、 disengagement 和沉默）以及显性信号（包括取消和重复）解释为自动推断的、嘈杂的潜在错误分类行为指标。这些信号驱动在线模式更新，无需手动标注，而用于离线评估的意图/非意图参考标签则由人工标注。它从先前的错误中进行泛化，在推理时应用纠正调整，并与自然语言执行并行持续更新。在一个包含3,667次交互的人类标注意图/非意图参考标签的专有数据集上进行评估，涵盖14种声学和上下文条件，ASCIL在从基线失败构建的会话不重叠子集上实现了54.27%的相对错误减少，并在问题标记的评估切片上在阈值0.90时实现了高达24.39%的相对错误减少。这些增益是在提高意图接受率的同时实现的，报告的基准中中位增加延迟低于60毫秒。

View on arXiv Download PDF AI Translation

cs.CL / 23 / 2609.12471

AMDKernelVault: Large-Scale Datasets and Agentic Training for AMD GPU Kernel Optimization

AMDKernelVault：用于AMD GPU内核优化的大规模数据集和智能训练

Liu, Ji, Majumder, Saptarshi, Huang, Yiqing, Ouyang, Wenwen, Pandey, Umang, Li, Zeping, Chen, Chushi, An, Zihao, Yang, Puyuan, Li, Zekai, Rafati, Sina, Liu, Ziqiong, Brahma, Pratik Prabhanjan, Li, Dong, Liu, Zicheng, Zhou, Sharon, Barsoum, Emad

Abstract

We introduce AMDKernelVault, an open HIP and Triton kernel corpus and training framework for recent AMD CDNA GPUs. Existing LLM-based kernel agents are largely CUDA/NVIDIA-centric and often depend on repeated frontier-LLM calls for generation, reflection, and optimization. To address this gap, we develop HIPKernelGen and TritonKernelGen, agent-driven pipelines that transform PyTorch references into HIP or Triton kernels, compile and validate candidates under ROCm, and latency-profile them on AMD hardware. The corpus contains 62,153 execution-verified HIP kernel samples, 2,377 production-grounded ROCm Libraries QA entries, and 39,893 Triton kernels. We further train Qwen3-8B with supervised fine-tuning and execution-aware reinforcement learning as a demonstration of the corpus's utility. Under fixed evaluation budgets, it achieves the highest correctness among the compared models on PyTorch-to-HIP (34.0% Pass@1), TritonBench-G (33.2% Corr@3), and ROCmBench (41.94% Corr@3), but does not uniformly lead compilation or speed metrics. The corpus and documentation are available at https://huggingface.co/datasets/amd/AIG-Datasets, and the associated training and kernel-generation code is available at https://github.com/AMD-AGI/hip_kernel_llm_lab.

Chinese Translation

我们介绍了AMDKernelVault，这是一个针对最新AMD CDNA GPU的开放HIP和Triton内核库及训练框架。现有的基于大型语言模型（LLM）的内核代理主要集中于CUDA/NVIDIA，并且通常依赖于重复的前沿LLM调用进行生成、反思和优化。为了解决这一问题，我们开发了HIPKernelGen和TritonKernelGen，这些代理驱动的管道将PyTorch参考转换为HIP或Triton内核，在ROCm下编译和验证候选项，并在AMD硬件上进行延迟分析。该库包含62,153个执行验证的HIP内核样本、2,377个基于生产的ROCm库质量保证条目和39,893个Triton内核。我们进一步通过监督微调和执行感知强化学习训练Qwen3-8B，以展示该库的实用性。在固定评估预算下，它在PyTorch到HIP（34.0% Pass@1）、TritonBench-G（33.2% Corr@3）和ROCmBench（41.94% Corr@3）上实现了与比较模型中最高的正确率，但并不总是领先于编译或速度指标。该库和文档可在https://huggingface.co/datasets/amd/AIG-Datasets获取，相关的训练和内核生成代码可在https://github.com/AMD-AGI/hip_kernel_llm_lab获取。

View on arXiv Download PDF AI Translation

cs.CL / 24 / 2609.12475

Zipbench: Low-Cost Framework for Compressing Comprehensive Benchmarks of Large Language Models

Zipbench：压缩大型语言模型综合基准的低成本框架

Huang, Zhongzhan, Li, Junxin, Ling, Guoming, Lin, Yupei, Zhong, Shanshan, Wu, Hefeng

Abstract

Comprehensive benchmark suites are essential for improving large language models (LLMs), but many widely used benchmarks are redundant, making evaluation unnecessarily expensive. Although recent benchmark compression methods (BCMs) can mitigate this cost, many strong BCMs rely on large collections of per-sample evaluation results from numerous LLMs to identify representative samples. Building such collections is also expensive unless they are already public, making these methods difficult to extend to newly released benchmarks. To address this challenge, we present ZipBench, a simple and low-cost BCM with theoretical error and rank-consistency guarantees. ZipBench evaluates only a small set of anchor LLMs, synthesizes pseudo evaluation results to broaden coverage, learns compact sample representations, and selects a small yet representative subset. Building on it, we create ZipBench Zoo, a collection of compact versions of 100+ benchmark proxies spanning text, multimodal, and agent tasks. These benchmark achieve mean absolute errors of 0.002--0.02 and average Spearman correlations of ~0.98 with the full benchmarks. Overall, ZipBench reduces the cost of both LLM evaluation and compact benchmark construction, lowering the barrier to broad LLM research for compute-constrained researchers. The code has been released in https://github.com/MilkThink-Lab/ZipBench.

Chinese Translation

综合基准套件对于提升大型语言模型（LLMs）至关重要，但许多广泛使用的基准存在冗余，使得评估成本不必要地高。尽管近期的基准压缩方法（BCMs）可以缓解这一成本，但许多有效的BCM依赖于来自多个LLM的大量每个样本评估结果，以识别代表性样本。除非这些集合已经公开，否则构建这样的集合也非常昂贵，这使得这些方法难以扩展到新发布的基准。为了解决这一挑战，我们提出了ZipBench，这是一种简单且低成本的BCM，具有理论误差和排名一致性保证。ZipBench仅评估一小组锚定LLM，合成伪评估结果以扩大覆盖范围，学习紧凑的样本表示，并选择一个小而具有代表性的子集。在此基础上，我们创建了ZipBench Zoo，这是一个包含100多个基准代理的紧凑版本的集合，涵盖文本、多模态和代理任务。这些基准的平均绝对误差为0.002--0.02，与完整基准的平均斯皮尔曼相关系数约为0.98。总体而言，ZipBench降低了LLM评估和紧凑基准构建的成本，为计算资源有限的研究人员降低了广泛进行LLM研究的门槛。代码已发布在https://github.com/MilkThink-Lab/ZipBench。

View on arXiv Download PDF AI Translation

cs.CL / 25 / 2609.12537

The House with a Million Windows: Interactive Fiction for Narrative Restorying

拥有百万窗户的房子：用于叙事重述的互动小说

Kommers, Cody, Immel, Sarah G, Hemment, Drew, Lee, Mina

Abstract

AI-assisted writing can flatten meaning in human storytelling, enabling the production of homogeneous outputs without the intentional effort and sense-making writing entails. To address this challenge, we present The House with a Million Windows (HWAMW), an LLM-based interactive fiction system designed to help users explore both the breadth and depth of potential meanings within their personal stories -- drawing on a psychological paradigm called the restorying intervention. In HWAMW, users play through a text-based narrative in which they tell a story, then encounter a set of LLM-generated "windows" reframing it according to different literary styles. Empirical evidence shows that HWAMW increases users' sense of narrative identity, while an expert review explores how this effect is achieved. Our findings suggest that HWAMW facilitates restorying and offers a valuable paradigm for AI-assisted writing, wherein LLMs do not tell our stories but rather help us see greater potential in the stories we tell.

Chinese Translation

AI辅助写作可能会削弱人类叙事中的意义，使得在没有意图努力和意义构建的情况下产生同质化的输出。为了解决这一挑战，我们提出了《拥有百万窗户的房子》（The House with a Million Windows，HWAMW），这是一个基于大型语言模型（LLM）的互动小说系统，旨在帮助用户探索个人故事中潜在意义的广度和深度——借助一种称为重述干预（restorying intervention）的心理学范式。在HWAMW中，用户通过文本叙事进行游戏，讲述一个故事，然后遇到一组根据不同文学风格重新框架的LLM生成的“窗户”。实证证据表明，HWAMW增强了用户的叙事身份感，而专家评审则探讨了这一效果是如何实现的。我们的研究结果表明，HWAMW促进了重述，并为AI辅助写作提供了一个有价值的范式，在这个范式中，LLM并不是讲述我们的故事，而是帮助我们看到我们所讲故事中的更大潜力。

View on arXiv Download PDF AI Translation

cs.CL / 26 / 2609.12541

Agent as Policy for Robotic Manipulation

代理作为机器人操作的策略

Jia, Mengzhao, Lin, Yang, Zhang, Xixin, Zhang, Zhihan, Liu, Xiaobai, Jiang, Meng

Abstract

We demonstrate that a general-purpose agent can directly drive a physical robot throughout task execution without any task-specific or environment-specific training. We introduce Agent as Policy (AGP), which places task planning and execution under the agent's control. Given a task and a robot interface, the agent interprets visual evidence, writes executable programs, issues motion commands, and revises its actions in response to physical outcomes. This brings the agent's reasoning and programming capabilities into continuous interaction with the physical world. We study AGP across multiple real-world manipulation tasks spanning precision manipulation, dynamic motions, and deformable objects. These include assembly from human videos, block construction from goal images, die reorientation, targeted throwing, and bimanual towel folding. AGP achieves success rates of 100%, 100%, and 80% on three block construction configurations. These findings establish a path for general-purpose agents to act as robotic policies, extending their autonomy to physical manipulation through runtime reasoning, programming, and interaction.

Chinese Translation

我们展示了一种通用代理可以在任务执行过程中直接驱动物理机器人，而无需任何特定于任务或环境的训练。我们引入了代理作为策略（Agent as Policy, AGP），将任务规划和执行置于代理的控制之下。在给定任务和机器人接口的情况下，代理能够解读视觉证据，编写可执行程序，发出运动指令，并根据物理结果修正其行动。这使得代理的推理和编程能力与物理世界进行持续互动。我们在多个真实世界的操作任务中研究了AGP，这些任务涵盖了精确操作、动态运动和可变形物体。这些任务包括从人类视频中进行组装、根据目标图像进行积木构建、模具重新定向、定向投掷和双手折叠毛巾。AGP在三种积木构建配置中取得了100%、100%和80%的成功率。这些发现为通用代理作为机器人策略的应用奠定了基础，通过运行时推理、编程和互动，扩展其在物理操作中的自主性。

View on arXiv Download PDF AI Translation

cs.CL / 27 / 2609.12544

Meddies-PII: A Multilingual Framework for Personally Identifiable Information Extraction in Clinical De-identification

Meddies-PII：一个用于临床去标识化中个人可识别信息提取的多语言框架

Le, Linh Uyen, Hoang, Christian, Ha, Huy Hoang

Abstract

Clinical de-identification relies on accurately identifying personally identifiable information (PII). However, manually annotated datasets are costly to construct, while existing synthetic alternatives often provide limited details about their generation process or rely on relatively simple synthesis strategies. We introduce Meddies-PII-Dataset, a corpus of one million synthetic clinical documents spanning seventeen languages and nine PII labels. The documents are generated using attribute-conditioned prompts and validated through thirteen deterministic gates that enforce structural and annotation consistency. To evaluate the dataset's utility, we train Meddies-PII-Model, a BIOES token classifier, and compare it with existing PII extraction systems using exact-match entity-level F1. Meddies-PII-Model achieves the highest performance among the evaluated systems on all reported benchmarks, with a mean F1 of 0.827 across fifteen external benchmarks, compared with 0.658 for the strongest baseline. Upon acceptance, we will publicly release the dataset, benchmark suite, model, generation framework, and evaluation code to support research on multilingual clinical de-identification.

Chinese Translation

临床去标识化依赖于准确识别个人可识别信息（PII）。然而，手动标注的数据集构建成本高，而现有的合成替代品通常对其生成过程提供有限的细节或依赖于相对简单的合成策略。我们介绍了Meddies-PII-Dataset，这是一个包含一百万份合成临床文档的语料库，涵盖十七种语言和九个PII标签。这些文档是通过属性条件提示生成的，并通过十三个确定性门进行验证，以确保结构和标注的一致性。为了评估数据集的实用性，我们训练了Meddies-PII-Model，一个BIOES标记分类器，并使用精确匹配实体级F1与现有的PII提取系统进行比较。Meddies-PII-Model在所有报告的基准测试中实现了最高的性能，在十五个外部基准上的平均F1为0.827，而最强基线为0.658。接受后，我们将公开发布数据集、基准套件、模型、生成框架和评估代码，以支持多语言临床去标识化的研究。

View on arXiv Download PDF AI Translation

cs.CL / 28 / 2609.12575

Calibrated Ambiguity in Multimodal Language Models: Humans reach for cultural references, while models describe the picture

多模态语言模型中的校准模糊性：人类追寻文化参考，而模型描述图像

Kommers, Cody, Ye, Mingrui, Gius, Evelyn, Mihai, Daniela, Long, Hoyt, Yuan, Zheng, Hemment, Drew

Abstract

Ambiguity is often treated as a bug for AI systems to resolve---but in human communication and culture, ambiguity can also be a generative resource. From humour to politics to art, people express themselves in words and images that are open enough to invite different interpretations, yet constrained enough to be interpretable. We operationalise this notion of calibrated ambiguity with a task drawn from the parlour game Dixit. We compare differences in clues generated by human vs multimodal language models, based on a novel coding rubric for calibrated ambiguity, and find that models consistently exhibit ambiguity collapse (i.e., their outputs are over-specified, leaving no room for multiple legitimate interpretations). Unlike human clues, AI-generated clues also exhibit cultural flattening; they almost never make reference to culturally-situated knowledge, even when prompted to use allusion and figurative language.

Chinese Translation

模糊性通常被视为人工智能系统需要解决的缺陷——但在人的沟通和文化中，模糊性也可以成为一种生成资源。从幽默到政治再到艺术，人们用足够开放的语言和图像表达自己，以邀请不同的解读，同时又足够受限以便于解读。我们通过一个源自桌游Dixit的任务来操作化这种校准模糊性的概念。我们基于一种新颖的校准模糊性编码标准，比较了人类与多模态语言模型生成的线索之间的差异，发现模型始终表现出模糊性崩溃（即，它们的输出过于具体，无法容纳多种合法的解读）。与人类线索不同，人工智能生成的线索还表现出文化扁平化；即使在被提示使用暗示和比喻语言时，它们几乎从不提及文化背景知识。

View on arXiv Download PDF AI Translation

cs.CL / 29 / 2609.12653

SWARM: A Multilingual Human-Annotated Dataset for Russian Propaganda Detection in Search Engine Results

SWARM：一个用于搜索引擎结果中俄罗斯宣传检测的多语言人工标注数据集

Tonneau, Manuel, Dubey, Abhinav, Shaikh, Farhan, Vitulano, Ilaria, Stolze, Martha, Dedeoglu, Hale, Riechert, Clara, Kuka, Ella, Sydorova, Maryna, Makhortykh, Mykola, Kuznetsova, Elizaveta

Abstract

Russian state propaganda spreads across many languages and online spaces. Yet, most computational work examines only one such space, usually social media, in one or two languages, and analyses sources rather than content. We introduce SWARM (Search-Web documents Annotated for Russian propaganda, Multilingual), a dataset of 2,183 search engine results across nine languages and diverse web domains (e.g., news, blogs, government sites), each annotated by trained coders for whether it supports a recurring Russian propaganda narrative. We benchmark a source-based blocklist, supervised classifiers, and zero-shot LLMs against these labels. The blocklist misses most propaganda-supporting documents, because such content is not confined to flagged "propaganda" outlets but also appears on mainstream ones. Content-level analysis helps, though how much depends on the model: the strongest LLM reaches a positive-class F1 of 0.73, whereas the supervised classifiers reach only about 0.5, with the smaller LLMs over-predicting support, mistaking topical relevance for endorsement. Detecting search-borne propaganda thus requires per-language, content-level evaluation, which we hope SWARM and our evaluation code enable.

Chinese Translation

俄罗斯国家宣传在多种语言和在线空间中传播。然而，大多数计算工作仅研究一个这样的空间，通常是社交媒体，且仅限于一两种语言，并且分析的是来源而非内容。我们介绍了SWARM（Search-Web documents Annotated for Russian propaganda, Multilingual），这是一个包含2183条搜索引擎结果的数据集，涵盖九种语言和多样的网络领域（例如，新闻、博客、政府网站），每条结果均由经过培训的编码员标注，以判断其是否支持反复出现的俄罗斯宣传叙事。我们对基于来源的黑名单、监督分类器和零样本大语言模型（LLMs）进行了基准测试。黑名单错过了大多数支持宣传的文档，因为此类内容不仅限于被标记为“宣传”的渠道，还出现在主流渠道上。内容层面的分析有所帮助，但效果取决于模型的性能：最强的大语言模型的正类F1值达到0.73，而监督分类器仅达到约0.5，较小的大语言模型则过度预测支持，将主题相关性误认为是认可。因此，检测搜索引擎中的宣传需要针对每种语言进行内容层面的评估，我们希望SWARM及我们的评估代码能够实现这一目标。

View on arXiv Download PDF AI Translation

cs.CL / 30 / 2609.12655

LifeMem: Enabling Lifelong Experience Reuse for LLM Agents

LifeMem：为大型语言模型代理启用终身经验重用

Qiu, Yuli, Li, Yutong, Su, Wei, Liu, Zeming, Che, Wanxiang, Huang, Heyan, Wang, Haifeng, Guo, Yuang

Abstract

Large language model agents are expected to continuously adapt to new tasks and environments over their lifetime by reusing past experience. However, existing memory-based agents struggle to transfer reusable experience across environments and suffer from catastrophic forgetting as experience accumulated. To address these challenges, we propose LifeMem, a lifelong learning framework that enables agents to transfer knowledge across multiple environments. During learning, LifeMem clusters accumulated interaction trajectories based on underlying workflows to extract reusable skills. When solving a new task at inference time, the agent recalls relevant skills and trajectories to guide actions. To validate our method, we conduct experiments across 10 environments and over 13k tasks with 2k newly annotated interaction trajectories. Results show that LifeMem enables effective experience reuse in lifelong learning, achieving both reduced forgetting on learned tasks and superior cross-task transfer. Further analysis reveals that task streaming impacts learning, while consolidating structurally similar trajectories within memory boosts performance.

Chinese Translation

大型语言模型代理预计将在其生命周期内通过重用过去的经验不断适应新任务和新环境。然而，现有的基于记忆的代理在不同环境之间转移可重用经验方面存在困难，并且随着经验的积累，容易遭遇灾难性遗忘。为了解决这些挑战，我们提出了LifeMem，这是一种终身学习框架，使代理能够在多个环境之间转移知识。在学习过程中，LifeMem根据潜在的工作流程对累积的交互轨迹进行聚类，以提取可重用的技能。在推理时解决新任务时，代理会回忆相关的技能和轨迹以指导行动。为了验证我们的方法，我们在10个环境和超过13,000个任务中进行了实验，并使用2,000个新标注的交互轨迹。结果表明，LifeMem在终身学习中实现了有效的经验重用，减少了对已学习任务的遗忘，并在跨任务转移中表现优越。进一步分析显示，任务流对学习有影响，而在记忆中巩固结构相似的轨迹则提升了性能。

View on arXiv Download PDF AI Translation

cs.CL / 31 / 2609.12674

Doc2FRC: Length-Consistent Document-Level Machine Translation via Fixed-Range Chunking

Doc2FRC：通过固定范围分块实现长度一致的文档级机器翻译

Wang, Xiaotian, Lin, Youyuan, Shen, Zhan, Yanaka, Hitomi

Abstract

Advanced large language models (LLMs) with long context windows can substantially reduce input truncation in document-level machine translation (DocMT). However, direct Doc2Doc translation remains prone to n-gram repetition and progressive quality degradation. A common remedy is to segment the document into finer-grained chunks. Nonetheless, conventional rule-based chunking approaches fail to handle the length distribution mismatch between training and inference. To address this, we introduce Fixed-Range Chunking (FRC), utilizing dynamic programming to partition documents into chunks within a predefined length interval. By consistently applying FRC during training and inference, the input documents of any length are mapped to the same length distribution, substantially reducing train-test length mismatch. Centered on FRC, we propose a lightweight dual-boundary matching algorithm for chunk alignment, alongside four distinct training strategies. Experimental results show that FRC-based fine-tuning substantially improves 7B LLMs over direct Doc2Doc fine-tuning and outperforms existing DocMT methods on IWSLT2017. We further construct GlobVDoc, a 10-language test set independent of mainstream DocMT training sources, and show that FRC improves out-of-distribution document translation.

Chinese Translation

具有长上下文窗口的先进大型语言模型（LLMs）可以显著减少文档级机器翻译（DocMT）中的输入截断。然而，直接的Doc2Doc翻译仍然容易出现n-gram重复和逐步质量下降。一个常见的解决方案是将文档细分为更小的块。然而，传统的基于规则的分块方法无法处理训练和推理之间的长度分布不匹配。为了解决这个问题，我们引入了固定范围分块（Fixed-Range Chunking, FRC），利用动态规划将文档划分为预定义长度区间内的块。通过在训练和推理过程中一致地应用FRC，任何长度的输入文档都被映射到相同的长度分布，从而显著减少训练与测试之间的长度不匹配。基于FRC，我们提出了一种轻量级双边界匹配算法用于块对齐，并提出四种不同的训练策略。实验结果表明，基于FRC的微调在7B LLMs上显著优于直接的Doc2Doc微调，并在IWSLT2017数据集上超越了现有的DocMT方法。我们进一步构建了GlobVDoc，这是一个独立于主流DocMT训练源的10种语言测试集，并显示FRC在分布外文档翻译中的改进效果。

View on arXiv Download PDF AI Translation

cs.CL / 32 / 2609.12791

Cognition on Graph: Navigating Massive Knowledge Space via Cognitive Cycles and Bidirectional Graph-Text Synergy

图上的认知：通过认知循环和双向图-文本协同导航大规模知识空间

Zhou, Gengxian, Xu, Jian, Tang, Zichen, Xiang, Shiming, E, Haihong, Liu, Cheng-Lin

Abstract

Retrieval-Augmented Generation (RAG) has empowered Large Language Models (LLMs) to tackle knowledge-intensive tasks. However, navigating global, heterogeneous knowledge bases (large-scale knowledge graphs and text corpora) for complex reasoning remains a challenge. Existing methods typically employ reactive, graph-driven exploration strategies, which blindly follow graph topology without adapting to the question context or evolving exploration progress, and lack deep bidirectional synergy between graph and text. To address these limitations, we propose CoG (Cognition on Graph), a cognitive-inspired, training-free framework for adaptive knowledge exploration. Drawing inspiration from human problem-solving, CoG performs a continuous plan-explore-reflect cycle, where it proactively formulates investigation plans, performs dual-source retrieval, and dynamically reflects on progress to adjust strategies. Crucially, it establishes deep bidirectional synergy between structured graph and unstructured text, where entities extracted from text dynamically guide graph exploration to bridge knowledge gaps. Extensive experiments on seven multi-hop QA benchmarks demonstrate that CoG significantly outperforms state-of-the-art methods while achieving superior exploration efficiency. Our code and datasets are available at https://github.com/zhougengxian/CoG.

Chinese Translation

检索增强生成（Retrieval-Augmented Generation, RAG）使大型语言模型（Large Language Models, LLMs）能够应对知识密集型任务。然而，对于复杂推理而言，导航全球异构知识库（大规模知识图谱和文本语料库）仍然是一项挑战。现有方法通常采用反应式的图驱动探索策略，盲目遵循图的拓扑结构，而未能适应问题上下文或不断演变的探索进展，并且缺乏图与文本之间的深度双向协同。为了解决这些局限性，我们提出了CoG（Cognition on Graph），一个灵感来源于认知的、无训练的自适应知识探索框架。CoG从人类问题解决中汲取灵感，执行持续的计划-探索-反思循环，其中主动制定调查计划，进行双源检索，并动态反思进展以调整策略。关键在于，它在结构化图和非结构化文本之间建立了深度的双向协同，其中从文本中提取的实体动态引导图的探索，以弥合知识空白。在七个多跳问答基准上的广泛实验表明，CoG显著优于最先进的方法，同时实现了更高的探索效率。我们的代码和数据集可在 https://github.com/zhougengxian/CoG 获取。

View on arXiv Download PDF AI Translation

cs.CL / 33 / 2609.12872

DuplexDrama: A Synthesized Dialogue Dataset with Scenarios, Full-Duplex Behaviors, Expressive Speech, and Sound Events

DuplexDrama：一个包含场景、全双工行为、表现性语音和声音事件的合成对话数据集

Guo, Qingxiang, Fan, Wenke, Zhao, Shuofeng, Yang, Dawei, Zhou, Zhiyang, Shang, Yingxin, Cai, Hongwei, Wang, Zhou, Wang, Weixu, Yang, Lin, Zhou, Shuran, Song, Yang

Abstract

We present DuplexDrama, the first synthesized spoken dialogue dataset that simultaneously covers four dimensions: (i) complete persona and scenario settings; (ii) three full-duplex behaviors (interruption, backchannel, incomplete); (iii) expressive speech with persona-aligned emotion labels; and (iv) script-aware sound events. DuplexDrama is built via a 4-stage pipeline; quality validation on both scripts and synthesized audio confirms its quality. We have produced more than 2,000 hours audio data with a 64-voice timbre pool spanning 13 personas and 5 age buckets; 3.8% of all turns carry at least one full-duplex behavior. This data has been validated through internal full-duplex model training. We will release a curated subset of 6,400 bilingual dialogues (800 h, Chinese ~500 h + English ~300 h) to advance full-duplex spoken dialogue model research. Data samples are available at our demo page and LLM-judge evaluation prompts will be released with the dataset.

Chinese Translation

我们提出了DuplexDrama，这是第一个同时涵盖四个维度的合成口语对话数据集：(i) 完整的人物和场景设置；(ii) 三种全双工行为（打断、回馈、未完成）；(iii) 带有人物对齐情感标签的表现性语音；以及(iv) 脚本感知的声音事件。DuplexDrama通过四个阶段的流程构建；对脚本和合成音频的质量验证确认了其质量。我们制作了超过2000小时的音频数据，包含64种音色，涵盖13个人物和5个年龄组；3.8%的所有对话轮次至少包含一种全双工行为。该数据已通过内部全双工模型训练进行了验证。我们将发布一个经过精心整理的6400个双语对话子集（800小时，中文约500小时 + 英文约300小时），以推动全双工口语对话模型研究。数据样本可在我们的演示页面获取，LLM-judge评估提示将与数据集一起发布。

View on arXiv Download PDF AI Translation

cs.CL / 34 / 2609.12884

MedSNIP: Building and Benchmarking Snippet-Level Granularity for Medical Fact Verification

MedSNIP：构建和基准测试医学事实验证的片段级粒度

Iqbal, Hasan, Ahmad, Sarfraz, Kim, Hyunjae, Park, Sihyeon, Liao, Junjie, Chen, Qingyu, Nakov, Preslav, Wang, Yuxia

Abstract

A medical claim's correctness often depends not on the claim alone, but on the clinical structure around it. A claim may require a lab reference range, a causal or conditional link, or patient-specific details to be judged correctly, and atom-level decomposition can fragment these dependencies, leaving the verifier with clinically incomplete claims. We reformulate medical fact-checking around snippet-level verification, where clause-grouped units preserve local clinical structure. We introduce MedSNIP-Bench, a human-annotated benchmark for snippet-level medical fact verification, and MedSNIP, an automatic snippet-generation pipeline. MedSNIP-Bench covers 276 consumer-health and clinical-vignette responses, segmented into 2,524 snippets with dual in-general and in-patient-context labels and six structural pattern codes. MedSNIP is evaluated against human snippet boundaries on MedSNIP-Bench and then used to generate snippet-level units for external corpora. Across MedSNIP-Bench, HealthFC, and MedHallu, snippet-level verification preserves or improves false-class F1, with gains concentrated where answers are long enough to fragment and where the verifier is strong enough to exploit the recovered structure. The largest merge-pattern gain is on causal-conditional clinical chains. It also reduces verifier calls by 24-73%, though the saving survives end-to-end only when decomposition is cheap, which an open-weight decomposer makes possible at no loss of chunking fidelity.

Chinese Translation

医学声明的正确性往往不仅依赖于声明本身，还依赖于其周围的临床结构。一个声明可能需要实验室参考范围、因果或条件链接，或特定于患者的细节才能被正确判断，而原子级分解可能会破坏这些依赖关系，使验证者面临临床不完整的声明。我们围绕片段级验证重新构建医学事实检查，其中按子句分组的单元保留了局部临床结构。我们引入了MedSNIP-Bench，这是一个经过人工标注的片段级医学事实验证基准，以及MedSNIP，一个自动片段生成管道。MedSNIP-Bench涵盖了276个消费者健康和临床案例响应，分割成2,524个片段，具有一般和患者背景的双重标签和六种结构模式代码。MedSNIP在MedSNIP-Bench上与人工片段边界进行评估，然后用于生成外部语料库的片段级单元。在MedSNIP-Bench、HealthFC和MedHallu中，片段级验证保持或提高了假阳性F1，增益集中在答案足够长以进行分割的地方，以及验证者足够强大以利用恢复的结构的地方。最大的合并模式增益出现在因果-条件临床链上。它还将验证者调用减少了24-73%，尽管这种节省只有在分解成本低时才能在端到端中生存，而开放权重分解器使得这一点成为可能，且不损失分块的保真度。

View on arXiv Download PDF AI Translation

cs.CL / 35 / 2609.12913

Parameter-Efficient Retrievers for Polish and European Languages

面向波兰语和欧洲语言的参数高效检索器

Dadas, Sławomir, Poświata, Rafał, Grębowiec, Małgorzata, Perełkiewicz, Michał

Abstract

Dense retrieval systems increasingly rely on multi-billion-parameter language models, whose memory and computational requirements make large-scale indexing, frequent corpus updates, and low-latency serving costly. We present a three-stage training pipeline for developing compact and efficient retrievers that remain competitive with substantially larger models. The pipeline combines cross-lingual alignment, relational knowledge distillation, and contrastive fine-tuning. It requires no original ground-truth relevance labels, relying exclusively on supervision generated by strong embedding models and rerankers utilised as teachers. Using this pipeline, we develop PolDense and EuroDense, both supporting contexts of up to 8,192 tokens. PolDense is a family of six Polish retrievers ranging from 17M to 1B parameters. EuroDense is a 435M-parameter retriever supporting nine European languages. We conduct an extensive evaluation covering 41 Polish and 150 multilingual retrieval tasks. The results demonstrate strong quality-efficiency trade-offs. PolDense-1B outperforms the evaluated retrievers with up to 9B parameters, while the PolDense family forms the Pareto frontier across model sizes. Among the evaluated models below 1B parameters, EuroDense ranks first in both task-averaged and language-averaged performance and leads in seven of nine languages. We release all models publicly.

Chinese Translation

密集检索系统日益依赖于数十亿参数的语言模型，这些模型的内存和计算需求使得大规模索引、频繁的语料库更新和低延迟服务的成本高昂。我们提出了一种三阶段训练流程，用于开发紧凑且高效的检索器，使其在性能上与显著更大的模型相竞争。该流程结合了跨语言对齐、关系知识蒸馏和对比微调。它不需要原始的真实相关性标签，而是完全依赖于由强嵌入模型和作为教师的重排序器生成的监督信息。通过该流程，我们开发了PolDense和EuroDense，均支持最多8,192个标记的上下文。PolDense是一个包含六个波兰检索器的系列，参数范围从1700万到10亿。EuroDense是一个435M参数的检索器，支持九种欧洲语言。我们进行了广泛的评估，涵盖了41个波兰语和150个多语言检索任务。结果显示出强大的质量-效率权衡。PolDense-1B在评估的检索器中表现优于参数高达90亿的模型，而PolDense系列在模型规模上形成了帕累托前沿。在评估的参数低于10亿的模型中，EuroDense在任务平均和语言平均性能上均排名第一，并在九种语言中的七种语言中领先。我们将所有模型公开发布。

View on arXiv Download PDF AI Translation

cs.CL / 36 / 2609.12915

LLM-Enhanced Dual-Branch Learning for Large-Scale Multi-Label Text Classification

基于LLM增强的双分支学习用于大规模多标签文本分类

Ye, Hui, Zhang, Jing, Yang, Xiulong, Sunderraman, Rajshekhar

Abstract

Large-scale multi-label text classification assigns a small subset of relevant labels to each document from a vocabulary containing thousands or tens of thousands of candidate labels. Although pretrained language models have improved semantic text representations, most representation-based approaches center their prediction pipelines on a primary encoder or combine auxiliary features within a single ranker. The complementarity between heterogeneous language models therefore remains insufficiently explored. We propose DualMLC, a dual-branch framework that processes the same document through an autoregressive decoder-only language model and a bidirectional encoder. Each branch maintains its own representation pathway and independently estimates relevance scores over the shared label space. DualMLC combines the two score vectors through late logit fusion, allowing shared evidence to reinforce relevant labels and branch-specific evidence to compensate for limitations in the other branch's representation. DualMLC achieves state-of-the-art results on three widely used large-scale multi-label text classification benchmarks. Ablation results further confirm that integrating the heterogeneous predictors produces stronger rankings than either branch alone. The source code is publicly available at https://github.com/huiyegit/DualMLC.

Chinese Translation

大规模多标签文本分类从包含数千或数万个候选标签的词汇中，为每个文档分配一小部分相关标签。尽管预训练语言模型改善了语义文本表示，但大多数基于表示的方法将其预测流程集中在一个主要编码器上，或在单一排序器中结合辅助特征。因此，异构语言模型之间的互补性尚未得到充分探索。我们提出了DualMLC，一个双分支框架，通过自回归解码器语言模型和双向编码器处理相同的文档。每个分支维护其自己的表示路径，并独立地在共享标签空间上估计相关性分数。DualMLC通过后期对数融合结合两个分数向量，使共享证据能够增强相关标签，而分支特定证据则弥补另一个分支表示的局限性。DualMLC在三个广泛使用的大规模多标签文本分类基准上实现了最先进的结果。消融实验进一步确认，整合异构预测器比单独使用任一分支产生更强的排名。源代码已公开，地址为 https://github.com/huiyegit/DualMLC。

View on arXiv Download PDF AI Translation

cs.CL / 37 / 2609.12960

Fewer Words, Not Fewer Tokens: Measuring the Sanskrit Tokenization Penalty per Proposition

更少的词，而不是更少的标记：每个命题的梵语标记化惩罚测量

Sharma, Devansh

Abstract

Sanskrit fuses case, number, person and tense into word endings and chains clauses into compounds, so it is information-dense per word. Whether that density survives subword tokenization is a separate question, to be asked per unit of meaning rather than per word. On identical FLORES-200 devtest content, Sanskrit costs 1.774-2.187 times the English tokens under deployed tokenizers with vocabularies of 200,019 ids or more, but only 1.325-1.353 times the Hindi tokens. Against a deployed English tokenizer, Sanskrit-trained BPE arms then look cheaper per proposition than English on contemporary prose (0.887). Against a matched English control, the same algorithm and vocabulary trained on the English side of the same corpus, that flip disappears: at 32,000 and 64,000 pieces all 8 matched pairs, each size-matched arm against both a pair-matched and a byte-matched control, sit above 1.0 on prose with 95% intervals excluding it. The gap closes as the vocabulary grows: at 128,000 pieces the BPE pair reads 0.983 in domain while staying above parity out of domain (1.025) and on FLORES (1.116). The ratio factorises into a character-length ratio and a tokens-per-character ratio, the second near 1 throughout: what survives matched tokenization is character-level length, which Sanskrit prose lacks over English in SLP1 (1.028) and Sanskrit verse has (0.596). The robust statement is about deployed practice: on contemporary prose and on FLORES, with the Sanskrit side in SLP1 against the deployed o200k English pivot, Sanskrit costs 1.831-2.899 English tokens per proposition under the tokenizers people actually ship. Code, the results snapshot and every table here are public.

Chinese Translation

梵语将格、数、人称和时态融合到词尾中，并将从句链入复合词，因此每个词的信息密度较高。该密度是否在子词标记化过程中得以保留是一个独立的问题，应根据意义单位而非单词进行考量。在相同的 FLORES-200 开发测试内容下，梵语在使用词汇量为 200,019 个 ID 或更多的已部署标记器时，其标记成本为英语的 1.774-2.187 倍，但仅为印地语的 1.325-1.353 倍。与已部署的英语标记器相比，经过梵语训练的 BPE（Byte Pair Encoding）模型在当代散文中每个命题的成本看起来比英语便宜（0.887）。与一个匹配的英语控制组相比，使用相同算法和词汇在同一语料库的英语部分进行训练时，这种反转消失：在 32,000 和 64,000 个标记的情况下，所有 8 对匹配组在与配对匹配和字节匹配的控制组比较中，均在散文上超过 1.0，且 95% 的置信区间不包括 1.0。随着词汇量的增加，差距缩小：在 128,000 个标记时，BPE 配对在领域内的读数为 0.983，而在领域外（1.025）和 FLORES（1.116）中均保持在平价以上。该比率可分解为字符长度比和每字符标记数比，后者在整个过程中接近 1：在匹配标记化中存活下来的特征是字符级长度，而在 SLP1 中，梵语散文在这一点上低于英语（1.028），而梵语诗歌则具有（0.596）。这一强有力的声明是关于已部署实践的：在当代散文和 FLORES 上，梵语部分在 SLP1 中与已部署的 o200k 英语基准相比，梵语在实际使用的标记器下每个命题的成本为 1.831-2.899 个英语标记。代码、结果快照和此处的每个表格均为公开可用。

View on arXiv Download PDF AI Translation

cs.CL / 38 / 2609.13003

Judging by the Cover: Cleaning LLM Truthfulness Benchmarks to Avoid Surface-Level Feature Leakage

以表面为准：清理大型语言模型的真实性基准以避免表面特征泄漏

Namjoo, Foad, Ogasawara, Remy, Abdullah, Amirali, Anderson, Cullen, Oozeer, Narmeen Fatimah, Phillips, Jeff M.

Abstract

Binary-choice truth benchmarks ask models to choose between a correct and an incorrect answer, but if the two answers differ systematically in surface-level features, models can exceed chance without performing the intended reasoning. We show that this failure mode is detectable and can be exploited by downstream classifiers. In TruthfulQA, a simple six-feature logistic classifier achieves substantial accuracy in separating correct from incorrect answers. We further show that similar surface-level artifacts are present in additional benchmarks. To counteract this, we developed a general mechanism to clean them by removing the most leakage-reinforcing pairs. We release a version of TruthfulQA with surface-feature leakage reduced close to chance and provide a mechanism, Audit-Prune, so that the datasets can be cleaned before release.

Chinese Translation

二元选择的真实性基准要求模型在正确答案和错误答案之间进行选择，但如果这两个答案在表面特征上系统性地存在差异，模型可能在没有进行预期推理的情况下就超过随机猜测的准确率。我们展示了这种失败模式是可检测的，并且可以被下游分类器利用。在 TruthfulQA 中，一个简单的六特征逻辑回归分类器在区分正确答案和错误答案方面取得了显著的准确率。我们进一步表明，类似的表面特征伪影在其他基准中也存在。为了应对这一问题，我们开发了一种通用机制，通过移除最具泄漏增强性的配对来清理这些基准。我们发布了一个版本的 TruthfulQA，其表面特征泄漏降低到接近随机水平，并提供了一种机制，Audit-Prune，以便在发布前对数据集进行清理。

View on arXiv Download PDF AI Translation

cs.CL / 39 / 2609.13005

Tasks over Application Manuals: Revealing Gaps in Long-Horizon Procedural Reasoning for Language Models

应用手册中的任务：揭示语言模型在长时间程推理中的不足

Soni, Utkarsh, Murtaza, Syed Shariyar, Nie, Yifan, Chandrasekhar, Sachin, Wen, Eugene

Abstract

Large language models (LLMs) have achieved strong performance on a wide range of natural language tasks, and recent benchmarks suggest that they are increasingly adept at multi-hop reasoning. However, these benchmarks are typically short-horizon, requiring only a small number of retrieval or inference steps, and provide limited evidence of reliability on real-world tasks that involve following manuals spanning hundreds of pages with complex, interdependent guidelines. In this paper, we introduce Tasks over Application Manuals (TAM), a benchmark for evaluating long-horizon procedural reasoning. We construct TAM by curating real-world tasks from two domains: ICD-10-CM clinical coding (mapping medical conditions to diagnostic codes) and U.S. federal sentencing (computing crime sentencing guideline outcomes, specifically offense levels), with human-validated labels. Each task requires following an authoritative manual with tens of thousands of rules and executing a sequence of interdependent steps across different sections to produce an exact answer. We evaluate general-purpose prompting approaches, including retrieval-augmented generation, ReAct-style prompting, and an agent-harness baseline on GPT-5, and find that the best exact-match performance remains extremely low: 1% on ICD-10-CM coding and 15.5% on sentencing tasks. These results show that current benchmarks may overestimate LLM reasoning ability and miss a key challenge: reliably following long, rule-based procedures. The complete TAM data and code are publicly available.

Chinese Translation

大型语言模型（LLMs）在广泛的自然语言任务中表现出色，最近的基准测试表明它们在多跳推理方面日益娴熟。然而，这些基准通常是短时间程的，仅需少量的检索或推理步骤，并且在涉及遵循数百页的手册、具有复杂相互依赖指南的现实世界任务中，提供的可靠性证据有限。在本文中，我们介绍了应用手册中的任务（Tasks over Application Manuals, TAM），这是一个用于评估长时间程程序推理的基准。我们通过从两个领域策划真实世界任务来构建TAM：ICD-10-CM临床编码（将医疗状况映射到诊断代码）和美国联邦量刑（计算犯罪量刑指南结果，特别是犯罪等级），并附上经过人工验证的标签。每个任务都需要遵循一个包含数万个规则的权威手册，并在不同部分执行一系列相互依赖的步骤以产生准确答案。我们评估了通用提示方法，包括检索增强生成、ReAct风格提示和在GPT-5上的代理基础线，并发现最佳的准确匹配表现仍然极低：ICD-10-CM编码任务为1%，量刑任务为15.5%。这些结果表明，当前基准可能高估了LLM的推理能力，并忽视了一个关键挑战：可靠地遵循长时间程的基于规则的程序。完整的TAM数据和代码已公开提供。

View on arXiv Download PDF AI Translation

cs.CL / 40 / 2609.13045

Kraken: LLM-based Speech-to-Speech Translation via Low-bitrate VQ and Dual-path Source Conditioning

Kraken：基于LLM的低比特率向语音翻译的语音到语音翻译

Futami, Hayato, Shahmohammadi, Hassan, Dhyani, Tushar, Koudounas, Alkis, Lafargue, Raphaël, Kashiwagi, Yosuke, Jodelet, Quentin, Tsunoo, Emiru

Abstract

Speech-to-speech translation (S2ST) has advanced significantly with speech LLMs, offering the potential for joint optimization and preserving non-linguistic information. However, these models struggle with predicting high-bitrate speech tokens in LLMs, and face the challenge of relying on S2ST training data with ideally aligned speaker identity and prosody. We propose using low-bitrate tokens based on single-layer vector quantization, trained to reconstruct self-supervised learning (SSL) features. We also employ a separate token-to-waveform decoder named Autowave-X, which is also conditioned on the source speech to improve non-linguistic transfer, thereby relaxing the training data constraints. With the integration of these techniques, we propose an S2ST model named Kraken, which augments a pre-trained LLM with speech feature inputs and the low-bitrate token outputs, followed by Autowave-X vocoder. We built the model upon Qwen3-8B and trained it using 150k hours of multilingual and multitask speech data. We demonstrated that our model exhibited better translation quality than SeamlessM4T-Large v2 and Qwen2.5-Omni, along with improved speaker and prosody transfer capabilities.

Chinese Translation

语音到语音翻译（S2ST）在语音大语言模型（LLM）的推动下取得了显著进展，提供了联合优化和保留非语言信息的潜力。然而，这些模型在LLM中预测高比特率语音标记时面临困难，并且依赖于理想对齐的说话者身份和韵律的S2ST训练数据。我们提出使用基于单层向量量化的低比特率标记，训练以重构自监督学习（SSL）特征。我们还采用了一个名为Autowave-X的独立标记到波形解码器，该解码器也基于源语音进行条件处理，以改善非语言信息的传递，从而放宽训练数据的约束。通过整合这些技术，我们提出了一个名为Kraken的S2ST模型，该模型增强了预训练的LLM，结合了语音特征输入和低比特率标记输出，随后使用Autowave-X声码器。我们在Qwen3-8B的基础上构建了该模型，并使用150k小时的多语言和多任务语音数据进行了训练。我们证明我们的模型在翻译质量上优于SeamlessM4T-Large v2和Qwen2.5-Omni，并且在说话者和韵律传递能力上也有所改善。

View on arXiv Download PDF AI Translation

cs.CL / 41 / 2609.13058

Expert-Space Exploration in MoE Reinforcement Learning

MoE强化学习中的专家空间探索

He, Hongyi, Lin, Zhenghao, Liu, Xiao, Cheng, Peng, Lu, Yan, Gong, Yeyun

Abstract

Reinforcement learning (RL) has become central to post-training of large language models. Recent advances in RL for Mixture-of-Experts (MoE) models have primarily focused on improving optimization stability and training efficiency, while treating the expert selection as a fixed component. Since routing determines the sparse computation paths that induce output distributions, expert selection offers an additional source of rollout diversity. Through empirical analysis, we find that perturbing expert routing effectively alters model output and increases rollout diversity, which is similar to increasing the decoding temperature. However, direct perturbation can activate unsuitable experts and substantially degrade rollout quality. Motivated by these observations, we introduce Expert-Space Exploration Reinforcement Learning (ESRL), an architecture-aware framework that explicitly explores the expert-routing space of MoE models. ESRL preserves high-confidence experts as anchors, and restricts stochastic routing to a plausible candidate pool, thereby retaining reliable computation paths. The perturbation strength is further adapted according to router entropy to avoid over-perturbation. To mitigate the routing mismatch introduced by perturbation, ESRL records the expert paths used during rollout and replays them during policy optimization. Experiments demonstrate that ESRL achieves the best performance across MoE backbones with top-K, top-1, and shared-expert routing, as well as across mathematics, science, and code tasks without additional sampling or computational cost. Specifically, ESRL on Qwen3-30B-A3B achieves the best among all compared methods, improving average Pass@1 and Pass@8 over GRPO by 3.2 and 4.5 percentage points, respectively. Further analyses of expert utilization and training dynamics provide insights into how exploiting MoE-specific routing structure benefits RL training.

Chinese Translation

强化学习（RL）已成为大型语言模型后训练的核心。近期在混合专家（Mixture-of-Experts, MoE）模型中强化学习的进展主要集中在提高优化稳定性和训练效率，同时将专家选择视为固定组件。由于路由决定了稀疏计算路径，从而引发输出分布，专家选择提供了额外的回滚多样性。通过实证分析，我们发现扰动专家路由有效地改变模型输出并增加回滚多样性，这类似于增加解码温度。然而，直接扰动可能激活不合适的专家，并显著降低回滚质量。基于这些观察，我们提出了专家空间探索强化学习（Expert-Space Exploration Reinforcement Learning, ESRL），这是一个架构感知框架，明确探索MoE模型的专家路由空间。ESRL将高置信度专家作为锚点，并将随机路由限制在合理的候选池中，从而保留可靠的计算路径。扰动强度进一步根据路由器熵进行调整，以避免过度扰动。为了减轻扰动引入的路由不匹配，ESRL记录回滚过程中使用的专家路径，并在策略优化期间重放这些路径。实验表明，ESRL在使用top-K、top-1和共享专家路由的MoE骨干网络中实现了最佳性能，并且在数学、科学和代码任务中没有额外的采样或计算成本。具体而言，ESRL在Qwen3-30B-A3B上在所有比较方法中表现最佳，分别提高了平均Pass@1和Pass@8相较于GRPO的3.2和4.5个百分点。对专家利用率和训练动态的进一步分析提供了关于如何利用MoE特定路由结构来促进RL训练的见解。

View on arXiv Download PDF AI Translation

cs.CL / 42 / 2609.13117

Continue, Adapt, or Yield: In-Turn Adaptation to Overlapping Speech in Full-Duplex Agents

继续、适应或让步：全双工代理中的轮流适应重叠语音

Lu, Yunqi, Baumgartner, Tyler, Johri, Nikhil, Tai, Brandon, Fan, Candice, Debaupte, Luc, Aguilar, Ruben, Wang, Bill, Zhong, Yi

Abstract

Full-duplex evaluation often emphasizes whether an agent keeps speaking or stops. That binary cannot express a third response humans use routinely: continuing to speak while incorporating what the listener just contributed. The contribution may be a missing word, a correction or a clarification. We introduce Duplex Cue, an evaluation of this \emph{in-turn adaptation} in full-duplex voice agents. Duplex Cue separates listener intent (backchannel, collaboration, or interruption) from speaker behavior: continuing unchanged, adapting within the turn, or yielding. Adaptation includes acknowledgment as well as content revision. In a single-model case study using 300 human-confirmed cues from unscripted English conversations, we compare recorded human responses with PersonaPlex continuations generated while replaying the listener's audio. We retain 208 pairs with the ongoing speaker active at cue onset and a scorable response in each condition. On the 66 collaborative pairs, recorded speakers adapt in 68.2\% of cases, compared with 34.8\% for PersonaPlex. The model otherwise continues unchanged (42.4\%) or yields (22.7\%). These findings show why evaluating natural voice interaction requires measuring how an agent responds to a listener's contribution as well as whether it keeps speaking.

Chinese Translation

全双工评估通常强调代理是继续说话还是停止。这种二元选择无法表达人类常用的第三种反应：在继续说话的同时融入听者刚刚提供的信息。这种贡献可能是缺失的词语、修正或澄清。我们引入了Duplex Cue，这是对全双工语音代理中这种 extit{轮流适应}的评估。Duplex Cue将听者意图（回馈、协作或打断）与说话者行为（保持不变、在轮流中适应或让步）分开。适应包括确认以及内容修订。在一个使用300个经过人类确认的提示的单模型案例研究中，我们比较了录制的人类反应与在重放听者音频时生成的PersonaPlex延续。我们保留了208对在提示开始时正在发言的说话者和每种条件下都有可评分反应的配对。在66个协作配对中，录制的说话者在68.2\%的情况下进行了适应，而PersonaPlex的适应率为34.8\\%。该模型在其他情况下保持不变（42.4\\%）或让步（22.7\\%）。这些发现表明，评估自然语音交互需要测量代理如何响应听者的贡献，以及它是否继续说话。

View on arXiv Download PDF AI Translation

cs.CL / 43 / 2609.13141

SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking

SAS：通过端到端优化上下文排名实现简单注意力稀疏化

Li, Zhiwei, Zhu, Lei, Gu, Hao, Hu, Xiang, Wang, Yan, Mi, Haitao, Han, Sirui, Liang, Leo, Guo, Zhijiang

Abstract

Post-training attention sparsification reduces the quadratic cumulative attention cost of pretrained Transformers by selecting a small set of context units (tokens or blocks) for each query. Existing trainable methods usually use a lightweight selector to score context units, followed by hard Top-K selection that blocks gradients from the language modeling loss. Consequently, these methods commonly distill layer-wise dense attention distributions. Although this encourages the selector to rank context units by dense attention weights in the original model, the ranking is not directly aligned with their impact on predictions under a fixed attention budget (i.e., the number of attended context units per query), potentially wasting the limited budget on less useful units. To address this misalignment, we propose Simple Attention Sparsification (SAS), a gated sparse attention mechanism that optimizes context ranking end-to-end with the language modeling loss. The key idea is to inject the selector's continuous scores into attention logits during training, allowing the loss to update the selector through standard backpropagation. We identify several choices crucial for this simple design to work well in practice: placing the gate inside the attention softmax in log form, using normalized softmax gates to calibrate historical context against the always-retained current block, and preserving continuous selector scores so the model learns relative priorities rather than only hard selections. To support long-sequence training, we implement a memory-efficient Triton kernel that integrates SAS into FlashAttention-style computation. Across reasoning, long-context understanding, and agentic tasks, SAS consistently outperforms trainable sparse attention baselines across attention budgets, with especially large gains under tight budgets, demonstrating more effective context ranking for downstream tasks.

Chinese Translation

后训练注意力稀疏化通过为每个查询选择一小组上下文单元（标记或块），降低了预训练变换器的二次累积注意力成本。现有的可训练方法通常使用轻量级选择器对上下文单元进行评分，随后进行硬性Top-K选择，这会阻碍语言建模损失的梯度传播。因此，这些方法通常提炼层级稠密注意力分布。尽管这鼓励选择器根据原始模型中的稠密注意力权重对上下文单元进行排名，但这种排名并未与在固定注意力预算下（即每个查询的关注上下文单元数量）对预测的影响直接对齐，可能会浪费有限的预算在不太有用的单元上。为了解决这种不对齐问题，我们提出了简单注意力稀疏化（SAS），这是一种门控稀疏注意力机制，通过语言建模损失端到端优化上下文排名。其关键思想是在训练期间将选择器的连续评分注入注意力logits中，使损失能够通过标准反向传播更新选择器。我们确定了几个对该简单设计在实践中良好运作至关重要的选择：将门控放置在注意力softmax的对数形式中，使用归一化softmax门控来校准历史上下文与始终保留的当前块，并保留连续选择器评分，使模型学习相对优先级而不仅仅是硬性选择。为了支持长序列训练，我们实现了一个内存高效的Triton内核，将SAS集成到FlashAttention风格的计算中。在推理、长上下文理解和代理任务中，SAS在各种注意力预算下始终优于可训练稀疏注意力基线，尤其在紧张预算下获得了显著提升，展示了对下游任务更有效的上下文排名。

View on arXiv Download PDF AI Translation

cs.CL / 44 / 2609.13144

Type Diversity Enables Transformers to Generalise Compositionally

类型多样性使变压器能够进行组合性泛化

Moisio, Anssi, Creutz, Mathias, Kurimo, Mikko

Abstract

Compositional generalisation has been divided into lexical and structural generalisation. Previous work has found that structural generalisation is harder than lexical for Transformers. We propose that this difference is not inherent to Transformers, but due to the high diversity of lexical types and low diversity of structural types in the specific datasets of these previous works. By type diversity we mean the number of different constructors of that type, instead of, for example, the specific word combinations that might populate the structure. To test this, we vary the amounts of type diversity of lexical and structural types in previously published datasets. We create linguistically diverse variants of the COGS and SLOG datasets using Grammatical Framework. We find that type diversity correlates with compositional generalisation equally in lexical and structural test cases, supporting our hypothesis. We note a contradiction with the proposition in previous work that compound divergence explains the difficulty in compositional generalisation tasks. We further investigate the effects of other dataset properties on compositional generalisation, such as the diversity of types other than the novel test structure, and surface properties of the logical semantics format.

Chinese Translation

组合性泛化被分为词汇泛化和结构泛化。以往的研究发现，对于变压器而言，结构泛化比词汇泛化更为困难。我们提出，这一差异并非变压器固有的，而是由于以往研究中特定数据集中词汇类型的高多样性和结构类型的低多样性所致。我们所说的类型多样性是指该类型不同构造器的数量，而不是例如可能填充结构的特定词组合。为了验证这一点，我们在以往发布的数据集中改变词汇和结构类型的多样性。我们使用语法框架（Grammatical Framework）创建了COGS和SLOG数据集的语言多样性变体。我们发现，类型多样性在词汇和结构测试案例中的组合性泛化上具有相同的相关性，支持了我们的假设。我们注意到与以往研究中提出的复合发散解释组合性泛化任务困难的观点存在矛盾。我们进一步探讨了其他数据集属性对组合性泛化的影响，例如除了新颖测试结构之外的类型多样性，以及逻辑语义格式的表面属性。

View on arXiv Download PDF AI Translation

arXiv Papers

Scenario-Independent Criticality Assessment and Prediction for Vulnerable Road Users in Autonomous Driving

Pelican-Sim 1.0: A General World Model Simulator for Embodied Intelligence

MoPA: Coordinated Mobile Manipulation via Subsystem-Specific Perception Alignment

Uncertainty-Aware Conflict Detection Against Operator-Conditioned Weather Hazards

RodForesight: A World Model Enhanced Diffusion Policy for Slender and Material Agnostic Rod Insertion

Multi-Objective Agent-Based Model Predictive Controller for Plug-and-Play Vehicle Control

A Data-Driven Distributed Control Scheme: Learning Multi-Objective Agent-Based MPC for Path-Tracking

A Physics-Based Closed-Loop Robotic Bioprinting Framework Towards Volumetric Muscle Loss Treatment

Battery-Aware Predictive Trajectory Planning and Control for Multirotors Under Disturbances

An Automated Thickness Evaluation Procedure Using an Integrated Structured Light 3D Camera in a Robotic Bioprinting Framework

Guardrailed Meta-Agent Loops: Stress-Testing Policy Pinning, Budget Bounds, and Crash Recovery

Chain-SLAM: Globally Consistent Backend for Multi-Session LiDAR SLAM via Chained Loop Closure

DIA: Denoising Intermediate Advantage for Diffusion Policy Optimization

Trajectory Bundle Method in SE(3) for Black-Box Fixed-Wing Aircraft Trajectory Optimization

Pneumatic neurons for soft robots enable inflate-and-fire networks for rhythmic motion

AnchorVLN: Geometry-Anchored Vision-Language Grounding Reasoning for Open-Vocabulary Navigation

Mission Performance: Automatic and Adaptive Race Pace Progression for Autonomous Racing

DATAFARM: Distribution-Aligned Task and Motion Planning for Fine-Tuning Vision-Language-Action Models

DWMP: Leveraging Dual World Models for Humanoid Obstacle Traversal

A Deployable Architecture for Robot-Mediated Tasks (DART): Evaluation in Socially Assistive Robot-Guided Cognitive Behavioral Therapy Exercises

READ: Learning Risk-Informed Fields for End-to-End Autonomous Driving

Understanding Whole-Body Robot Teleoperation Strategies Under Diverse Task Objectives and Constraints

FoldNet++: a Large-Scale Synthetic Dataset for Robotic T-Shirt Folding and Unfolding

IMPLY: Physically Anchored Consistency for World-Model Rollouts

PATH: Continuous Target Sensing among Autonomous Cooperative Drones

ArtManip: Category-Level Articulated In-Hand Manipulation

Communication-Constrained Multi-Robot Exploration With Adaptive Communication Windows

Autonomous Precision Milling of Biological Structures via Generic Anatomical Priors and Active Boundary Perception

STAR: Sparse Tactile Representation Learning in Vision-Tactile-Language-Action Models for Dexterous Manipulation

A Hierarchical Coverage Path Planning Algorithm for Unknown Environments

Quantifying Spectral Differences in Vehicle Between Production Autonomous and Human-Driven Vehicles Across Driving Scenarios

ProClosure: Hierarchical Room-Object Assignment using Progressive Boundary Closure from Monocular Video

Online Material Estimation for Conditioned Diffusion Policy in Shaping Deformable Linear Objects

Breaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Models

Driving Context-guided Model Predictive Planning and Control for Autonomous Car Racing at the Limit and Beyond

Size Doesn't Matter: Material-State Reinforcement Learning for Excavator Transferable Soil Manipulation

Improving Imitation Learning Efficiency for Manipulation through Geometric Prior Pretraining

Control Architecture for Safe Grasping of Fragile Objects Using a Coarse Position-Controlled Gripper

High-Fidelity Multi-Body Simulator for Autonomous Racing

VertexCBF: Improving Neural Control Barrier Functions via Vertex-Restricted Control Search

Parameter Sensitivity Analysis for Aerial LiDAR-Inertial Odometries in low-altitude flights

A Multi-Vehicle Dataset with Camera, LiDAR, and Radar Sensors and Scanned 3D Models for Custom Auto-Annotation using RTK-GNSS

From Transportation to Manipulation: Enabling Grasping in Magnetic Robotics

Before the Tipping Point: Force-Guided Active Perception for Shape-Agnostic Estimation of 3D Centers of Mass

Robust Underwater Grasping of Sloped Objects with a Waterproof Passive Adaptive Gripper

ARC: Autonomous Robotics Compliance A Three-Layer Governance Architecture for Deployed Autonomous Systems

A Robot Among People:From Social Imitation to the Social Becoming of Human Groups

Distributed Stochastic Optimal Control for Pattern-Oriented Swarms

Tuning ROS 2 for Energy-Efficient Navigation: Empirical Insights from Costmap 2D Configurations

Comfort by Construction: Adaptive, Comfort-Bounded Action Spaces for Learned Driving Policies

Global Path Planner with Multi-Model Switching

Dynin-Robotics: Omnimodal Unified Diffusion Vision-Language-Action Model

ASTRIL-MPC: Autonomous Traversal Framework of Articulated Tracked Robots with Language-Guided Neural-Kinematic MPC

Feature Recovery for Object Understanding After Irreversible Fire Damage

Does Video Memory Use What It Retrieves? A Causal Audit of Memory Specificity

Beyond Argmax: A Mechanistic Study of Semantic Retention in Frozen Foundation-Model Composition for Generalized Few-Shot 3D Segmentation

HSI-Road Relabeled: Surface-Aware Road-Scene Segmentation

Single-Query Person-Centric Bimanual Hand-Object Interaction Detection

USPLIT-VQA: U-Shaped Split Learning for Visual Question Answering with Contribution-Aware Weighted Aggregation

When Ground-Truth Fidelity Matters: An Orchestrated UAS Framework for Wheat Streak Mosaic Virus Detection Using Vision Transformers and Machine Learning

Physics as the label for measuring and correcting materials reasoning in multimodal models

QuPAINT: Physics-Aware Multimodal Reasoning for Quantum Material Characterization

Revisiting Multi-Object Tracking Baselines: Hyperparameter Optimization with Multi-Fidelity Greedy Coordinate Search

UniMo: Unifying Human and Animal Motion Generation

VS-Splat: Voxel-Selective feed-forward Gaussian Splatting for end-to-end 3D object reconstruction from sparse-views

EgoMaize: A First-Person Maize Instance Segmentation Benchmark under Severe Field Occlusion

Context-Aware Causal Gaze Forecasting for Human-Vehicle Interaction During In-Cabin Tracking Dropouts

UFO: Chain-of-Evaluation for Omni-Condition Alignment in Multi-Modal Image Generation

GSO-Net: Visual State Machines for Hazardous Freight Transfer Compliance at Petrochemical Logistics Nodes

OphBiWSSD: Scaling Temporal Action Localization in Ophthalmic Surgeries with Bidirectional Weight-tied State Space Duality

A Multimodal Explainable Deep Learning Framework for Alzheimer's Disease Diagnosis using 3D Magnetic Resonance Imaging and Clinical Data

DERA: Detached Edge-Residual Adaptation for Prohibited item Detection

Spectral Consistency-Guided Multiview Point Cloud Registration for Low-Overlap Scenes

An End-to-End Automated Pipeline for Controllable Crack Data Synthesis

Bridging Vision Foundation Model Priors with CLIP for Spatial-aware Few-shot Anomaly Detection in Medical Images

Partition-Invariant Tuning for 3D Scene Understanding

PhysioAI: Clinical Knowledge-Guided Semantic Supervision for Skeleton-Based Physiotherapy Action Recognition

RoES: Rotational Equivariant Selective-frequency Fusion for Multimodal Images

LettuceVisSim: A Simulator That Generates Lettuce Image Time-series for Vision-Based Reinforcement Learning