cs.RO / 1 / 2609.01662
Not All Agreement Counts as Corroboration: Provenance-Conserving Multi-View Fusion for Typed Action Admission in Human-Robot Collaboration
并非所有的协议都算作证实：保留来源的多视角融合用于人机协作中的类型化动作接纳
Abstract
For embodied systems, predictive agreement alone does not determine whether evidence warrants action; evidential origin matters. Repeated inference over one observation can multiply agreement without adding evidence, while source-local values do not reveal whether outputs have separately countable origins. PACT treats evidence countability as a relational variable for provenance-conserving fusion and typed action admission. A supplied provenance partition defines countable units. PACT retains coordinatewise support shared within each unit, accumulates only across units, and maps unmet release conditions to hold, confirm, or fallback. Under the stated assumptions, source-local values cannot identify countability; the coordinatewise meet is the greatest budget satisfying singleton fidelity and insertion non-amplification, with coarsening monotonicity and fixed-partition stability. Across 31,200 evaluations in 48 scene clusters, PACT attains a common-support normalized risk-coverage area (ncsAURC) of 0.0861. Excluding the constructed adversarial-consensus arm, provenance-partition aggregation reduces ncsAURC by 0.0557 relative to singleton aggregation, while the corroboration contrast vanishes. On complete-source records, native scores favor PACT, but a common posterior-peak score narrows its difference from nested Dirichlet and favors product fusion. Reassigning provenance over unchanged predictions moves evidence budgets as predicted. In offline human-robot collaboration, eightfold within-camera duplication leaves 720 typed responses per checkpoint unchanged; camera-grouped PACT admits 47 of 57 Qwen3-VL-32B reference-consistent candidates with no observed reference-inconsistent admission in 60 episodes. PACT separates computational from evidential multiplicity: agreement constitutes corroboration only when provenance permits separate accumulation.
Chinese Translation
对于具身系统，仅有预测一致性并不能决定证据是否足以支持行动；证据的来源同样重要。对同一观察的重复推理可能会增加一致性而不增加证据，而源本地值并不能揭示输出是否具有可单独计数的来源。PACT将证据的可计数性视为保留来源融合和类型化动作接纳的关系变量。提供的来源划分定义了可计数单元。PACT保留每个单元内共享的坐标支持，仅在单元之间进行累积，并将未满足的释放条件映射为保持、确认或回退。在所述假设下，源本地值无法识别可计数性；坐标相遇是满足单例保真度和插入非放大性的最大预算，具有粗化单调性和固定划分稳定性。在48个场景集群中的31,200次评估中，PACT达到了0.0861的共同支持归一化风险覆盖面积（ncsAURC）。排除构造的对抗共识臂后，来源划分聚合相较于单例聚合减少了0.0557的ncsAURC，而证实对比消失。在完整来源记录中，原生分数偏向PACT，但共同后验峰值分数缩小了其与嵌套Dirichlet的差距，并偏向产品融合。对不变预测重新分配来源按预期移动证据预算。在离线人机协作中，八倍的摄像头内重复使每个检查点的720个类型化响应保持不变；摄像头分组的PACT在60个回合中接纳了57个Qwen3-VL-32B参考一致候选中的47个，未观察到参考不一致的接纳。PACT将计算多重性与证据多重性分开：只有在来源允许单独累积时，一致性才构成证实。
cs.RO / 2 / 2609.01731
TriSAR: Task Coordination and Collision Avoidance for Aerial Robot Teams in Disaster Response
TriSAR：灾害响应中空中机器人团队的任务协调与碰撞避免
Abstract
Multi-Unmanned Aerial Vehicle (UAV) disaster-response systems require coordinated task assignment and local trajectory control, yet the individual and combined contributions of these coordination layers to mission efficiency and operational safety remain insufficiently characterised under controlled experimental conditions. TriSAR is evaluated as a five-UAV coordination system operating in a physics-based Gazebo simulation of an earthquake-damaged urban environment. A 2 x 2 factorial design compares two task-allocation strategies (Genetic Algorithm and greedy fitness-based allocation) with reactive collision avoidance enabled or disabled. Each of the four configurations was evaluated over 30 stochastic episodes in a common scenario of five UAVs and eight targets. Under greedy allocation, enabling repulsion eliminated recorded collision-threshold violations, confirmed by a Mann-Whitney test (U = 885, p = 4.03 x 10^-12, rank-biserial r = 0.97). Under GA allocation, the same protective effect was confirmed (U = 675, p = 1.26 x 10^-5, rank-biserial r = 0.50). For mission-efficiency metrics, GA-based allocation showed no statistically detectable advantage over greedy allocation when repulsion was enabled, but a significant advantage in steps, path length, and energy when repulsion was disabled (Welch's t-tests, |g| between 0.92 and 1.76). These results show that reactive repulsion provides a substantial, allocation-dependent safety benefit, while the additional computational complexity of GA-based task allocation yields a detectable mission-efficiency benefit only when repulsion is disabled.
Chinese Translation
多无人机（UAV）灾害响应系统需要协调的任务分配和局部轨迹控制，但在受控实验条件下，这些协调层对任务效率和操作安全的个体及综合贡献尚未得到充分表征。TriSAR作为一个五架无人机的协调系统，在基于物理的Gazebo模拟器中评估了一个地震损坏城市环境。采用2 x 2因子设计比较了两种任务分配策略（遗传算法和贪婪适应度分配），并在启用或禁用反应式碰撞避免的情况下进行评估。四种配置在五架无人机和八个目标的共同场景中进行了30次随机实验评估。在贪婪分配下，启用排斥机制消除了记录的碰撞阈值违规，经过Mann-Whitney检验确认（U = 885, p = 4.03 x 10^-12, rank-biserial r = 0.97）。在遗传算法分配下，同样的保护效果得到了确认（U = 675, p = 1.26 x 10^-5, rank-biserial r = 0.50）。在任务效率指标方面，当启用排斥机制时，基于遗传算法的分配未显示出相较于贪婪分配的统计显著优势，但在禁用排斥机制时，在步骤、路径长度和能量方面显示出显著优势（Welch's t检验，|g|在0.92到1.76之间）。这些结果表明，反应式排斥提供了显著的、依赖于分配的安全益处，而基于遗传算法的任务分配所带来的额外计算复杂性仅在禁用排斥机制时才会产生可检测的任务效率益处。
cs.RO / 3 / 2609.01799
Designing Versatile Samples for Learned Trajectory Scoring
为学习轨迹评分设计多功能样本
Abstract
Many current end-to-end driving policies emit a pool of candidate trajectories and select one, which makes selection a separable component: a scorer can be retrained while the planner, its backbone, and its trajectory generator all stay frozen. However, many strong planners concentrate their proposals around safe mode, providing limited supervision near decision boundaries. In this work, we design a training dataset that provides more informative supervision for the scorer. In particular, we construct two generators that perturb the logged human trajectory along the two axes a vehicle can be displaced: laterally toward the drivable boundary and longitudinally toward a leading vehicle. The designed dataset produces more informative positive and negative samples than the base planner's proposal pool. We attach a transformer-based scorer to two frozen generative planners, DiffusionDrive and MeanFuser, and train it on the NAVSIM navtrain dataset. The results of the experiments show that we achieve 90.1 EPDMS on DiffusionDrive and 90.4 EPDMS on MeanFuser when using ResNet-34, with 0.4 and 0.3 EPDMS respectively, from the designed training dataset.
Chinese Translation
许多当前的端到端驾驶策略会生成一组候选轨迹并选择其中一个，这使得选择成为一个可分离的组件：评分器可以在规划器（其主干）和轨迹生成器保持不变的情况下进行重新训练。然而，许多强大的规划器将其提议集中在安全模式附近，导致在决策边界附近提供的监督有限。在本研究中，我们设计了一个训练数据集，为评分器提供更具信息性的监督。具体而言，我们构建了两个生成器，沿着车辆可以位移的两个轴对记录的人类轨迹进行扰动：横向朝向可驾驶边界和纵向朝向前方车辆。所设计的数据集产生的正负样本比基础规划器的提议池更具信息性。我们将基于变换器的评分器附加到两个冻结的生成规划器，DiffusionDrive 和 MeanFuser，并在 NAVSIM navtrain 数据集上进行训练。实验结果表明，当使用 ResNet-34 时，我们在 DiffusionDrive 上达到了 90.1 EPDMS，在 MeanFuser 上达到了 90.4 EPDMS，分别来自设计的训练数据集的 EPDMS 为 0.4 和 0.3。
cs.RO / 4 / 2609.01938
One Demonstration, Many Objects: Generalizing Manipulation via Local Contact Geometry
一次演示，多种物体：通过局部接触几何实现操控的泛化
Abstract
Dexterous manipulation with multi-fingered robot hands promises human-level dexterity, but collecting large-scale dexterous robot hand data remains difficult. Learning from human demonstrations has emerged as a scalable alternative to robot teleoperation, providing strong priors on object interaction and contact strategies. Recent sim-to-real RL methods incorporate such priors, but often (i) omit rewards that explicitly incentivize precise contact, yielding weak real-world performance, and/or (ii) generalize poorly to unseen object instances. We propose DemoMimic (Dexterous Motion Mimic), a policy that manipulates objects by focusing on their geometry local to the contact points. Its contact-centric rewards encourage precise contact and improve sim-to-real consistency, yielding a single real-world policy that transfers across objects of varying shape, scale, mass, and friction wherever local contact structure is preserved. Real-world ablations show that DemoMimic achieves 71% success across 16 objects, four tasks, and two robot-hand embodiments, with the smallest sim-to-real drop compared to baselines.
Chinese Translation
多指机器人手的灵巧操控承诺实现人类水平的灵巧性，但收集大规模的灵巧机器人手数据仍然困难。从人类演示中学习已成为机器人遥操作的可扩展替代方案，为物体交互和接触策略提供了强有力的先验知识。最近的模拟到现实（sim-to-real）强化学习（RL）方法融入了这些先验，但通常（i）忽略了明确激励精确接触的奖励，导致现实世界性能较弱，和/或（ii）对未见物体实例的泛化能力较差。我们提出了DemoMimic（灵巧运动模仿），一种通过关注接触点局部几何来操控物体的策略。其以接触为中心的奖励鼓励精确接触并改善模拟到现实的一致性，产生一个能够在形状、规模、质量和摩擦不同的物体之间转移的单一现实世界策略，只要局部接触结构得以保留。现实世界的消融实验表明，DemoMimic在16个物体、四个任务和两个机器人手的实现中达到了71%的成功率，且与基线相比，模拟到现实的性能下降最小。
cs.RO / 5 / 2609.01961
MACAW: Reliable And Efficient Surgical Debridement Using Monocular Adaptive Compact Attention Windows
MACAW：使用单目自适应紧凑注意窗口的可靠高效外科清创
Abstract
Augmenting the dexterity of human surgeons has the potential to free them from tedious subtasks. We consider debridement (removal of diseased or dead tissue fragments), which is challenging due to imprecision in spatial perception and cable actuation. We develop an augmented dexterity system for surgical debridement that uses visual servoing to align the cable-driven gripper with the target position in the image plane, and then introduces a novel approach to depth control, MACAW: Monocular Adaptive Compact Attention Windows. Across 100 physical trials using the da Vinci Research Kit (dVRK) robot, camera-frame servoing reduced average gripper position offset from 37 to fewer than 5 pixels within 4 optimization steps, taking an average of only 0.39s. MACAW significantly outperforms procedural and learned VLA baselines, achieving a 93% success rate at 11 seconds per fragment, yielding a throughput of 304 fragments per hour. Extending MACAW to a bimanual debridement setup maintains a 92% success rate at an average of 7 seconds per fragment, increasing the throughput to 473 fragments per hour.
Chinese Translation
增强人类外科医生的灵活性有潜力使他们摆脱繁琐的子任务。我们考虑清创（去除病变或死亡组织碎片），由于空间感知和电缆驱动的精确性不足，这一过程具有挑战性。我们开发了一种用于外科清创的增强灵活性系统，该系统利用视觉伺服技术将电缆驱动的抓手与图像平面中的目标位置对齐，并引入了一种新的深度控制方法，MACAW（Monocular Adaptive Compact Attention Windows）。在使用达芬奇研究工具包（dVRK）机器人进行的100次物理试验中，摄像头帧伺服将平均抓手位置偏移从37个像素减少到不到5个像素，仅需4个优化步骤，平均耗时仅为0.39秒。MACAW的性能显著优于程序性和学习的VLA基线，在每个碎片11秒的时间内实现93%的成功率，产出为每小时304个碎片。将MACAW扩展到双手清创设置时，成功率保持在92%，每个碎片的平均时间为7秒，产出提高至每小时473个碎片。
cs.RO / 6 / 2609.02003
Design and Validation of a Lightweight, Low-Profile Powered Knee Prosthesis with Quasi-Direct Drive Actuation
轻量化、低轮廓的准直接驱动膝关节假肢的设计与验证
Abstract
Fully-powered knee prostheses, unlike traditional passive knees, can perform controlled positive work, reducing the need for compensatory behaviors by users during energy-intensive activities. While quasi-direct drive (QDD) actuators provide superior torque control, backdrivability, and acoustic noise properties compared to traditional highly-geared actuators, prior QDD prototypes have been too heavy and bulky for commercial translation. In this work, we present the design and validation of a new lightweight (2.6 kg) and low-profile (24.5 cm tip-to-tip build height) QDD knee prosthesis. By optimizing an 18 to 1 two-stage transmission alongside thermal and structural finite-element analyses, we significantly reduce device mass while enabling a peak torque of 145 Nm. Through benchtop tests, we validate the device's high output torque, low backdrive torque (1 Nm), and its precision position and torque control capabilities. We also demonstrate biomimetic kinematics and peak knee extension torques (within one standard deviation of able-bodied references) during both level-ground walking and sit-stand transitions performed by three participants with transfemoral amputation and varying K-levels. By meeting or improving upon the mass, build height, peak torque, and acoustic noise of a leading commercial powered knee, this work establishes the clinical viability of emerging QDD prostheses that promise improved dynamic performance for their users.
Chinese Translation
与传统被动膝关节不同，完全电动膝关节假肢能够执行受控的正功，从而减少用户在高能耗活动中的补偿行为需求。虽然准直接驱动（QDD）执行器在扭矩控制、反驱动能力和声学噪声特性方面优于传统的高齿比执行器，但以往的QDD原型过于笨重，不适合商业化应用。在本研究中，我们提出了一种新的轻量化（2.6 kg）和低轮廓（24.5 cm尖端到尖端的高度）QDD膝关节假肢的设计与验证。通过优化18:1的两级传动系统，并结合热和结构有限元分析，我们显著降低了设备的质量，同时实现了145 Nm的峰值扭矩。通过台架测试，我们验证了该设备的高输出扭矩、低反驱动扭矩（1 Nm）及其精确的位置和扭矩控制能力。我们还展示了仿生运动学以及在三名不同K级别的股骨上肢截肢参与者进行平地行走和坐立转换时的峰值膝关节伸展扭矩（在健全人参考值的一标准差范围内）。通过满足或改善领先商业电动膝关节的质量、构建高度、峰值扭矩和声学噪声，本研究确立了新兴QDD假肢的临床可行性，这些假肢承诺为用户提供更好的动态性能。
cs.RO / 7 / 2609.02020
Real-Time Dynamics-Based Torque-Sampling MPPI for Compliant and Force Aware Manipulation
基于实时动力学的扭矩采样模型预测路径积分（MPPI）用于顺应性和力感知操控
Abstract
This study proposes a novel Model Predictive Path Integral (MPPI)-based task-space control framework. The proposed framework explicitly solves rigid-body dynamics within a real-time MPC formulation and enforces safety constraints, enabling accurate motion and force control that yields compliant behaviors for safe and effective physical interaction of robotic manipulators in unstructured environments. By leveraging MPPI, the proposed framework efficiently handles nonlinear dynamics that are difficult to solve with conventional MPC approaches in real-time. Furthermore, we develop a torque-sampling-based control architecture that enables efficient exploitation of GPU-based parallelization, resulting in effective compliant and force-aware behaviors. As a result, the proposed framework achieves a solver update rate of over 166 Hz with a 0.18 s prediction horizon, and its performance is validated through real-world experiments on a 7-DoF manipulator.
Chinese Translation
本研究提出了一种新颖的基于模型预测路径积分（MPPI）的任务空间控制框架。该框架在实时模型预测控制（MPC）形式中显式求解刚体动力学，并强制执行安全约束，从而实现准确的运动和力控制，产生顺应性行为，以确保机器人操控器在非结构化环境中的安全和有效物理交互。通过利用MPPI，所提出的框架有效处理了传统MPC方法在实时中难以求解的非线性动力学。此外，我们开发了一种基于扭矩采样的控制架构，使得能够高效利用基于GPU的并行化，从而实现有效的顺应性和力感知行为。因此，所提出的框架在0.18秒的预测视野下达到了超过166 Hz的求解器更新率，并通过在7自由度操控器上的实际实验验证了其性能。
cs.RO / 8 / 2609.02046
Modeling What Changes: Sparse, Residual World Models for Object-Centric Manipulation
建模变化：用于物体中心操作的稀疏残差世界模型
Abstract
Monolithic world models predict the entire next state at every step, spending capacity re-predicting the static majority of a scene and injecting error into it. We ask whether explicitly modeling change (a per-object change gate plus a residual delta head that perturbs only the objects the gate flags) is a more effective and interpretable bias for physical prediction and control. On a MuJoCo tabletop pushing benchmark scaling from 3 to 8 objects, the sparse/residual model predicts next-state poses 2.5 to 4.6 times more accurately than a dense multilayer perceptron at 8.6 to 11.1 times fewer parameters, sustains change-detection F1 of 0.80 to 0.87 where the dense baseline is degenerate, transfers across object counts with zero retraining (99.4 percent F1 retention), and reaches about 90 percent of its full-data accuracy with a quarter of the data. In autoregressive rollout it compounds far less error, hugging the no-motion floor while the dense model drifts. Finally, inside a sampling-based planner, prediction-only models fail (though a true-simulator oracle solves the task with the identical planner, confirming the planner is sound), but once featurized and trained for the states a planner visits, the sparse model begins to plan (0.23 plus or minus 0.06 success over three seeds) while the dense monolith stays at zero at every seed. Modeling what changes, rather than re-predicting the whole world, is a simple, effective bias for object-centric physical AI; code, data generators, and all checkpoints will be released upon publication.
Chinese Translation
单一的世界模型在每一步预测整个下一个状态，消耗能力重新预测场景中静态的大多数部分并引入误差。我们探讨是否明确建模变化（每个物体的变化门加上仅扰动被门标记的物体的残差增量头）是一种更有效且可解释的物理预测和控制偏差。在MuJoCo桌面推送基准测试中，从3个物体扩展到8个物体，稀疏/残差模型在8.6到11.1倍更少的参数下，预测下一个状态的姿态准确度比密集的多层感知器高出2.5到4.6倍，保持变化检测F1值为0.80到0.87，而密集基线则表现不佳，且在不同物体数量之间迁移时无需重新训练（99.4%的F1保留率），在使用四分之一数据时达到其全数据准确度的约90%。在自回归展开中，它累积的误差远低于密集模型，保持在无运动的底线，而密集模型则出现漂移。最后，在基于采样的规划器中，仅预测模型失败（尽管一个真实模拟器的oracle使用相同的规划器解决了任务，确认该规划器是合理的），但一旦对规划器访问的状态进行特征化和训练，稀疏模型开始进行规划（在三个种子上成功率为0.23±0.06），而密集模型在每个种子上都保持在零。建模变化，而不是重新预测整个世界，是一种简单有效的物体中心物理人工智能偏差；代码、数据生成器和所有检查点将在发表时发布。
cs.RO / 9 / 2609.02079
Koopman-Based Robust Model Predictive Control for Nonlinear Systems with Stochastic Intermittent Measurements
基于Koopman的鲁棒模型预测控制用于具有随机间歇性测量的非线性系统
Abstract
Intermittent state measurements pose fundamental challenges to model predictive control of constrained nonlinear systems because prediction uncertainty grows during feedback outages and measurement-triggered resets disrupt nominal state propagation, potentially compromising closed-loop stability and recursive feasibility. This paper develops a Koopman-based stochastic MPC framework with probabilistically truncated soft constraints. Specifically, a Lipschitz-constrained deep Koopman model provides a linear latent predictor, enabling computationally efficient online optimization. The intermittent measurement process is modeled as a two-mode discrete-time Markov chain, yielding a unified Markov jump error model for open-loop propagation and measurement-triggered resets. Under numerically verifiable sufficient conditions, the prediction error is shown to be mean-square ultimately bounded, and an explicit uniform second-moment bound is obtained. A distribution-free probabilistic error radius is then constructed for a prescribed confidence level and used to truncate dropout-dependent constraint tightening. An exact-penalty soft-constraint mechanism accommodates reset-induced jumps and prolonged dropouts. Under the stated terminal compatibility and bounded-disturbance conditions, recursive feasibility and mean-square ultimate boundedness of the closed-loop regulation error are established. Numerical simulations on a visual-servoing tracking task corroborate these theoretical results and demonstrate effective tracking under stochastic measurement unavailability.
Chinese Translation
间歇性状态测量对受约束的非线性系统的模型预测控制提出了基本挑战，因为在反馈中断期间预测不确定性会增加，而测量触发的重置会干扰名义状态传播，可能会危及闭环稳定性和递归可行性。本文提出了一种基于Koopman的随机模型预测控制（MPC）框架，采用概率截断的软约束。具体而言，具有Lipschitz约束的深度Koopman模型提供了一个线性潜在预测器，从而实现计算上高效的在线优化。间歇性测量过程被建模为一个双模式离散时间马尔可夫链，产生一个统一的马尔可夫跳跃误差模型，用于开环传播和测量触发的重置。在可数值验证的充分条件下，预测误差被证明是均方最终有界的，并获得了一个明确的均匀二阶矩界限。然后构建了一个无分布的概率误差半径，以满足规定的置信水平，并用于截断依赖于掉落的约束收紧。一个精确惩罚的软约束机制适应于重置引起的跳跃和延长的掉落。在所述的终端兼容性和有界干扰条件下，建立了闭环调节误差的递归可行性和均方最终有界性。在一个视觉伺服跟踪任务上的数值仿真验证了这些理论结果，并展示了在随机测量不可用情况下的有效跟踪。
cs.RO / 10 / 2609.02134
Unified Motion Retargeting for Humanoids with Learned Point Cloud Correspondence
统一的人形机器人运动重定向与学习的点云对应
Abstract
Humanoid learning increasingly relies on transforming vast and diverse human motion data into high-quality robot reference trajectories. However, retargeting human motion to humanoid robots is challenging due to substantial differences in morphology, degrees of freedom, joint ranges, and kinematic constraints between humans and robots. Existing retargeting methods typically address these differences by defining human-robot correspondence through hand-crafted sparse keypoints or body-part pairs. As a result, retargeting quality depends heavily on manual semantic design, limiting scalability across motion sources and robot morphologies and providing only sparse guidance for reproducing detailed poses and interactions. In this paper, we present Unified Motion Retargeting (UMR), a framework that learns dense point cloud correspondence without requiring manually designed human-robot mappings. By treating exterior point clouds as a unified interface between human motion and humanoid robots, UMR decouples retargeting from source-specific skeletal semantics and robot-specific topology. The learned dense correspondence provides fine-grained geometric anchors for constrained point cloud matching optimization, enabling surface-level pose alignment and direct transfer of interaction contacts. Experiments demonstrate that UMR unifies retargeting across heterogeneous motion sources, robot embodiments, and downstream scenarios ranging from locomotion to interaction, while achieving higher motion fidelity and plausibility than state-of-the-art methods. UMR therefore provides a scalable foundation for transforming large-scale human motion references into robot-ready training data.
Chinese Translation
人形学习日益依赖于将大量多样的人类运动数据转化为高质量的机器人参考轨迹。然而，由于人类与机器人在形态、自由度、关节范围和运动约束等方面存在显著差异，将人类运动重定向到人形机器人面临挑战。现有的重定向方法通常通过手工设计的稀疏关键点或身体部位对来定义人机对应，从而解决这些差异。因此，重定向质量在很大程度上依赖于人工语义设计，限制了在运动源和机器人形态之间的可扩展性，并仅提供稀疏的指导以重现详细的姿势和交互。在本文中，我们提出了统一运动重定向（Unified Motion Retargeting, UMR）框架，该框架学习密集的点云对应，而无需手动设计的人机映射。通过将外部点云视为人类运动与人形机器人之间的统一接口，UMR将重定向与源特定的骨骼语义和机器人特定的拓扑解耦。学习到的密集对应为受限的点云匹配优化提供了细粒度的几何锚点，使得表面级姿势对齐和交互接触的直接转移成为可能。实验表明，UMR在异构运动源、机器人表现形式以及从运动到交互的下游场景中统一了重定向，同时在运动保真度和合理性方面超越了现有最先进的方法。因此，UMR为将大规模人类运动参考转化为适合机器人训练的数据提供了可扩展的基础。
cs.RO / 11 / 2609.02157
Towards Effective Physical Reservoir Computing with a Pneumatic Soft Robot
朝向有效的物理水库计算：以气动软机器人为例
Abstract
Physical reservoir computing (PRC) refers to the use of a physical dynamical system as a computational resource for tasks such as state estimation and control, but there has been a lack of formal study of design rules towards more effective design of such physical reservoirs. Using a pneumatic soft arm with a five-pouch sensing column, this work studies how the pouch interconnection topology, robot stiffness, and the number of instrumented sensors affect bending-angle estimation performance. Across 36 matched trials spanning waveform, baseline pressure of the sensing column, and actuation range, all designs are evaluated under the same-time bending-angle estimation benchmark using 0.2 s of pressure history and a fixed ridge estimator. Our analysis of the experimental results leads to three design guidelines. First, independently sealed pouches preserve a much richer observable state than a shared manifold. Second, increasing the baseline pressure of the sensing column makes the pouch responses more redundant and increases estimation error most strongly in the coupled topology. Third, in the sealed topology, two strategically placed sensors already recover most of the attainable benefit, three capture essentially all of it, and additional sensors provide little or no additional value. In summary, the results suggest that topology, stiffness, and number of instrumented sensors should be co-designed for accurate PRC of soft robot states; stronger excitation alone cannot recover the diversity that poor design choices have already removed.
Chinese Translation
物理水库计算（PRC）是指利用物理动态系统作为计算资源来执行状态估计和控制等任务，但对于如何更有效地设计这些物理水库的设计规则缺乏正式研究。本研究使用一个具有五个囊袋传感器的气动软臂，探讨了囊袋互连拓扑、机器人刚度和仪器传感器数量如何影响弯曲角度估计性能。在涵盖波形、传感柱基线压力和驱动范围的36个匹配试验中，所有设计在相同的弯曲角度估计基准下进行评估，使用0.2秒的压力历史和固定的脊估计器。我们对实验结果的分析得出了三条设计指导原则。首先，独立密封的囊袋保留了比共享流形更丰富的可观测状态。其次，增加传感柱的基线压力使囊袋响应更加冗余，并在耦合拓扑中最显著地增加了估计误差。第三，在密封拓扑中，两个战略性放置的传感器已经恢复了大部分可获得的收益，三个传感器几乎捕获了所有收益，而额外的传感器提供的附加价值很小或没有。因此，结果表明，拓扑、刚度和仪器传感器数量应共同设计，以实现软机器人状态的准确PRC；单靠更强的激励无法恢复由于不良设计选择而已被移除的多样性。
cs.RO / 12 / 2609.02219
Hardware-Accelerated Instance Segmentation for Resource-Constrained Space Robotics with Criticality Analysis
针对资源受限的空间机器人进行硬件加速的实例分割及其关键性分析
Abstract
Autonomous lunar missions require real-time per- ception under three coupled constraints: extreme low-light conditions, limited onboard compute, and radiation-induced hardware faults that can silently corrupt inference. We present a deployment-oriented instance segmentation framework for resource-constrained lunar robotics that jointly addresses quan- tization calibration and system-level fault exposure under strict compute constraints. First, we introduce Activation Variance Informative Sampling (AVIS), a label-free calibration strategy that deterministically selects calibration samples based on activation variance statistics. Second, we deploy a YOLO-based segmentation model on a Deep Learning Processor Unit (DPU) with architectural modifications that reduce CPU fallback paths and enable statically compiled execution with bounded latency in low-lighting conditions. We further introduce a software-level criticality analysis to estimate fault exposure and guide mitigation under radiation-constrained operation. On a lunar micro-rover platform, AVIS with bias correction recovers 69.8% of quantization-induced accuracy loss while achieving 309 ms inference latency and 5.7 W power consumption. Targeted mitigation reduces global criticality by 31.7%. The results demonstrate an integrated approach and a blueprint for a reliable and safe AI perception framework under space deployment constraints.
Chinese Translation
自主月球任务需要在三个相互关联的约束条件下进行实时感知：极低光照条件、有限的机载计算能力以及可能静默破坏推理的辐射诱发硬件故障。我们提出了一种面向部署的实例分割框架，旨在为资源受限的月球机器人解决量化校准和系统级故障暴露问题，同时满足严格的计算约束。首先，我们引入了激活方差信息采样（Activation Variance Informative Sampling, AVIS），这是一种无标签的校准策略，通过激活方差统计确定性地选择校准样本。其次，我们在深度学习处理单元（Deep Learning Processor Unit, DPU）上部署了基于YOLO的分割模型，并对其架构进行了修改，以减少CPU回退路径，并在低光照条件下实现具有界限延迟的静态编译执行。我们进一步引入了一种软件级关键性分析，以评估故障暴露并指导在辐射约束操作下的缓解措施。在月球微型探测器平台上，结合偏差校正的AVIS恢复了69.8%的量化引起的准确性损失，同时实现了309毫秒的推理延迟和5.7瓦的功耗。针对性的缓解措施使全球关键性降低了31.7%。结果展示了一种综合方法，并为在空间部署约束下建立可靠安全的人工智能感知框架提供了蓝图。
cs.RO / 13 / 2609.02222
FOCUS: Foot Observation Confidence for Robust Humanoid Proprioceptive Odometry
FOCUS：用于稳健类人机器人本体感觉里程计的足部观察置信度
Abstract
Foot forward kinematics (FK) is widely used to improve proprioceptive legged odometry by providing reliable velocity constraints during foot support. Existing contact-aided estimators generally rely on binary contact decisions to determine whether the FK measurements of an entire foot should be trusted. However, contact does not necessarily imply FK reliability. Dynamic locomotion often involves partial support, toe dragging, and foot slip, causing binary contact decisions to accumulate significant drift over long trajectories. To address this limitation, we propose FOCUS (Foot Observation Confidence from Unannotated Simulation), which predicts a continuous FK reliability weight for each foot instead of estimating binary foot contact. Rather than replacing the model-based estimator, the predicted reliability weights are used to blend FK velocity observations with IMU-propagated body velocity and to adapt the observation covariance of an extended Kalman filter (EKF), enabling smooth reliability-aware fusion without hard contact switching. The network is trained from automatically generated simulation signals using an FK-weighted velocity consistency loss with lightweight simulator-contact regularization, without manually annotated continuous FK-reliability labels. The deployed model relies only on IMU and joint kinematic measurements, making it suitable for hardware platforms with unreliable torque sensing. Experiments demonstrate that FOCUS reduces absolute trajectory error (ATE) by 83.7% on simulated walking episodes, preserves simulated dynamic-motion fidelity in motion scale and spectral energy, reduces ATE by 70.8% across 19 real walking segments, and reduces mean ATE by 42.7% across four real dynamic-motion routines.
Chinese Translation
足部前向运动学（FK）广泛用于通过在足部支撑期间提供可靠的速度约束来改善本体感觉腿部里程计。现有的接触辅助估计器通常依赖于二元接触决策来判断整个足部的 FK 测量是否值得信赖。然而，接触并不一定意味着 FK 的可靠性。动态运动通常涉及部分支撑、脚趾拖曳和足部滑动，导致二元接触决策在长轨迹上累积显著漂移。为了解决这一限制，我们提出了 FOCUS（未标注仿真中的足部观察置信度），它为每只足部预测一个连续的 FK 可靠性权重，而不是估计二元足部接触。预测的可靠性权重用于将 FK 速度观测与 IMU 传播的身体速度融合，并调整扩展卡尔曼滤波器（EKF）的观测协方差，从而实现平滑的可靠性感知融合，而无需硬接触切换。该网络通过使用 FK 加权的速度一致性损失和轻量级仿真接触正则化，从自动生成的仿真信号中训练，而无需手动标注连续的 FK 可靠性标签。所部署的模型仅依赖于 IMU 和关节运动学测量，使其适用于具有不可靠扭矩传感的硬件平台。实验表明，FOCUS 在模拟步态实验中将绝对轨迹误差（ATE）降低了 83.7%，在运动尺度和谱能量上保持了模拟动态运动的保真度，在 19 个真实步态段中将 ATE 降低了 70.8%，并在四个真实动态运动例程中将平均 ATE 降低了 42.7%。
cs.RO / 14 / 2609.02252
DiffuSearch: How Hybrid Trajectory Planning Benefits from Aligned Objectives in Diffusion and Action Space
DiffuSearch：混合轨迹规划如何受益于扩散与动作空间中对齐的目标
Abstract
In trajectory planning for autonomous driving, hybrid planning architectures are often realized as a collection of disparate modules, each with its own objectives. This lack of a unifying principle can lead to inconsistencies between the initial and refined trajectory, resulting in suboptimal behavior. We address this by introducing DiffuSearch, a novel hybrid planner that uses a unified set of objectives across generation and refinement. Our model encourages all components to follow the same shared driving goals: collision avoidance, drivable area compliance, comfort, and progress. DiffuSearch employs a two-stage architecture. First, a guided diffusion model generates a scene-consistent, joint trajectory prediction, using our driving objectives as differentiable guidance functions to implicitly steer the denoising process. Second, a Monte Carlo Tree Search (MCTS) in a discretized action space performs an explicit, local refinement of this proposal, leveraging the same driving objectives as its reward function. This synergistic design leverages the diffusion model's strength in finding scene-consistent solutions combined with the explainable, constraint-aware refinement of MCTS. Experiments on nuPlan and interPlan reactive closed-loop benchmarks demonstrate that DiffuSearch achieves strong and often state-of-the-art performance, substantially reducing collisions and improving comfort, particularly in complex, interactive scenarios. Our ablation studies indicate that MCTS refinement is the main mechanism behind the gains, while sharing objectives between implicit guidance and explicit search provides further consistent improvements.
Chinese Translation
在自动驾驶的轨迹规划中，混合规划架构通常被实现为一组各自具有不同目标的模块。这种缺乏统一原则的情况可能导致初始轨迹与优化轨迹之间的不一致，从而导致次优行为。我们通过引入DiffuSearch来解决这个问题，这是一种新颖的混合规划器，使用统一的目标集进行生成和优化。我们的模型鼓励所有组件遵循相同的共享驾驶目标：避免碰撞、符合可行驶区域、舒适性和进展。DiffuSearch采用两阶段架构。首先，一个引导扩散模型生成场景一致的联合轨迹预测，利用我们的驾驶目标作为可微分的引导函数，隐式地引导去噪过程。其次，在离散化的动作空间中，蒙特卡洛树搜索（Monte Carlo Tree Search, MCTS）对该提案进行显式的局部优化，利用相同的驾驶目标作为其奖励函数。这种协同设计结合了扩散模型在寻找场景一致解决方案方面的优势，以及MCTS在约束感知优化中的可解释性。我们在nuPlan和interPlan反应闭环基准上的实验表明，DiffuSearch实现了强大且常常是最先进的性能，显著减少了碰撞并提高了舒适性，尤其是在复杂的交互场景中。我们的消融研究表明，MCTS优化是性能提升的主要机制，而隐式引导与显式搜索之间共享目标则提供了进一步一致的改进。
cs.RO / 15 / 2609.02270
CrashDiffuser: VLM-Guided Collision Intent Reasoning for Fine-Grained Safety-Critical Traffic Scenario Generation
CrashDiffuser：基于VLM的碰撞意图推理用于细粒度安全关键交通场景生成
Abstract
Generating safety-critical scenarios is essential for evaluating autonomous driving systems. However, existing generators primarily focus on inducing collisions and offer limited control over where contact occurs on the target vehicle. In this paper, we study fine-grained safety-critical scenario generation, where success requires both a target collision and a specified head, rear, or side contact region. We propose CrashDiffuser, a closed-loop VLM-guided diffusion framework that decouples semantic collision reasoning from continuous trajectory synthesis through a hierarchical collision-intent interface derived from the requested target contact region. At initialization, the VLM extracts reusable scene-level context; at each replanning step, it predicts a structured action tuple describing speed change, turning behavior, and collision stage. This intent conditions a diffusion model to generate executable adversarial trajectories, while collision-guided sampling, candidate selection, and short-horizon replanning adapt generation to the target vehicle's evolving behavior. On WOMD-derived closed-loop scenarios, CrashDiffuser achieves a target-collision rate of 50.33% in a single attempt and 67.98% after three attempts, together with a contact-region control success rate of 40.05% and competitive trajectory naturalness. Component ablations further support the proposed design.
Chinese Translation
生成安全关键场景对于评估自动驾驶系统至关重要。然而，现有的生成器主要集中于诱发碰撞，并且对目标车辆上接触发生的位置控制有限。本文研究了细粒度安全关键场景生成，其中成功不仅需要目标碰撞，还需要指定前部、后部或侧面接触区域。我们提出了CrashDiffuser，一个闭环的基于VLM的扩散框架，通过从请求的目标接触区域派生的分层碰撞意图接口，将语义碰撞推理与连续轨迹合成解耦。在初始化时，VLM提取可重用的场景级上下文；在每个重新规划步骤中，它预测一个结构化的动作元组，描述速度变化、转向行为和碰撞阶段。该意图为扩散模型提供条件，以生成可执行的对抗轨迹，同时碰撞引导的采样、候选选择和短期重新规划将生成适应于目标车辆不断变化的行为。在基于WOMD的闭环场景中，CrashDiffuser在一次尝试中实现了50.33%的目标碰撞率，三次尝试后达到67.98%，同时接触区域控制成功率为40.05%，并且轨迹自然性具有竞争力。组件消融实验进一步支持了所提出的设计。
cs.RO / 16 / 2609.02306
Contact-Constrained Lower-Limb Joint-Offset Calibration for Humanoid Robots
面向类人机器人接触约束的下肢关节偏移校准
Abstract
Accurate joint encoder offsets are essential for kinematic consistency in humanoid lower limbs, yet existing calibration methods typically require external motion-capture systems or fiducial targets. We present a self-contained calibration framework exploiting only onboard joint encoders and a pelvis-mounted IMU during static double-support contact. The inter-foot transform from forward kinematics must stay constant when both feet are fixed; minimizing its posture-dependent dispersion yields a nonlinear least-squares problem over the 12-dimensional offset vector. A Hessian eigenstructure analysis shows that parallel pitch axes induce a rotational coupling. Orientation residuals then observe only the pitch-offset sum, while translation and posture diversity set the remaining numerical observability. For the A3 pitch-to-roll-to-yaw ordering, hip-roll and hip-yaw excitation reduce hip-pitch coupling. A standing-posture knee prior then anchors the remaining weak pitch-chain decomposition. Simulation and real-machine injection tests show consistent recovery, and on held-out recordings calibration reduces foot-height RMS residuals from 4.26 to 2.20 mm on A3 and from 8.03 to 1.43 mm on A2. An independent LiDAR-inertial reference checks the pitch-coupled channel. Removing an injected pitch offset moves the leg-odometry vertical drift back toward the LiDAR trajectory. A few static double-support stances thus provide contact-consistent corrections for well-excited directions. Individual offsets in the weak pitch chain remain prior-dependent.
Chinese Translation
准确的关节编码器偏移对于类人下肢的运动学一致性至关重要，然而现有的校准方法通常需要外部运动捕捉系统或基准目标。我们提出了一种自包含的校准框架，仅利用机载关节编码器和安装在骨盆上的惯性测量单元（IMU），在静态双支撑接触期间进行校准。当双脚固定时，前向运动学中的双足变换必须保持不变；最小化其姿态依赖的分散性会导致一个关于12维偏移向量的非线性最小二乘问题。Hessian特征结构分析表明，平行的俯仰轴会引起旋转耦合。方向残差仅观察俯仰偏移的总和，而平移和姿态多样性则设定了其余的数值可观测性。对于A3的俯仰-滚转-偏航顺序，髋关节的滚转和偏航激励减少了髋关节的俯仰耦合。站立姿势下的膝关节先验则固定了剩余的弱俯仰链分解。仿真和真实机器注入测试显示出一致的恢复效果，在保留的录音中，校准将A3的足高均方根残差从4.26 mm降低到2.20 mm，将A2的足高均方根残差从8.03 mm降低到1.43 mm。独立的激光雷达-惯性参考检查了俯仰耦合通道。去除注入的俯仰偏移使腿部里程计的垂直漂移回归激光雷达轨迹。因此，几个静态双支撑姿势提供了接触一致的校正，以应对良好激励的方向。弱俯仰链中的各个偏移仍然依赖于先验。
cs.RO / 17 / 2609.02319
From Multi-Fisheye Sensing to Panoramic Perception: A Parallax-Aware Onboard Platform for Ultra-Low-Altitude UAVs
从多鱼眼传感到全景感知：一种针对超低空无人机的视差感知机载平台
Abstract
Ultra-low-altitude unmanned aerial vehicles (UAVs) require surround vision near buildings, vegetation, and other obstacles. We present a parallax-aware onboard platform that converts four synchronized fisheye streams into an open 1280x640 equirectangular panorama (ERP) interface. A purpose-built carbon-fiber airframe integrates the cameras, NVIDIA Jetson Orin NX, a flight controller, and a global navigation satellite system (GNSS) receiver. The formation pipeline selects projection depth per overlap and combines controlled seams and photometric fusion. Its accuracy profile adds content-adaptive seam search and a validation-gated residual mesh, whereas its deployed profile retains margin-gated Per-seam updates for sensor-rate operation. Evaluation uses more than 50,000 four-view groups from 18 field sequences. Relative to Fixed Depth, the accuracy profile reduces far-field P90 feature misalignment by 41.6%; the deployed Per-seam profile achieves the lowest aggregate geometric errors across held-out sites. Under a paced 20 Hz replay, the deployed profile sustains 19.99 frames/s at 13.29W mean module-input power. Eight-sector ERP sampling reaches 90.8% mean daytime visual-place-recognition Recall@5. Together, these results validate an integrated onboard panoramic-perception architecture that unifies parallax-aware formation, sensor-rate embedded execution, and reusable downstream vision interfaces for ultra-low-altitude UAVs. The project has been open-sourced at https://github.com/DUNDAI1998/parallax-aware-uav-panorama.
Chinese Translation
超低空无人机（UAV）在靠近建筑物、植被和其他障碍物时需要全景视觉。我们提出了一种视差感知机载平台，将四个同步的鱼眼视频流转换为开放的1280x640等矩形全景（ERP）接口。专门设计的碳纤维机身集成了摄像头、NVIDIA Jetson Orin NX、飞行控制器和全球导航卫星系统（GNSS）接收器。形成管道根据重叠选择投影深度，并结合受控接缝和光度融合。其精度配置增加了内容自适应接缝搜索和验证门控残差网格，而其部署配置则保留了边缘门控的每接缝更新以实现传感器速率操作。评估使用了来自18个现场序列的超过50,000个四视图组。相较于固定深度，精度配置将远场P90特征错位减少了41.6%；部署的每接缝配置在保留站点中实现了最低的总几何误差。在20 Hz的节奏重放下，部署配置在平均模块输入功率为13.29W时维持了19.99帧/秒。八扇区ERP采样在白天的视觉位置识别中达到了90.8%的平均召回率@5。这些结果共同验证了一个集成的机载全景感知架构，该架构统一了视差感知形成、嵌入式传感器速率执行和可重用的下游视觉接口，适用于超低空无人机。该项目已开源，网址为 https://github.com/DUNDAI1998/parallax-aware-uav-panorama。
cs.RO / 18 / 2609.02358
Humanoid Safe Stop via Learned Stoppability Value
通过学习的可停止性值实现类人机器人安全停靠
Abstract
Humanoid robots responding to emergency stop commands typically execute a fixed maneuver, without reasoning about whether a safe stop is actually feasible from the current state. We cast emergency stopping as a reach-avoid problem and propose Safe-Stop, a task-agnostic framework that pairs a learned stop policy with learned stoppability estimators. The estimators are complementary: a stop-probability estimator supervised by the actual outcomes of the fixed stop policy, and a reach-avoidance estimator supervised by a Hamilton-Jacobi backup over physical state. The first captures emergent stopping behavior of the learned controller; the second provides a complementary recoverability signal. Because the stop policy and estimators do not depend on the behavior policy that preceded the stop command, they transfer across diverse upstream tasks without retraining. At deployment, the two estimates are combined: Safe-Stop commits to the stop only when both estimators indicate that stopping remains feasible, otherwise it hands off to a fall policy, instantiated as a damping fallback. This agreement check yields decisions that are robust without sacrificing reactivity.
Chinese Translation
类人机器人在响应紧急停车命令时通常执行固定的操作，而不考虑从当前状态出发是否真的可行安全停靠。我们将紧急停车视为一个到达-避免问题，并提出了Safe-Stop，这是一个与任务无关的框架，将学习的停车策略与学习的可停止性估计器相结合。这些估计器是互补的：一个由固定停车策略的实际结果监督的停车概率估计器，以及一个由物理状态上的Hamilton-Jacobi备份监督的到达-避免估计器。第一个捕捉了学习控制器的突现停车行为；第二个提供了一个互补的可恢复性信号。由于停车策略和估计器不依赖于停车命令之前的行为策略，因此它们可以在不同的上游任务中转移而无需重新训练。在部署时，这两个估计值被结合起来：Safe-Stop仅在两个估计器都指示停车仍然可行时才承诺停车，否则将转交给一个以阻尼回退形式实例化的跌倒策略。这种一致性检查产生了稳健的决策，而不牺牲反应性。
cs.RO / 19 / 2609.02402
A Physics-Consistent Benchmark for Contact-Rich Human-Robot Interaction in Assistive Care
一个物理一致的基准用于助理护理中的接触丰富人机交互
Abstract
Conventional task-level evaluation asks whether a robot policy completes a specified action, but can miss failures that emerge only during physical human contact. This limitation is critical in contact-rich assistive tasks, where meaningful evaluation requires a physically responsive human, interaction-quality assessment beyond task success, and a leak-free observer-scorer protocol. We introduce a physics-consistent benchmark for contact-rich human-robot interaction, instantiated in robot-assisted bathing. The benchmark combines a deformable, passively responding human, physics-aware scores alongside task-level success, and a frozen vision-only / scorer-only evaluation protocol. To establish physical validity, region-wise simulated responses are calibrated against force-indentation measurements from Franka impedance pushes on a medical-care manikin. Under a frozen T1-T7 protocol with 140 runs per method, an LLM-augmented state machine (State Machine) achieves 72.9% task success but drops to 56.4% after correct-region and force-safety screening; VoxPoser produces lighter and more stable contact but completes only 27.9% of trials; and zero-shot pi0.5 achieves 0.7% task success with no correct-region or safety-gated successes. These results show that task completion alone does not imply physically valid contact and motivate physics-aware screening before deployment of contact-rich assistive robot policies.
Chinese Translation
传统的任务级评估关注机器人策略是否完成特定动作，但可能会忽略仅在物理人机接触中出现的失败。这一局限性在接触丰富的助理任务中尤为关键，因为有意义的评估需要一个物理响应的人类、超越任务成功的交互质量评估，以及无泄漏的观察者评分协议。我们引入了一个物理一致的基准，用于接触丰富的人机交互，具体体现在机器人辅助洗澡中。该基准结合了一个可变形、被动响应的人类、物理感知评分以及任务级成功，并采用了冻结的视觉仅/评分仅评估协议。为了建立物理有效性，区域级模拟响应与来自Franka阻抗推力在医疗护理人偶上的力-压痕测量进行了校准。在冻结的T1-T7协议下，每种方法进行了140次实验，增强型LLM状态机（State Machine）实现了72.9%的任务成功率，但在正确区域和力安全筛选后降至56.4%；VoxPoser产生了更轻且更稳定的接触，但仅完成了27.9%的试验；而零-shot pi0.5在没有正确区域或安全门成功的情况下，仅实现了0.7%的任务成功率。这些结果表明，仅仅完成任务并不意味着物理接触有效，并促使在部署接触丰富的助理机器人策略之前进行物理感知筛选。
cs.RO / 20 / 2609.02487
An Adaptive Control Architecture for Slope and Terrain Compensation in Autonomous Navigation in Mediterranean Greenhouses
一种用于地中海温室自主导航中的坡度和地形补偿的自适应控制架构
Abstract
The ability to move stably over terrain with varying slopes and textures is essential for mobile agricultural robots operating in complex and dynamic environments such as greenhouses, where small terrain irregularities can lead to significant navigation errors. This article presents a novel terrain-adaptation strategy based on the carried payload, ensuring accurate and robust trajectory tracking. The proposed approach is based on: (i) the experimental characterization of the most common types of greenhouse soil, concrete, compacted sand, and gravel, and (ii) the direct measurement of terrain slope using the IMU, in order to estimate the force with which this angle affects the motor input. Based on this information, a cascade trajectory-tracking scheme has been designed, consisting of a model-based predictive controller (MPC) in the outer loop and a PI controller in the inner loop. The system incorporates an adaptive feedforward control through gain scheduling approach, capable of adjusting to disturbances caused by variations in slope and terrain type. Simulation results demonstrate that the differential-drive robot achieves a significant improvement both in error indices and in control signal efficiency, highlighting the effectiveness and robustness of the proposed approach.
Chinese Translation
在复杂动态环境中，如温室，移动农业机器人在不同坡度和纹理的地形上稳定移动的能力至关重要，因为微小的地形不规则性可能导致显著的导航误差。本文提出了一种基于载荷的创新地形适应策略，确保精确且稳健的轨迹跟踪。所提方法基于：(i) 对最常见的温室土壤类型（如混凝土、压实沙子和碎石）的实验特性描述，以及 (ii) 通过惯性测量单元（IMU）直接测量地形坡度，以估算该角度对电机输入的影响力。基于这些信息，设计了一种级联轨迹跟踪方案，外环为基于模型的预测控制器（MPC），内环为比例-积分（PI）控制器。该系统通过增益调度方法集成了自适应前馈控制，能够调整因坡度和地形类型变化而引起的干扰。仿真结果表明，差分驱动机器人在误差指标和控制信号效率方面均取得了显著改善，突显了所提方法的有效性和稳健性。
cs.RO / 21 / 2609.02493
MS-MEM: Multi-Skill Manipulation-Enhanced Mapping via Uncertainty- and Disturbance-Aware Action Selection
MS-MEM：通过不确定性和干扰感知的多技能操作增强映射
Abstract
Accurate scene understanding in confined, cluttered spaces such as shelves is essential for service robots, as many everyday tasks require them to locate and retrieve objects reliably. Yet, it remains challenging due to severe occlusions, restricted accessibility, and the need to avoid excessive scene changes. In this paper, we propose Multi-Skill Manipulation-Enhanced Mapping (MS-MEM), an evidential framework for uncertainty-aware mapping that integrates active viewpoint selection, object pushing, and grasping. MS-MEM combines scene-level metric-semantic evidential belief estimators with an uncertainty-aware grasp representation. This representation is learned using a novel full-evidential grasp estimator that models both grasp affordance and orientation uncertainty. In our framework, candidate perception and manipulation actions are evaluated within a unified action selection pipeline using a common information gain criterion. For manipulation actions, we further introduce a collateral disturbance constraint (CDC) that discourages excessive changes to confident regions of the scene belief. This enables MS-MEM to select actions that effectively reduce map uncertainty while limiting collateral scene changes. Experimental results show that, compared with single-skill and unconstrained baselines that ignore scene disturbance, MS-MEM achieves higher mapping accuracy while substantially reducing scene disturbance, highlighting the synergistic effects of active viewpoint selection, push, and grasp actions.
Chinese Translation
在如货架等封闭、杂乱的空间中，准确的场景理解对于服务机器人至关重要，因为许多日常任务要求它们可靠地定位和获取物体。然而，由于严重的遮挡、受限的可达性以及避免过度场景变化的需求，这一任务仍然具有挑战性。本文提出了多技能操作增强映射（MS-MEM），这是一个针对不确定性感知映射的证据框架，集成了主动视点选择、物体推动和抓取。MS-MEM将场景级度量-语义证据信念估计器与不确定性感知抓取表示相结合。该表示通过一种新颖的全证据抓取估计器学习，该估计器同时建模抓取适应性和方向不确定性。在我们的框架中，候选感知和操作动作在统一的动作选择管道中使用共同的信息增益标准进行评估。对于操作动作，我们进一步引入了一个附带干扰约束（CDC），以抑制对场景信念中自信区域的过度变化。这使得MS-MEM能够选择有效减少地图不确定性同时限制附带场景变化的动作。实验结果表明，与忽略场景干扰的单技能和无约束基线相比，MS-MEM在显著减少场景干扰的同时实现了更高的映射准确性，突显了主动视点选择、推动和抓取动作的协同效应。
cs.RO / 22 / 2609.02542
World-Model-Augmented Visual Locomotion for Humanoids on Foothold-Constrained Terrain
基于世界模型增强的视觉运动控制在足部接触受限地形上的应用
Abstract
Foothold-constrained terrain is characterized by sparse, discontinuous, or geometrically restricted feasible foot contacts, as encountered on stepping stones, across gaps, and on narrow stair treads. On such terrain, a single misstep often leaves little room to recover, so policies that base foot-placement decisions primarily on the immediately visible terrain are prone to failure. We ask whether a learned predictive summary of near-future observations and rewards can provide the anticipatory information required in such settings. We present World-Model-Augmented Visual Locomotion (WM-LOCO), which jointly trains a recurrent world model and a PPO policy. Conditioned on proprioception and a single onboard depth image, the world model produces a predictive recurrent feature that guides the policy, without explicit foothold labels. In simulation, WM-LOCO succeeds on gaps and stepping stones where a matched baseline fails completely, and matches the baseline's success rate on stairs while improving stride efficiency and reducing pelvis acceleration. We deploy the same policy onboard a physical Unitree G1 humanoid using onboard proprioception and a single depth stream; it traverses all three terrain classes with an average success rate of 93.3%.
Chinese Translation
足部接触受限的地形特征为稀疏、不连续或几何受限的可行足部接触点，例如在踏脚石、跨越间隙和狭窄的楼梯踏板上。在这样的地形上，单次失误往往几乎没有恢复的余地，因此主要基于即时可见地形做出足部放置决策的策略容易失败。我们探讨了是否可以通过学习的对近未来观察和奖励的预测性总结，提供在此类环境中所需的预见性信息。我们提出了基于世界模型增强的视觉运动控制（World-Model-Augmented Visual Locomotion, WM-LOCO），该方法联合训练一个递归世界模型和一个PPO策略。在本体感知和单个机载深度图像的条件下，世界模型生成一个预测性的递归特征，指导策略，而无需显式的足部接触标签。在仿真中，WM-LOCO在跨越间隙和踏脚石时成功，而匹配的基线则完全失败，并且在楼梯上与基线的成功率相当，同时提高了步幅效率并减少了骨盆加速度。我们在物理Unitree G1人形机器人上部署相同的策略，利用机载本体感知和单个深度流；它在三种地形类别上以93.3%的平均成功率完成穿越。
cs.RO / 23 / 2609.02546
ZETA: A Controlled Study of Zero-Shot Cross-Embodiment VLA Transfer for Tabletop Manipulation
ZETA：针对桌面操作的零样本跨体现视觉-语言-动作转移的受控研究
Abstract
Zero-shot generalization to unseen embodiments is important for generalizable vision-language-action (VLA) models as robot hardware evolves and task-specific data collection remains costly. However, a systematic understanding of this problem remains limited, in part because the literature lacks a unified zero-shot transfer definition and controlled evaluation settings that isolate embodiment changes from differences in tasks, scenes, or protocols. To address this gap, we first distinguish strict zero-shot transfer, where the target embodiment is absent from all training data, from pretrain-exposed zero-shot transfer, where it appears only during pretraining. We then introduce a controlled benchmark spanning 14 held-out target embodiments across simulation and real-world validation. Within this framework, we conduct a controlled analysis of four factors: state-action representations, pretraining embodiment diversity, auxiliary co-training objectives, and target-embodiment exposure. Experimental results show that local end-effector (EEF) state-action representations, the source embodiment diversity, and auxiliary co-training improve cross-embodiment transfer by around 15, 18, and 7 percentage points, respectively. We further find that adding only 5% target-embodiment data during pretraining improves average target-embodiment progress by 13.4 percentage points, showing that strict and pretrain-exposed zero-shot transfer are distinct and should be reported separately. Together, these findings provide practical guidance for evaluating and improving cross-embodiment VLA transfer in stationary tabletop manipulation with two-finger grippers, while motivating future investigation of broader settings including mobile-base control, dexterous hands, and long-horizon tasks.
Chinese Translation
对未见体现的零样本泛化对于可推广的视觉-语言-动作（VLA）模型至关重要，因为机器人硬件不断发展，而特定任务的数据收集仍然成本高昂。然而，对这一问题的系统理解仍然有限，部分原因在于文献缺乏统一的零样本转移定义和受控评估设置，这些设置能够将体现变化与任务、场景或协议的差异隔离开来。为了解决这一空白，我们首先区分严格的零样本转移，即目标体现在所有训练数据中均不存在，与预训练暴露的零样本转移，即目标体现仅在预训练期间出现。然后，我们引入一个受控基准，涵盖14个保留的目标体现，涉及模拟和现实世界的验证。在这一框架内，我们对四个因素进行了受控分析：状态-动作表示、预训练体现多样性、辅助共同训练目标和目标体现暴露。实验结果表明，局部末端执行器（EEF）状态-动作表示、源体现的多样性和辅助共同训练分别提高了跨体现转移约15、18和7个百分点。我们进一步发现，在预训练期间仅添加5%的目标体现数据，平均目标体现进展提高了13.4个百分点，表明严格的和预训练暴露的零样本转移是不同的，应该单独报告。综合来看，这些发现为评估和改善静态桌面操作中使用双指夹持器的跨体现VLA转移提供了实用指导，同时激励未来对包括移动底盘控制、灵巧手和长时间任务等更广泛设置的研究。
cs.RO / 24 / 2609.02575
Pre-Lane-change Signal in Transitional Autonomous Vehicles: Results from Controlled Experiments
过渡性自动驾驶车辆中的预变道信号：来自受控实验的结果
Abstract
This paper investigates how a production transitional autonomous vehicle (tAV) develops and executes mandatory lane-change decisions. Using 150 controlled mandatory lane changes from the NC-tALC experiments, the study examines whether the eventual target gap is observable before lateral movement begins and how the tAV progresses longitudinally from that pre-lane-change state to lane-change start. Signal time (SigT) is defined as an operational pre-lane-change-start reference point. A Firth logistic regression predicts whether the tAV eventually merges in front of or behind its nearest target-lane vehicle using relative position and relative speed at SigT. Longitudinal progression from SigT to lane-change start is then examined separately for in-position and repositioning cases. The traffic state at SigT contains substantial information about eventual target-gap choice and provides meaningful lead time before lateral movement begins. The proposed formulation predicts whether the tAV remains with its current gap or repositions to a neighboring gap by moving forward or dropping back, including cases with longitudinal overlap and ambiguous current-gap geometry. The model achieves an average five-fold cross-validated accuracy of 0.89. Results also provide preliminary evidence that in-position and repositioning cases follow different longitudinal pathways from SigT to lane-change start. These findings support a two-stage conjecture of the observable lane-change process: longitudinal preparation from SigT to lane-change start, followed by lateral maneuver execution. The formulation applies to in-position, repositioning, and longitudinally overlapping cases, and can support lane-change models that distinguish target-gap choice from lateral-onset timing while representing longitudinal preparation before lateral movement begins.
Chinese Translation
本文研究了一种生产型过渡性自动驾驶车辆（tAV）如何制定和执行强制变道决策。通过分析150次来自NC-tALC实验的受控强制变道，研究探讨了在横向移动开始之前，最终目标间隙是否可观察，以及tAV如何从该预变道状态纵向推进到变道开始。信号时间（SigT）被定义为一个操作性的预变道开始参考点。使用Firth逻辑回归预测tAV最终是在其最近目标车道车辆前方还是后方合并，依据SigT时的相对位置和相对速度。然后分别对在位和重新定位的情况考察从SigT到变道开始的纵向进展。SigT时的交通状态包含关于最终目标间隙选择的重要信息，并在横向移动开始之前提供了有意义的提前时间。所提出的模型预测tAV是保持当前间隙还是通过向前移动或后退来重新定位到相邻间隙，包括纵向重叠和当前间隙几何形状模糊的情况。该模型的平均五折交叉验证准确率为0.89。结果还提供了初步证据，表明在位和重新定位的情况在从SigT到变道开始的过程中遵循不同的纵向路径。这些发现支持了可观察变道过程的两阶段假设：从SigT到变道开始的纵向准备，随后是横向机动执行。该模型适用于在位、重新定位和纵向重叠的情况，并可以支持区分目标间隙选择与横向开始时机的变道模型，同时表示在横向移动开始之前的纵向准备。
cs.RO / 25 / 2609.02605
Advancing Accessible Underwater Robotics: The Mini-Girona I-AUV at RAMI 2025
推进可及的水下机器人技术：RAMI 2025上的迷你吉罗纳I-AUV
Abstract
The Mini-Girona Intervention Autonomous Underwater Vehicle (I-AUV) represents an advancement in accessible underwater robotics, designed to bridge the gap between costly, specialized research AUVs and basic Remotely Operated Vehicles (ROVs). Developed with a focus on affordability and usability, the Mini-Girona, priced at approximately $50,000, integrates advanced components such as a 5-DOF manipulator arm, stereo vision, and AI-driven processing for autonomous navigation and intervention tasks. This paper presents the design and development of the Mini-Girona, detailing its performance during the RAMI 2025 student competition. Despite challenges such as thermal management issues and restricted team access, the Mini-Girona achieved second place overall, excelling in vision-based perception and intervention tasks. This work highlights the platform's potential as a tool for underwater robotics research and education, fostering innovation in real-world underwater applications.
Chinese Translation
迷你吉罗纳干预自主水下航行器（I-AUV）代表了可及水下机器人技术的进步，旨在填补昂贵的专业研究AUV与基本遥控水下机器人（ROV）之间的差距。迷你吉罗纳的开发侧重于经济性和可用性，定价约为50,000美元，集成了先进的组件，如5自由度机械臂、立体视觉和基于人工智能的处理，用于自主导航和干预任务。本文介绍了迷你吉罗纳的设计与开发，详细描述了其在RAMI 2025学生竞赛中的表现。尽管面临热管理问题和团队访问受限等挑战，迷你吉罗纳在整体上获得了第二名，在基于视觉的感知和干预任务中表现出色。这项工作突显了该平台作为水下机器人研究和教育工具的潜力，促进了现实世界水下应用中的创新。
cs.RO / 26 / 2609.02634
Latent Cluster Analysis for Vision-Language-Action Models
用于视觉-语言-动作模型的潜在聚类分析
Abstract
Vision-Language-Action (VLA) Models are increasingly used in robotics for their ability to ground language and perception into action, yet the internal representations driving their behaviour remain poorly understood. We propose LAVLA, a framework for latent cluster analysis of VLA models, and conduct a layer-wise study of the state-of-the-art GR00T N1.5 model, with particular focus on its action decoder. To better characterise the latent space during action diffusion, we introduce a cross-attention-based embedding-weighting method that amplifies relevant features while suppressing less informative ones. Quantitative evaluation shows that weighted clustering consistently outperforms the baseline. To improve interpretability, we extract human-interpretable concepts for each cluster, linking latent representations to semantic descriptions. Our analysis shows that latent clusters progressively disentangle spatiotemporal and kinematic features, with representations becoming more refined in the middle layers and stabilising toward the output. As such, LAVLA advances the interpretability of language-driven robotic systems.
Chinese Translation
视觉-语言-动作（VLA）模型因其将语言和感知与动作结合的能力而在机器人领域得到越来越多的应用，但驱动其行为的内部表示仍然不够清晰。我们提出了LAVLA，一个用于VLA模型潜在聚类分析的框架，并对最先进的GR00T N1.5模型进行了逐层研究，特别关注其动作解码器。为了更好地表征动作扩散过程中的潜在空间，我们引入了一种基于交叉注意力的嵌入加权方法，该方法在放大相关特征的同时抑制不太信息丰富的特征。定量评估表明，加权聚类始终优于基线。为了提高可解释性，我们为每个聚类提取人类可解释的概念，将潜在表示与语义描述联系起来。我们的分析表明，潜在聚类逐渐解开时空和运动学特征，表示在中间层变得更加精细，并在输出阶段趋于稳定。因此，LAVLA推动了基于语言的机器人系统的可解释性。
cs.RO / 27 / 2609.02653
HINT: Human-Intent Inception for Long-Horizon Robot Manipulation
HINT：基于人类意图的长时间机器人操作的起始框架
Abstract
Humans can perform complex manipulations given a simple intent through an overall instruction, while continuously adapting to evolving visual observations. However, current vision-language action (VLA) models and other action policies struggle to realize this high-level intelligent behavior under dense, evolving visual inputs and sparse language guidance. Visual correlations can then dominate semantic intent, leading actions to follow visual shortcuts rather than human goals. We present HINT (Human-INTent INcepTion), an agentic framework inspired by the human manipulation principles: semantic intent changes sparsely at manipulation-pattern transitions, whereas continuous control primarily depends on the evolving object-hand relationship. HINT invokes semantic reasoning only at pattern transitions to resolve the current subtask and target, then maintains this commitment through multi-view grounding and visual tracking. We explore two visual interfaces-image-space semantic highlighting and attention-prior injection-to communicate the tracked intent to the action policy without introducing additional trainable parameters into the foundation action model. Experiments across three long-horizon tasks and out-of-distribution variants show that HINT substantially improves intent understanding, task progress, and end-to-end success across two foundation policies while preserving low-latency control.
Chinese Translation
人类能够通过整体指令在简单意图的基础上执行复杂的操作，同时不断适应不断变化的视觉观察。然而，当前的视觉-语言动作（VLA）模型和其他动作策略在密集、不断变化的视觉输入和稀疏的语言指导下难以实现这种高级智能行为。视觉关联可能主导语义意图，导致动作遵循视觉捷径而非人类目标。我们提出了HINT（Human-INTent INcepTion），这是一个受人类操作原则启发的代理框架：在操作模式转换时，语义意图变化稀疏，而连续控制主要依赖于不断变化的物体-手关系。HINT仅在模式转换时调用语义推理，以解决当前子任务和目标，然后通过多视角定位和视觉跟踪保持这一承诺。我们探索了两种视觉接口——图像空间语义高亮和注意力优先注入——以在不向基础动作模型引入额外可训练参数的情况下，将跟踪的意图传达给动作策略。在三个长时间任务和分布外变体的实验中，HINT显著提高了意图理解、任务进展和端到端成功率，同时保持了低延迟控制。
cs.RO / 28 / 2609.02688
From Proxy Learning to Driving Decisions: A Transfer-Based Framework for Evaluating Future-Aware Autonomous Driving Planners
从代理学习到决策驱动：一个基于转移的框架用于评估未来感知的自主驾驶规划者
Abstract
Future-aware representations and world models are increasingly used in proposal-based autonomous-driving planners to improve trajectory selection. However, improvements in proxy objectives or restricted subsets are often interpreted as planning gains without verifying proposal ordering, selected trajectories, full-scale utility, and critical driving components. We propose the Proxy-to-Decision Transfer (PDT) Framework, an analysis framework that evaluates when learned future information supports a reliable driving-performance improvement claim. Its Decision-Transfer Decomposition Module localizes value loss through score margins, switch-conditioned utility, and support-versus-selection regret. Its Reliability-Constrained Validation Module requires exact pairing, a minimum meaningful effect, scale-expanded confirmation, safety non-compensation, sequential comparability, and family-level robustness. On a representative future-aware planner evaluated with NAVSIM-v1, component BCE decreases from 0.705 to 0.530 while held selected PDM decreases from 0.963 to 0.961. A separate candidate improves a 512-record prefix by 0.00909, with a scene-bootstrap 95% interval of [0.000744, 0.0177], but its 2048-record and complete-support intervals include zero. A proposal-level replay further confirms the switch-utility decomposition, yet none of 432 screened configurations passes the two-half, two-seed robustness gate. PDT therefore identifies where decision transfer fails or remains indeterminate across proxy, subset, aggregate, and selection evidence.
Chinese Translation
未来感知的表征和世界模型在基于提案的自主驾驶规划者中被越来越多地使用，以改善轨迹选择。然而，代理目标或限制子集的改进常常被解读为规划收益，而未验证提案排序、选定轨迹、全尺度效用和关键驾驶组件。我们提出了代理到决策转移（Proxy-to-Decision Transfer, PDT）框架，这是一个分析框架，用于评估何时学习到的未来信息支持可靠的驾驶性能提升声明。其决策转移分解模块通过得分边际、切换条件效用和支持与选择的遗憾来定位价值损失。其可靠性约束验证模块要求精确配对、最小有意义效应、规模扩展确认、安全非补偿、序列可比性和家族级鲁棒性。在使用NAVSIM-v1评估的代表性未来感知规划者上，组件BCE从0.705降至0.530，而保持选定的PDM从0.963降至0.961。一个单独的候选者将512条记录的前缀提高了0.00909，场景引导的95%区间为[0.000744, 0.0177]，但其2048条记录和完全支持区间包含零。提案级重放进一步确认了切换效用分解，但432个筛选配置中没有一个通过两个半、两个种子的鲁棒性门。因此，PDT识别出决策转移在代理、子集、聚合和选择证据中失败或保持不确定的地方。
cs.RO / 29 / 2609.02811
Do Better Imagined Rollouts Mean Better Robot Control? A Controlled Study of World-Model Evaluation Under Feedback
更好的想象滚动是否意味着更好的机器人控制？在反馈下的世界模型评估控制研究
Abstract
Predictive models are increasingly used in robotics for state estimation, planning, control, and policy evaluation, yet they are often judged by open-loop prediction accuracy over a fixed horizon. In closed-loop operation, a robot repeatedly acts, receives new measurements, updates its state estimate, and recomputes control. We study this difference in a differential-drive path-tracking task with biased odometry and intermittent landmark sensing. Six state estimators are evaluated across 24 sensing conditions using trajectory replay, a 20-step measurement-free rollout, and closed-loop tracking. Replay position RMSE correlates more strongly with closed-loop cross-track RMSE than rollout error (Spearman rho = 0.923 vs. 0.774) and selects a different estimator from the closed-loop optimum in 5/24 conditions, compared with 18/24 for the rollout metric. We then vary rollout horizon and measurement-update interval. With H=20, rank agreement decreases from rho = 0.916 with measurements at every step to rho = 0.774 with no measurements. A horizon-update grid shows that long prediction horizons remain informative when regular corrections are retained, whereas long rollouts without correction can produce rankings that differ substantially from closed-loop behavior. We also test recurrent estimators trained on longer sensing outages. This improves the EKF-anchored models under combined sensing degradation, reducing GRU-EKF cross-track RMSE from 1.72 m to 1.06 m, but the gain is not consistent across isolated outages or estimator architectures. These results show that predictive-model evaluation in robotics should specify both prediction horizon and measurement-update schedule. For models used in feedback, offline rollouts are most informative when their sensing and correction pattern reflects closed-loop operation. Code is available at https://github.com/rdharini2001/Robot_World_Model
Chinese Translation
预测模型在机器人技术中越来越多地用于状态估计、规划、控制和策略评估，但它们通常仅通过固定时间范围内的开放式预测准确性来评判。在闭环操作中，机器人反复执行动作，接收新的测量值，更新其状态估计，并重新计算控制。我们在一个具有偏置里程计和间歇性地标感知的差分驱动路径跟踪任务中研究了这一差异。通过轨迹重放、20步无测量滚动和闭环跟踪，评估了六种状态估计器在24种感知条件下的表现。重放位置的均方根误差（RMSE）与闭环横向跟踪RMSE的相关性比滚动误差更强（Spearman rho = 0.923 vs. 0.774），并且在24种条件中，有5种条件选择了与闭环最优解不同的估计器，而滚动指标则为18种。随后，我们改变了滚动时间范围和测量更新间隔。当H=20时，随着每一步都有测量，排名一致性从rho = 0.916下降到没有测量时的rho = 0.774。一个时间范围更新网格显示，当保留定期修正时，较长的预测时间范围仍然具有信息性，而没有修正的长滚动可能会产生与闭环行为显著不同的排名。我们还测试了在较长感知中断上训练的递归估计器。这在结合感知降级的情况下改善了以扩展卡尔曼滤波器（EKF）为基础的模型，将GRU-EKF的横向RMSE从1.72米降低到1.06米，但在孤立中断或估计器架构之间的增益并不一致。这些结果表明，机器人中的预测模型评估应明确预测时间范围和测量更新计划。对于用于反馈的模型，当其感知和修正模式反映闭环操作时，离线滚动最具信息性。代码可在 https://github.com/rdharini2001/Robot_World_Model 获取。
cs.RO / 30 / 2609.02830
Toward Robust LiDAR Semantic Segmentation for Real-World Deployment: Evaluation under Coarse Labels, Adverse Conditions, and Domain Shifts
面向实际部署的鲁棒LiDAR语义分割：在粗略标签、不利条件和领域转移下的评估
Abstract
LiDAR-based semantic segmentation is a core perception module for autonomous vehicles and mobile robots. Despite the strong performance of recent state-of-the-art methods on standard benchmarks, existing evaluation protocols remain focused on clean, single-domain settings and fine-grained label taxonomies, leaving deployment readiness largely unassessed. Real-world systems must handle safety-critical label semantics, degraded sensing conditions, and cross-domain variability, yet no unified protocol currently addresses all three aspects together. In this paper, we propose a structured evaluation protocol that assesses the deployment readiness of LiDAR semantic segmentation models along three complementary dimensions: (i) coarse-label evaluation aligned with autonomous driving safety priorities, revealing how label granularity affects different methods; (ii) robustness under eight types of LiDAR corruptions designed to emulate real-world atmospheric, geometric, and sensor degradations; and (iii) domain generalization across datasets without adaptation. The evaluation includes inference speed measured on an embedded Jetson AGX Orin platform, directly reflecting deployment constraints. Our results show that fine-grained benchmark rankings do not always reflect safety-relevant performance, that all methods experience substantial degradation under corruptions with architecture-dependent robustness characteristics, and that current domain generalization remains insufficient for reliable deployment. These findings expose concrete gaps between benchmark performance and deployment readiness, and provide a reference protocol for more practically grounded evaluation of LiDAR semantic segmentation.
Chinese Translation
基于LiDAR的语义分割是自动驾驶汽车和移动机器人核心的感知模块。尽管近期最先进的方法在标准基准测试中表现强劲，但现有的评估协议仍然集中于干净的单一领域设置和细粒度标签分类，导致部署准备情况未得到充分评估。现实世界的系统必须处理安全关键的标签语义、退化的感知条件和跨领域的变异性，但目前没有统一的协议能够同时解决这三个方面。在本文中，我们提出了一种结构化的评估协议，从三个互补的维度评估LiDAR语义分割模型的部署准备情况：(i) 与自动驾驶安全优先事项对齐的粗略标签评估，揭示标签粒度如何影响不同方法；(ii) 在八种LiDAR损坏类型下的鲁棒性评估，这些损坏旨在模拟现实世界中的大气、几何和传感器退化；(iii) 在不进行适应的情况下跨数据集的领域泛化。评估还包括在嵌入式Jetson AGX Orin平台上测量的推理速度，直接反映部署约束。我们的结果表明，细粒度基准排名并不总是反映与安全相关的性能，所有方法在损坏下都经历了显著的性能下降，且具有架构依赖的鲁棒性特征，而当前的领域泛化仍不足以实现可靠的部署。这些发现揭示了基准性能与部署准备之间的具体差距，并为更实际的LiDAR语义分割评估提供了参考协议。
cs.RO / 31 / 2609.02861
Towards Trustworthy Autonomous Robots: An Explainable AI-Based Decision Framework
迈向可信赖的自主机器人：基于可解释人工智能的决策框架
Abstract
Autonomous robots powered by deep learning face a fundamental auditability challenge: when incidents occur, investigators cannot reconstruct why the system made specific decisions. This paper presents TRACE (Transparent Reasoning Architecture for Credible Execution), a decision framework that ensures every autonomous action can be traced back to sensor evidence through documented causal chains. The framework organizes decision-making into four auditable layers: Semantic Perception for evidence-grounded entity recognition, Belief Reasoning for probabilistic state estimation with causal graphs, Action Synthesis for constraint-aware planning with counterfactual documentation, and Execution Verification for compliance monitoring. TRACE is model-agnostic yet designed to integrate learning-based perception modules (CNNs, transformers) while preserving decision-level auditability. We evaluate the framework using three objective metrics: Evidence Traceability (sensor-to-decision linkage), Decision Reconstructability (post-hoc analysis capability), and Temporal Continuity (audit trail completeness). Experimental evaluation on warehouse robot navigation demonstrates that TRACE achieves 98.6% evidence traceability, 99.0% temporal continuity, and 98.1% decision reconstructability across 500 simulated decision cycles. Post-hoc methods like LIME provide feature attributions but lack the artifact structure needed for decision-level reconstruction. The framework addresses EU AI Act requirements for high-risk system transparency and contributes to Explainable AI for safety-critical autonomous systems.
Chinese Translation
由深度学习驱动的自主机器人面临一个基本的可审计性挑战：当事件发生时，调查人员无法重建系统为何做出特定决策的原因。本文提出了TRACE（透明推理架构以实现可信执行），这是一个决策框架，确保每个自主行动都可以通过记录的因果链追溯到传感器证据。该框架将决策过程组织为四个可审计层次：语义感知用于基于证据的实体识别，信念推理用于利用因果图进行概率状态估计，行动合成用于考虑约束的规划及反事实文档，执行验证用于合规监测。TRACE是模型无关的，但设计上能够集成基于学习的感知模块（卷积神经网络CNN、变换器transformers），同时保持决策级别的可审计性。我们使用三个客观指标评估该框架：证据可追溯性（传感器与决策的关联）、决策可重构性（事后分析能力）和时间连续性（审计轨迹的完整性）。在仓库机器人导航的实验评估中，TRACE在500个模拟决策周期中实现了98.6%的证据可追溯性、99.0%的时间连续性和98.1%的决策可重构性。事后方法如LIME提供特征归因，但缺乏决策级重构所需的文档结构。该框架满足欧盟人工智能法案对高风险系统透明度的要求，并为安全关键的自主系统的可解释人工智能做出了贡献。
cs.CV / 1 / 2609.01659
Beyond Textual Chain-of-Thought: A Survey on Action-Grounded Reasoning in Autonomous Driving
超越文本链式思维：关于自动驾驶中基于行动的推理的调查
Abstract
Chain-of-thought (CoT) reasoning powers generative models by eliciting intermediate steps before producing an answer. In autonomous driving, the answer is a continuous action. Thus its reasoning must share the same spatiotemporal structure as the physical world. This survey studies the resulting shift from textual CoT to action-grounded reasoning. Surveying 171 papers, including 130 method papers and 41 benchmarks, datasets, surveys, and analysis papers, we propose a representation-centered taxonomy that treats the form of the intermediate state as the organizing axis. We systematize the 130 methods into four categories: language-based, visual-spatial, latent-dynamic, and externalized reasoning, further divided into 13 subtypes tied to distinct regions of interests. Our synthesis shows that the open frontier of reasoning in driving agents lies in intermediate representations that can be grounded in the real world, coupled to real-time action, and verified under safety-critical systems. Project page: https://github.com/tangzhengxu/awesome-av-cot.
Chinese Translation
链式思维（Chain-of-thought, CoT）推理通过引导中间步骤来增强生成模型的能力，从而产生答案。在自动驾驶中，答案是一个连续的动作。因此，其推理必须与物理世界共享相同的时空结构。本调查研究了从文本链式思维到基于行动的推理的转变。我们调查了171篇论文，包括130篇方法论文和41篇基准、数据集、调查和分析论文，提出了一种以表示为中心的分类法，将中间状态的形式作为组织轴。我们将130种方法系统化为四个类别：基于语言的、视觉空间的、潜在动态的和外部化的推理，进一步细分为与不同兴趣区域相关的13个子类型。我们的综合表明，驾驶代理推理的开放前沿在于能够扎根于现实世界的中间表示，结合实时动作，并在安全关键系统下进行验证。项目页面：https://github.com/tangzhengxu/awesome-av-cot。
cs.CV / 2 / 2609.01683
FORGE: Forward-Only Test-Time Adaptation for Integer-Only Vision Models on Microcontrollers
FORGE：微控制器上仅支持整数运算的视觉模型的前向测试时适应
Abstract
Vision models deployed on microcontrollers (MCUs) are quantized to integer-only arithmetic and run in inference-only runtimes that do not carry the machinery backpropagation needs: the standard tool for adapting a model to the distribution shift (sensor noise, blur, lighting) it meets in the field. Existing forward-only test-time adaptation (TTA) methods either run only on server- or edge-GPU-class models (not true microcontroller integer execution), or require the batch-normalization (BN) layers that integer deployment fuses away. We present a forward-only TTA method that operates on deployed, BN-folded, integer-only convolutional networks. The key observation is that fusing BN into the preceding convolution, a mandatory step for integer inference, destroys the statistics that normalization-based adaptation relies on. We restore adaptation by re-normalizing each folded convolution's per-channel output to its clean training statistics, using only forward-pass estimates. The method (i) recovers most of gradient-based TENT's accuracy gain (+20.9 vs. +24.9 points) and matches forward-only BN adaptation, while being the only method that runs on a folded integer-only model; (ii) needs to adapt only 3 of 21 layers (selected without seeing the test corruptions) to recover 93% of the benefit; (iii) survives single-sample streaming with a batch-size-scaled momentum; and (iv) generalizes across three datasets (up to 200 classes) and two architectures. We validate bit-exact int8 convolution execution and deploy on an ESP32-S3, where, measured with a Nordic PPK2 power profiler, the forward-only adaptation (a lightweight fp32 recalibration around the int8 convolutions) costs only 8.3 mJ (6.8% of inference energy) and 21.9 ms on the deployed SIMD-optimized model: forward-only adaptation is cheap on a real microcontroller.
Chinese Translation
部署在微控制器（MCUs）上的视觉模型被量化为仅支持整数运算，并在不具备反向传播所需机制的推理运行时中运行：反向传播是适应模型在实际应用中遇到的分布变化（传感器噪声、模糊、光照）的标准工具。现有的仅前向测试时适应（TTA）方法要么仅在服务器或边缘GPU级别的模型上运行（而非真正的微控制器整数执行），要么需要整数部署中已融合的批归一化（BN）层。我们提出了一种仅前向的TTA方法，适用于已部署的、经过BN折叠的仅支持整数运算的卷积网络。关键观察是，将BN融合到前面的卷积中是整数推理的强制步骤，但这会破坏基于归一化的适应所依赖的统计信息。我们通过重新归一化每个折叠卷积的每通道输出到其干净的训练统计，使用仅前向传递的估计来恢复适应。该方法（i）恢复了基于梯度的TENT的准确性增益的大部分（+20.9与+24.9点相比），并与仅前向的BN适应相匹配，同时是唯一在折叠的仅支持整数模型上运行的方法；（ii）仅需适应21层中的3层（在未见测试损坏的情况下选择）即可恢复93%的收益；（iii）在单样本流式处理中能够存活，具有按批量大小缩放的动量；（iv）在三个数据集（最多200个类别）和两种架构上具有良好的泛化能力。我们验证了位精确的int8卷积执行，并在ESP32-S3上部署，使用Nordic PPK2功率分析仪测量，前向适应（围绕int8卷积的轻量级fp32重新校准）仅消耗8.3 mJ（占推理能量的6.8%）和21.9 ms，在部署的SIMD优化模型上：前向适应在真实微控制器上成本低廉。
cs.CV / 3 / 2609.01691
FairLens: Benchmarking Fairness in Vision-Language Models for High-Stakes Decision-Making
FairLens：高风险决策中视觉-语言模型公平性的基准评估
Abstract
Vision-language models (VLMs) are increasingly used to make decisions from visual inputs. We introduce FAIRLENS, a benchmark and evaluation framework for measuring both the fairness and the validity of VLM responses in three high-stakes domains: hiring, legal, and healthcare. FAIRLENS pairs real face images spanning gender, race, and age groups with closed- and open-ended questions, giving more than 100K image-question pairs per model, and evaluates responses from four complementary views: demographic parity over adverse outcome rates, soundness, demographic association over unsupported roles and statuses, and bias in free-text generation. Soundness is the central validity criterion: a response is sound when it follows the evidence stated in the question and abstains when the image cannot support an answer. Evaluating eight VLMs, we find that the primary failure is unwarranted inference rather than unequal treatment. Models routinely infer qualifications, threat, illness, or professional role from a face instead of abstaining, and the weakest model does so on 99% of the questions its input cannot answer. These failures are most severe in legal and healthcare, where recognizing insufficient evidence matters most, and disparity metrics alone would miss them: parity gaps are small in absolute terms, yet when baseline adverse rates are low the same gap means one demographic group receives adverse labels several times as often as another, and a small gap can equally reflect a model that treats every group unsafely. Bias in free-text responses is only loosely coupled to multiple-choice accuracy, so correct structured answers do not imply safe generation. FAIRLENS shows that fair high-stakes VLM behavior requires similar treatment across groups and refusal to infer high-stakes attributes from appearance, and its question suite transfers to any face corpus with demographic annotations.
Chinese Translation
视觉-语言模型（VLMs）越来越多地用于从视觉输入中做出决策。我们介绍了FAIRLENS，一个用于测量VLM响应的公平性和有效性的基准和评估框架，涵盖三个高风险领域：招聘、法律和医疗。FAIRLENS将涵盖性别、种族和年龄群体的真实面部图像与封闭式和开放式问题配对，为每个模型提供超过10万个图像-问题对，并从四个互补的视角评估响应：不良结果率的人口统计平等、有效性、对不支持角色和状态的人口统计关联，以及自由文本生成中的偏见。有效性是中心有效性标准：当响应遵循问题中陈述的证据并在图像无法支持答案时选择不作答时，该响应被视为有效。评估八个VLM时，我们发现主要的失败是无根据的推断，而非不平等待遇。模型通常从面部推断资格、威胁、疾病或职业角色，而不是选择不作答，最弱的模型在99%的无法回答的问题中都进行了这种推断。这些失败在法律和医疗领域尤为严重，因为识别不足的证据至关重要，仅依靠差异指标无法捕捉到这些问题：绝对差距较小，但当基线不良率较低时，相同的差距意味着一个人口群体获得不良标签的频率是另一个的几倍，而小差距也可能反映出一个模型对每个群体的处理都不安全。自由文本响应中的偏见与多项选择准确性之间的关系较为松散，因此正确的结构化答案并不意味着安全的生成。FAIRLENS表明，高风险VLM行为的公平性要求对各群体进行类似的处理，并拒绝根据外貌推断高风险属性，其问题集可迁移至任何带有人口统计注释的面部数据集。
cs.CV / 4 / 2609.01725
From Visual Cues to Spoken Narration: Rethinking Audio Description
从视觉线索到口头叙述：重新思考音频描述
Abstract
Audio Description (AD) provides spoken narration of visual events during dialogue gaps, making movies accessible to visually impaired audiences. The problem requires determining both what (which visual event) and when (position for inserting the AD) to narrate, to achieve the best user experience. Prior work has largely reduced the problem to video captioning of pre-segmented video clips, i.e., what is largely predefined and when is ignored entirely. We propose Cue2Narrate, a two-stage pipeline that jointly predicts what and when to narrate in longer untrimmed movie clips. A dual-head audio-visual localizer predicts two temporally distinct windows per AD utterance: a visual cue window and a spoken narration window. A LoRA-adapted VLM then generates concise ADs from the predicted visual evidence, trained with a Description Ranking Loss that ranks captions (negative samples) of the same frames lower than the GT AD. To benchmark this new problem statement, we introduce the LongLSMDC benchmark with up to 8-min movie clips (~6.5min on average). On LongLSMDC, Cue2Narrate outperforms video-only and audio-only localization baselines by 5--12 points in avg. mAP. Under both predicted- and GT-window evaluation, Cue2Narrate improves AD generation over the corresponding fine-tuned base VLM. These results establish the first benchmark for multi-segment AD generation on long-form clips. Data & Code: https://github.com/multimodal-ai-lab/Cue2Narrate
Chinese Translation
音频描述（Audio Description, AD）在对话间隙提供视觉事件的口头叙述，使得电影对视力障碍观众可访问。这个问题需要确定叙述的内容（哪个视觉事件）和时机（插入音频描述的位置），以实现最佳用户体验。以往的研究大多将问题简化为对预分段视频片段的视频字幕，即内容大多是预定义的，而时机则完全被忽略。我们提出了Cue2Narrate，一个两阶段的管道，联合预测在较长未剪辑电影片段中叙述的内容和时机。一个双头音视频定位器为每个音频描述话语预测两个时间上不同的窗口：一个视觉线索窗口和一个口头叙述窗口。经过LoRA适配的视觉语言模型（VLM）然后根据预测的视觉证据生成简洁的音频描述，训练时采用描述排名损失（Description Ranking Loss），使同一帧的字幕（负样本）排名低于真实音频描述（GT AD）。为了基准测试这一新问题声明，我们引入了LongLSMDC基准，包含最长8分钟的电影片段（平均约6.5分钟）。在LongLSMDC上，Cue2Narrate在平均mAP上比仅视频和仅音频的定位基线提高了5到12个点。在预测窗口和真实窗口的评估下，Cue2Narrate在音频描述生成上优于相应的微调基础视觉语言模型。这些结果为长片段的多段音频描述生成建立了首个基准。数据与代码：https://github.com/multimodal-ai-lab/Cue2Narrate
cs.CV / 5 / 2609.01738
UAV Thermal Imagery for Inert Ordnance Screening: Multi Campaign Dataset Development,Object Detection, and Practical Recommendations
无人机热成像在惰性弹药筛查中的应用：多次活动数据集开发、目标检测及实用建议
Abstract
Unexploded ordnance (UXO) continues to restrict civilian access, agricultural activity, infrastructure recovery, and environmental remediation in contaminated areas around the world. This study created a multi campaign UAV thermal image data set of inert ordnance, developed a labeled image set from collected imagery, tested object detection models, and identified practical considerations for humanitarian mine action and demining applications. Data were collected during four field campaigns in Tennessee under summer and winter conditions using inert mines, munitions, and other ordnance placed in short grass, tall vegetation, gravel, mulch, rock, compost, and compacted surfaces. Thermal imagery was collected under flight altitutes of 33 m and 15 m. The final source inventory contained 5,855 thermal image label pairs, including 918 positive images and 4,937 background images. After retaining all positive images and downsampling background images, the 33 m dataset contained 420 training and 106 validation images, while the 15 m dataset contained 629 training and 157 validation images. YOLOV11l and RT-DETR-R50 algorithms were trained and evaluated to develop an automated candidate detection model. Practical recommendations include collecting thermal and RGB imagery together, incorporating varied surfaces and background only imagery, considering periods following changes in solar exposure, balancing survey coverage against target pixel representation, calibrating models with representative local data, and retaining qualified human review. The intended use is screening and prioritization for follow on technical survey or EOD assessment, and not a standalone clearance.
Chinese Translation
未爆炸弹药（UXO）继续限制全球受污染地区的民用访问、农业活动、基础设施恢复和环境修复。本研究创建了一个多次活动的无人机热成像数据集，涵盖惰性弹药，开发了从收集的图像中生成的标注图像集，测试了目标检测模型，并确定了人道主义排雷行动和扫雷应用的实际考虑因素。数据是在田纳西州的四次实地活动中收集的，涵盖夏季和冬季条件，使用了放置在短草、高植被、砾石、覆盖物、岩石、堆肥和压实表面上的惰性地雷、弹药和其他弹药。热成像是在33米和15米的飞行高度下收集的。最终的数据源清单包含5,855对热成像标签，包括918张正面图像和4,937张背景图像。在保留所有正面图像并对背景图像进行下采样后，33米数据集包含420张训练图像和106张验证图像，而15米数据集包含629张训练图像和157张验证图像。对YOLOV11l和RT-DETR-R50算法进行了训练和评估，以开发自动候选检测模型。实用建议包括同时收集热成像和RGB图像，结合多样的表面和仅背景图像，考虑在太阳照射变化后的时间段，平衡调查覆盖与目标像素表示，使用具有代表性的本地数据对模型进行校准，并保留合格的人为审核。预期用途是对后续技术调查或爆炸物处置评估进行筛查和优先排序，而不是单独的清理。
cs.CV / 6 / 2609.01740
ZipTok3D: High-Fidelity 3D Tokenization with Compact Token Prefixes
ZipTok3D：具有紧凑令牌前缀的高保真3D令牌化
Abstract
Compact token sequences are essential for efficient 3D generation. However, existing 3D tokenizers typically organize latent representations either over spatial regions or as fixed-size sets of global tokens, both suffering sharp reconstruction degradation when compressed to extremely low token budgets. In this paper, we present ZipTok3D, a 3D tokenizer designed for high-fidelity reconstruction from extremely short token sequences. Its key idea is to organize object geometry into progressively informative global-token prefixes and unfold these compact representations through iterative decoding. Specifically, nested dropout randomly truncates the latent sequence after encoding during training and requires each retained prefix to reconstruct the complete object, thereby prioritizing essential geometric information in the leading tokens. The decoder then repeatedly applies a parameter-shared Transformer block to recover fine-grained geometry from each prefix without a separate generative sampling stage. With the same token dimension, ZipTok3D achieves reconstruction quality comparable to the 32-token COD-VAE baseline using only one token on ShapeNet and four on TRELLIS, yielding $32\times$ and $8\times$ shorter token sequences, respectively.
Chinese Translation
紧凑的令牌序列对于高效的3D生成至关重要。然而，现有的3D令牌化方法通常将潜在表示组织为空间区域或固定大小的全局令牌集合，在压缩到极低的令牌预算时，都会出现明显的重建质量下降。本文提出了ZipTok3D，这是一种旨在从极短令牌序列中实现高保真重建的3D令牌化方法。其关键思想是将物体几何形状组织为逐步信息丰富的全局令牌前缀，并通过迭代解码展开这些紧凑表示。具体而言，嵌套丢弃在训练期间在编码后随机截断潜在序列，并要求每个保留的前缀重建完整的物体，从而优先考虑前导令牌中的重要几何信息。解码器随后重复应用一个参数共享的Transformer模块，从每个前缀中恢复细粒度几何形状，而无需单独的生成采样阶段。在相同的令牌维度下，ZipTok3D在ShapeNet上仅使用一个令牌，在TRELLIS上使用四个令牌，达到了与32令牌COD-VAE基线相当的重建质量，分别实现了$32 imes$和$8 imes$更短的令牌序列。
cs.CV / 7 / 2609.01742
Evidential Deep Learning for Multi-Modal Anti-UAV Detection
多模态反无人机检测的证据深度学习
Abstract
Anti-UAV systems increasingly fuse multiple sensors, yet their detection heads provide no per-modality reliability signal. This study evaluates whether evidential deep learning (EDL) heads, Dempster-Shafer (DS) evidence fusion, and uncertainty-driven temporal sensor gating improve anti-UAV detection through a controlled ablation on three benchmarks: thermal tracking (AntiUAV600), RGB-audio-RF classification (TRIDENT), and RGB-IR tracking (MM-UAV). The EDL training objective improves accuracy over retrained sigmoid baselines (+5.9 percentage points in accuracy and a tripled tracker-on-absent rate in E1; +4.8 percentage points in classification accuracy in E2, surviving a clip-clustered bootstrap, p = 0.011) and ranks classification errors substantially better (entropy UAUC approximately 0.94 vs. 0.51). The remaining components do not support their respective hypotheses. DS fusion does not outperform simple probability averaging. Dirichlet vacuity adds no ranking power beyond predictive entropy and inverts at the detection level, where extreme background imbalance causes it to encode class membership rather than error likelihood, a failure also observed for entropy and sigmoid confidence. Temporal gating preserves accuracy only when nearly inactive and yields no realised latency saving on shared-backbone hardware. The benefit of evidential learning therefore arises primarily from its training objective rather than its uncertainty estimate; a crop-level control further localises the detection-level breakdown to anchor-level evaluation rather than the learned representation.
Chinese Translation
反无人机系统越来越多地融合多种传感器，但其检测头并未提供每种模态的可靠性信号。本研究评估了证据深度学习（Evidential Deep Learning, EDL）头、邓普斯特-沙弗（Dempster-Shafer, DS）证据融合和基于不确定性的时间传感器门控是否通过对三个基准的控制消融实验改善反无人机检测：热成像跟踪（AntiUAV600）、RGB-音频-RF 分类（TRIDENT）和RGB-红外跟踪（MM-UAV）。EDL训练目标在准确性上优于重新训练的sigmoid基线（准确性提高5.9个百分点，E1中的缺失跟踪器率增加三倍；E2中的分类准确性提高4.8个百分点，经过剪辑聚类自助法检验，p = 0.011），并且在分类错误排名上表现显著更好（熵的无偏面积（UAUC）约为0.94，而0.51）。其余组件未能支持各自的假设。DS融合并未优于简单的概率平均。狄利克雷空洞性在预测熵之外未增加排名能力，并在检测层面上出现反转，极端背景不平衡导致其编码类别成员资格而非错误可能性，这一失败在熵和sigmoid置信度中也有观察到。时间门控仅在几乎不活动时保持准确性，并未在共享骨干硬件上实现延迟节省。因此，证据学习的好处主要来自其训练目标，而非其不确定性估计；作物级控制进一步将检测级的崩溃局限于锚级评估，而非学习到的表示。
cs.CV / 8 / 2609.01743
SCULPT: Training Edge Vision Models for Post-Training Quantization Readiness
SCULPT：为后训练量化准备训练边缘视觉模型
Abstract
Edge vision models are difficult to deploy on resource-constrained hardware, making low-bit post-training quantization (PTQ) attractive. In practice, standard FP32 training often produces heavy-tailed activation distributions whose outliers destabilize activation quantization: preserving the full range wastes quantization bins on rare extremes, while aggressive clipping causes information loss. Existing solutions typically rely on quantization-aware training (QAT), which adds training complexity and bit-width coupling, or advanced PTQ procedures that repair the model after training. We present SCULPT (Statistical Clipping and Uniform Loss for Post-Training), a training-time method that improves PTQ readiness during ordinary FP32 fine-tuning. SCULPT combines a topology-aware activation regularizer that suppresses quantization-hostile skewness and kurtosis with a stable percentile-based clipping mechanism that learns deployment-ready activation bounds. Unlike QAT, SCULPT does not simulate quantization during optimization; unlike post hoc outlier-repair PTQ methods, it does not require runtime activation transformations. The learned clipping bounds can be exported directly into a standard PTQ workflow for low-bit deployment, including INT8 and lower-bit settings such as W4A8.
Chinese Translation
边缘视觉模型在资源受限的硬件上部署困难，使得低位后训练量化（PTQ）变得具有吸引力。在实践中，标准的 FP32 训练往往产生重尾激活分布，其异常值会使激活量化不稳定：保留全范围会浪费在稀有极值上的量化箱，而激进的裁剪则会导致信息损失。现有解决方案通常依赖于量化感知训练（QAT），这增加了训练复杂性和位宽耦合，或依赖于在训练后修复模型的高级 PTQ 程序。我们提出了 SCULPT（后训练的统计裁剪和均匀损失），这是一种在普通 FP32 微调期间提高 PTQ 准备性的训练时间方法。SCULPT 结合了一种拓扑感知的激活正则化器，抑制量化不友好的偏斜度和峰度，以及一种基于稳定百分位的裁剪机制，学习部署就绪的激活边界。与 QAT 不同，SCULPT 在优化过程中不模拟量化；与事后异常值修复的 PTQ 方法不同，它不需要运行时激活转换。学习到的裁剪边界可以直接导出到标准 PTQ 工作流程中，以实现低位部署，包括 INT8 和更低位设置，如 W4A8。
cs.CV / 9 / 2609.01749
Swin Meets EfficientNet: Lightweight Architectures for GAN-Based Face Forensics
Swin与EfficientNet相结合：基于GAN的人脸取证轻量级架构
Abstract
Modern generative models, such as GANs, diffusion architectures, and autoregressive systems, now produce facial images that are nearly indistinguishable from authentic photographs. This capability makes detecting forged images increasingly difficult, raising serious concerns about identity theft, fraud, and misinformation campaigns. Our research focuses specifically on GAN-generated synthetic faces, which underpin many face-centric deepfakes, and investigates efficient detection approaches using image analysis alone. Existing detection systems rely heavily on either convolutional neural networks (CNNs) or global vision transformers. While CNNs excel at identifying texture-based local features, they struggle with broader contextual understanding. Traditional Vision Transformer (ViT) models can capture long-range structures effectively, but demand substantial computational resources. Our work explores Swin-Transformer-based architectures across three implementations: a compact Swin Transformer trained from the ground up, ImageNet-1K pre-trained Swin-Tiny and Swin-Small models adapted for binary classification, and a novel hybrid combining EfficientNet-B0's convolutional processing with a Swin Transformer backend. We evaluated all models using the 140K Real and Fake Faces dataset, which includes StyleGAN-generated fake faces alongside authentic images from Flickr and DFDC, with balanced splits for training, validation, and testing. The EfficientNetB0+Swin hybrid achieved 99% accuracy and a 99.44% recall on 5,000 test images, outperforming both pure Swin variants and a previous CNN-only baseline on this dataset. Our results suggest that combining hierarchical CNN features with shifted-window self-attention provides an efficient and computationally lightweight method for detecting GAN-generated synthetic faces.
Chinese Translation
现代生成模型，如生成对抗网络（GAN）、扩散架构和自回归系统，现已能够生成与真实照片几乎无法区分的面部图像。这一能力使得伪造图像的检测变得愈加困难，严重影响了身份盗窃、欺诈和虚假信息传播的安全性。我们的研究专注于GAN生成的合成面孔，这些面孔是许多以人脸为中心的深度伪造技术的基础，并探讨了仅通过图像分析进行高效检测的方法。现有的检测系统主要依赖卷积神经网络（CNN）或全局视觉变换器（Vision Transformers）。虽然CNN在识别基于纹理的局部特征方面表现优异，但在更广泛的上下文理解上存在困难。传统的视觉变换器（ViT）模型能够有效捕捉长距离结构，但需要大量的计算资源。我们的工作探索了基于Swin-Transformer的架构，涵盖三种实现方式：从零开始训练的紧凑型Swin Transformer、经过ImageNet-1K预训练的Swin-Tiny和Swin-Small模型（适用于二分类），以及一种新颖的混合模型，将EfficientNet-B0的卷积处理与Swin Transformer后端结合。我们使用包含StyleGAN生成的伪造面孔和来自Flickr及DFDC的真实图像的140K真实与伪造面孔数据集对所有模型进行了评估，数据集在训练、验证和测试中进行了平衡分割。EfficientNetB0+Swin混合模型在5,000张测试图像上达到了99%的准确率和99.44%的召回率，优于该数据集上的纯Swin变体和之前的仅CNN基线。我们的结果表明，将分层CNN特征与移动窗口自注意力相结合，提供了一种高效且计算轻量的方法来检测GAN生成的合成面孔。
cs.CV / 10 / 2609.01757
AlphaRAD: Grounded Zero-Shot Classification in Chest Radiology via $\alpha$-Corrected Binary Cross Entropy and Factorized Latent Supervision
AlphaRAD：通过 $eta$-修正的二元交叉熵和因子化潜在监督实现胸部放射学的基础零样本分类
Abstract
Vision-Language Pretrained Models (VLPMs) offer a scalable path to open-vocabulary chest radiology understanding, yet two aspects remain underexplored: how structured clinical semantics extracted from medical reports can reduce in-batch noise during contrastive learning, and how cross-modal fusion can be designed to produce more faithful spatial grounding without added complexity. We introduce AlphaRAD, addressing these opportunities through two contributions. First, we construct a large-scale structured medical concept space from medical reports parsed by a Large Language Model for training, thereby mitigating in-batch learning noise and removing heuristic pair matching in contrastive learning, and thus naturally positioning AlphaRAD as a medical concept discriminator trained via $\alpha$-Corrected Binary Cross-Entropy. Second, we propose FLaS (Factorized Latent Supervision), an extremely simple yet effective cross-modal feature fusion module that factorizes VLPM representations into independent subspaces, using dedicated alignment supervision to enhance the expressiveness of spatial grounding without introducing additional model parameters. Through extensive empirical validation, AlphaRAD shows strong zero-shot generalization across diverse chest radiology tasks. Notably, it establishes state-of-the-art average performance across 16 classification benchmarks, while achieving individual state-of-the-art results via distinct gains on 7 grounding/phrase grounding and 3 segmentation datasets.
Chinese Translation
视觉-语言预训练模型（VLPMs）为开放词汇的胸部放射学理解提供了一条可扩展的路径，但仍有两个方面未得到充分探索：从医学报告中提取的结构化临床语义如何在对比学习中减少批次内噪声，以及如何设计跨模态融合以在不增加复杂性的情况下产生更真实的空间基础。我们引入了AlphaRAD，通过两个贡献来解决这些机会。首先，我们从通过大型语言模型解析的医学报告中构建了一个大规模的结构化医学概念空间用于训练，从而减轻批次学习噪声，并消除对比学习中的启发式配对匹配，因此自然地将AlphaRAD定位为通过 $eta$-修正的二元交叉熵训练的医学概念鉴别器。其次，我们提出了FLaS（因子化潜在监督），这是一个极其简单但有效的跨模态特征融合模块，它将VLPM表示分解为独立的子空间，使用专门的对齐监督来增强空间基础的表现力，而不引入额外的模型参数。通过广泛的实证验证，AlphaRAD在多样的胸部放射学任务中显示出强大的零样本泛化能力。值得注意的是，它在16个分类基准上建立了最先进的平均性能，同时在7个基础/短语基础和3个分割数据集上通过不同的增益实现了个体最先进的结果。
cs.CV / 11 / 2609.01778
Allocate Before You Embed: Adaptive Visual Input Allocation for Video Embeddings
嵌入前的分配：视频嵌入的自适应视觉输入分配
Abstract
Large-scale video retrieval requires embedding models to encode long and diverse videos under tight visual-input and inference budgets. Existing methods typically sample a small, fixed set of frames at their original resolution, limiting temporal coverage and ignoring frame importance. Our empirical analysis shows that expanding temporal coverage improves retrieval even under a fixed visual-input budget. Gains are larger when the original per-frame resolution is preserved, highlighting the complementary roles of temporal coverage and spatial fidelity. Motivated by this finding, we propose AllocEmbed, an allocate-then-embed framework that reallocates a fixed visual-input budget across more frames. A lightweight allocator uses low-cost previews to assign frame-wise resolutions before the embedding backbone, preserving more detail where it most benefits retrieval while reducing visual cost elsewhere. We further introduce Retrieval-Driven Policy Optimization (RDPO), which learns the allocator directly from retrieval feedback using a rank-validated similarity gap and a confidence-guided efficiency incentive. Operating entirely before the backbone, AllocEmbed integrates with existing retrieval systems without modifying the embedding model or downstream pipeline. Experiments on the MMEB-V2 V-QA and V-RET tasks and our LongRet benchmark show that AllocEmbed achieves the best overall retrieval performance among the evaluated budget-matched methods and transfers across embedding backbones. Our code is publicly available at https://github.com/jinsong8/AllocEmbed.
Chinese Translation
大规模视频检索要求嵌入模型在严格的视觉输入和推理预算下对长且多样化的视频进行编码。现有方法通常在其原始分辨率下对一小组固定帧进行采样，这限制了时间覆盖并忽略了帧的重要性。我们的实证分析表明，扩展时间覆盖即使在固定视觉输入预算下也能改善检索效果。当保留原始每帧分辨率时，增益更大，突显了时间覆盖和空间保真度的互补作用。基于这一发现，我们提出了AllocEmbed，一个先分配后嵌入的框架，它在更多帧之间重新分配固定的视觉输入预算。一个轻量级的分配器利用低成本的预览在嵌入主干之前分配逐帧分辨率，在最有利于检索的地方保留更多细节，同时在其他地方减少视觉成本。我们进一步引入了检索驱动的策略优化（Retrieval-Driven Policy Optimization, RDPO），它通过使用排名验证的相似性差距和基于信心的效率激励直接从检索反馈中学习分配器。AllocEmbed完全在主干之前操作，能够与现有的检索系统集成，而无需修改嵌入模型或下游管道。在MMEB-V2 V-QA和V-RET任务以及我们的LongRet基准上的实验表明，AllocEmbed在评估的预算匹配方法中实现了最佳的整体检索性能，并能够在不同的嵌入主干之间迁移。我们的代码已公开发布在 https://github.com/jinsong8/AllocEmbed。
cs.CV / 12 / 2609.01786
Ten Architectures, One Error: Shared Failure Modes in Hyperspectral Classification under Spatially Disjoint Evaluation
十种架构，一个错误：空间上不相交评估下的高光谱分类共享失败模式
Abstract
Hyperspectral image classification still relies heavily on random pixel splits within a single scene. The Salinas dataset, randomly split, is among the most widely used datasets for comparing different architectures. However, under a random split method, a large fraction of test pixels fall immediately adjacent to a training pixel, which inflates reported accuracy. This work introduces a leakage-free evaluation protocol linking spatial separation to the model's receptive field. Applying this protocol across ten different architectures, including classical, spectral, spectral-spatial, transformer, vision-backbone, and state-space families, shows that Macro-F1 drops by 0.147 on average and model rankings change by as many as five places. Furthermore, leakage-free evaluation limits which architectures can be tested on a given benchmark. Since each partition supports patches only within a finite radius, reporting this radius alongside the receptive field is essential for fair comparison. In addition, this study reveals that all ten architectures misclassify largely the same pixels, pointing to a spectral ambiguity in the data that none of them resolves.
Chinese Translation
高光谱图像分类仍然在很大程度上依赖于单一场景内的随机像素划分。随机划分的Salinas数据集是比较不同架构时最常用的数据集之一。然而，在随机划分方法下，测试像素中有很大一部分紧邻训练像素，这导致报告的准确率被夸大。本文提出了一种无泄漏评估协议，将空间分离与模型的感受野联系起来。在十种不同的架构上应用该协议，包括经典架构、光谱架构、光谱-空间架构、变换器架构、视觉骨干架构和状态空间架构，结果显示Macro-F1平均下降0.147，模型排名变化多达五个名次。此外，无泄漏评估限制了在给定基准上可以测试的架构。由于每个划分仅支持有限半径内的块，因此在报告感受野时附上该半径对于公平比较至关重要。此外，本研究还揭示所有十种架构在很大程度上错误分类了相同的像素，指向数据中存在的光谱模糊性，而这些架构均未能解决这一问题。
cs.CV / 13 / 2609.01787
CoViT: Instance-Correspondence Contrastive Learning for Vision Transformer
CoViT：用于视觉变换器的实例对应对比学习
Abstract
Vision Transformers (ViT) excel in semantic understanding but fail to discriminate between object instances (e.g., identical embeddings for two dogs), limiting their use in instance-level tasks such as object detection and instance segmentation. We propose Contrastive Vision Transformer (CoViT), a self-supervised learning framework that injects instance-awareness into ViT through geometry-guided contrastive learning. CoViT uniquely coordinates ViT's attention maps and embeddings by constructing triplets: (1) Attention-guided masking: Refine multi-head attention via adaptive thresholding and morphological operations to generate instance masks, identifying foreground anchors; (2) Hardest contrastive mining: For each anchor, computing pairwise embedding similarities to select the intra-instance hardest positive (least similar patch within its mask) and inter-instance hardest negative (most similar patch from other instances), with intra-instance regions masked during negative search. These triplets drive a contrastive loss that simultaneously compresses intra-instance variance and expands inter-instance margins, forcing ViT to discern subtle geometric and appearance differences between instances. CoViT consistently achieves stable performance gains of over 2 AP points across multiple instance-level perception tasks by using ViT as backbone architecture. Notably, CoViT requires no extra decoders or labels, demonstrating that a pure ViT can learn instance-aware representations via inherent attention priors and targeted contrastive constraints. Code and models will be released.
Chinese Translation
视觉变换器（ViT）在语义理解方面表现出色，但在区分对象实例（例如，对于两只狗产生相同的嵌入）方面存在不足，这限制了它们在实例级任务（如目标检测和实例分割）中的应用。我们提出了对比视觉变换器（CoViT），这是一个自监督学习框架，通过几何引导的对比学习将实例意识注入ViT。CoViT通过构建三元组独特地协调ViT的注意力图和嵌入： (1) 基于注意力的掩蔽：通过自适应阈值和形态学操作来优化多头注意力，以生成实例掩膜，识别前景锚点； (2) 最难对比挖掘：对于每个锚点，计算成对的嵌入相似度，以选择实例内最难的正样本（其掩膜内最不相似的补丁）和实例间最难的负样本（来自其他实例的最相似补丁），在负样本搜索过程中对实例内区域进行掩蔽。这些三元组驱动对比损失，同时压缩实例内方差并扩大实例间边界，迫使ViT辨别实例之间微妙的几何和外观差异。通过使用ViT作为主干架构，CoViT在多个实例级感知任务中始终实现超过2个AP点的稳定性能提升。值得注意的是，CoViT不需要额外的解码器或标签，证明了纯ViT可以通过固有的注意力先验和针对性的对比约束学习到实例感知表示。代码和模型将会发布。
cs.CV / 14 / 2609.01795
DESA-TTA: Dynamic EMA and Source Anchoring for Test-Time Adaptation
DESA-TTA：动态EMA与源锚定用于测试时适应
Abstract
Vision-language object detectors (VLODs) achieve strong zero-shot performance but remain vulnerable to distribution shifts during deployment. Mean-teacher methods for test-time adaptation (TTA) can improve robustness by updating a student model using teacher-generated pseudo-labels. However, mean-teacher TTA is highly sensitive to the choice of a fixed exponential moving average (EMA) coefficient for teacher updates, and repeated optimization with noisy pseudo-labels can cause cumulative student drift. We propose Dynamic EMA and Source Anchoring for TTA (DESA-TTA), a low-overhead method that jointly regulates teacher updates and student drift through dynamic temporal averaging and source anchoring. Dynamic temporal averaging estimates teacher uncertainty from pseudo-label confidence and box density and uses it to select a sample-wise EMA coefficient within bounds determined by teacher parameter drift. Source anchoring partially restores the updated student parameters toward their pretrained values, with the anchoring strength increasing according to student drift. Experiments across diverse distribution shifts and two VLOD architectures show consistent improvements over existing TTA methods. On VOC-C, DESA-TTA improves AP$_{50}$ by 14.5 points over zero-shot inference while achieving 55\% higher inference throughput than the previous state-of-the-art TTA method for YOLO-World. Our code: https://github.com/imatif17/DESA-TTA
Chinese Translation
视觉-语言物体检测器（VLODs）在零样本性能上表现出色，但在部署过程中仍然容易受到分布变化的影响。基于均值教师的方法用于测试时适应（TTA），通过使用教师生成的伪标签更新学生模型，从而提高鲁棒性。然而，均值教师TTA对教师更新的固定指数移动平均（EMA）系数的选择高度敏感，且在噪声伪标签下的重复优化可能导致学生模型的累积漂移。我们提出了用于TTA的动态EMA与源锚定（DESA-TTA），这是一种低开销的方法，通过动态时间平均和源锚定共同调节教师更新和学生漂移。动态时间平均根据伪标签的置信度和框密度估计教师的不确定性，并利用该不确定性选择在教师参数漂移所确定的范围内的样本级EMA系数。源锚定部分恢复更新后的学生参数向其预训练值，锚定强度随着学生漂移的增加而增强。在不同的分布变化和两种VLOD架构上的实验表明，DESA-TTA在现有TTA方法上有一致的改进。在VOC-C数据集上，DESA-TTA在零样本推理的基础上提高了14.5个点的AP$_{50}$，同时实现了比之前的YOLO-World最先进的TTA方法高出55 ext{%}的推理吞吐量。我们的代码： https://github.com/imatif17/DESA-TTA
cs.CV / 15 / 2609.01800
Improved Automatic Target Recognition in Synthetic Aperture Sonar Imagery Using Large Deep Neural Networks
基于大型深度神经网络的合成孔径声纳图像自动目标识别的改进
Abstract
Automatic Target Recognition (ATR) in Synthetic Aperture Sonar (SAS) is a task largely dominated by deep neural networks (DNNs). Most SAS-ATR models use convolutional neural network (CNN) architectures whereas transformer-based architectures have had much less representation in the literature despite being state of the art in general computer vision (CV) research. Additionally, researchers have had mixed results in attempting to overcome challenges presented by a scarcity of labeled training data by using methods such as data augmentation and the use of pretrained weights from a variety of imaging modalities. In this work, we compare the performance of modern CNN and transformer-based DNNs to determine which architecture and training configurations elicit the highest performance in SAS-ATR. We investigate how network size, architecture, pretraining method, data augmentation and other forms of regularization affect SAS-ATR performance with a focus on producing the highest-performing model and providing a roadmap for training state-of-the-art SAS-ATR models.
Chinese Translation
合成孔径声纳（SAS）中的自动目标识别（ATR）任务主要由深度神经网络（DNN）主导。大多数SAS-ATR模型使用卷积神经网络（CNN）架构，而基于变换器的架构在文献中代表性较少，尽管在一般计算机视觉（CV）研究中已被证明是最先进的。此外，研究人员在尝试通过数据增强和使用来自多种成像模式的预训练权重等方法来克服标记训练数据稀缺所带来的挑战时，结果也不尽相同。在本研究中，我们比较了现代CNN和基于变换器的DNN的性能，以确定哪种架构和训练配置能够在SAS-ATR中实现最高性能。我们调查了网络规模、架构、预训练方法、数据增强及其他形式的正则化如何影响SAS-ATR性能，重点是生产最高性能的模型，并为训练最先进的SAS-ATR模型提供路线图。
cs.CV / 16 / 2609.01806
Consistency as Regularization for Unsupervised Shadow Removal
一致性作为无监督阴影去除的正则化
Abstract
Shadow removal is an important preprocessing step for many vision tasks, yet existing supervised methods require paired shadow and shadow-free images, while unsupervised approaches often still rely on shadow masks or shadow-free references. We propose ShadowCLR, an unsupervised framework that learns shadow removal directly from shadow images. Our key observation is that shadows vary across observations while the underlying scene content remains largely consistent. We therefore use consistency across shadow observations as regularization, encouraging the model to recover scene-consistent appearance while suppressing shadow-specific variations. Global and local consistency further enable us to explore visually related images, learn from imperfectly aligned observations, and focus the representation on shared scene information. Experiments on multiple benchmarks show that ShadowCLR achieves competitive and often superior performance over state-of-the-art unsupervised methods, demonstrating that consistency can provide regularization for shadow removal without shadow masks or shadow-free images.
Chinese Translation
阴影去除是许多视觉任务的重要预处理步骤，然而现有的监督方法需要成对的阴影图像和无阴影图像，而无监督方法往往仍依赖于阴影掩模或无阴影参考图像。我们提出了ShadowCLR，一个无监督框架，直接从阴影图像中学习阴影去除。我们的关键观察是，阴影在不同观察中变化，而基础场景内容基本保持一致。因此，我们利用阴影观察之间的一致性作为正则化，鼓励模型恢复场景一致的外观，同时抑制特定于阴影的变化。全局和局部一致性进一步使我们能够探索视觉相关图像，从不完美对齐的观察中学习，并将表示集中在共享的场景信息上。在多个基准测试中的实验表明，ShadowCLR在性能上与最先进的无监督方法具有竞争力，且通常表现优越，证明了一致性可以为阴影去除提供正则化，而无需阴影掩模或无阴影图像。
cs.CV / 17 / 2609.01808
Integrated Laser Scanning and Image-Based Topology Optimization Techniques for Detection and Quantification of Visible and Subsurface Structural Defects
集成激光扫描与基于图像的拓扑优化技术用于可见和地下结构缺陷的检测与量化
Abstract
Reliable characterization of structural defects requires methods capable of resolving both directly observable surface damage and damage that is not visible from the inspected surface. This study presents two complementary non-contact, vision-based approaches for the detection and quantitative characterization of defects in structural components. The first approach employs high-resolution laser scanning to generate three-dimensional (3D) point clouds of damaged steel specimens. Comparative processing of measured and reference point clouds is used to localize damaged regions, quantify geometric loss, and transfer the measured defect geometry to a finite element representation. The second approach combines full-field surface deformation measurements obtained using three-dimensional digital image correlation (3D-DIC) with finite element model updating and topology optimization. In this inverse framework, measured surface response is used to infer subsurface abnormalities through their influence on the spatial distribution of structural response. Experimental steel-beam specimens containing controlled smooth defects and randomly distributed defects are used to evaluate the approaches. Comparisons with milling-based ground-truth measurements demonstrate that both methods can identify and quantify defect geometry, while providing complementary information for visible and subsurface damage assessment. The combined framework establishes a pathway toward high-fidelity, non-contact structural condition assessment and model updating for components with complex and irregular damage.
Chinese Translation
可靠的结构缺陷表征需要能够解决直接可观察的表面损伤和从检查表面不可见的损伤的方法。本研究提出了两种互补的非接触视觉基础方法，用于检测和定量表征结构组件中的缺陷。第一种方法采用高分辨率激光扫描生成受损钢样本的三维（3D）点云。通过对测量点云和参考点云的比较处理，定位受损区域，量化几何损失，并将测得的缺陷几何形状转移到有限元表示中。第二种方法结合了使用三维数字图像相关（3D-DIC）获得的全场表面变形测量与有限元模型更新和拓扑优化。在这一反演框架中，测得的表面响应用于推断地下异常，通过其对结构响应空间分布的影响。使用包含受控光滑缺陷和随机分布缺陷的实验钢梁样本来评估这些方法。与基于铣削的真实测量结果的比较表明，这两种方法均能识别和量化缺陷几何形状，同时为可见和地下损伤评估提供互补信息。该综合框架为高保真、非接触的结构状态评估和复杂不规则损伤组件的模型更新建立了一条途径。
cs.CV / 18 / 2609.01816
Video2Reaction: Training Foundation Video Models to Predict Audience Reaction
Video2Reaction：训练基础视频模型以预测观众反应
Abstract
We introduce Video2Reaction, a multimodal dataset that maps short movie segments to the induced emotional reactions of viewers in the wild, as expressed through social media comments. Video2Reaction captures the natural diversity of emotional responses by aggregating reactions from online comments at scale, modeling labels as distributions over categorical emotions to better reflect the subjective and ambiguous nature of emotional perception. We benchmark two vision-language models (VLMs) finetuned with LoRA, showing that VLMs learn effectively from Video2Reaction and outperform specialized baselines on dominant reaction prediction. We further demonstrate that VLMs pre-finetuned on Video2Reaction transfer effectively to VCE, another induced emotion dataset with a different taxonomy and video domain. Notably, LLaVA-NeXT-Video-7B pre-finetuned on Video2Reaction and adapted on only 1% of VCE training data achieves a top-3 accuracy of 0.682, on par with the best reported VCE performance trained on the full dataset. The dataset is available at https://huggingface.co/datasets/infofusionlab/Video2Reaction
Chinese Translation
我们介绍了Video2Reaction，一个多模态数据集，将短片段电影与观众在社交媒体评论中表达的情感反应相映射。Video2Reaction通过大规模聚合在线评论的反应，捕捉情感反应的自然多样性，将标签建模为分类情感的分布，以更好地反映情感感知的主观性和模糊性。我们基准测试了两个经过LoRA微调的视觉语言模型（VLMs），结果显示VLMs能够有效地从Video2Reaction中学习，并在主导反应预测上超越专业基线。我们进一步证明，预先在Video2Reaction上微调的VLMs能够有效转移到VCE，这是另一个具有不同分类法和视频领域的诱发情感数据集。值得注意的是，预先在Video2Reaction上微调并仅在1%的VCE训练数据上适应的LLaVA-NeXT-Video-7B，达到了0.682的前3名准确率，与在完整数据集上训练的最佳报告VCE性能相当。数据集可在https://huggingface.co/datasets/infofusionlab/Video2Reaction获取。
cs.CV / 19 / 2609.01823
Kirin: Animal Motion Generation from In-the-Wild Video
Kirin：来自野外视频的动物运动生成
Abstract
Understanding animal motion is fundamental to modeling animal behavior and biomechanics, yet progress in this area lags far behind human motion research due to the scarcity of high-quality motion data. While human motion can be captured in controlled environments, it is impractical for most animal species, resulting in small, domain-limited datasets that restrict downstream applications such as animation. To address this challenge, we introduce Kirin, a framework that reconstructs motion from video, learns motion priors at scale, and generates realistic motion that can be directly applied to animated assets. Using large collections of in-the-wild animal videos, we reconstruct 3D motion sequences and pair them with captions to create AiM3D, the first large-scale dataset offering aligned video-text-motion tuples for quadruped animals. Building on this dataset, we develop a visual-guided motion generation model that conditions on both text and image to guide the generation of realistic motion across diverse animal species. Finally, by leveraging an off-the-shelf image-to-3D model, we automatically rig and animate 3D meshes using generated motion, producing ready-to-render animated animals. Together, our dataset and framework establish a new foundation for large-scale, text and image conditioned animal motion generation and animation. Project page: https://kirin-ani.github.io/.
Chinese Translation
理解动物运动对于建模动物行为和生物力学至关重要，但由于高质量运动数据的稀缺，该领域的进展远远落后于人类运动研究。尽管人类运动可以在受控环境中捕捉，但对于大多数动物物种而言，这种方法并不切实际，导致数据集规模小且领域有限，从而限制了动画等下游应用。为了解决这一挑战，我们提出了Kirin，一个从视频重建运动、在大规模上学习运动先验并生成可直接应用于动画资产的真实运动的框架。通过使用大量野外动物视频，我们重建了3D运动序列，并将其与字幕配对，创建了AiM3D，这是第一个提供四足动物视频-文本-运动元组对齐的大规模数据集。在此数据集的基础上，我们开发了一个视觉引导的运动生成模型，该模型同时以文本和图像为条件，引导不同动物物种的真实运动生成。最后，通过利用现成的图像到3D模型，我们自动为生成的运动装配和动画3D网格，制作出可直接渲染的动画动物。我们的数据集和框架共同建立了一个用于大规模、以文本和图像为条件的动物运动生成和动画的新基础。项目页面：https://kirin-ani.github.io/
cs.CV / 20 / 2609.01827
SliceBridge: context-consistent repair of corrupted slice intervals in T1-weighted MRI
SliceBridge：在T1加权MRI中对损坏切片区间进行上下文一致性修复
Abstract
Structural magnetic resonance imaging (MRI) images are sometimes corrupted over a contiguous set of slices, where acquisition, motion, hardware, or reconstruction effects leave a single slice or short interval inconsistent with its neighbors while the rest of the image remains usable. Such localized corruption can bias downstream morphometric analysis, yet discarding or reacquiring an otherwise usable image is costly. We formulate this as an image restoration problem: given the location of the affected interval, reconstruct those slices from the surrounding anatomical and imaging context. We propose SliceBridge, a framework for restoring corrupted slice intervals in T1-weighted MRI using rectified flow matching conditioned on the surrounding intact slices and their relative slice positions. Through-plane consistency is encouraged by coupling the slices within the interval through interval-correlated initial noise, a shared flow time, and synchronized sampling. The restored interval is then inserted back, leaving all other slices unchanged. We trained and validated the model on 9,877 T1-weighted brain MRI volumes from four datasets and evaluated it on 581 external subjects using clean interval withholding and controlled corruptions. Compared with a matched model that reconstructed target slices independently, SliceBridge reduced error in slice-to-slice changes within repaired intervals by 32.9%-41.3% across interval lengths and achieved higher SSIM at every interval length. In controlled-corruption cases, SliceBridge reduced the median error in regional brain volume estimates produced by a downstream segmentation model from 1.95% in corrupted volumes to 1.05%.
Chinese Translation
结构性磁共振成像（MRI）图像有时会在一组连续的切片中受到损坏，其中采集、运动、硬件或重建效应导致单个切片或短区间与其邻近切片不一致，而图像的其余部分仍然可用。这种局部损坏可能会对下游的形态测量分析产生偏差，但丢弃或重新获取一幅本可用的图像是成本高昂的。我们将此问题表述为图像修复问题：给定受影响区间的位置，从周围的解剖和成像上下文中重建这些切片。我们提出了SliceBridge，一个用于修复T1加权MRI中损坏切片区间的框架，利用基于周围完整切片及其相对切片位置的校正流匹配。通过通过区间相关的初始噪声、共享流时间和同步采样，将区间内的切片耦合在一起，从而鼓励切片间的一致性。然后将修复后的区间插回，保持其他切片不变。我们在来自四个数据集的9,877个T1加权脑MRI体积上训练和验证了该模型，并在581个外部受试者上进行了评估，采用干净区间保留和控制损坏。与一个独立重建目标切片的匹配模型相比，SliceBridge在修复区间内的切片间变化错误减少了32.9%-41.3%，并在每个区间长度上实现了更高的结构相似性指数（SSIM）。在控制损坏的情况下，SliceBridge将下游分割模型产生的区域脑体积估计的中位误差从损坏体积中的1.95%降低到1.05%。
cs.CV / 21 / 2609.01840
Cross-Model Distillation of a Human-Pose Foundation Model from Unannotated Infant Video for Markerless 3D Pose Estimation
从未标注婴儿视频中跨模型蒸馏人类姿态基础模型以实现无标记3D姿态估计
Abstract
Spontaneous movement is one of the earliest windows onto an infant's neuromotor health, and structured clinical instruments that score it are validated early predictors of cerebral-palsy risk. However, they require specially trained raters, are time-consuming, and carry inter-rater variability. This motivates automated, video-based markerless assessment, especially as marker-based motion capture is impractical in infants. Yet the foundation models that make markerless capture possible are trained almost entirely on adults: our recent multi-view infant study found that no single model is jointly best, with strong 2D keypoint accuracy and direct 3D body recovery split across different models. While that study identifies this trade-off, it does not resolve it. Here, we perform cross-model distillation from the Sapiens 2 pose model into the SAM 3D Body model, using unannotated infant video alone. A frozen teacher supplies dense pseudo-labels, and a differentiable renderer aligns the predicted mesh to them in the training loop. On eleven held-out infants (18 sessions, 173 recordings) under our prior study's multi-view protocol, fine-tuning improves same-view 2D keypoint agreement with the Sapiens reference (median body percentage of correct keypoints @ 10px 0.22 -> 0.42, face 0.22 -> 0.42) and Procrustes-aligned mean per joint 3D position error (25.5 -> 22.2 mm). This demonstrates how cross-model distillation improves SAM 3D Body model performance on infants.
Chinese Translation
自发运动是观察婴儿神经运动健康的最早窗口之一，而对其进行评分的结构化临床工具已被验证为脑瘫风险的早期预测指标。然而，这些工具需要经过专门培训的评估者，耗时且存在评估者之间的变异性。这促使我们开发基于视频的自动化无标记评估，尤其是因为基于标记的运动捕捉在婴儿中并不实用。然而，使无标记捕捉成为可能的基础模型几乎完全是在成人身上训练的：我们最近的多视角婴儿研究发现，没有单一模型在强2D关键点准确性和直接3D身体恢复方面表现最佳，这些性能分散在不同模型之间。虽然该研究识别了这种权衡，但并未解决它。在此，我们从Sapiens 2姿态模型向SAM 3D身体模型进行跨模型蒸馏，仅使用未标注的婴儿视频。一个冻结的教师模型提供密集的伪标签，而一个可微分的渲染器在训练循环中将预测的网格与这些标签对齐。在我们之前研究的多视角协议下，对11名被试婴儿（18个会话，173个录音）进行微调，提高了与Sapiens参考模型的同视图2D关键点一致性（正确关键点的中位身体百分比@10px从0.22提高到0.42，面部从0.22提高到0.42）以及Procrustes对齐的每个关节3D位置误差的均值（从25.5毫米降低到22.2毫米）。这表明跨模型蒸馏如何改善SAM 3D身体模型在婴儿上的表现。
cs.CV / 22 / 2609.01876
RAFT-DVC: Resolution-Aware Machine Learning-Based Digital Volume Correlation
RAFT-DVC：基于分辨率感知的机器学习数字体积相关
Abstract
Digital volume correlation (DVC) provides three-dimensional full-field displacement measurements from volumetric images, but how the internal resolution of a machine-learning-based DVC model affects accuracy and operating range remains poorly understood. Here, we present RAFT-DVC, a resolution-aware family of recurrent all-pairs field transforms (RAFT)-based DVC solvers with encoder downsampling factors s = 2, 4, and 8. Using a matched design, we find that the three solvers localize displacement to approximately 0.017 feature-grid voxel, giving an empirical raw-volume error scaling of approximately 0.017s voxel. The solvers exhibit complementary operating regimes governed jointly by displacement reach and volumetric-texture compatibility. Synthetic benchmarks show that RAFT-DVC achieves errors of the same order as tuned classical DVC under fine-texture, small-to-moderate-displacement conditions and becomes competitive or advantageous under coarse-texture, large-displacement conditions. Frequency-swept tests quantify deformation spatial resolution, while tiled inference enables dense estimation on large volumes. Evaluation on confocal volumetric images acquired during indentation illustrates the importance of matching solver operating regime to deformation magnitude and image texture. Tests on micro-CT images of elastomeric foam, despite training only on particle-labeled synthetic data, provide evidence of cross-texture transfer. We also identify coordinate-order inconsistencies in three-dimensional RAFT correlation sampling and introduce a non-cubic impulse test to verify sampler geometry independently of network training. Correcting the sampler improves native-input accuracy and generalization to unseen volume dimensions. Together, these results establish RAFT-DVC as a fast, resolution-aware framework for dense DVC with characterized accuracy and operating regimes.
Chinese Translation
数字体积相关（DVC）从体积图像中提供三维全场位移测量，但机器学习基础的 DVC 模型的内部分辨率如何影响准确性和操作范围仍然不甚了解。在此，我们提出了 RAFT-DVC，一种基于递归全对场变换（RAFT）的分辨率感知 DVC 求解器系列，具有编码器下采样因子 s = 2、4 和 8。通过匹配设计，我们发现这三种求解器能够将位移定位到大约 0.017 特征网格体素，给出大约 0.017s 体素的经验原始体积误差缩放。求解器表现出互补的操作模式，由位移范围和体积纹理兼容性共同决定。合成基准测试表明，在细纹理、小到中等位移条件下，RAFT-DVC 的误差与调优后的经典 DVC 相当，而在粗纹理、大位移条件下则变得具有竞争力或优势。频率扫描测试量化了变形空间分辨率，而平铺推理则实现了对大体积的密集估计。在压痕过程中获取的共聚焦体积图像的评估显示了将求解器操作模式与变形幅度和图像纹理匹配的重要性。尽管仅在粒子标记的合成数据上进行训练，但对弹性泡沫的微型 CT 图像的测试提供了跨纹理转移的证据。我们还识别了三维 RAFT 相关采样中的坐标顺序不一致性，并引入了一种非立方脉冲测试，以独立于网络训练验证采样器几何形状。修正采样器提高了原生输入的准确性，并增强了对未见体积维度的泛化能力。综合来看，这些结果确立了 RAFT-DVC 作为一个快速、分辨率感知的密集 DVC 框架，具有明确的准确性和操作模式。
cs.CV / 23 / 2609.01886
SignMatch: Matching Dictionary Signs to Continuous Sign Language Video
SignMatch：将词典手势与连续手语视频匹配
Abstract
The objective of this paper is to match dictionary sign videos to corresponding signs in continuous signing videos, where a match is defined by the visual similarity alone - the handshape and motion relative to the body. To achieve this, we learn a prototype-structured sign embedding space from continuous video annotated with signs, where each learnable prototype corresponds to a sign class. Isolated dictionary videos are then mapped into this sign space, enabling the matching between dictionary exemplars and continuous sign instances. This design supports direct dictionary-guided sign matching through embedding similarity and naturally extends to unseen signs using only dictionary exemplars. Experiments on ASL-Citizen dictionary retrieval, ChaLearn OSLWL dictionary-to-continuous sign matching, and using BOBSL's CSLR2 evaluation for automatic sign annotation demonstrate strong generalisation across datasets, tasks and sign languages. Without benchmark-specific supervision, the learned representation transfers effectively across American, British, and Spanish Sign Languages, outperforming prior methods on all three benchmarks. Project page: https://www.robots.ox.ac.uk/~vgg/research/signmatch/
Chinese Translation
本文的目的是将词典手势视频与连续手语视频中的相应手势进行匹配，其中匹配仅通过视觉相似性来定义——手形和相对于身体的动作。为了实现这一目标，我们从带有手势标注的连续视频中学习一个原型结构的手势嵌入空间，其中每个可学习的原型对应一个手势类别。然后，将孤立的词典视频映射到这个手势空间，从而实现词典示例与连续手势实例之间的匹配。这一设计支持通过嵌入相似性进行直接的词典引导手势匹配，并自然地扩展到仅使用词典示例的未见手势。在ASL-Citizen词典检索、ChaLearn OSLWL词典到连续手势匹配以及使用BOBSL的CSLR2评估进行自动手势标注的实验中，展示了在数据集、任务和手语之间的强泛化能力。在没有特定基准监督的情况下，学习到的表示在美国、英国和西班牙手语之间有效迁移，在所有三个基准上均优于先前的方法。项目页面：https://www.robots.ox.ac.uk/~vgg/research/signmatch/
cs.CV / 24 / 2609.01888
Does Playing it Safe Count as Faithfulness? Reassessing LVLM Hallucination Mitigation Methods
安全游戏算不算忠实性？重新评估大型视觉语言模型的幻觉缓解方法
Abstract
Recent inference-time hallucination mitigation methods for large vision-language models (LVLMs) report strong gains on hallucination benchmarks. However, it remains unclear whether lower hallucination scores reflect improved multimodal grounding or more conservative generation. We evaluate six mitigation methods across three LVLMs and four benchmarks, including hallucination-focused evaluation and the diverse capability benchmark MMStar. Our analysis reveals two consistent patterns. First, hallucination reduction is often coupled with reduced informativeness: methods that lower hallucination rates also reduce object recall, visual coverage, or response detailedness. Second, improvements on hallucination benchmarks do not reliably transfer to broader multimodal capabilities, with methods showing inconsistent or degraded performance on fine-grained perception and reasoning tasks. Our findings suggest that current evaluation protocols may overestimate progress by rewarding conservative generation. We argue that hallucination mitigation should be evaluated as a faithfulness--informativeness--capability trade-off rather than through hallucination scores alone.
Chinese Translation
近期针对大型视觉语言模型（LVLM）的推理时幻觉缓解方法在幻觉基准测试中报告了显著的提升。然而，较低的幻觉得分是否反映了多模态基础的改善，还是更为保守的生成方式，仍然不清楚。我们评估了三种LVLM和四个基准下的六种缓解方法，包括专注于幻觉的评估和多样化能力基准MMStar。我们的分析揭示了两个一致的模式。首先，幻觉减少通常伴随着信息量的降低：降低幻觉率的方法也减少了对象回忆、视觉覆盖或响应的详细程度。其次，在幻觉基准上的改进并不可靠地转移到更广泛的多模态能力上，这些方法在细粒度感知和推理任务上表现不一致或退化。我们的发现表明，当前的评估协议可能通过奖励保守生成来高估进展。我们认为，幻觉缓解应作为忠实性-信息量-能力的权衡进行评估，而不仅仅依赖幻觉得分。
cs.CV / 25 / 2609.01899
TAPVid-MV: A Benchmark for Tracking Any Point in 3D Across Multiple Views
TAPVid-MV：一个用于在多个视角中跟踪任意三维点的基准
Abstract
Multi-camera systems are increasingly practical for robotics, AR/VR, and autonomous driving because complementary views reduce depth ambiguity and preserve visibility under occlusion. Existing point-tracking benchmarks, however, focus on a single video or static multi-camera rigs. None test long-term 3D point tracking across several synchronized views under camera motion. We introduce TAPVid-MV (Tracking Any Point in Video across Multiple Views), the first benchmark for this setting. It contains a curated set of 284 sequences, 1,142 calibrated camera streams, and 109,769 point tracks across seven subsets spanning indoor and outdoor domains, from robotics and human activity to driving and synthetic procedural scenes. We obtain these trajectories using dataset-specific auxiliary modalities: sensor depth, LiDAR, SLAM and SfM points, human meshes, posed object meshes, and simulation. Every sequence and trajectory is visually verified by human annotators. Across more than 30 baselines, no method comes close to solving the task. Surprisingly, existing multi-view point trackers do not consistently outperform monocular point trackers. By evaluating reconstruction and point tracking on the same datasets, TAPVid-MV helps distinguish errors in recovered geometry from errors in point correspondence. Through this joint analysis, we identify geometry recovery as a major bottleneck for accurate 3D point tracking. Beyond multi-view 3D point tracking, our released annotations support monocular 2D and 3D point tracking, future-trajectory prediction, and 4D reconstruction.
Chinese Translation
多摄像头系统在机器人技术、增强现实/虚拟现实和自动驾驶等领域越来越实用，因为互补视角可以减少深度歧义并在遮挡情况下保持可见性。然而，现有的点跟踪基准主要集中在单一视频或静态多摄像头设备上。没有任何基准测试在相机运动下跨多个同步视角进行长期三维点跟踪。我们提出了TAPVid-MV（Tracking Any Point in Video across Multiple Views），这是第一个针对这种设置的基准。它包含了一组经过精心挑选的284个序列、1,142个校准的摄像头流和109,769个点轨迹，涵盖了室内和室外领域的七个子集，从机器人技术和人类活动到驾驶和合成程序场景。我们使用数据集特定的辅助模态获取这些轨迹：传感器深度、激光雷达、SLAM和SfM点、人类网格、姿态物体网格以及仿真。每个序列和轨迹都经过人工标注者的视觉验证。在超过30个基线方法中，没有任何方法接近解决这一任务。令人惊讶的是，现有的多视角点跟踪器并没有始终优于单目点跟踪器。通过在相同数据集上评估重建和点跟踪，TAPVid-MV帮助区分恢复几何体中的错误与点对应中的错误。通过这种联合分析，我们确定几何体恢复是准确的三维点跟踪的主要瓶颈。除了多视角三维点跟踪外，我们发布的注释还支持单目二维和三维点跟踪、未来轨迹预测和四维重建。
cs.CV / 26 / 2609.01921
Automated Maize Ear Phenotyping Using 3D Reconstructions
基于三维重建的玉米穗自动表型分析
Abstract
Maize kernel traits such as row number, kernels per row, and kernel size vary largely for genetic reasons and are consistently associated with regions of the genome that influence yield. Manual measurement of these traits, however, cannot keep pace with the volume of maize generated in a breeding program. To address this, we developed and validated a fully automated pipeline for extracting these traits from 3D point clouds of corn ears, built on a recently developed video-to-point-cloud platform. Raw video frames are processed through COLMAP and NeRF, the ear is isolated via density-based separation, and the point cloud is distance-calibrated to physical units. The calibrated ear point cloud was Z-axis aligned via PCA and cylindrically unwrapped to a 2D image. We enhanced contrast and performed zero-fine-tuning instance segmentation using Cellpose-SAM. A triple-juxtaposed unwrap strategy was used to prevent double-counting at the seam. The pipeline achieved kernel count R^2 = 0.921 (MAPE = 10.33%) and kernel row number within +-2 rows for 95.2% of ears (MAE = 0.75 rows) on a 168-ear held-out set from the 268-ear labeled dataset. The resulting multi-trait dataset has known genotype identity for each ear, positioning it for phenotype-to-genotype association analyses.
Chinese Translation
玉米籽粒性状如行数、每行籽粒数和籽粒大小因遗传原因而存在较大差异，并且与影响产量的基因组区域密切相关。然而，手动测量这些性状无法跟上育种项目中生成的玉米数量。为了解决这一问题，我们开发并验证了一种完全自动化的管道，从玉米穗的三维点云中提取这些性状，该管道基于最近开发的视频到点云平台。原始视频帧通过COLMAP和NeRF进行处理，穗通过基于密度的分离进行隔离，点云被距离标定为物理单位。经过标定的穗点云通过主成分分析（PCA）进行Z轴对齐，并以圆柱形展开为二维图像。我们增强了对比度，并使用Cellpose-SAM进行了零微调的实例分割。采用三重并排展开策略以防止在接缝处重复计数。该管道在268个标记数据集中，168个穗的保留集上实现了籽粒计数R^2 = 0.921（平均绝对百分比误差 = 10.33%），95.2%的穗的籽粒行数在±2行范围内（平均绝对误差 = 0.75行）。生成的多性状数据集为每个穗提供了已知的基因型身份，为表型与基因型关联分析奠定了基础。
cs.CV / 27 / 2609.01928
Learning with Volterra Neural Networks: A System Theoretic Perspective
基于沃尔泰拉神经网络的学习：系统理论视角
Abstract
Higher-order interaction components are important for signal, image, and video modeling, but explicit high-order operators often suffer from rapidly increasing parameter and computational costs. This paper presents kVNN, a learnable kernelized Volterra Neural operator for compact higher-order filtering. The motivation is to use kernelization to improve the efficiency of Volterra-type neural operators while providing a structured interpretation of their higher-order components. The proposed formulation combines the order-wise structure of Volterra filtering with learnable polynomial-kernel atoms, allowing different interaction orders to be represented by separate learnable centers and coefficients. This order-decoupled representation avoids explicit high-order tensor parameterization and can be implemented as a CNN-compatible layer. Experiments on representative vision tasks show that kVNN achieves a favorable accuracy--efficiency trade-off.
Chinese Translation
高阶交互成分在信号、图像和视频建模中至关重要，但显式的高阶算子往往面临参数和计算成本快速增加的问题。本文提出了kVNN，一种可学习的核化沃尔泰拉神经算子，用于紧凑的高阶滤波。其动机在于利用核化技术提高沃尔泰拉类型神经算子的效率，同时提供其高阶成分的结构化解释。所提出的公式结合了沃尔泰拉滤波的阶次结构与可学习的多项式核原子，使得不同的交互阶次可以通过独立的可学习中心和系数来表示。这种阶次解耦的表示避免了显式的高阶张量参数化，并可以实现为与卷积神经网络（CNN）兼容的层。对代表性视觉任务的实验表明，kVNN在准确性与效率之间实现了良好的权衡。
cs.CV / 28 / 2609.01963
Aggregating Neighbor Embedding Projection and Rank-Based Manifold Learning for Image Retrieval
邻域嵌入投影与基于排名的流形学习在图像检索中的聚合
Abstract
Content-based image retrieval (CBIR) has advanced significantly with deep learning, yet effectively ranking similar images remains challenging, particularly in high-dimensional feature spaces, where pairwise distances often fail to capture contextual relationships and the semantic gap between visual features and high-level concepts persists. Manifold learning and rank-based refinement methods have emerged as complementary strategies, respectively improving feature representations and exploiting contextual information embedded in ranked lists, such as neighborhood relationships among images. However, combining these projection-based and rank-based strategies to exploit their complementary properties remains a challenging research problem. To address this, we propose a framework that combines neighbor embedding projections with rank-based manifold learning through rank aggregation. Uniform Manifold Approximation and Projection (UMAP) generates alternative low-dimensional feature representations, and ranked lists obtained from UMAP projections and rank-based re-ranking methods are combined using the Borda Count aggregation strategy. Experiments were conducted on several public datasets using deep learning features extracted from ResNet152, Swin Transformer, and DINOv2 models. Results show that the proposed approach improves retrieval effectiveness in several scenarios, particularly when the baseline representation struggles to achieve high precision. The aggregation strategy also often improves the quality of top-ranked positions, leading to competitive Mean Average Precision (MAP) and Precision values across different datasets and feature extractors. These findings suggest that combining projection-based and rank-based manifold learning strategies through rank aggregation can provide complementary contextual information for image retrieval tasks.
Chinese Translation
基于内容的图像检索（CBIR）在深度学习的推动下取得了显著进展，但在高维特征空间中有效地对相似图像进行排序仍然具有挑战性。在这些空间中，成对距离往往无法捕捉上下文关系，并且视觉特征与高层次概念之间的语义差距依然存在。流形学习和基于排名的优化方法已成为互补策略，分别改善特征表示和利用嵌入在排名列表中的上下文信息，例如图像之间的邻域关系。然而，如何将这些基于投影和基于排名的策略结合起来以利用其互补特性仍然是一个具有挑战性的研究问题。为此，我们提出了一个框架，通过排名聚合将邻域嵌入投影与基于排名的流形学习相结合。统一流形近似与投影（UMAP）生成替代的低维特征表示，来自UMAP投影和基于排名的重新排序方法获得的排名列表通过博尔达计数聚合策略进行结合。在多个公共数据集上进行了实验，使用从ResNet152、Swin Transformer和DINOv2模型提取的深度学习特征。结果表明，所提出的方法在多个场景中提高了检索效果，特别是在基线表示难以达到高精度时。聚合策略还通常改善了排名前列位置的质量，使得在不同数据集和特征提取器上获得了具有竞争力的平均精度（MAP）和精度值。这些发现表明，通过排名聚合结合基于投影和基于排名的流形学习策略可以为图像检索任务提供互补的上下文信息。
cs.CV / 29 / 2609.01987
Morphology signal in whole slide image foundation models can automatically triage slides
全切片图像基础模型中的形态信号可以自动分流切片
Abstract
Patient exams in the cancer diagnosis and staging process typically generate several whole slide images (WSIs). One of the initial steps in training models on WSI data is identifying one or a few slides containing tumor or other diagnostic biomarkers necessary for downstream prediction tasks such as estimating recurrence risk or progression-free survival. This step requires tedious manual curation by experienced pathologists. Many published datasets make the artificial assumption of 1 slide per patient. Alternatively, all slides per patient may be used for model training, which may dilute the signal from the few slides containing tumor or other relevant information. In this paper, we present a pipeline to overcome these challenges using publicly available WSI foundation models (FMs). Our evaluations show that ranking WSIs based on predictions from zero-shot classification using WSI FMs accurately identifies slides with the most tumor, indicating that WSI FMs contain sufficient morphology signal to automatically triage slides. We also present a formulation for ranked evaluation to benchmark FM performance in slide triage. We show, on multiple datasets, that tumor slides are identified in the top-2 ranked slides for patients with up to 43 slides.
Chinese Translation
在癌症诊断和分期过程中，患者检查通常会生成多个全切片图像（WSIs）。在对WSI数据进行模型训练的初始步骤之一是识别包含肿瘤或其他诊断生物标志物的一个或几个切片，这些生物标志物对于后续的预测任务（如估计复发风险或无进展生存期）是必要的。此步骤需要经验丰富的病理学家进行繁琐的人工整理。许多已发布的数据集假设每位患者仅有1个切片。另一种选择是使用每位患者的所有切片进行模型训练，这可能会稀释包含肿瘤或其他相关信息的少数切片所传递的信号。在本文中，我们提出了一种利用公开可用的WSI基础模型（FMs）来克服这些挑战的流程。我们的评估表明，基于使用WSI FMs进行零-shot分类的预测对WSIs进行排名，能够准确识别出含有最多肿瘤的切片，表明WSI FMs包含足够的形态信号以自动分流切片。我们还提出了一种排名评估的公式，以基准FM在切片分流中的表现。我们在多个数据集上展示，肿瘤切片在最多43个切片的患者中被识别为排名前2的切片。
cs.CV / 30 / 2609.01997
Linear Fusion MultiDiffusion for Fast Training-Free Spherical Panorama Generation
线性融合多重扩散用于快速无训练球形全景生成
Abstract
We propose LF-MultiDiffusion, a training-free panorama generation method that extends MultiDiffusion to support linear projections between target and reference image spaces. Our key idea is to reformulate latent aggregation as a regularized least-squares problem and solve it efficiently with a Krylov-based iterative solver inside the denoising loop. This formulation enables denser and more natural mappings than prior training-free methods, yielding more stable generation with far fewer perspective views. As a result, LF-MultiDiffusion reduces the number of image generator evaluations during denoising and significantly improves inference efficiency. Experiments show that LF-MultiDiffusion achieves better visual quality, text alignment, and panoramic consistency than the strongest training-free baseline, while providing a 15.36$\times$ speedup. Our project page is available at: https://ahykw.github.io/lfmd.
Chinese Translation
我们提出了LF-MultiDiffusion，这是一种无训练的全景生成方法，扩展了多重扩散（MultiDiffusion），以支持目标和参考图像空间之间的线性投影。我们的关键思想是将潜在聚合重新表述为一个正则化的最小二乘问题，并在去噪循环中使用基于Krylov的迭代求解器高效求解。该表述能够实现比先前的无训练方法更密集和更自然的映射，从而在更少的视角下产生更稳定的生成结果。因此，LF-MultiDiffusion减少了去噪过程中图像生成器的评估次数，并显著提高了推理效率。实验表明，LF-MultiDiffusion在视觉质量、文本对齐和全景一致性方面优于最强的无训练基线，同时提供了15.36倍的加速。我们的项目页面可访问： https://ahykw.github.io/lfmd。
cs.CV / 31 / 2609.02000
Who Drives the Probability Game of VLMs? A Temporal Causal Drive Evaluation Framework
谁驱动了视觉语言模型的概率游戏？一个时间因果驱动评估框架
Abstract
Vision-language models (VLMs) are increasingly evaluated on complex image and video understanding tasks, yet conventional metrics primarily assess final-answer quality and reveal little about how different information sources shape the generation process. We propose a causal and temporal evaluation framework that traces the evolving roles of visual input, question text, and generated prefixes during autoregressive decoding. Grounded in a Structural Causal Model, we use interventions and backdoor adjustment to derive three step-indexed causal-drive metrics---Visual Causal Drive (VCD), Question Causal Drive (QCD), and Prefix Causal Drive (PCD)---for characterizing source-specific generation patterns without requiring reference answers. Experiments on Qwen3-VL-8B-Instruct across MAVIS, LLaVA-Video-178K, and MiraData, together with cross-model validation on InternVL2-8B, reveal a consistent transition from stronger early question and visual guidance toward increasing reliance on generated prefixes. Randomized-intervention validation shows that QCD and PCD reduce recovery error over observational PMI baselines by 34.8\% and 47.1\%, respectively. On VLMBias, the prefix--visual imbalance score achieves 0.767 AUROC and 0.873 AUPRC for distinguishing prior-driven from visually grounded generations. These results show that causal-drive trajectories provide complementary source-level diagnostics for multimodal generation.
Chinese Translation
视觉语言模型（VLMs）越来越多地在复杂的图像和视频理解任务中进行评估，但传统指标主要评估最终答案的质量，几乎没有揭示不同信息源如何影响生成过程。我们提出了一种因果和时间评估框架，追踪自回归解码过程中视觉输入、问题文本和生成前缀的演变角色。基于结构因果模型，我们使用干预和后门调整推导出三种逐步索引的因果驱动指标——视觉因果驱动（Visual Causal Drive, VCD）、问题因果驱动（Question Causal Drive, QCD）和前缀因果驱动（Prefix Causal Drive, PCD）——用于表征特定源的生成模式，而无需参考答案。在MAVIS、LLaVA-Video-178K和MiraData上对Qwen3-VL-8B-Instruct的实验，以及在InternVL2-8B上的跨模型验证，揭示了从早期更强的问题和视觉指导向对生成前缀的依赖逐渐增强的持续转变。随机干预验证表明，QCD和PCD分别比观察性PMI基线减少了34.8%和47.1%的恢复误差。在VLMBias上，前缀-视觉不平衡分数达到了0.767的AUROC和0.873的AUPRC，用于区分由先验驱动和视觉基础生成的结果。这些结果表明，因果驱动轨迹为多模态生成提供了互补的源级诊断。
cs.CV / 32 / 2609.02002
InsightSeg: Reusing Correction Insights for Guideline-Consistent Segmentation
InsightSeg：重用纠正洞察以实现一致的分割指导
Abstract
Guideline-consistent semantic segmentation requires more than category recognition, as real-world labeling policies demand fine-grained, task-specific decisions. Recent multi-agent refinement systems improve compliance with such textual guidelines by detecting and correcting errors. However, they are stateless: feedback from the critiquing agent is discarded, causing the same guideline-specific mistakes to be repeatedly rediscovered and corrected across the dataset at the cost of additional refinement. We introduce InsightSeg, an episodic memory mechanism that converts successful correction episodes into reusable, visually grounded insights. A meta-analyzer distills each qualifying episode into directive natural-language insights and anchors them to the local image regions that caused the error using patch-level visual concept vectors. On subsequent images, these concepts are matched against dense patch embeddings to retrieve relevant insights, which condition the segmenting agent before making its first prediction. This shifts the system from correcting recurring errors to preventing them, improving segmentation quality before any refinement occurs. Across Waymo and Cityscapes, InsightSeg improves both first-pass and final guideline-consistent segmentation performance while requiring fewer refinement steps, demonstrating that multi-agent refinement can become more accurate and efficient by drawing on past correction experience.
Chinese Translation
一致的语义分割不仅需要类别识别，因为现实世界的标注政策要求细粒度、任务特定的决策。近期的多代理精细化系统通过检测和纠正错误来提高对这些文本指导的遵从性。然而，它们是无状态的：批评代理的反馈被丢弃，导致相同的指导特定错误在数据集中被反复发现和纠正，从而增加了额外的精细化成本。我们提出了InsightSeg，一种情节记忆机制，将成功的纠正情节转化为可重用的、视觉基础的洞察。元分析器将每个合格的情节提炼为指令性自然语言洞察，并使用补丁级视觉概念向量将其锚定到导致错误的局部图像区域。在后续图像中，这些概念与密集补丁嵌入进行匹配，以检索相关洞察，从而在分割代理进行第一次预测之前对其进行调节。这使得系统从纠正重复错误转变为防止错误，提高了分割质量，而无需进行任何精细化。在Waymo和Cityscapes数据集上，InsightSeg提高了首次和最终一致的分割性能，同时减少了精细化步骤，证明多代理精细化可以通过借鉴过去的纠正经验变得更加准确和高效。
cs.CV / 33 / 2609.02004
InstEditSeg: Instruction-Driven Image Editing for Polyp and Skin Lesion Segmentation
InstEditSeg：基于指令的息肉和皮肤病变分割图像编辑
Abstract
Accurate segmentation of polyps and skin lesions is pivotal for clinical diagnosis, yet existing methods struggle with low contrast, ambiguous boundaries, and cross-domain distribution discrepancies. Discriminative networks and most diffusion-based segmentation approaches predict standalone binary masks, leaving the visual priors of large-scale pretrained generative models largely unexploited. We propose InstEditSeg, a unified generative framework that reformulates medical segmentation as an instruction-driven image editing problem. Instead of emitting a mask, the model renders a color-coded overlay on the original image, conditioned on a textual instruction, so that the edited output aligns with the natural image distribution learned by latent diffusion models and mitigates the domain gap between natural and medical imagery. To recover fine anatomical structures, we introduce DINOv3 as an auxiliary visual encoder and a DINO Feature Guidance Block that builds a multi-scale feature pyramid. The pyramid is fused into the diffusion U-Net by channel concatenation and zero-initialized convolution so that hierarchical discriminative priors can be injected without perturbing the pretrained weights. A dual-branch classifier-free guidance strategy requiring only two forward passes per denoising step reduces inference cost. On polyp and skin lesion benchmarks the framework achieves accuracy competitive with strong discriminative baselines, and it further demonstrates concrete advantages of the generative formulation: notably better cross-domain generalization on unseen data, more complete multi-lesion segmentation, instruction-conditioned task control, and sampling flexibility. We also analyze the strengths and limitations of the paradigm, including its color sensitivity and unsupported attribute-conditioned selection. Code is available at: https://github.com/wincharm001/InstEditSeg.
Chinese Translation
息肉和皮肤病变的准确分割对临床诊断至关重要，但现有方法在低对比度、模糊边界和跨域分布差异方面面临挑战。区分网络和大多数基于扩散的分割方法预测独立的二进制掩膜，未能充分利用大规模预训练生成模型的视觉先验。我们提出了InstEditSeg，一个统一的生成框架，将医学分割重新定义为基于指令的图像编辑问题。模型不是输出掩膜，而是在原始图像上渲染一个基于文本指令的彩色覆盖层，从而使编辑后的输出与潜在扩散模型学习的自然图像分布一致，并减小自然图像与医学图像之间的领域差距。为了恢复细微的解剖结构，我们引入了DINOv3作为辅助视觉编码器，并构建了一个多尺度特征金字塔的DINO特征引导模块。该金字塔通过通道连接和零初始化卷积融合到扩散U-Net中，以便在不干扰预训练权重的情况下注入层次化的区分先验。双分支无分类器引导策略仅需每个去噪步骤进行两次前向传递，从而降低推理成本。在息肉和皮肤病变基准测试中，该框架的准确性与强大的区分基线相当，并进一步展示了生成形式的具体优势：在未见数据上显著更好的跨域泛化能力、更完整的多病变分割、基于指令的任务控制和采样灵活性。我们还分析了该范式的优缺点，包括其对颜色的敏感性和不支持的属性条件选择。代码可在以下链接获取：https://github.com/wincharm001/InstEditSeg。
cs.CV / 34 / 2609.02012
GeoStore: Finding Small Storefronts in Large Scenes -- A Fine-Grained POI Localization Benchmark with Global-to-Local Asymmetric Matching
GeoStore：在大场景中寻找小型店面——一个全球到本地不对称匹配的细粒度POI定位基准
Abstract
Point-of-interest (POI) localization -- matching a user's close-up storefront photograph against large-scale geo-tagged street-view imagery -- underpins map construction, POI verification, and location-based services. Its closest existing paradigm, visual place recognition (VPR), assumes symmetric, whole-image matching of the same scene at a comparable scale; POI localization instead must match a close-up query, in which the target fills the frame, against wide references in which the same POI occupies only a small, off-center region among visually similar shops, under a substantial capture-domain gap. We introduce GeoStore, to our knowledge the first benchmark dedicated to this asymmetric, fine-grained, open-set formulation, and show that global-descriptor methods tuned for symmetric VPR are systematically limited on it, since a single global vector dilutes the small target. We further propose GLAM (Global-to-Local Asymmetric Matching), which couples a retrieval-anchoring global descriptor with an asymmetric local pathway: each reference is kept as a compact set of pooled region tokens and matched against a single query probe through a learnable soft late interaction; at inference, the same tokens enable a lightweight mutual-nearest-neighbor re-ranking. GLAM surpasses strong global and two-stage baselines on Recall@1/5/10 and mAP, with ~5x smaller re-ranking features and ~two orders of magnitude lower per-pair matching cost than prior local re-ranking. The benchmark and code will be publicly released.
Chinese Translation
兴趣点（POI）定位——将用户的近距离店面照片与大规模地理标记的街景图像进行匹配——是地图构建、POI验证和基于位置的服务的基础。其最接近的现有范式是视觉地点识别（VPR），该方法假设在可比规模下对同一场景进行对称的全图匹配；而POI定位则必须将近距离查询（目标填满画面）与宽幅参考（同一POI仅占据视觉上相似商店中一个小的偏离中心区域）进行匹配，并且存在显著的捕获域差距。我们介绍了GeoStore，这是我们所知的首个专门针对这种不对称、细粒度、开放集形式的基准，并展示了针对对称VPR调优的全局描述符方法在此方面的系统性局限，因为单一的全局向量稀释了小目标。我们进一步提出了GLAM（全球到本地不对称匹配），它将检索锚定的全局描述符与不对称的局部路径相结合：每个参考保持为一组紧凑的池化区域标记，并通过可学习的软晚期交互与单个查询探针进行匹配；在推理时，相同的标记使得轻量级的互最近邻重新排序成为可能。GLAM在Recall@1/5/10和mAP上超越了强大的全局和两阶段基准，具有约5倍更小的重新排序特征和约两个数量级更低的每对匹配成本。该基准和代码将公开发布。
cs.CV / 35 / 2609.02028
Detecting Object Hallucinations in Large Vision-Language Models via Cross-Modal Attention Drifts and Mask-Based Verification
通过跨模态注意力漂移和基于掩码的验证检测大型视觉-语言模型中的对象幻觉
Abstract
Despite recent advances in large vision-language models (LVLMs), object hallucination remains a major barrier to their reliable deployment. Existing detection methods often characterize visual grounding using attention from individual layers, leaving its evolution across layers underexplored. We propose CADMP, a lightweight object hallucination detection framework that combines adjacent-layer cross-modal attention drift with prediction sensitivity to targeted visual masking. During decoding, CADMP quantifies distributional changes between consecutive cross-modal attention maps to capture abrupt transitions in visual grounding. It then selects the transition with the largest drift, locates the corresponding visually relevant regions, and measures the change in prediction probability after masking these regions. These two signals provide complementary evidence: attention drift characterizes the stability of internal visual grounding, while probability variation verifies whether a prediction truly depends on the identified visual evidence. A lightweight detector integrates both signals to identify hallucinated predictions. Experiments on multiple benchmarks and representative open-source LVLMs demonstrate that CADMP achieves consistently competitive detection performance. Ablation studies further confirm the complementary contributions of adjacent-layer drift modeling and mask-based grounding verification.
Chinese Translation
尽管大型视觉-语言模型（LVLMs）最近取得了进展，但对象幻觉仍然是其可靠部署的主要障碍。现有的检测方法通常通过单层的注意力来表征视觉基础，导致其在各层之间的演变未得到充分探索。我们提出了CADMP，一个轻量级的对象幻觉检测框架，结合了相邻层之间的跨模态注意力漂移与对目标视觉掩码的预测敏感性。在解码过程中，CADMP量化连续跨模态注意力图之间的分布变化，以捕捉视觉基础中的突变。然后，它选择漂移最大的过渡，定位相应的视觉相关区域，并测量在掩码这些区域后预测概率的变化。这两个信号提供了互补的证据：注意力漂移表征内部视觉基础的稳定性，而概率变化验证预测是否真正依赖于识别的视觉证据。一个轻量级检测器整合这两个信号以识别幻觉预测。在多个基准和代表性的开源LVLMs上的实验表明，CADMP实现了一致的竞争性检测性能。消融研究进一步确认了相邻层漂移建模和基于掩码的基础验证的互补贡献。
cs.CV / 36 / 2609.02036
SelfLift: Accelerating Few-Step Diffusion via Self-Recovering Resolution Transition
SelfLift：通过自我恢复分辨率过渡加速少步扩散
Abstract
Few-step diffusion models substantially compress temporal computation, making the spatial cost of each model evaluation an increasingly dominant source of inference latency. Progressive-resolution inference reduces this cost by performing early denoising at low resolution and reserving high-resolution computation for refinement. However, existing methods typically lift intermediate latents directly and rely on subsequent steps to absorb the induced distribution mismatch. In the few-step regime, the limited recovery budget leaves these errors as visible artifacts, constraining how late the transition can occur and, consequently, how efficiently it can be performed. We introduce SelfLift, a self-recovering progressive-resolution framework that derives both transition-repair signals and trajectory-aligned supervision from the generative model itself. SelfLift-zero proposes a training-free Artifact-Aware Consistency Lift, using disagreement between direct latent lifting and pixel-VAE re-encoding as both a localized artifact-risk signal and a model-native correction direction. It enables reliable late transitions without external super-resolution, extra denoiser evaluations, or sampling-schedule modifications. Building on this robust transition, SelfLift-rich performs On-Policy Self Recovery on student-visited states, transferring dense high-resolution guidance from an internal self-teacher while remaining aligned with the altered progressive-resolution dynamics. Across FLUX.2-Klein and Z-Image-Turbo, SelfLift reduces end-to-end latency by 41.5% and 44.1%, respectively. Combined with timestep distillation, it delivers overall speedups of 29.61x and 19.21x over the corresponding 50-step models while preserving competitive generation quality, establishing a stronger speed-quality frontier for few-step diffusion.
Chinese Translation
少步扩散模型显著压缩了时间计算，使得每次模型评估的空间成本成为推理延迟的一个日益主导的来源。渐进分辨率推理通过在低分辨率下进行早期去噪，并将高分辨率计算保留用于精细化，从而降低了这一成本。然而，现有方法通常直接提升中间潜变量，并依赖后续步骤来吸收引起的分布不匹配。在少步情况下，有限的恢复预算使这些错误成为可见的伪影，限制了过渡发生的时间以及因此的执行效率。我们提出了SelfLift，一种自我恢复的渐进分辨率框架，它从生成模型本身推导出过渡修复信号和轨迹对齐的监督。SelfLift-zero提出了一种无训练的伪影感知一致性提升，利用直接潜变量提升和像素-VAE重新编码之间的不一致性作为局部伪影风险信号和模型本身的修正方向。它能够在没有外部超分辨率、额外去噪器评估或采样计划修改的情况下实现可靠的晚期过渡。在这一稳健过渡的基础上，SelfLift-rich在学生访问状态上执行策略内自我恢复，从内部自教师转移密集的高分辨率指导，同时与改变的渐进分辨率动态保持一致。在FLUX.2-Klein和Z-Image-Turbo上，SelfLift分别减少了端到端延迟41.5%和44.1%。结合时间步蒸馏，它在保持竞争生成质量的同时，相较于对应的50步模型提供了29.61倍和19.21倍的整体加速，建立了少步扩散的更强速度-质量前沿。
cs.CV / 37 / 2609.02039
Test-Time Logit Prompting for Source-Free Missing Modality Adaptation
测试时逻辑提示用于无源缺失模态适应
Abstract
Vision-language models (VLMs) have achieved remarkable performance by leveraging complementary information from large-scale image-text pairs. However, missing-modality inputs are commonly encountered during real-world deployment, often leading to significant performance degradation. Existing methods primarily enhance model robustness by learning modality compensation strategies from source training data. However, their reliance on source training data makes them difficult to apply when original data are unavailable due to privacy, storage, or accessibility constraints, such as clinical applications and personalized AI services. This raises an important yet underexplored question: can VLMs be efficiently adapted at test time for visual recognition with missing modalities without accessing source training data? To this end, we propose Test-Time Logit Prompting (TLP), a lightweight source-free test-time adaptation framework for visual recognition with missing modalities. To address missing-induced prediction shifts, TLP optimizes logit prompts with uncertainty-aware adjustment and modality-complete consistency regularization, adaptively adjusting prediction confidence while preserving semantic consistency. Extensive experiments across diverse vision-language benchmarks demonstrate that TLP consistently enhances recognition performance under missing-modality scenarios, achieving up to 8\% improvements while requiring only hundreds of tunable parameters and a few test-time optimization steps.
Chinese Translation
视觉语言模型（VLMs）通过利用大规模图像-文本对中的互补信息，取得了显著的性能。然而，在实际应用中，常常会遇到缺失模态输入，这通常会导致性能显著下降。现有方法主要通过从源训练数据中学习模态补偿策略来增强模型的鲁棒性。然而，它们对源训练数据的依赖使得在由于隐私、存储或可访问性限制（如临床应用和个性化人工智能服务）而无法获得原始数据时，应用变得困难。这引发了一个重要但尚未深入探讨的问题：在测试时，是否可以在不访问源训练数据的情况下，针对缺失模态进行视觉识别的有效适应？为此，我们提出了测试时逻辑提示（Test-Time Logit Prompting, TLP），这是一个轻量级的无源测试时适应框架，用于处理缺失模态的视觉识别。为了解决因缺失引起的预测偏移，TLP通过不确定性感知调整和模态完整一致性正则化来优化逻辑提示，自适应地调整预测置信度，同时保持语义一致性。在多种视觉语言基准上的广泛实验表明，TLP在缺失模态场景下始终提升识别性能，最高可实现8%的提升，同时仅需数百个可调参数和少量测试时优化步骤。
cs.CV / 38 / 2609.02063
LaST-SR: Laplace-Inspired Steady-Transient Complex-Frequency Decomposition for Single Image Super-Resolution
LaST-SR：基于拉普拉斯启发的稳态-瞬态复频率分解用于单幅图像超分辨率
Abstract
Single-image super-resolution (SISR) requires global context modeling for structurally consistent reconstruction. Fourier operators are increasingly adopted for global feature modeling. However, their periodic spectral bases constrain the representation of localized aperiodic variations, limiting the recovery of irregular structures and fine details. In dynamical systems, the Laplace neural operator extends Fourier modes to complex frequencies and decomposes the output signal into complementary steady-state and transient responses to jointly model periodic and aperiodic information. We derive, for the first time, an approximate steady-transient decomposition for two-dimensional feature maps, providing an analytical basis for the proposed complex-frequency decomposition. Accordingly, we propose LaST-SR, centered on a Complex-Frequency Decomposition module that couples a global full-spectrum Fourier branch for image-wide dependencies and long-range structural consistency with a window-conditioned local complex-frequency branch for localized, content-dependent aperiodic variations. To fuse the resulting features, we further design a Steady-Transient Collaborative Aggregation module for cross-branch interaction and joint aggregation. Experiments on five benchmarks show that LaST-SR achieves the best PSNR/SSIM among the compared methods for $\times2$ and $\times4$ SISR. Ablation studies further validate the effectiveness of the proposed architecture and its key modeling mechanisms.
Chinese Translation
单幅图像超分辨率（SISR）需要全局上下文建模以实现结构一致的重建。傅里叶算子越来越多地被用于全局特征建模。然而，它们的周期性谱基限制了对局部非周期性变化的表征，从而限制了不规则结构和细节的恢复。在动态系统中，拉普拉斯神经算子将傅里叶模式扩展到复频率，并将输出信号分解为互补的稳态和瞬态响应，以共同建模周期性和非周期性信息。我们首次推导出二维特征图的近似稳态-瞬态分解，为所提出的复频率分解提供了分析基础。因此，我们提出了LaST-SR，核心是一个复频率分解模块，该模块结合了一个全局全谱傅里叶分支，用于图像范围内的依赖关系和长距离结构一致性，以及一个窗口条件下的局部复频率分支，用于局部的、内容依赖的非周期性变化。为了融合生成的特征，我们进一步设计了一个稳态-瞬态协同聚合模块，以实现跨分支的交互和联合聚合。在五个基准测试上的实验表明，LaST-SR在$ imes2$和$ imes4$ SISR中实现了与比较方法相比最佳的PSNR/SSIM。消融研究进一步验证了所提出架构及其关键建模机制的有效性。
cs.CV / 39 / 2609.02075
DPA: Decoupling Product-Agnostic Anomaly Representations for Zero-shot Anomaly Generation
DPA：解耦产品无关异常表示的零-shot异常生成
Abstract
Industrial anomaly detection benefits from anomaly samples, yet newly deployed products typically provide only normal images, making anomaly samples difficult to collect. Zero-shot anomaly generation offers a promising solution which avoids collection of target-product anomalies. However, existing methods mainly rely on texture images or text descriptions as anomaly sources, which often produce unrealistic anomalies. Observing that similar anomalies can recur across different products, we propose anomaly transfer-based zero-shot generation, which reuses real anomalies from existing source products, making target-product anomalies no longer necessary to generate realistic anomalious samples for unseen target products. Since not every anomaly type suits the target product, an anomaly type filtering mechanism first selects plausible source types. To transfer selected anomaly, we propose DPA, a diffusion-based framework that decouples product-agnostic anomaly representations. Instead of directly extracting anomaly representations, DPA learns product-irrelevant anomaly embeddings through training with the mismatched data pair, enabling transferable anomaly concept learning across products. Furthermore, we design an adaptive mask-guided pipeline that leverages adaptive masks to control the positional and geometric plausibility of generated anomalies during generation. A training-free anomaly labeling module is further introduced to produce pixel-level annotations aligned with generated anomalies. Extensive experiments on MVTec-AD, VisA, and a dedicated anomaly-transfer benchmark demonstrate that the proposed setting and DPA generate more realistic anomalies and significantly improve downstream anomaly detection performance under both zero-shot and few-shot settings. Source code and models will be released.
Chinese Translation
工业异常检测受益于异常样本，但新部署的产品通常仅提供正常图像，使得收集异常样本变得困难。零-shot异常生成提供了一种有前景的解决方案，避免了目标产品异常的收集。然而，现有方法主要依赖纹理图像或文本描述作为异常来源，这往往会产生不现实的异常。我们观察到相似的异常可以在不同产品中重复出现，因此提出了一种基于异常转移的零-shot生成方法，该方法重用来自现有源产品的真实异常，使得生成未见目标产品的真实异常样本不再需要目标产品的异常。由于并非每种异常类型都适合目标产品，异常类型过滤机制首先选择合适的源类型。为了转移所选异常，我们提出了DPA，一个基于扩散的框架，解耦产品无关的异常表示。DPA并不是直接提取异常表示，而是通过与不匹配的数据对进行训练，学习与产品无关的异常嵌入，从而实现跨产品的可转移异常概念学习。此外，我们设计了一个自适应掩码引导的流程，利用自适应掩码在生成过程中控制生成异常的位置信息和几何合理性。进一步引入了一个无训练的异常标注模块，以生成与生成异常对齐的像素级注释。在MVTec-AD、VisA和一个专门的异常转移基准上的大量实验表明，所提出的设置和DPA生成了更真实的异常，并在零-shot和少-shot设置下显著提高了下游异常检测性能。源代码和模型将会发布。
cs.CV / 40 / 2609.02077
KSG-Net: Key-Sparse and Global-Context Learning for Maritime 3D Ship Detection
KSG-Net：用于海洋三维船舶检测的关键稀疏与全局上下文学习
Abstract
Accurate 3D ship detection in maritime environments is critical for autonomous navigation, yet remains challenging due to large-scale vessel variations, sparse point clouds of small vessels, and severe sea-clutter interference. Existing methods, primarily based on 2D features or dense representations, struggle to balance detection accuracy and computational efficiency, while sparse 3D detectors designed for road scenes generalize poorly to maritime scenarios. This paper focuses on two key challenges in maritime LiDAR perception: weak feature representation for small and sparse vessels, and insufficient global structural modeling for large vessels due to the limited receptive field of local sparse convolutions. To address these issues, we propose KSG-Net, a Key-Sparse and Global-Context learning network for maritime 3D ship detection. The core idea is to jointly enhance local discriminative features and global structural awareness within a unified fully sparse detection framework. Specifically, a Key Sparse Multi-scale Aggregation (KSMA) module is designed to enhance the representation of small and sparse vessels by selecting informative key voxels and aggregating cross-scale neighborhood features. Furthermore, a Global Context Aggregation (GCA) module is introduced to capture long-range geometric dependencies through scene-level context modeling with gated residual interactions, thereby improving the representation of large vessels. Extensive experiments on the Thames River vessel dataset and simulated datasets demonstrate that KSG-Net consistently outperforms existing methods in multi-scale vessel detection and exhibits strong robustness in complex maritime environments.
Chinese Translation
在海洋环境中，准确的三维船舶检测对自主导航至关重要，但由于船舶的规模变化、较小船舶的稀疏点云以及严重的海洋杂波干扰，这一任务仍然具有挑战性。现有方法主要基于二维特征或密集表示，难以在检测精度和计算效率之间取得平衡，而为公路场景设计的稀疏三维检测器在海洋场景中的泛化能力较差。本文聚焦于海洋激光雷达感知中的两个关键挑战：对于小型稀疏船舶的特征表示不足，以及由于局部稀疏卷积的感受野有限导致的大型船舶的全局结构建模不足。为了解决这些问题，我们提出了KSG-Net，一种用于海洋三维船舶检测的关键稀疏与全局上下文学习网络。其核心思想是在统一的全稀疏检测框架内共同增强局部判别特征和全局结构意识。具体而言，我们设计了一个关键稀疏多尺度聚合（KSMA）模块，通过选择信息量丰富的关键体素并聚合跨尺度邻域特征来增强小型稀疏船舶的表示。此外，引入了一个全局上下文聚合（GCA）模块，通过带门控残差交互的场景级上下文建模来捕捉长距离几何依赖，从而改善大型船舶的表示。在泰晤士河船舶数据集和模拟数据集上的大量实验表明，KSG-Net在多尺度船舶检测中始终优于现有方法，并在复杂的海洋环境中展现出强大的鲁棒性。
cs.CV / 41 / 2609.02088
Rendering-in-the-Loop: An Execution-Driven Agent for Interactive Web Development
循环渲染：一种用于交互式网页开发的执行驱动代理
Abstract
Multimodal large language models have achieved remarkable progress in front-end web development, generating interactive webpages from multimodal references such as screenshots and interaction videos. However, existing work largely emphasizes visual metrics such as aesthetics and layout similarity, while overlooking the more critical validation of interactive functionality. We present RILA, an execution-driven agent that puts browser rendering in the loop, iteratively editing generated code from runtime interaction feedback. RILA introduces an Action Interaction Verification (AIV) module that replays the reference interaction trajectory on the generated webpage to collect grounded execution-aware observations, and an Execution-aware Rendering Score (ERS) that jointly measures interaction correctness and visual fidelity to guide iterative optimization. We further build an execution-verified data synthesis pipeline that produces diverse, high-quality training data, offering gains complementary to inference-time optimization. On IWR-Bench, RILA consistently improves both interaction and visual fidelity across foundation models. Notably, with our training pipeline, RILA lifts the compact Qwen3.5-9B backbone from 40.40% to 57.52%, surpassing far larger one-shot generators, including the 1T-parameter Kimi-K2.6 (55.61%) and the proprietary GPT-5.5 (55.74%).
Chinese Translation
多模态大型语言模型在前端网页开发中取得了显著进展，能够从多模态参考（如截图和交互视频）生成交互式网页。然而，现有研究主要强调视觉指标，如美学和布局相似性，而忽视了更为关键的交互功能验证。我们提出了RILA，一种将浏览器渲染纳入循环的执行驱动代理，它通过运行时交互反馈迭代编辑生成的代码。RILA引入了一个动作交互验证（Action Interaction Verification, AIV）模块，该模块在生成的网页上重放参考交互轨迹，以收集基于执行的观察结果，并引入了一个执行感知渲染评分（Execution-aware Rendering Score, ERS），该评分共同衡量交互正确性和视觉保真度，以指导迭代优化。我们进一步构建了一个执行验证的数据合成管道，生成多样化的高质量训练数据，为推理时优化提供互补收益。在IWR-Bench上，RILA在基础模型上持续提升了交互和视觉保真度。值得注意的是，通过我们的训练管道，RILA将紧凑型Qwen3.5-9B主干的性能从40.40%提升至57.52%，超越了包括1T参数的Kimi-K2.6（55.61%）和专有的GPT-5.5（55.74%）等更大规模的一次性生成器。
cs.CV / 42 / 2609.02097
Evidence-Guided Detection, Localization and Explanation for Text-Centric Image Forensics
基于证据的文本中心图像取证检测、定位与解释
Abstract
The rapid progress of AIGC has made text-centric image manipulation increasingly accessible, creating new forensic challenges that require not only authenticity detection but also spatial grounding and evidence-based explanation. This paper presents our solution to the GenText-Forensics Challenge at ACM Multimedia 2026. We propose an evidence-guided detector-localizer-reasoner system, where an image-level detector provides a global authenticity prior, a dedicated localizer extracts tampered regions as spatial grounding evidence, and an MLLM-based reasoner generates structured forensic reports grounded in this expert forensic evidence. These modules are connected through a cascaded evidence flow: the detector gates the subsequent localization and prompting process, the localizer converts tamper responses into grounding boxes, and the reasoner is trained to synthesize the detector decision and localized evidence into the final report. As a key part of our method, we introduce iterative difficulty-aware mining to improve localization quality and apply report-mask consistency post-processing to align report grounding with predicted masks. On the official hidden test set, our system achieves a final score of 0.638 and ranks second in the challenge, validating the effectiveness of the proposed evidence-guided system. The code is available at https://github.com/peifengLiu42/ACMMM26-evidence-guided-detector-localizer-reasoner-system.
Chinese Translation
AIGC的快速发展使得文本中心图像处理变得愈加容易，这带来了新的取证挑战，不仅需要真实性检测，还需要空间定位和基于证据的解释。本文提出了我们在2026年ACM多媒体GenText-Forensics挑战赛中的解决方案。我们提出了一种基于证据的检测-定位-推理系统，其中图像级检测器提供全局真实性先验，专用定位器提取篡改区域作为空间定位证据，基于多模态大语言模型（MLLM）的推理器生成基于这些专家取证证据的结构化取证报告。这些模块通过级联的证据流连接：检测器控制后续的定位和提示过程，定位器将篡改响应转换为定位框，推理器则被训练以综合检测器的决策和定位证据生成最终报告。作为我们方法的关键部分，我们引入了迭代的难度感知挖掘以提高定位质量，并应用报告-掩膜一致性后处理以对齐报告定位与预测掩膜。在官方隐藏测试集上，我们的系统最终得分为0.638，在挑战中排名第二，验证了所提出的基于证据的系统的有效性。代码可在https://github.com/peifengLiu42/ACMMM26-evidence-guided-detector-localizer-reasoner-system获取。
cs.CV / 43 / 2609.02111
Disease Burden over Skin Tone: Decomposing the Dermatology-AI Generalization Gap
疾病负担与肤色的关系：解构皮肤病学人工智能的泛化差距
Abstract
Dermatology artificial intelligence (AI) models are predominantly trained on light-skinned, cancer-focused image collections, yet they are increasingly proposed for deployment in resource-constrained settings where patients differ from training populations along two confounded axes: skin tone and disease distribution. We investigate whether poor generalization is primarily caused by skin-tone underrepresentation or disease-distribution shift. We evaluate a cancer-trained baseline (ResNet-50 fine-tuned on HAM10000 and ISIC 2019), two dermatology foundation models (DermLIP and MONET), and a general-purpose vision model (DINOv3) as frozen feature extractors. Models are evaluated on a tone-stratified disease-matched dataset (Diverse Dermatology Images, DDI) and a disease-shifted tone-diverse dataset (Skin Condition Image Network, SCIN). Our results show that disease-distribution shift contributes more than skin tone in the evaluated settings. The cancer baseline decreases from 0.62 to 0.21 balanced accuracy when transferred to unfamiliar clinical conditions, while the within-disease skin-tone gap is smaller (0.10-0.18) and inconsistent. Label-free representation analysis shows that this failure reflects a representational limitation rather than only missing output labels: cancer-specialized features poorly cluster unfamiliar conditions (kNN purity lift +0.06 over chance), whereas dermatology-pretrained features retain stronger transferable structure (+0.23). Finally, we show that representation quality predicts recoverable performance under lightweight adaptation. Starting from dermatology foundation models, approximately ten labeled examples per clinical category recover most attainable performance. We release the evaluation protocol and code to support reproducible auditing of dermatology AI generalization.
Chinese Translation
皮肤病学人工智能（AI）模型主要在以浅色皮肤为主的癌症图像集合上进行训练，但它们越来越多地被提议用于资源有限的环境中，而这些环境中的患者在肤色和疾病分布两个相互关联的维度上与训练人群存在差异。我们研究了泛化能力差的主要原因是肤色代表性不足还是疾病分布的变化。我们评估了一个以癌症为训练基础的模型（在HAM10000和ISIC 2019上微调的ResNet-50）、两个皮肤病学基础模型（DermLIP和MONET）以及一个通用视觉模型（DINOv3）作为固定特征提取器。模型在一个按肤色分层且与疾病匹配的数据集（多样化皮肤病图像，DDI）和一个疾病分布变化且肤色多样化的数据集（皮肤状况图像网络，SCIN）上进行评估。我们的结果表明，在评估的环境中，疾病分布的变化对模型性能的影响大于肤色。癌症基线模型在转移到不熟悉的临床条件时，平衡准确率从0.62下降至0.21，而同一疾病内的肤色差距较小（0.10-0.18）且不一致。无标签表示分析显示，这一失败反映了表示能力的局限，而不仅仅是缺少输出标签：癌症专用特征在不熟悉的条件下聚类效果较差（kNN纯度提升+0.06），而皮肤病学预训练特征保持了更强的可转移结构（+0.23）。最后，我们展示了表示质量可以预测在轻量适应下可恢复的性能。从皮肤病学基础模型出发，每个临床类别大约需要十个标记示例即可恢复大部分可达到的性能。我们发布了评估协议和代码，以支持对皮肤病学AI泛化能力的可重复审计。
cs.CV / 44 / 2609.02118
Synergistic Information Disentanglement for Omni-modal Slide Representation Learning in Computational Pathology
计算病理学中全模态幻灯片表示学习的协同信息解耦
Abstract
In computational pathology (CPath), developing omni-modal self-supervised learning (SSL) models that integrate histology, genomics, and clinical reports enables transferable representation learning for whole slide images (WSIs). Existing approaches implicitly force heterogeneous modalities into a uniform latent space by contrastive alignment, causing modality collapse where unique, synergistic diagnostic signals (termed as $\mathrm{\Phi}$) are discarded in favor of trivial redundancy. We hypothesize that the strongest task-agnostic SSL training signal stems from distilling the synergistic interactions over merely aligning shared redundancy. To this end, we introduce \textsc{$\mathrm{\Phi}$-Omni}, a synergistic information disentanglement framework grounded in Partial Information Decomposition (PID) theory for slide representation learning. Unlike standard contrastive approaches, \textsc{$\mathrm{\Phi}$-Omni} employs a Synergistic Information Bottleneck (SIB) regulated by the proposed $\mathrm{\Phi}\text{ID}$ objective, which explicitly suppresses marginal redundancy while maximizing irreducible synergy, thereby distilling high-order cross-modal interactions. Following pretraining on breast ($n$=1031) and lung ($n$=919) cohorts, \textsc{$\mathrm{\Phi}$-Omni} demonstrates superior few-shot performance across five independent external datasets spanning eight tasks compared to supervised and SSL baselines. Source code is available here.
Chinese Translation
在计算病理学（CPath）中，开发整合组织学、基因组学和临床报告的全模态自监督学习（SSL）模型，使得对全幻灯片图像（WSIs）的可迁移表示学习成为可能。现有方法通过对比对齐隐式地将异构模态强制映射到统一的潜在空间，从而导致模态崩溃，独特的协同诊断信号（称为 $ ext{ extit{ extbf{Φ}}}$）被舍弃，取而代之的是琐碎的冗余。我们假设，最强的任务无关的 SSL 训练信号源于提炼协同交互，而不仅仅是对齐共享冗余。为此，我们提出了 extsc{$ ext{ extit{ extbf{Φ}}}$-Omni}，这是一个基于部分信息分解（PID）理论的协同信息解耦框架，用于幻灯片表示学习。与标准对比方法不同， extsc{$ ext{ extit{ extbf{Φ}}}$-Omni} 采用了由提出的 $ ext{ extit{ extbf{Φ}}} ext{ID}$ 目标调节的协同信息瓶颈（SIB），该目标明确抑制边际冗余，同时最大化不可简化的协同作用，从而提炼高阶跨模态交互。在对乳腺（$n$=1031）和肺（$n$=919）队列进行预训练后， extsc{$ ext{ extit{ extbf{Φ}}}$-Omni} 在五个独立的外部数据集上展示了优于监督和 SSL 基线的少样本性能。源代码可在此处获取。
cs.CV / 45 / 2609.02159
World-Coherent Decoding: Self-Verifying Test-Time Planning for World Action Models
世界一致性解码：自验证的测试时间规划用于世界行动模型
Abstract
World Action Models (WAMs) aim to control robots by stochastically generating visual futures and then decoding actions, but empirical observations indicate that the results can strongly depend on which future is selected. We propose World-Coherent-Decoding (WCD), a self-verifying test-time planning framework that treats WAM rollouts as falsifiable future--action hypotheses. At each decision step, WCD samples multiple candidates from a frozen WAM and ranks them using internal generative signals: flow-based video surprisal for visual plausibility and action path effort for action-generation stability. After execution, the realized observation audits the selected imagination, yielding an imagination--reality mismatch that trains a lightweight online predictor for future candidate selection. Thus, WCD converts delayed self-verification into pre-execution reliability estimation without updating the backbone model. On RoboTwin 2.0, WCD improves Hard success under limited randomized-scene supervision from $55.80\%$ to $60.90\%$, with a $+16.43$ gains on Horizon-3 tasks, and shows qualitative robustness on real Franka visual-shift tests. These results highlight a simple principle: test-time scaling for WAMs depends less on sampling more futures than on selecting reliable ones.
Chinese Translation
世界行动模型（WAMs）旨在通过随机生成视觉未来并解码动作来控制机器人，但实证观察表明，结果可能强烈依赖于选择的未来。我们提出了世界一致性解码（World-Coherent-Decoding，WCD），这是一种自验证的测试时间规划框架，将WAM的展开视为可证伪的未来-动作假设。在每个决策步骤中，WCD从一个冻结的WAM中采样多个候选项，并使用内部生成信号对其进行排名：基于流的视觉惊讶度用于视觉合理性，动作路径努力用于动作生成的稳定性。执行后，实际观察审核所选的想象，产生想象-现实不匹配，从而训练一个轻量级的在线预测器用于未来候选项的选择。因此，WCD将延迟自验证转化为执行前的可靠性估计，而无需更新主干模型。在RoboTwin 2.0上，WCD在有限随机场景监督下将困难成功率从55.80%提高到60.90%，在Horizon-3任务上获得了+16.43的增益，并在真实的Franka视觉偏移测试中显示出定性鲁棒性。这些结果突显了一个简单的原则：WAM的测试时间扩展不太依赖于采样更多的未来，而是依赖于选择可靠的未来。
cs.CV / 46 / 2609.02171
Progressive Pseudo-Label Optimization for Point-Supervised Change Detection
渐进式伪标签优化用于点监督变化检测
Abstract
Point-supervised change detection (PS-CD) aims to identify pixel-level changes between bi-temporal images using only sparsely annotated points. Although point annotations substantially reduce labeling costs, their limited spatial coverage often results in incomplete and noisy pseudo-labels. To address this issue, we propose a two-stage framework that introduces SAM2 priors into PS-CD and progressively adapts them to the target task. In Stage I, SAM2 generates object-aware candidate masks from point annotations on the bi-temporal images, and a bi-temporal mask selection strategy is designed to convert generic segmentation responses into more reliable change pseudo-labels. Subsequently, a lightweight CNN refinement module with an uncertainty-aware loss is employed to improve boundary quality and local structural consistency. In Stage II, we construct a teacher-student self-training framework in which the teacher is updated by exponential moving average and periodically refreshes the pseudo-labels. This design establishes a closed-loop optimization process that alternates between pseudo-label refinement and model re-optimization. Experiments on three benchmark datasets, including WHU-CD, LEVIR-CD, and SYSU-CD, demonstrate that the proposed method outperforms previous weakly supervised approaches on most benchmarks and remains competitive with several fully supervised methods.
Chinese Translation
点监督变化检测（PS-CD）旨在仅使用稀疏标注的点识别双时相图像之间的像素级变化。尽管点标注显著降低了标注成本，但其有限的空间覆盖范围往往导致伪标签的不完整和噪声。为了解决这一问题，我们提出了一个两阶段框架，将SAM2先验引入PS-CD，并逐步将其适应于目标任务。在第一阶段，SAM2根据双时相图像上的点标注生成对象感知候选掩膜，并设计了一种双时相掩膜选择策略，将通用分割响应转换为更可靠的变化伪标签。随后，采用一个轻量级的卷积神经网络（CNN）精炼模块，结合不确定性感知损失，以提高边界质量和局部结构一致性。在第二阶段，我们构建了一个教师-学生自训练框架，其中教师通过指数移动平均进行更新，并定期刷新伪标签。该设计建立了一个闭环优化过程，在伪标签精炼和模型再优化之间交替进行。在WHU-CD、LEVIR-CD和SYSU-CD等三个基准数据集上的实验表明，所提出的方法在大多数基准上优于之前的弱监督方法，并与几种完全监督的方法保持竞争力。
cs.CV / 47 / 2609.02184
CC-4DGS: Computational Deformation and Point-Cloud Compression for Storage-Efficient Dynamic Gaussian Splatting
CC-4DGS：用于存储高效动态高斯溅射的计算变形与点云压缩
Abstract
Dynamic four-dimensional (4D) Gaussian Splatting has emerged as a powerful explicit representation for high-quality view synthesis, yet existing methods still require tens to hundreds of megabytes per scene due to their heavy reliance on large multi-resolution hash tables and high-dimensional Gaussian attributes. This paper presents CC-4DGS, a storage-efficient and scalable framework that rethinks both deformation modeling and canonical attribute storage. First, we introduce a computational deformation field (CDF) that replaces large multi-resolution learnable hash tables with deterministic dense hash encoding and compact neural decoders, enabling on-the-fly synthesis of deformation features while reducing deformation storage to only 1--3 MB per scene. Second, we propose a compression of canonical point-cloud attributes (CCA) pipeline that compresses high-dimensional spherical harmonic appearance terms and auxiliary Gaussian attributes via conditional autoencoding, selective quantization, and residual codebooks, achieving 3--5$\times$ point-cloud reduction with negligible quality loss. Together, these components yield a unified representation that preserves real-time rendering performance while reducing total storage to 20--30 MB. Extensive experiments across the N3DV and Technicolor Light Field datasets demonstrate that CC-4DGS achieves reconstruction accuracy comparable to state-of-the-art methods such as Swift4D, while offering significantly improved storage efficiency and favorable runtime-memory trade-offs.
Chinese Translation
动态四维（4D）高斯溅射已成为高质量视图合成的强大显式表示，然而现有方法仍需每个场景占用数十到数百兆字节的存储空间，原因在于它们对大型多分辨率哈希表和高维高斯属性的高度依赖。本文提出了CC-4DGS，一个存储高效且可扩展的框架，重新思考了变形建模和典型属性存储。首先，我们引入了一种计算变形场（CDF），它用确定性稠密哈希编码和紧凑的神经解码器替代了大型多分辨率可学习哈希表，从而实现了变形特征的即时合成，同时将每个场景的变形存储减少到仅1-3 MB。其次，我们提出了一种典型点云属性压缩（CCA）管道，通过条件自编码、选择性量化和残差代码本压缩高维球谐外观项和辅助高斯属性，实现了3-5倍的点云减少且几乎没有质量损失。结合这些组件，形成了一个统一的表示，保持了实时渲染性能，同时将总存储减少到20-30 MB。在N3DV和Technicolor光场数据集上的广泛实验表明，CC-4DGS在重建精度上可与Swift4D等最先进的方法相媲美，同时显著提高了存储效率和运行时内存的权衡。
cs.CV / 48 / 2609.02187
Lightweight Adaptation of General-Purpose VLMs for Multispectral and SAR Image Understanding
通用视觉语言模型的轻量级适应用于多光谱和合成孔径雷达图像理解
Abstract
General-purpose vision-language models (VLMs) now support strong visual recognition, instruction following, and generation. However, most pretrained visual encoders are built around three-channel natural images and do not directly accommodate observations such as native multispectral measurements or synthetic aperture radar (SAR). Adapting VLMs to these sensors typically requires dedicated encoders and domain pretraining, slowing the reuse of stronger general-purpose checkpoints. We show that the multi-image interface of general-purpose VLMs offers a lightweight alternative. Our protocol renders each observation as five optical views and one SAR view, names them in the prompt, and adapts the language network and selected visual transformer blocks with LoRA. This exposes band composites, spectral indices, and radar backscatter through an existing visual interface. For land-cover recognition, structured supervision couples predicted classes with sensor evidence. We further construct preference pairs in which a true label is omitted while its supporting evidence is retained, encouraging complete predictions that remain consistent with the observations. On a balanced six-class land-cover benchmark derived from BigEarthNet-v2, the adapted Qwen3-VL reaches 0.8275 micro F1. The same input and adaptation protocol improves all four tested VLM architectures and transfers to Sen1Floods11 flood verification and BigEarthNet.txt captioning. Image removal and mismatch controls show that the adapted models use the supplied sensor observations. Together, these results demonstrate that VLMs can be repurposed for multispectral and SAR tasks through rendered inputs and compact LoRA adaptation, without training a new foundation model.
Chinese Translation
通用视觉语言模型（VLMs）现已支持强大的视觉识别、指令跟随和生成。然而，大多数预训练的视觉编码器是围绕三通道自然图像构建的，并不直接适应诸如原生多光谱测量或合成孔径雷达（SAR）等观测。将VLMs适应于这些传感器通常需要专用编码器和领域预训练，这减缓了更强通用检查点的重用。我们展示了通用VLMs的多图像接口提供了一种轻量级的替代方案。我们的协议将每个观测渲染为五个光学视图和一个SAR视图，在提示中命名它们，并使用LoRA适应语言网络和选定的视觉变换器模块。这通过现有的视觉接口暴露了波段复合、光谱指数和雷达后向散射。对于土地覆盖识别，结构化监督将预测类别与传感器证据结合在一起。我们进一步构建了偏好对，其中真实标签被省略，而其支持证据被保留，鼓励与观测一致的完整预测。在从BigEarthNet-v2派生的平衡六类土地覆盖基准上，适应后的Qwen3-VL达到了0.8275的微F1。相同的输入和适应协议改善了所有四种测试的VLM架构，并转移到Sen1Floods11洪水验证和BigEarthNet.txt字幕生成。图像移除和不匹配控制显示适应后的模型使用了提供的传感器观测。综合这些结果表明，VLMs可以通过渲染输入和紧凑的LoRA适应重新用于多光谱和SAR任务，而无需训练新的基础模型。
cs.CV / 49 / 2609.02204
TAME: Temporal-Aware Mixture-of-Experts for Text-Video Retrieval
TAME：用于文本-视频检索的时间感知专家混合模型
Abstract
Text-Video Retrieval (TVR) retrieves videos that match a natural-language query, but extending image-text models such as CLIP to videos is fundamentally limited by the lack of temporal modeling. Videos exhibit frame-wise heterogeneity in appearance and motion, and compressing all frames into a single representation often obscures temporal structure and semantic transitions. To address this, we propose Temporal-Aware Mixture-of-Experts for Text-Video Retrieval (TAME), a CLIP-based framework that jointly models frame-level structure and temporal relations. First, we integrate sparse Mixture-of-Experts (MoE) layers into both CLIP encoders and apply frame-consistent routing on the vision branch so that experts specialize according to frame-level visual patterns while preserving the original vision-language alignment. Second, we introduce Frame-Temporal (FT) tokens that aggregate global cross-frame information and feed it back to each frame, enabling the visual encoder to capture long-range temporal dependencies without harming local details. Third, we design a Cross-Temporal Interaction and Aggregation (CTIA) module that refines frame-wise sentence-video similarities through staged temporal filtering and fusion. Experiments on standard TVR benchmarks show that TAME consistently improves over CLIP-based baselines. On MSR-VTT, it improves R@1 by 4.0 over CLIP4Clip, and also achieves consistent gains on DiDeMo, MSVD, LSMDC, and ActivityNet. The code is available at https://github.com/sejong-rcv/TAME.
Chinese Translation
文本-视频检索（TVR）旨在检索与自然语言查询匹配的视频，但将图像-文本模型（如CLIP）扩展到视频的根本限制在于缺乏时间建模。视频在外观和运动上表现出逐帧的异质性，而将所有帧压缩为单一表示往往会模糊时间结构和语义转换。为了解决这个问题，我们提出了用于文本-视频检索的时间感知专家混合模型（TAME），这是一个基于CLIP的框架，能够联合建模帧级结构和时间关系。首先，我们将稀疏的专家混合（MoE）层集成到CLIP编码器中，并在视觉分支上应用帧一致的路由，使专家根据帧级视觉模式进行专业化，同时保持原始的视觉-语言对齐。其次，我们引入帧-时间（FT）标记，聚合全局跨帧信息并反馈给每一帧，使视觉编码器能够捕捉长程时间依赖关系而不损害局部细节。第三，我们设计了一个跨时间交互与聚合（CTIA）模块，通过阶段性时间过滤和融合来细化逐帧的句子-视频相似度。在标准TVR基准上的实验表明，TAME在CLIP基础上持续改进。在MSR-VTT上，它比CLIP4Clip提高了4.0的R@1，并在DiDeMo、MSVD、LSMDC和ActivityNet上也取得了一致的提升。代码可在https://github.com/sejong-rcv/TAME获取。
cs.CV / 50 / 2609.02207
LeakageBench: Document-Level Leakage Risk for Redacting Personally Identifiable Information in Document Images
LeakageBench：文档级泄露风险评估用于在文档图像中编辑个人可识别信息
Abstract
Real-world personally identifiable information (PII) redaction often operates on document images---scans, screenshots, and PDF renderings---where OCR errors, layout structure, and visual noise determine whether sensitive information is actually removed. Existing PII benchmarks are mostly text-centric and do not measure document-level redaction risk: a page remains unsafe if even one identifier is missed. We introduce LeakageBench, a challenge set of 500 document images with 11,954 GDPR-aligned PII annotations spanning direct identifiers, linkage keys, and contextual re-identification surfaces. We evaluate generic OCR pipelines, commercial and task-adapted OCR-dependent detectors, and OCR-free vision-language models using entity-level F1, group-wise leakage, and document-level leakage metrics. Code Interpreter raises GPT-5.5 localization F1 from 0.090 to 0.249, but critical page-level leakage remains 0.968. These results show that stronger detection and tool assistance improve localization without making most pages safe for release. LeakageBench provides a diagnostic benchmark for high-recall, spatially grounded PII redaction in document images.
Chinese Translation
现实世界中的个人可识别信息（PII）编辑通常在文档图像上进行——扫描件、截图和PDF渲染——其中OCR错误、布局结构和视觉噪声决定了敏感信息是否被实际移除。现有的PII基准大多以文本为中心，未能衡量文档级的编辑风险：如果遗漏了一个标识符，则该页面仍然不安全。我们引入了LeakageBench，这是一个包含500个文档图像的挑战集，具有11,954个符合GDPR的PII注释，涵盖直接标识符、链接键和上下文重识别表面。我们评估了通用OCR管道、商业和任务适应的OCR依赖检测器，以及使用实体级F1、组级泄露和文档级泄露指标的无OCR视觉语言模型。代码解释器将GPT-5.5的定位F1从0.090提高到0.249，但关键的页面级泄露仍为0.968。这些结果表明，增强的检测和工具辅助提高了定位能力，但并未使大多数页面在发布时安全。LeakageBench为文档图像中的高召回率、空间基础的PII编辑提供了诊断基准。
cs.CV / 51 / 2609.02210
Asymmetric Paired-Annotation Learning for Multi-Structure ULF Pediatric Brain MRI Segmentation
用于多结构超低场儿童脑MRI分割的非对称配对注释学习
Abstract
Portable ultra-low-field (ULF) MRI can expand access to pediatric neuroimaging, but segmentation at 0.064 T remains challenging because anatomical boundaries are weakly delineated, small structures may be only partially visible, and high-field references can be locally misregistered. The LISA 2026 Challenge provides two non-equivalent annotations reflecting different sources of anatomical evidence: a highfield-derived (HF) mask defining the scored target and a low-field-edited (LF) mask aligned with visible ULF anatomy. In this challenge report, we describe AURA, an nnU-Net-based asymmetric supervision strategy that treats these annotations as distinct observations rather than interchangeable ground truths. AURA anchors training to the HF mask and incorporates the LF mask through a bounded reliability gate based on label disagreement, boundaries, predictive uncertainty, class reliability, and training stage. On a 16-case development split, the HF-supervised baseline, AURA, and their ensemble achieved Dice scores of 0.7984, 0.7950, and 0.7988, respectively, while the ensemble achieved an HD95 of 1.8892 and an ASSD of 0.7855. These results provide a preliminary evaluation of AURA within the LISA 2026 Challenge and motivate further assessment on the hidden test set and external ULF cohorts. Our code and pretrained models are available at https://github.com/minhdang050806/ A-nnU-Net-based-asymmetric-supervision-strategy.
Chinese Translation
便携式超低场（ULF）MRI可以扩大对儿童神经影像学的获取，但在0.064 T下的分割仍然具有挑战性，因为解剖边界的划分较弱，小结构可能仅部分可见，并且高场参考可能存在局部错配。LISA 2026挑战提供了两种不等价的注释，反映了不同来源的解剖证据：一种高场衍生（HF）掩模定义了评分目标，另一种低场编辑（LF）掩模与可见的ULF解剖结构对齐。在本挑战报告中，我们描述了AURA，这是一种基于nnU-Net的非对称监督策略，将这些注释视为不同的观测而非可互换的真实标签。AURA将训练锚定于HF掩模，并通过基于标签不一致性、边界、预测不确定性、类别可靠性和训练阶段的有界可靠性门 Incorporate LF掩模。在16个案例的开发分割中，HF监督基线、AURA及其集成分别获得了0.7984、0.7950和0.7988的Dice分数，而集成的HD95为1.8892，ASSD为0.7855。这些结果为AURA在LISA 2026挑战中的初步评估提供了依据，并激励对隐藏测试集和外部ULF队列进行进一步评估。我们的代码和预训练模型可在https://github.com/minhdang050806/A-nnU-Net-based-asymmetric-supervision-strategy获取。
cs.CV / 52 / 2609.02212
FuDU: A Fuzzy Dual-dimensional Uncertainty Framework for Streaming Active Learning in Industrial Defect Detection
FuDU：一种用于工业缺陷检测的流式主动学习模糊双维不确定性框架
Abstract
Ensuring the reliability of deep learning models in real-time industrial defect detection is critical for high-stakes quality inspection. To mine uncertain samples within continuous industrial media streams, thereby enhancing the reliability of the detection system, this paper proposes a streaming active learning method based on the Fuzzy Dual-dimensional Uncertainty (FuDU) framework. Specifically, we first design a Prototype-based Global Uncertainty Quantification (PGUQ) module on the backbone to evaluate image-level uncertainty via normal/defective feature prototypes. A Dual-entropy defect Uncertainty Evaluator (DeUE) is then integrated into the detection head to quantify box-level uncertainty. Finally, by modeling uncertainty as systematic error, we propose a fuzzy dual-dimensional uncertainty-aware strategy that leverages fuzzy inference to fuse dual-dimensional uncertainties, enabling expert knowledge-driven adaptive sampling decisions. Comprehensive experiments demonstrate that FuDU is efficient and flexible, making it well-suited for challenging industrial inspection tasks such as the detection of nuclear fuel rod defects. Our code is publicly available at: https://github.com/wangzhaoyang-508/FuDU.
Chinese Translation
确保深度学习模型在实时工业缺陷检测中的可靠性对于高风险的质量检查至关重要。为了在连续的工业媒体流中挖掘不确定样本，从而增强检测系统的可靠性，本文提出了一种基于模糊双维不确定性（FuDU）框架的流式主动学习方法。具体而言，我们首先在主干网络上设计了一个基于原型的全局不确定性量化（PGUQ）模块，通过正常/缺陷特征原型评估图像级不确定性。然后，将双熵缺陷不确定性评估器（DeUE）集成到检测头中，以量化框级不确定性。最后，通过将不确定性建模为系统误差，我们提出了一种模糊双维不确定性感知策略，该策略利用模糊推理融合双维不确定性，从而实现基于专家知识的自适应采样决策。全面的实验表明，FuDU高效且灵活，适用于挑战性的工业检测任务，如核燃料棒缺陷检测。我们的代码已公开发布，链接为：https://github.com/wangzhaoyang-508/FuDU。
cs.CV / 53 / 2609.02224
Signal or Noise? Auditing Rotation-Induced Saliency Drift in Medical and Aerial Imaging
信号还是噪声？审计旋转引起的医学和航空影像中的显著性漂移
Abstract
Post-hoc saliency maps such as Grad-CAM are increasingly used to audit why a deployed vision model made a decision, yet the heatmap drifts when the input is rotated, even when the prediction is unchanged. In domains with no canonical orientation, such as histopathology and aerial imagery, this undermines using saliency as evidence. We ask whether that drift is faithful signal or noise introduced by the CAM operator, and answer it by measuring equivariance at every stage of the operator rather than inferring it from the network's output. The instability is not where one would guess: the channel weights are the most rotation-stable stage, and on ResNet-50 exactly stable, because a GAP+linear head makes the class gradient field spatially constant. What moves is the spatial activation tensor, and the classifier's own pooling discards that movement. A causal test confirms the consequence: occluding the pixels whose saliency drifts costs the model less than occluding random pixels, at either orientation. The drift is carried by degrees of freedom the classifier throws away, which is what makes removing it faithful rather than destructive. EquiGrad-CAM is a training-free wrapper that takes T rotated views, inverse-rotates each view's saliency into a common canonical frame, and averages. On the full ImageNet-1K validation set it raises equivariance over single-view Grad-CAM by +36.0% (ResNet-50), +87.5% (VGG-16) and +247% (ViT-B/16); a scale-matched ablation isolates alignment before averaging, not the locus of aggregation, as the driver. It beats rotation-augmented training without retraining, lifts zero-shot CLIP by +145%, and yields rotation-consistent explanations on PatchCamelyon and RESISC45. Its by-product PEUM ranks explanations by how reproducible they are, at no cost beyond the views already taken. Code: https://github.com/Khawaja-Murad/EquiGrad-CAM
Chinese Translation
后验显著性图（如 Grad-CAM）越来越多地用于审计已部署视觉模型做出决策的原因，但当输入被旋转时，即使预测不变，热图也会漂移。在没有规范方向的领域，如组织病理学和航空影像，这削弱了将显著性作为证据的有效性。我们探讨这种漂移是忠实信号还是由 CAM 操作引入的噪声，并通过在操作的每个阶段测量等变性来回答这一问题，而不是从网络的输出推断。这个不稳定性并不是人们所猜测的：通道权重是最稳定的旋转阶段，在 ResNet-50 中完全稳定，因为 GAP+线性头使得类别梯度场在空间上保持恒定。移动的是空间激活张量，而分类器自身的池化则丢弃了这种运动。因果测试证实了这一结果：遮挡那些显著性漂移的像素对模型的影响小于遮挡随机像素，无论方向如何。漂移是由分类器抛弃的自由度所携带的，这使得去除漂移变得忠实而非破坏性。EquiGrad-CAM 是一个无训练的包装器，它接受 T 个旋转视图，将每个视图的显著性逆旋转到一个共同的规范框架中，并进行平均。在完整的 ImageNet-1K 验证集上，它使单视图 Grad-CAM 的等变性提高了 +36.0%（ResNet-50）、+87.5%（VGG-16）和 +247%（ViT-B/16）；一个规模匹配的消融实验将对齐隔离在平均之前，而不是聚合的地点，作为驱动因素。它在不重新训练的情况下超越了旋转增强训练，使零-shot CLIP 提升了 +145%，并在 PatchCamelyon 和 RESISC45 上提供了旋转一致的解释。其副产品 PEUM 根据解释的可重复性对其进行排名，除了已经采取的视图外没有额外成本。代码： https://github.com/Khawaja-Murad/EquiGrad-CAM
cs.CV / 54 / 2609.02233
InfraPatch: Cross-Task Targeted Grayscale Patch Attacks on Infrared-Adapted Vision-Language Models
InfraPatch：针对红外适应视觉-语言模型的跨任务目标灰度补丁攻击
Abstract
Infrared vision-language models (IR-VLMs) have emerged as a promising paradigm for multimodal perception under low-visibility conditions, yet their robustness to targeted adversarial attacks remains poorly understood. Existing adversarial patch methods mainly study RGB-based models or a single downstream task and do not characterize whether localized perturbations can induce an intended semantic target in IR-VLMs. We propose InfraPatch, a white-box, per-instance framework for targeted digital grayscale patch attacks against IR-VLMs. InfraPatch optimizes a compact single-channel patch within an approximately 5% local-area budget, combines proxy-guided placement with task-adaptive semantic objectives, and induces target behaviors in image classification, image captioning, and binary visual question answering. We evaluate ten infrared-adapted model variants on 300 synthetic infrared-style images generated by applying DiffV2IR to a fixed 30-category COCO subset, using clean-conditioned targeted success criteria. InfraPatch achieves targeted attack success rates from 86.00% to 100% across the ten variants. On CLIP and BLIP-2, proxy location search improves success by 6.67 and 10.33 percentage points over optimized random placement, respectively; LLaVA-1.5 remains saturated near 100% under both settings. Patch-area and objective ablations further expose substantial differences in vulnerability across architectures and task formats. These results show that small grayscale patches can inject chosen target semantics across IR-VLM families under a controlled digital threat model, motivating stronger robustness evaluation for infrared multimodal systems.
Chinese Translation
红外视觉-语言模型（IR-VLMs）作为一种在低可见度条件下进行多模态感知的有前景的范式，然而它们对目标对抗攻击的鲁棒性仍然了解不足。现有的对抗补丁方法主要研究基于RGB的模型或单一下游任务，并未表征局部扰动是否能够在IR-VLMs中引发预期的语义目标。我们提出了InfraPatch，这是一个针对IR-VLMs的白盒、每实例框架，用于目标数字灰度补丁攻击。InfraPatch在大约5%的局部区域预算内优化一个紧凑的单通道补丁，结合代理引导的放置与任务自适应的语义目标，并在图像分类、图像描述和二元视觉问答中引发目标行为。我们在300张通过将DiffV2IR应用于固定的30类COCO子集生成的合成红外风格图像上评估了十种红外适应模型变体，使用干净条件下的目标成功标准。InfraPatch在这十种变体中实现了86.00%到100%的目标攻击成功率。在CLIP和BLIP-2上，代理位置搜索分别比优化的随机放置提高了6.67和10.33个百分点；而LLaVA-1.5在这两种设置下仍然保持在接近100%的饱和状态。补丁面积和目标的消融实验进一步揭示了不同架构和任务格式之间的脆弱性存在显著差异。这些结果表明，小型灰度补丁能够在受控的数字威胁模型下在IR-VLM家族中注入选定的目标语义，促使对红外多模态系统进行更强的鲁棒性评估。
cs.CV / 55 / 2609.02247
SAUF-Net: Structure--Appearance Representation Learning with Uncertainty Feedback for Semi-Supervised Medical Image Segmentation
SAUF-Net：带有不确定性反馈的结构-外观表征学习用于半监督医学图像分割
Abstract
Semi-supervised learning has shown great potential for reducing annotation costs in medical image segmentation. However, most existing methods mainly exploit unlabeled data through prediction-level consistency, while the reliability of internal feature representations is often overlooked. In medical images, target-related structural cues are easily entangled with unstable appearance variations, which may lead to unreliable pseudo labels and error accumulation during training. To address these issues, we propose SAUF-Net, a Structure--Appearance Representation Learning with Uncertainty Feedback Network for semi-supervised medical image segmentation. SAUF-Net uses the Structure--Appearance Decomposition Module (SADM) to separate bottleneck features into structural and appearance representations. The Disentangled Guidance Module (DGM) injects these representations into the decoding process to enhance structure-aware segmentation. Meanwhile, the Auxiliary Decoder produces branch-specific predictions for reliability estimation and a fused prediction for appearance-swapped consistency. Furthermore, we introduce an Appearance-Swapped Consistency branch to encourage structural representations to remain stable under appearance variations. We also introduce a reliability-map-guided dual-head discriminator with a Validity Head and an Uncertainty Head to provide feature-level uncertainty feedback. Extensive experiments on ISIC-2016 and Kvasir-SEG demonstrate that SAUF-Net outperforms state-of-the-art semi-supervised methods, especially under low-label settings.
Chinese Translation
半监督学习在降低医学图像分割中的标注成本方面展现出巨大潜力。然而，现有大多数方法主要通过预测级一致性来利用未标记数据，而内部特征表征的可靠性往往被忽视。在医学图像中，目标相关的结构线索容易与不稳定的外观变化纠缠在一起，这可能导致不可靠的伪标签和训练过程中的错误积累。为了解决这些问题，我们提出了SAUF-Net，一种带有不确定性反馈的结构-外观表征学习网络，用于半监督医学图像分割。SAUF-Net使用结构-外观分解模块（Structure--Appearance Decomposition Module, SADM）将瓶颈特征分离为结构和外观表征。解耦引导模块（Disentangled Guidance Module, DGM）将这些表征注入解码过程，以增强结构感知分割。同时，辅助解码器生成特定于分支的预测以进行可靠性评估，并生成融合预测以实现外观交换一致性。此外，我们引入了一个外观交换一致性分支，以鼓励结构表征在外观变化下保持稳定。我们还引入了一个可靠性图引导的双头判别器，包含有效性头（Validity Head）和不确定性头（Uncertainty Head），以提供特征级不确定性反馈。在ISIC-2016和Kvasir-SEG上的大量实验表明，SAUF-Net在低标签设置下优于最先进的半监督方法。
cs.CV / 56 / 2609.02251
Handwriting Trajectory Recovery via Autoregressive Ordered Stroke Instance Prediction
通过自回归有序笔画实例预测进行手写轨迹恢复
Abstract
Handwriting trajectory recovery aims to infer the dynamic writing process hidden behind a static handwritten image. Since offline handwriting preserves only the final spatial ink pattern, temporal information such as stroke order, writing direction, and pen-tip motion is lost, making recovery inherently ambiguous. Existing learning-based methods often directly predict the complete character trajectory without explicitly exploiting the stroke-level organization of handwriting. We argue that recovering the writing process should follow the writing process itself. Accordingly, we propose a two-stage framework that first recovers ordered stroke instances and then reconstructs continuous within-stroke motion. The first stage integrates stroke extraction and stroke-order recovery through autoregressive ordered stroke prediction, while direction-related structural cues further support within-stroke trajectory generation. Experiments on Chinese handwriting show that the proposed ordered prediction is more effective than post-hoc stroke ordering. Even without trajectory simplification, our full-point model achieves numerically better results than those reported by all compared baselines, while a controlled analysis shows that trajectory sampling density substantially affects measured recovery performance. Additional experiments demonstrate generalization to unseen Chinese character categories and cross-language extensibility to English and Tamil handwriting.
Chinese Translation
手写轨迹恢复旨在推断隐藏在静态手写图像背后的动态书写过程。由于离线手写仅保留最终的空间墨迹模式，笔画顺序、书写方向和笔尖运动等时间信息丢失，使得恢复过程本质上存在模糊性。现有的基于学习的方法通常直接预测完整的字符轨迹，而未明确利用手写的笔画级组织。我们认为，恢复书写过程应遵循书写过程本身。因此，我们提出了一个两阶段框架，首先恢复有序的笔画实例，然后重建笔画内的连续运动。第一阶段通过自回归有序笔画预测整合了笔画提取和笔画顺序恢复，而与方向相关的结构线索进一步支持笔画内轨迹的生成。在中文手写实验中，所提出的有序预测比事后笔画排序更有效。即使在没有轨迹简化的情况下，我们的全点模型在数值上也取得了比所有比较基线更好的结果，而控制分析表明，轨迹采样密度对测量的恢复性能有显著影响。额外的实验展示了对未见中文字符类别的泛化能力以及对英语和泰米尔语手写的跨语言扩展性。
cs.CV / 57 / 2609.02255
T2LSC-Bench: Benchmarking Localized Semantic Control in Text-to-Image Generation
T2LSC-Bench：文本到图像生成中的局部语义控制基准测试
Abstract
Recent text-to-image models have become increasingly capable of rendering explicit text, but reliable localized text control requires more than generating the correct string. In applications such as product labeling, signage, and interface design, target text should be rendered within a designated text-bearing region without altering the predefined subject identity or surrounding scene semantics. We refer to violations of this requirement as target-text-associated semantic leakage, in which target-text semantics are expressed through non-textual visual content beyond the designated anchor. Existing visual-text benchmarks primarily evaluate readability, spelling accuracy, and layout, leaving this form of semantic leakage largely unexamined. We introduce T2LSC-Bench, a controlled diagnostic benchmark comprising 50 seed subjects and 1,200 prompt cases per model, yielding 7,160 evaluated images across six models. Its factorized design varies semantic relation, scene openness, prompt mode, and language. A dual-branch protocol combines OCR-VLM text verification with structured VLM semantic judgments to measure Text-at-Anchor Accuracy (TAA), Semantic Subject Preservation (SSP), Semantic Leakage Rate (SLR), and Conditional Semantic Leakage Rate (cSLR). Under stress-test conditions, SLR increases from 1.2% to 18.1% and cSLR from 1.3% to 18.2%, whereas TAA decreases only from 91.4% to 90.9%. Anti-leakage prompting reduces SLR from 16.6% to 8.4% without degrading rendering accuracy. Human validation on 420 images shows strong agreement between automatic and adjudicated annotations. These results show that accurate text rendering does not guarantee local containment of target-text semantics.
Chinese Translation
近期的文本到图像模型在渲染明确文本方面的能力日益增强，但可靠的局部文本控制不仅仅依赖于生成正确的字符串。在产品标签、标识和界面设计等应用中，目标文本应在指定的承载文本区域内渲染，而不改变预定义的主体身份或周围场景的语义。我们将这一要求的违反称为目标文本相关的语义泄漏，其中目标文本的语义通过超出指定锚点的非文本视觉内容表达。现有的视觉-文本基准主要评估可读性、拼写准确性和布局，未能充分考察这种形式的语义泄漏。我们引入了T2LSC-Bench，这是一个受控的诊断基准，包含50个种子主体和每个模型1200个提示案例，共评估了六个模型的7160张图像。其因子化设计变化了语义关系、场景开放性、提示模式和语言。双分支协议结合了OCR-VLM文本验证与结构化VLM语义判断，以测量锚点文本准确性（Text-at-Anchor Accuracy, TAA）、语义主体保留（Semantic Subject Preservation, SSP）、语义泄漏率（Semantic Leakage Rate, SLR）和条件语义泄漏率（Conditional Semantic Leakage Rate, cSLR）。在压力测试条件下，SLR从1.2%增加到18.1%，cSLR从1.3%增加到18.2%，而TAA仅从91.4%下降到90.9%。反泄漏提示将SLR从16.6%降低到8.4%，而不降低渲染准确性。对420张图像的人类验证显示自动标注与裁定标注之间具有较强一致性。这些结果表明，准确的文本渲染并不能保证目标文本语义的局部包含。
cs.CV / 58 / 2609.02266
MAOL: Morphology-Aware Ordinal Learning for Fine-Grained Industrial Defect Severity Grading
MAOL：面向形态学的序数学习用于细粒度工业缺陷严重性分级
Abstract
Fine-grained defect severity grading is essential for industrial inspection, yet remains challenging due to the ordinal nature of severity labels, the strong dependence on morphology-related cues, and the train-test discrepancy between clean annotated instances and noisy predicted instances in two-stage pipelines. We propose MAOL, a Morphology-Aware Ordinal Learning framework for fine-grained industrial defect severity grading. MAOL formulates severity grading as an instance-level ordinal learning task, incorporates explicit morphological features to enhance representation learning, introduces class-conditional adaptive ordinal thresholds to model defect-specific grading boundaries, and employs prediction-aware training via localization perturbation to improve robustness to imperfect predicted instances. Extensive experiments under both clean-ROI and predicted-instance settings demonstrate that MAOL consistently outperforms rule-based methods, nominal classification models, and existing ordinal baselines, especially in the predicted-instance setting. The proposed approach ranked third in the IDA 2026 Challenge on Fine-Grained Severity Grading for High-Precision Manufacturing.
Chinese Translation
细粒度缺陷严重性分级对于工业检测至关重要，但由于严重性标签的序数特性、对形态相关线索的强依赖，以及在两阶段流程中干净标注实例与噪声预测实例之间的训练-测试差异，这一任务仍然具有挑战性。我们提出了MAOL，一个面向形态学的序数学习框架，用于细粒度工业缺陷严重性分级。MAOL将严重性分级形式化为实例级序数学习任务，结合显式的形态特征以增强表示学习，引入类条件自适应序数阈值以建模缺陷特定的分级边界，并通过定位扰动采用预测感知训练以提高对不完美预测实例的鲁棒性。在干净区域和预测实例设置下的广泛实验表明，MAOL在性能上始终优于基于规则的方法、名义分类模型和现有的序数基线，尤其是在预测实例设置中。该方法在2026年高精度制造细粒度严重性分级挑战赛中排名第三。
cs.CV / 59 / 2609.02282
RouteGraph-Mona: Confusion-Aware Routing Fine-Tuning for Mineral Image Classification
RouteGraph-Mona：基于混淆感知的矿物图像分类路由微调
Abstract
Mineral image classification is important for geological exploration and resource development, but it remains challenging due to substantial intra-class variations in appearance and high inter-class visual similarity. Multi-cognitive Visual Adapter (Mona) is a vision-oriented parameter-efficient adapter that adapts pre-trained visual models by tuning only a few parameters. However, Mona statically aggregates responses from multiple scales, limiting its ability to accommodate sample-specific scale preferences and model confusion among visually similar mineral categories. To address this issue, we propose \textbf{RouteGraph-Mona}, a lightweight route-space regularization method built on Mona. Specifically, we replace Mona's static multi-scale aggregation with sample-adaptive routing. The resulting branch-selection behavior defines a compact routing space that captures each image's scale preferences. We then regularize the resulting routing signatures with class-wise route anchors and confusion-weighted margins. The route anchors encourage class-consistent routing patterns, while the margins promote greater separation between visually similar categories in the routing space. Experiments on three public mineral image datasets with two visual backbones show that RouteGraph-Mona consistently outperforms Mona in mean accuracy and remains competitive with representative fine-tuning methods and mineral image classification baselines.
Chinese Translation
矿物图像分类对于地质勘探和资源开发至关重要，但由于外观上显著的类内变异和高类间视觉相似性，这一任务仍然具有挑战性。多认知视觉适配器（Mona）是一种面向视觉的参数高效适配器，通过仅调整少量参数来适应预训练的视觉模型。然而，Mona 静态地聚合来自多个尺度的响应，限制了其适应样本特定尺度偏好的能力，并在视觉上相似的矿物类别之间造成模型混淆。为了解决这一问题，我们提出了 extbf{RouteGraph-Mona}，一种基于 Mona 的轻量级路由空间正则化方法。具体而言，我们用样本自适应路由替换了 Mona 的静态多尺度聚合。由此产生的分支选择行为定义了一个紧凑的路由空间，捕捉每个图像的尺度偏好。然后，我们用类别路由锚和混淆加权边界对生成的路由特征进行正则化。路由锚鼓励类别一致的路由模式，而边界则促进视觉上相似类别在路由空间中的更大分离。在三个公共矿物图像数据集上进行的实验表明，RouteGraph-Mona 在平均准确率上始终优于 Mona，并与代表性的微调方法和矿物图像分类基线保持竞争力。
cs.CV / 60 / 2609.02288
Diffusion-Encoding Gaussian Field for Joint k-q dMRI Reconstruction
用于联合 k-q dMRI 重建的扩散编码高斯场
Abstract
Diffusion MRI requires repeated k-space acquisitions over multiple diffusion-encoding directions, making acquisition time dependent on both spatial and angular sampling. Existing joint k-q methods either associate directional parameters with fixed voxels or separate spatial reconstruction from angular completion. However, diffusion-weighted images acquired under different directions share the same anatomical organization, while their local signal intensities vary with diffusion encoding. Existing formulations do not fully exploit the complementarity between shared anatomy and direction-dependent signal variation. Consequently, residual spatial errors may be misinterpreted as genuine angular variation and propagated to unobserved directions. We propose a subject-specific spatial-angular Gaussian field for self-supervised joint k-q dMRI reconstruction. Shared 3D Gaussian primitives provide local spatial support, with each primitive carrying a continuous q-conditioned tensor-residual response. The signal at each location is synthesized from multiple overlapping primitive responses, coupling neighboring spatial regions and diffusion directions. The field is progressively optimized from undersampled k-space measurements of observed directions, without fully sampled targets or held-out-direction supervision. Experiments on three HCP diffusion shells under multiple acceleration settings demonstrated consistent improvements in missing-direction DWI reconstruction, tensor-derived metrics, and principal diffusion orientation estimation.
Chinese Translation
扩散 MRI 需要在多个扩散编码方向上进行重复的 k 空间采集，这使得采集时间依赖于空间和角度采样。现有的联合 k-q 方法要么将方向参数与固定体素关联，要么将空间重建与角度补全分开。然而，在不同方向下获得的扩散加权图像共享相同的解剖结构，而其局部信号强度则随着扩散编码而变化。现有的公式未能充分利用共享解剖结构与方向依赖信号变化之间的互补性。因此，残余的空间误差可能被误解为真实的角度变化，并传播到未观察到的方向。我们提出了一种特定于受试者的空间-角度高斯场，用于自监督的联合 k-q dMRI 重建。共享的 3D 高斯原语提供局部空间支持，每个原语携带一个连续的 q 条件张量残差响应。每个位置的信号由多个重叠的原语响应合成，耦合相邻的空间区域和扩散方向。该场从观察方向的欠采样 k 空间测量中逐步优化，而无需完全采样的目标或未观察方向的监督。在多个加速设置下对三个 HCP 扩散层的实验表明，在缺失方向的 DWI 重建、张量导出的度量和主扩散方向估计方面均有一致的改进。
cs.CV / 61 / 2609.02289
If It Moves, Radar Knows: A Physics-Aware Radar Transformer for Class-Agnostic Moving-Object Detection
如果它移动，雷达就知道：一种物理感知雷达变换器用于类别无关的移动物体检测
Abstract
Detectors trained on closed-set annotations can miss rare moving objects outside the training taxonomy. Automotive radar provides category-independent Doppler motion cues and is less affected by adverse illumination and weather, but sparse, noisy returns hinder class-aware 3D box detection. Surface location and velocity remain useful for motion reasoning and collision avoidance when full box geometry is difficult to recover. We present the Physics-Aware Radar Transformer (PART), a fully sparse radar-only detector that predicts existence confidence, a representative surface point, and 2D ground-plane velocity for each moving-object hypothesis. Doppler-Aware Query Initialization (DAQI) replaces scene-independent learned queries with input-dependent proposals by clustering radar returns in position and velocity, easing query-object assignment in sparse scenes. Physics-Guided Cross-Attention (PGCA) incorporates radial-Doppler consistency and radar cross section (RCS) into query-point association. Uncertainty-aware supervision randomly masks ground-truth objects and assigns soft existence targets to ambiguous radar-supported queries, reducing reliance on exhaustive annotations. With only 1.1 million parameters, PART achieves a class-agnostic average precision (CA-AP) of 0.8827, a mean average surface translation error (mASTE) of 0.3188 m, and a mean average velocity error (mAVE) of 0.8084 m/s on nuScenes. It attains 0.9203 recall on rare and safety-relevant categories excluded from the standard evaluation and remains effective at night, in rain, and under severe occlusion. Inspection of apparent false positives shows that some predictions correspond to moving objects absent from the nuScenes annotations. Code and pretrained model weights will be publicly available at https://github.com/sunyinghao-uestc/PART.
Chinese Translation
在封闭集注释上训练的检测器可能会错过训练分类法之外的稀有移动物体。汽车雷达提供类别无关的多普勒运动线索，并且不易受到不良光照和天气的影响，但稀疏和噪声的回波阻碍了基于类别的三维框检测。当完整的框几何形状难以恢复时，表面位置和速度仍然对运动推理和碰撞避免有用。我们提出了物理感知雷达变换器（PART），这是一种完全稀疏的仅雷达检测器，为每个移动物体假设预测存在置信度、代表性表面点和二维地面平面速度。多普勒感知查询初始化（DAQI）通过在位置和速度上聚类雷达回波，用输入依赖的提议替换场景无关的学习查询，从而简化稀疏场景中的查询-对象分配。物理引导的交叉注意力（PGCA）将径向多普勒一致性和雷达截面（RCS）纳入查询点关联。具有不确定性感知的监督随机掩盖真实对象，并将软存在目标分配给模糊的雷达支持查询，从而减少对详尽注释的依赖。在仅有110万个参数的情况下，PART在nuScenes上实现了0.8827的类别无关平均精度（CA-AP）、0.3188米的平均表面平移误差（mASTE）和0.8084米/秒的平均速度误差（mAVE）。它在标准评估中排除的稀有和安全相关类别上达到了0.9203的召回率，并在夜间、雨天和严重遮挡情况下仍然有效。对明显的假阳性进行检查显示，一些预测对应于nuScenes注释中缺失的移动物体。代码和预训练模型权重将公开发布于https://github.com/sunyinghao-uestc/PART。
cs.CV / 62 / 2609.02291
VoRTeC: Taming Foundation Flow for One-step Real time Video Compression
VoRTeC：驯化基础流以实现一步实时视频压缩
Abstract
Ultra-low bitrate video compression still faces critical challenges: traditional neural video compression inevitably introduces blurring artifacts, while diffusion-based generative video compression suffers from excessive decoding latency and poor temporal consistency. To address these issues, we propose $\mathtt{VoRTeC}$, a Video Compression framework built upon a foundational flow model (Wan2.1). By compactly encoding latent video representations, predicting the positions of compressed representations along flow trajectories, and integrating multi-scale priors, $\mathtt{VoRTeC}$ enables the compressor to harness generative video flow priors effectively. Without accessing the parameters or gradients of flow matching networks, our framework achieves one-step decoding and reconstructions with high perceptual fidelity. Meanwhile, we maintain consistency across frame groups via tail-frame reuse and prior caching. Extensive experiments demonstrate that our method reduces bit consumption by 58\% compared to prior diffusion-based approaches, with decoding speed boosted by 3 to 197 times: $\mathtt{VoRTeC}$ achieves a decoding speed of 13 FPS at 720p and 32 FPS at 480p.
Chinese Translation
超低比特率视频压缩仍面临关键挑战：传统的神经视频压缩不可避免地引入模糊伪影，而基于扩散的生成视频压缩则遭受过高的解码延迟和较差的时间一致性。为了解决这些问题，我们提出了$ exttt{VoRTeC}$，这是一个基于基础流模型（Wan2.1）的视频压缩框架。通过紧凑地编码潜在视频表示，预测沿流轨迹压缩表示的位置，并整合多尺度先验，$ exttt{VoRTeC}$使压缩器能够有效利用生成视频流先验。在不访问流匹配网络的参数或梯度的情况下，我们的框架实现了一步解码和高感知保真度的重建。同时，我们通过尾帧重用和先验缓存保持帧组之间的一致性。大量实验表明，我们的方法相比于先前的基于扩散的方法减少了58%的比特消耗，解码速度提升了3到197倍：在720p时，$ exttt{VoRTeC}$实现了13 FPS的解码速度，在480p时实现了32 FPS的解码速度。
cs.CV / 63 / 2609.02299
Domain shift-robust object detection with GenAI image editing
基于生成性人工智能图像编辑的领域转移鲁棒物体检测
Abstract
Object detectors often degrade under domain shifts such as changes in lighting, weather, or occlusion. These shifts alter object appearance and expose a reliance on visual shortcuts learned from the training distribution that do not generalize across domains. Acquiring sufficient real-world samples to capture such domain variation is particularly difficult in specialized, low-data settings. Recent advances in diffusion-based generative image editing have shown promise for improving the in-domain performance of object detectors through synthetic data augmentation. However, their potential to improve out-of-domain robustness remains largely unexplored. We hypothesize that generative image editing can simulate a controlled domain shift in training data, effectively bridging the gap between source and target domains. To test this, we studied camouflaged military vehicle detection as a challenging domain shift scenario. Detectors trained on uncamouflaged data demonstrate substantial degradation on real test imagery containing foliage, netting, and multi-spectral camouflage across 15 vehicle classes in close-up, ground-level imagery. We used two diffusion-based editing models, Qwen Image Edit 2509 and Flux.2 Dev, to synthetically add camouflage to the training data, alongside a LoRA fine-tuned version of Qwen. A non-generative black-bar occlusion baseline served as a lower bound on augmentation quality. Using a GroundingDINO detector trained on real and synthetic data, generative camouflage augmentation yielded substantial mAP improvements for foliage (+20.1) and netting (+14.4) camouflage. Generating multi-spectral camouflage proved more challenging, but LoRA fine-tuning improved performance by 4.4 mAP over the uncamouflaged baseline.
Chinese Translation
物体检测器在领域转移（如光照、天气或遮挡变化）下往往表现不佳。这些转移改变了物体的外观，并暴露出对从训练分布中学习的视觉捷径的依赖，而这些捷径在不同领域之间并不具有普适性。在专业的低数据环境中，获取足够的真实世界样本以捕捉这种领域变化尤其困难。最近，基于扩散的生成图像编辑技术在通过合成数据增强来提高物体检测器的领域内性能方面显示出良好的前景。然而，它们在提高领域外鲁棒性方面的潜力仍然未被充分探索。我们假设生成性图像编辑可以在训练数据中模拟受控的领域转移，有效地弥合源领域和目标领域之间的差距。为了验证这一假设，我们研究了伪装军事车辆检测作为一个具有挑战性的领域转移场景。基于未伪装数据训练的检测器在包含树叶、网状物和多光谱伪装的真实测试图像上显示出显著的性能下降，涉及15个车辆类别的近距离地面图像。我们使用了两个基于扩散的编辑模型，Qwen Image Edit 2509和Flux.2 Dev，合成性地将伪装添加到训练数据中，同时使用了经过LoRA微调的Qwen版本。一个非生成性的黑条遮挡基线作为增强质量的下限。使用在真实和合成数据上训练的GroundingDINO检测器，生成性伪装增强在树叶（+20.1）和网状物（+14.4）伪装方面带来了显著的mAP提升。生成多光谱伪装则更具挑战性，但LoRA微调使性能比未伪装基线提高了4.4 mAP。
cs.CV / 64 / 2609.02318
YesTrack: Referring Multi-Object Tracking via MLLM-based Yes/No Verification
YesTrack：基于MLLM的是/否验证的指称多目标跟踪
Abstract
Referring multi-object tracking (RMOT) aims to track every instance in a video that matches a given language expression. Despite the recent integration of multimodal large language models (MLLMs) to enhance generalization, existing methods predominantly relegate them to the role of caption generators, necessitating external modules for final decision-making. This paradigm not only introduces extra latency but also severely underutilizes the inherent vision-language alignment capabilities of MLLMs. To address these limitations, we propose YesTrack, a novel two-stage RMOT method that reformulates referring as a discriminative task, directly leveraging MLLMs for Yes/No verification without explicit text generation. To further enhance the reliability and efficiency of this MLLM-based verification, we introduce two lightweight temporal consistency constraints: Temporal Confidence Prior (TCP) and Temporal Reference Propagation (TRP). We further validate the generality of this discriminative paradigm by proposing YesTrack-MOT, a straightforward yet highly effective instantiation for generic multi-object tracking (MOT). Experiments on Refer-KITTI and Refer-KITTI-V2 show that YesTrack significantly outperforms existing state-of-the-art methods while maintaining high efficiency, even when implemented with the smallest variant of Qwen3-VL. Code is released at https://github.com/ggbondrighthere24/YesTrack.
Chinese Translation
指称多目标跟踪（RMOT）旨在跟踪视频中与给定语言表达相匹配的每个实例。尽管最近将多模态大型语言模型（MLLMs）整合以增强泛化能力，但现有方法主要将其 relegated 为字幕生成器，需依赖外部模块进行最终决策。这一范式不仅引入了额外的延迟，还严重低估了MLLMs固有的视觉-语言对齐能力。为了解决这些限制，我们提出了YesTrack，一种新颖的两阶段RMOT方法，将指称重新表述为一个判别任务，直接利用MLLMs进行是/否验证，而无需显式的文本生成。为了进一步提高这种基于MLLM的验证的可靠性和效率，我们引入了两个轻量级的时间一致性约束：时间置信先验（Temporal Confidence Prior, TCP）和时间参考传播（Temporal Reference Propagation, TRP）。我们通过提出YesTrack-MOT，进一步验证了这一判别范式的通用性，这是一种简单但极为有效的通用多目标跟踪（MOT）实例。在Refer-KITTI和Refer-KITTI-V2上的实验表明，YesTrack显著优于现有的最先进方法，同时保持高效，即使在使用最小变体的Qwen3-VL时也是如此。代码已发布在 https://github.com/ggbondrighthere24/YesTrack。
cs.CV / 65 / 2609.02333
ORB-SVM : An Innovative Hybrid Framework for Efficient Brain Tumor Detection from MRI Scans
ORB-SVM：一种用于高效脑肿瘤检测的创新混合框架，基于MRI扫描
Abstract
Brain cancer remains one of the most significant challenges in modern medicine, where the accuracy of early stage diagnosis is a decisive factor in patient survival and treatment efficacy. Although Magnetic Resonance Imaging (MRI) is the established gold standard for visualizing neurological structures, the interpretation of these high dimensional scans is often complicated by subjective variability among practitioners and the inherent noise present in complex medical images. While contemporary approaches frequently rely on high parameter deep learning architectures, such models often involve significant computational costs and require extensive data for effective training. This study introduces a hybrid framework that utilizes the Oriented FAST and Rotated BRIEF (ORB) algorithm for precise feature extraction and a Support Vector Machine (SVM) for classification [1], [2]. The proposed approach achieves a sub- stantial data reduction of approximately 99.5%, which effectively minimizes the influence of non informative background data while preserving critical diagnostic patterns essential for tumor identification. By balancing feature sparsity with a robust kernel based classifier, this methodology addresses the limitations of over parameterized systems while maintaining high diagnostic integrity. Experimental evaluations conducted on the Br35H dataset demonstrate that the framework attains a classification accuracy of 97.5%. The findings suggest that the integration of localized feature representation and optimized classification provides a reliable and resource efficient alternative for medical image analysis, offering a structured solution that maintains per- formance without the need for extensive computational overhead.
Chinese Translation
脑癌仍然是现代医学中最重要的挑战之一，早期诊断的准确性是影响患者生存和治疗效果的决定性因素。尽管磁共振成像（MRI）是可视化神经结构的公认金标准，但这些高维扫描的解读常常受到从业者主观差异和复杂医学图像中固有噪声的影响。尽管当代方法通常依赖于高参数深度学习架构，但此类模型往往涉及显著的计算成本，并且需要大量数据以实现有效训练。本研究提出了一种混合框架，利用定向FAST和旋转BRIEF（ORB）算法进行精确特征提取，并采用支持向量机（SVM）进行分类。所提出的方法实现了约99.5%的显著数据减少，有效地最小化了非信息背景数据的影响，同时保留了肿瘤识别所必需的关键诊断模式。通过在特征稀疏性与强健的基于核的分类器之间取得平衡，该方法解决了过度参数化系统的局限性，同时保持了高诊断完整性。在Br35H数据集上进行的实验评估表明，该框架达到了97.5%的分类准确率。研究结果表明，局部特征表示与优化分类的结合为医学图像分析提供了一种可靠且资源高效的替代方案，提供了一种结构化解决方案，能够在不需要大量计算开销的情况下保持性能。
cs.CV / 66 / 2609.02341
Towards Zero-Shot Transfer Across Embodiments For Driving VLAs
面向跨体态的零样本迁移驾驶视觉-语言-动作模型
Abstract
Vision-Language-Action models (VLAs) have shown strong potential in autonomous driving by leveraging multimodal pretraining for instruction following, visual reasoning, and scene-level generalization. In robotic manipulation, scaling VLA fine-tuning across multiple robot setups--especially when unifying representations across embodiments--has been shown to improve in-dataset performance and cross-embodiment generalization; in autonomous driving, however, VLAs remain largely trained on individual datasets and are rarely evaluated for zero-shot transfer to unseen datasets and camera rigs; furthermore naively adding more datasets to the training data does not necessarily lead to better performance within seen embodiments. To address these problems, we study multi-dataset training for the driving task and BEV-Forcing, an auxiliary objective that transfers ground-plane object-layout information from a specialized Bird's-Eye-View model into the VLA backbone. By encouraging the model to represent object position through a shared BEV spatial interface, we show that an auxiliary task such as BEV-Forcing can improve both in-distribution and out-of-distribution performance when training on a small number of camera rigs. As the number of training embodiments increases, however, the benefits of the auxiliary task are reduced; we present this as evidence that new techniques in the literature may see their benefits diminish when simply scaling up training diversity, which motivates presenting results taking into account data scaling.
Chinese Translation
视觉-语言-动作模型（VLAs）通过利用多模态预训练在指令遵循、视觉推理和场景级泛化方面展现了在自动驾驶中的强大潜力。在机器人操作中，跨多个机器人设置进行VLA微调的规模化——特别是在跨体态统一表示时——已被证明可以提高数据集内性能和跨体态泛化；然而，在自动驾驶中，VLAs仍然主要在单个数据集上进行训练，并且很少评估其在未见数据集和摄像机设备上的零样本迁移能力；此外，简单地将更多数据集添加到训练数据中并不一定会导致在已见体态中的更好性能。为了解决这些问题，我们研究了驾驶任务的多数据集训练和BEV-Forcing，这是一种辅助目标，将来自专门的鸟瞰图模型的地面平面物体布局信息转移到VLA主干中。通过鼓励模型通过共享的BEV空间接口表示物体位置，我们表明，像BEV-Forcing这样的辅助任务可以在少量摄像机设备上训练时提高数据集内和数据集外的性能。然而，随着训练体态数量的增加，辅助任务的益处会减少；我们将此视为证据，表明文献中的新技术在简单扩大训练多样性时可能会看到其益处减弱，这促使我们在考虑数据扩展的情况下呈现结果。
cs.CV / 67 / 2609.02348
Structured-Prior-Guided Diffusion Inpainting with Physical Consistency for Traffic Sign Augmentation
基于结构先验引导的物理一致性扩散修复用于交通标志增强
Abstract
Traffic sign detection faces a long-tailed data distribution. Many rare signs matter as much as common ones from a regulatory standpoint, yet they have very few samples. Generative data augmentation is one way out. General-purpose inpainting models, however, distort digits, deform geometry and perspective, and shift colours when applied directly to sign regions. We trace this to a single gap: the conditioning signal is too abstract for the physical composition of a sign. We propose a structured-prior-guided diffusion inpainting framework with physical consistency. It injects the semantic, appearance and geometric priors of a sign through three orthogonal pathways: a JSON-formatted text prompt, a front-view vector template rendered with measured dominant colours (via IP-Adapter), and an affine-aligned vector template (via ControlNet). Two physical consistency losses constrain colour with a CIELAB chromaticity $L_1$ term and edge structure with a Sobel gradient term. We train by self-supervised reconstruction on a large set of images collected in-house at AMAP, then evaluate zero-shot on the public TT100K-2021 dataset, a different source. Our method uses a Stable Diffusion 1.5 backbone of about 1.4B parameters. It beats seven representative competitors on every metric of reconstruction fidelity, physical consistency and semantic controllability. Its OCR exact-match rate reaches 91.1\%, against 44.2\% for the 12B industrial model FLUX.1 Fill [dev], and it needs only $1/14$ of that model's inference time. Leave-one-out ablations confirm that each of the three prior pathways and both loss terms contribute on their own. In downstream detection, the synthetic data raises the group-pooled AP50 of rare classes by $1.23\times$ to $7.40\times$ over a real-data-only baseline. Code and pre-trained models are available at https://github.com/52hz-whale/TrafficSignInpaint.
Chinese Translation
交通标志检测面临长尾数据分布的问题。许多稀有标志在监管方面与常见标志同样重要，但样本数量却非常有限。生成数据增强是一种解决方案。然而，通用的修复模型在直接应用于标志区域时，会扭曲数字、变形几何和透视，并且改变颜色。我们将此归因于一个单一的缺口：条件信号对于标志的物理组成过于抽象。我们提出了一种基于结构先验引导的扩散修复框架，具有物理一致性。该框架通过三条正交路径注入标志的语义、外观和几何先验：一个JSON格式的文本提示、一个使用测量主导颜色（通过IP-Adapter）渲染的正视向量模板，以及一个仿射对齐的向量模板（通过ControlNet）。两个物理一致性损失通过CIELAB色度$L_1$项约束颜色，通过Sobel梯度项约束边缘结构。我们在AMAP内部收集的大量图像上进行自监督重建训练，然后在公共的TT100K-2021数据集上进行零-shot评估，该数据集来源不同。我们的方法使用约14亿参数的Stable Diffusion 1.5骨干网络。在重建保真度、物理一致性和语义可控性等各项指标上超越了七个代表性竞争对手。其OCR精确匹配率达到91.1\%，而12B工业模型FLUX.1 Fill [dev]的匹配率为44.2\%，且只需该模型推理时间的$1/14$。逐一剔除的消融实验确认了三条先验路径和两个损失项各自的贡献。在下游检测中，合成数据将稀有类别的组池AP50提高了$1.23 imes$至$7.40 imes$，相较于仅使用真实数据的基线。代码和预训练模型可在https://github.com/52hz-whale/TrafficSignInpaint获取。
cs.CV / 68 / 2609.02349
GlyphAnchor: Enhancing Visual Text Rendering via Position-Anchored Glyph Priors
GlyphAnchor：通过位置锚定的字形先验增强视觉文本渲染
Abstract
Rendering accurate text remains difficult for image generation and editing models, especially when the target contains long, complex, and densely arranged text or rare characters. Existing approaches either improve native text rendering through stronger backbones and data-centric training without explicit glyph priors, or incorporate glyph priors through specialized designs that remain insufficiently accurate and robust under challenging scenarios. We introduce GlyphAnchor, a novel text-rendering enhancement method for both text-to-image and image-editing diffusion transformer models. GlyphAnchor enhances the backbone with lightweight glyph patch conditions whose positions are anchored to the target image through the model's native positional encoding. We train this capability with staged supervised finetuning and further refine it with text-aware post-training to improve robustness. We also introduce InfoTextBench, a benchmark for evaluating text-rich visual text rendering in both generation and editing settings. Experiments across multiple backbones and benchmarks, including long, complex, and densely arranged text and rare character scenarios, show that GlyphAnchor consistently improves text fidelity while preserving overall image quality.
Chinese Translation
准确渲染文本仍然是图像生成和编辑模型面临的挑战，尤其是在目标包含长文本、复杂文本、密集排列的文本或稀有字符时。现有方法要么通过更强大的骨干网络和以数据为中心的训练来改善本地文本渲染，而没有明确的字形先验，要么通过专门设计来结合字形先验，但在困难场景下的准确性和鲁棒性仍然不足。我们提出了GlyphAnchor，一种新颖的文本渲染增强方法，适用于文本到图像和图像编辑扩散变换器模型。GlyphAnchor通过轻量级的字形补丁条件增强骨干网络，这些条件的位置通过模型的本地位置编码锚定到目标图像。我们通过分阶段的监督微调训练这一能力，并通过文本感知的后训练进一步优化，以提高鲁棒性。我们还引入了InfoTextBench，一个用于评估生成和编辑环境中文本丰富的视觉文本渲染的基准。针对多个骨干网络和基准的实验，包括长文本、复杂文本、密集排列的文本和稀有字符场景，结果表明GlyphAnchor在保持整体图像质量的同时，始终提高了文本的保真度。
cs.CV / 69 / 2609.02350
LookStep: Efficient Vision-Language Navigation with Linguistic Foresight and Event Driven Memory
LookStep：具有语言前瞻性和事件驱动记忆的高效视觉-语言导航
Abstract
Vision-Language Navigation (VLN) requires an embodied agent to follow natural-language instructions in unseen environments. Recent progress has been largely driven by Multimodal Large Language Models (MLLMs). Existing methods follow a next-step action prediction paradigm, supervising only the expert action, which requires a high quantity of data for training. They also rely on cognitive maps, accumulated historical frames, or external 3D tools to maintain states, leading to high computational and memory overhead. To realize resource efficiency VLN, we propose LookStep, a unified end-to-end framework that combines Language Centric Future State Modeling and Event Driven Rolling Memory that uses language labels to generate coarse-grained navigation progress and future states for each candidate action, while autonomously deciding whether to write each observation into a bounded rolling memory with a semantic role. We validate LookStep empirically. On VLN-CE tasks, LookStep outperforms existing methods under the same training settings, achieving a 49.7\% success rate on R2R-CE Val-Unseen with better memory efficiency and less data usage. Code and model is available at https://github.com/kunyang-YU/LookStep.
Chinese Translation
视觉-语言导航（VLN）要求具身代理在未见环境中遵循自然语言指令。最近的进展主要得益于多模态大型语言模型（MLLMs）。现有方法遵循下一步动作预测范式，仅对专家动作进行监督，这需要大量的数据进行训练。它们还依赖于认知地图、累积的历史帧或外部3D工具来维护状态，导致高计算和内存开销。为了实现资源高效的VLN，我们提出了LookStep，一个统一的端到端框架，结合了以语言为中心的未来状态建模和事件驱动的滚动记忆，利用语言标签生成每个候选动作的粗粒度导航进度和未来状态，同时自主决定是否将每个观察写入具有语义角色的有界滚动记忆中。我们对LookStep进行了实证验证。在VLN-CE任务中，LookStep在相同训练设置下优于现有方法，在R2R-CE Val-Unseen上实现了49.7%的成功率，同时具有更好的内存效率和更少的数据使用。代码和模型可在 https://github.com/kunyang-YU/LookStep 获取。
cs.CV / 70 / 2609.02359
TempoGround: State-Aware Streaming Visual Grounding with Vision-Language Models
TempoGround：基于状态感知的视觉定位流媒体处理与视觉语言模型
Abstract
Visual grounding maps language referents to spatial targets and is central to open-vocabulary perception with vision-language models. Existing methods have made substantial progress on single-frame and video-based visual grounding, yet under streaming inputs they still suffer from identity drift, cross-frame inconsistency, and fragile localization under partial occlusion. To address these issues, we present TempoGround, a VLM-native framework that detects cross-frame object correspondence and explicitly models object presence states, thereby enabling accurate and consistent visual grounding under streaming inputs. The key is a curriculum prediction mechanism guided by state-aware cross-frame correspondence: TempoGround resolves 2D instance association, predicts whether each object newly enters, continues in, or leaves the view, decodes the 2D box, and then lifts it to a camera-frame 3D box. As token-level supervision alone cannot capture the geometric objectives of streaming grounding, we further introduce Streaming Grounding Reinforcement (SGR), which optimizes TempoGround with verifiable Grounding, Identity, and Consistency rewards, jointly reinforcing persistent localization and temporally consistent predictions. We carefully design a three-stage training strategy and train TempoGround on large-scale data. We evaluate visual grounding under causally streaming inputs on multiple challenging benchmarks: TempoGround improves
[email protected] and
[email protected] by 4.4 and 0.5 on average, and
[email protected] and AP_3D by 6.2 and 7.5, respectively. These results demonstrate that TempoGround provides a practical foundation for visual grounding under streaming inputs.
Chinese Translation
视觉定位将语言指称与空间目标关联，是使用视觉语言模型进行开放词汇感知的核心。现有方法在单帧和基于视频的视觉定位上取得了显著进展，但在流媒体输入下仍然面临身份漂移、跨帧不一致和在部分遮挡下的脆弱定位等问题。为了解决这些问题，我们提出了TempoGround，这是一个原生于视觉语言模型的框架，能够检测跨帧对象对应关系，并明确建模对象存在状态，从而在流媒体输入下实现准确且一致的视觉定位。其关键在于一个由状态感知的跨帧对应关系引导的课程预测机制：TempoGround解决了2D实例关联，预测每个对象是新进入、持续存在还是离开视野，解码2D框，然后将其提升为相机帧的3D框。由于仅靠令牌级监督无法捕捉流媒体定位的几何目标，我们进一步引入了流媒体定位强化（Streaming Grounding Reinforcement, SGR），通过可验证的定位、身份和一致性奖励来优化TempoGround，联合强化持久定位和时间一致的预测。我们精心设计了三阶段的训练策略，并在大规模数据上训练TempoGround。我们在多个具有挑战性的基准上评估了因果流媒体输入下的视觉定位：TempoGround在
[email protected]和
[email protected]上平均提高了4.4和0.5，在
[email protected]和AP_3D上分别提高了6.2和7.5。这些结果表明，TempoGround为流媒体输入下的视觉定位提供了一个实用的基础。
cs.CV / 71 / 2609.02369
Information Density Imbalance in Visual Object Detection
视觉目标检测中的信息密度不平衡
Abstract
In object detection, the number of instances is typically used to determine whether a dataset exhibits a long-tailed distribution, implicitly assuming that the model will perform poorly on categories with fewer instances. This assumption has led to extensive research on category bias in datasets with imbalanced instance numbers. However, even in datasets where instance numbers are relatively balanced, models still exhibit category bias, indicating that instance count alone cannot explain this phenomenon. In this work, we first introduce the concept and measurement of information density. We then observe a significant negative correlation between a category's information density and its accuracy, and we investigate how the training process impacts this relationship. Empirical studies suggest that information density imbalance may be a potential source of category bias. To preliminarily validate the potential of information density, we made simple improvements to three advanced object detection loss functions using this concept. Experiments on the Pascal VOC, COCO-LT, and LVIS datasets demonstrate that information density can significantly reduce model bias while effectively enhancing the overall performance of existing loss functions. This study provides a new perspective for understanding the generalized bias phenomenon in object detection models and offers new tools for designing fairer loss functions and training strategies.
Chinese Translation
在目标检测中，实例数量通常用于判断数据集是否呈现长尾分布，隐含假设模型在实例较少的类别上表现较差。这一假设导致了对实例数量不平衡的数据集中的类别偏差进行广泛研究。然而，即使在实例数量相对平衡的数据集中，模型仍然表现出类别偏差，这表明仅靠实例数量无法解释这一现象。在本研究中，我们首先引入信息密度的概念及其测量。然后，我们观察到类别的信息密度与其准确性之间存在显著的负相关关系，并探讨训练过程如何影响这种关系。实证研究表明，信息密度不平衡可能是类别偏差的潜在来源。为了初步验证信息密度的潜力，我们利用这一概念对三种先进的目标检测损失函数进行了简单改进。在 Pascal VOC、COCO-LT 和 LVIS 数据集上的实验表明，信息密度可以显著减少模型偏差，同时有效提升现有损失函数的整体性能。本研究为理解目标检测模型中的广义偏差现象提供了新的视角，并为设计更公平的损失函数和训练策略提供了新工具。
cs.CV / 72 / 2609.02377
ProSR: Semantic-Prototype-Guided Discrete Modeling for Physically Consistent SAR Super-Resolution
ProSR：基于语义原型引导的物理一致性SAR超分辨率离散建模
Abstract
High-resolution Synthetic Aperture Radar (SAR) imagery is critical for precision analysis such as automatic target recognition, yet its acquisition is costly. Although generative image super-resolution (ISR) models offer a promising alternative, current smooth-approximation based diffusion frameworks often struggle to preserve the coherent scattering statistics, causing stochastic structural distortions that are less consistent with real SAR physics. To address this, we propose Semantic Prototype-Guided Super-Resolution (ProSR), reformulating SAR ISR as a semantically-guided discrete token prediction task within a quantized latent space. By mapping signal features to discrete scattering primitives, ProSR preserves the impulsive nature of SAR without over-smoothing. Furthermore, we integrate a Self-Supervised Learning backbone into SAR ISR to extract label-free semantic priors, overcoming label scarcity. Guided by these priors, we introduce Semantic-Aligned Detail Encoding to decouple high-frequency signals into discrete scattering primitives. In parallel, the Semantic Prototype Map Generator explicitly constructs semantic prototype maps, allowing Prototype-Map-Guided Attention to route the information flows within identical categories and mitigate inter-class interference. To validate our approach, we present a large-scale 0.25m resolution benchmark from the Umbra Open Dataset. Experimental results show ProSR achieves superior visual quality while preserving essential scattering characteristics required for practical SAR applications.
Chinese Translation
高分辨率合成孔径雷达（SAR）图像对于自动目标识别等精确分析至关重要，但其获取成本高昂。尽管生成图像超分辨率（ISR）模型提供了一个有前景的替代方案，但当前基于平滑近似的扩散框架往往难以保持相干散射统计特性，导致随机结构失真，与真实的SAR物理特性不够一致。为了解决这一问题，我们提出了语义原型引导超分辨率（ProSR），将SAR ISR重新构建为量化潜在空间中的语义引导离散标记预测任务。通过将信号特征映射到离散散射原语，ProSR在不进行过度平滑的情况下保留了SAR的冲击特性。此外，我们将自监督学习骨干网络集成到SAR ISR中，以提取无标签的语义先验，从而克服标签稀缺的问题。在这些先验的引导下，我们引入了语义对齐细节编码，将高频信号解耦为离散散射原语。同时，语义原型图生成器明确构建语义原型图，允许原型图引导注意力在相同类别内引导信息流并减轻类间干扰。为了验证我们的方法，我们从Umbra开放数据集中提供了一个大规模的0.25米分辨率基准。实验结果表明，ProSR在保持实际SAR应用所需的基本散射特性方面，取得了优越的视觉质量。
cs.CV / 73 / 2609.02401
CA-OPD: Confidence-Aware On-Policy Distillation for Structured Visual Prediction
CA-OPD：基于置信度的在线蒸馏用于结构化视觉预测
Abstract
Autoregressive vision language models unify heterogeneous perception tasks but are highly susceptible to compounding errors. On-policy distillation (OPD) bridges the training-inference mismatch by training students on their own rollouts. However, unreliable student predictions, especially early in training, can derail the trajectory and degrade the quality of teacher supervision. While recent interleaved distillation methods allow the teacher to verify and replace student tokens, they primarily rely on rigid ranking metrics rather than exact teacher confidence, and they overlook how intervention decisions can inform token-level supervision. To address this, we introduce Confidence-Aware On-Policy Distillation (CA-OPD), a framework that couples reliable rollout construction with adaptive supervision. CA-OPD utilizes teacher confidence to selectively correct unreliable student transitions, gradually transferring rollout control to the student via a strict-to-relaxed schedule. Crucially, CA-OPD aligns knowledge transfer with these intervention decisions: corrected positions receive direct cross-entropy supervision from the teacher's prediction, while retained positions benefit from the teacher's full predictive distribution. Evaluated in a multi-teacher setting for GUI grounding and optical character recognition, CA-OPD substantially improves the Qwen3.5-0.8B baseline across all six target benchmarks, including gains of $9.50$ points on ScreenSpot-Pro and $6.72$ points on OCRBench-v2 English. Controlled studies further show that the gains depend on intervention placement, progressive rollout control, and intervention-aligned supervision, rather than intervention frequency alone.
Chinese Translation
自回归视觉语言模型统一了异构感知任务，但对累积错误高度敏感。在线蒸馏（OPD）通过在学生自己的回滚上进行训练来弥合训练与推理之间的差距。然而，不可靠的学生预测，尤其是在训练初期，可能会干扰轨迹并降低教师监督的质量。尽管最近的交错蒸馏方法允许教师验证和替换学生的标记，但它们主要依赖于严格的排名指标，而不是准确的教师置信度，并且忽视了干预决策如何影响标记级监督。为了解决这个问题，我们提出了基于置信度的在线蒸馏（CA-OPD），这是一个将可靠的回滚构建与自适应监督相结合的框架。CA-OPD利用教师置信度选择性地纠正不可靠的学生过渡，通过严格到放松的时间表逐步将回滚控制转移给学生。重要的是，CA-OPD将知识转移与这些干预决策对齐：被纠正的位置直接接受来自教师预测的交叉熵监督，而保留的位置则受益于教师的完整预测分布。在GUI定位和光学字符识别的多教师设置中评估，CA-OPD在所有六个目标基准上显著提高了Qwen3.5-0.8B基线，包括在ScreenSpot-Pro上提高了$9.50$分，在OCRBench-v2 English上提高了$6.72$分。控制研究进一步表明，增益依赖于干预位置、渐进的回滚控制和与干预对齐的监督，而不仅仅是干预频率。
cs.CV / 74 / 2609.02411
The Diagnosis a Reporter Leaves Unspoken: Surfacing Frozen Tumor Features for Brain-Tumor MRI Reporting
报告者未言明的诊断：揭示脑肿瘤MRI报告中的冻结肿瘤特征
Abstract
A capable brain-MRI report generator can still be, in effect, diagnostically silent. When a multi-chain chain-of-thought (CoT) reporter built on a medical Mistral-7B backbone is evaluated on held-out cohorts, it names most meningiomas and almost all metastases "glioma" (diagnosis recall 0.44/0.07). Yet the answer is not absent from the model: a supervised linear probe applied to its frozen segmentation features recovers the three tumour cohorts at 0.82 macro-F$_1$ (5-fold cross-validation; chance $\approx$0.33). We introduce NeuroFusion, an assistive reporter that surfaces this latent signal rather than overriding it: discriminative field-classifier heads over per-lesion features condition a fast, single-pass draft-then-review decoder on their committed outputs. Built on the identical Mistral backbone, this restores the diagnosis (meningioma 0.92, metastasis 0.75) and wins 8 of 9 prose-content comparisons across three held-out cohorts (RaTEScore, RadGraph-F$_1$, GREEN; Holm-corrected paired BCa), with no significant loss on the ninth, at 5-6x lower latency ($\approx$80 vs. 457 s/case). A controlled negative result sharpens the mechanism: a learned diagnosis pin that overrides the decoder instead of merely informing it collapses out-of-distribution metastasis recall to 0.03. Grammar-constrained decoding keeps 92.3% of records schema-valid, making every sentence entailment-checkable (7.5% contradicted vs. 36.8% for the direct baseline). In a blinded nine-case pilot, two board-certified neurologists independently rated NeuroFusion highest in every tumour type, the only system with zero critical errors, and gave it the top-rated sign-off in eight of nine cases (six outright, two ties).
Chinese Translation
一个有效的脑MRI报告生成器在实际诊断中仍然可能是沉默的。当基于医疗Mistral-7B骨干的多链思维链（CoT）报告者在保留的队列上进行评估时，它将大多数脑膜瘤和几乎所有转移性肿瘤称为“胶质瘤”（诊断召回率为0.44/0.07）。然而，模型中并非缺少答案：对其冻结分割特征应用的监督线性探针以0.82的宏F$_1$恢复了三种肿瘤队列（5折交叉验证；机会约为0.33）。我们引入了NeuroFusion，这是一种辅助报告者，它揭示了这一潜在信号，而不是覆盖它：在每个病变特征上施加的区分场分类器头根据其承诺的输出条件化一个快速的单次草拟-再审解码器。基于相同的Mistral骨干，这恢复了诊断（脑膜瘤0.92，转移瘤0.75），并在三个保留队列的9项文体内容比较中赢得了8项（RaTEScore，RadGraph-F$_1$，GREEN；经过Holm校正的配对BCa），在第九项上没有显著损失，延迟降低5-6倍（约80秒对比457秒/病例）。一个受控的负结果锐化了机制：一个覆盖解码器的学习诊断针而不仅仅是告知它，导致分布外转移召回率降至0.03。语法约束解码保持92.3%的记录符合模式，使每个句子都可进行蕴涵检查（反驳率为7.5%，而直接基线为36.8%）。在一个盲测的九病例试点中，两位获得认证的神经科医生独立评估NeuroFusion在每种肿瘤类型中均为最高，且是唯一没有关键错误的系统，在九个病例中有八个获得最高评级（六个明确，两个平局）。
cs.CV / 75 / 2609.02434
Uncertainty-Guided Adverse Weather Restoration via Gated Transformer Network
基于不确定性引导的恶劣天气图像恢复通过门控变换器网络
Abstract
Restoring images degraded by adverse weather remains challenging due to spatially heterogeneous degradations. Many existing weather-specific restoration models rely on weather-agnostic global aggregation, naive cross-scale fusion, and deterministic objectives, which struggle to handle heterogeneous degradations in all-in-one adverse-weather settings. To address these limitations, we propose an Uncertainty-guided Adverse-weather Restoration Network (UAR-Net), a weather-specific AiO framework that integrates a gated transformer with balanced multi-scale skip connections. Specifically, we employ Gated Dual-scale Transformer Blocks (GDTB) to jointly model selective global interactions and multi-scale local structures, a progressive Balanced Multi-scale Skip Connection (BMSC) for balanced multi-scale feature integration, and an Uncertainty-Aware Refinement Head (URH) that performs artifact removal, detail enhancement, and predictive uncertainty estimation. The model is supervised by a Brightness-Aware Energy Loss (BAE-Loss) to encourage accurate reconstruction with well-calibrated uncertainty. Extensive experiments demonstrate that our method achieves state-of-the-art performance across multiple adverse-weather benchmarks. The codes will open source upon acceptance.
Chinese Translation
由于空间异质性退化，恢复受到恶劣天气影响的图像仍然具有挑战性。许多现有的特定天气恢复模型依赖于与天气无关的全局聚合、简单的跨尺度融合和确定性目标，这些方法在一体化恶劣天气场景中难以处理异质性退化。为了解决这些局限性，我们提出了一种不确定性引导的恶劣天气恢复网络（UAR-Net），这是一个特定天气的全方位框架，结合了门控变换器和均衡的多尺度跳跃连接。具体而言，我们采用门控双尺度变换器块（GDTB）共同建模选择性全局交互和多尺度局部结构，采用渐进式均衡多尺度跳跃连接（BMSC）进行均衡的多尺度特征集成，以及一个不确定性感知细化头（URH），该头执行伪影去除、细节增强和预测不确定性估计。该模型通过亮度感知能量损失（BAE-Loss）进行监督，以鼓励在良好校准不确定性的情况下进行准确重建。大量实验表明，我们的方法在多个恶劣天气基准测试中实现了最先进的性能。代码将在接受后开源。
cs.CV / 76 / 2609.02448
Adapting a Foundation Model for Lunar Surface Height Estimation
适应基础模型进行月球表面高度估计
Abstract
Digital elevation models (DEMs) can provide accurate height information, making it invaluable for analyzing the lunar surface. As the European Space Agency (ESA) prepares for future lunar missions that aim to land on the Moon, a precise method for height estimation will be essential for hazardous terrain that could endanger the landing approach. Traditional approaches to generate DEMs from imagery, such as shape from shading (SfS) and stereophotogrammetry (SPG) have been proven highly valuable for this task. However, due to advancements in machine learning, especially computer vision, the focus has shifted towards monocular depth estimation via deep learning. The lunar surface is covered by rocks and craters, and classic hazard detection methods rely solely on 2D image data. Our goal is to address this issue by developing a relative lunar surface height estimator that can provide additional information for hazard localization. In this letter, we present a methodology that builds on the well-known zero-shot relative depth estimation model Depth Anything V2 (DAV2). Other works have been using it as a state-of-the-art comparison for their proposed lunar DEM estimation method, but without adaptations to the target domain. Thus, it may underperform. Therefore, we propose a fine-tuning strategy with publicly available SPG-derived DEM data of the lunar surface. Our results demonstrate a significant improvement in performance compared to the zero-shot model, effectively transforming DAV2 into a reliable relative depth estimator of the lunar surface.
Chinese Translation
数字高程模型（DEMs）能够提供准确的高度信息，这对于分析月球表面至关重要。随着欧洲航天局（ESA）为未来的月球任务做准备，这些任务旨在实现月球着陆，精确的高度估计方法对于可能危及着陆过程的危险地形将是必不可少的。传统的从影像生成DEMs的方法，如基于阴影的形状重建（SfS）和立体摄影测量（SPG），已被证明在这一任务中极具价值。然而，随着机器学习，特别是计算机视觉的进步，研究重点已转向通过深度学习进行单目深度估计。月球表面布满岩石和陨石坑，经典的危险检测方法仅依赖于二维图像数据。我们的目标是通过开发一个相对月球表面高度估计器来解决这一问题，以提供额外的信息用于危险定位。在本信中，我们提出了一种方法论，该方法基于著名的零样本相对深度估计模型Depth Anything V2（DAV2）。其他研究已将其作为其提出的月球DEMs估计方法的最先进比较，但未对目标领域进行适应。因此，它的表现可能不尽如人意。因此，我们提出了一种微调策略，利用公开可用的基于SPG的月球表面DEMs数据。我们的结果表明，与零样本模型相比，性能有显著提升，有效地将DAV2转变为可靠的月球表面相对深度估计器。
cs.CV / 77 / 2609.02452
WiFlow: Estimating Optical Flow using WiFi Channel State Information
WiFlow：利用WiFi信道状态信息估计光流
Abstract
Knowing where and how fast objects are moving within a scene is important across various domains. Usually, cameras are used to capture the data necessary for this task, but adding cameras often raises privacy concerns, and the quality of captured frames is heavily influenced by lighting conditions. In this work, we explore using WiFi channel state information (CSI) instead of camera frames for optical flow estimation. We propose WiFlow, a CSI based flow estimator, a preprocessor evaluation for CSI, and three model architectures that offer different trade-offs between accuracy and complexity. Further, we create the first dataset for training and evaluating CSI-based optical flow estimators, and our experiments provide insights into key design elements for this task. Code and data are available at https://visinf.github.io/wiflow.
Chinese Translation
了解场景中物体的运动位置和速度在多个领域中都至关重要。通常，使用摄像头来捕捉执行此任务所需的数据，但增加摄像头往往会引发隐私问题，并且捕获帧的质量受到光照条件的严重影响。在本研究中，我们探索使用WiFi信道状态信息（CSI）代替摄像头帧进行光流估计。我们提出了WiFlow，一种基于CSI的流估计器，以及CSI的预处理评估和三种模型架构，这些架构在准确性和复杂性之间提供了不同的权衡。此外，我们创建了第一个用于训练和评估基于CSI的光流估计器的数据集，我们的实验为此任务的关键设计元素提供了见解。代码和数据可在 https://visinf.github.io/wiflow 获取。
cs.CV / 78 / 2609.02462
VIPS: Vehicle-Infrastructure Cooperative Planning Benchmark via Pseudo-Simulation
VIPS：基于伪仿真的车辆-基础设施协同规划基准
Abstract
End-to-end autonomous driving in urban environments requires robust decision-making under partial observability and complex multi-agent interactions. Severe occlusions and dense traffic at intersections limit the perception capability of single-agent systems, motivating recent efforts on Vehicle-to-Infrastructure (V2I) cooperation for perception and planning. However, existing evaluation protocols face a fundamental trade-off: open-loop evaluation fails to capture error accumulation and recovery from deviations, while closed-loop evaluation is costly, difficult to scale, and often relies on simulated environments that may suffer from domain gaps. To bridge this gap, we propose VIPS, a benchmark for cooperative autonomous driving in V2I settings based on pseudo-simulation. VIPS extends pseudo-simulation by integrating vehicle and infrastructure observations. This enables scalable yet realistic evaluation of robustness and error propagation without full simulation. We further present CoS-V2X, a cooperative planning framework based on sparse representations. CoS-V2X models vehicle-infrastructure interactions using compact features for efficient communication and robust decision-making under heterogeneous observations. Code and dataset are available at https://vips2026.github.io.
Chinese Translation
在城市环境中实现端到端的自动驾驶需要在部分可观测性和复杂的多智能体交互下进行稳健的决策。交叉口的严重遮挡和密集交通限制了单智能体系统的感知能力，这促使了近期对车辆与基础设施（Vehicle-to-Infrastructure, V2I）合作进行感知和规划的研究。然而，现有的评估协议面临着一个基本的权衡：开环评估无法捕捉错误积累和从偏差中恢复，而闭环评估成本高、难以扩展，并且通常依赖于可能存在领域差距的模拟环境。为了解决这一问题，我们提出了VIPS，一个基于伪仿真的V2I环境中的协同自动驾驶基准。VIPS通过整合车辆和基础设施的观测来扩展伪仿真。这使得在不进行全面仿真的情况下，能够对稳健性和错误传播进行可扩展且现实的评估。我们进一步提出了CoS-V2X，一个基于稀疏表示的协同规划框架。CoS-V2X使用紧凑特征建模车辆与基础设施的交互，以实现高效的通信和在异构观测下的稳健决策。代码和数据集可在 https://vips2026.github.io 获取。
cs.CV / 79 / 2609.02471
Learning to Track from Privileged Target Appearances
从特权目标外观中学习跟踪
Abstract
Target templates define what a visual tracker searches for, yet the templates available at inference trade off localization certainty with appearance freshness: the initial ground-truth template is exact but becomes stale, whereas recent templates better reflect the current appearance but are cropped from uncertain predictions. We quantify this bottleneck with a non-deployable oracle that supplies an exact current-frame target crop, improving AUC on LaSOT by 15.2 percentage points. This gap reveals a training-only opportunity: frame-level ground truths provide exact current- and future-frame target crops, although such crops are unavailable at deployment. We introduce Privileged Appearance Transfer for Tracking (PATT), a teacher-student training framework that transfers these privileged appearances to a deployable tracker through multi-level representation prediction. The privileged teacher observes exact target crops from past, current, and future frames, whereas the student receives only past-frame templates and learns to predict the teacher's search representations. To avoid transferring unreliable teacher signals, PATT weights this transfer by the teacher's relative localization advantage over the student and its absolute localization accuracy. After training, the teacher, latent predictor, reliability weights, and privileged crops are removed, leaving standard student-only inference. Across seven benchmarks at two model scales, PATT achieves consistent gains under both long- and short-term tracking protocols.
Chinese Translation
目标模板定义了视觉跟踪器搜索的内容，但在推理时可用的模板在定位确定性与外观新鲜度之间存在权衡：初始的真实模板是准确的，但会变得过时，而最近的模板更好地反映了当前外观，但是从不确定的预测中裁剪而来的。我们通过一个不可部署的oracle量化了这一瓶颈，该oracle提供了当前帧目标的精确裁剪，使LaSOT上的AUC提高了15.2个百分点。这一差距揭示了一个仅限于训练的机会：帧级真实值提供了当前和未来帧目标的精确裁剪，尽管在部署时这些裁剪是不可用的。我们引入了用于跟踪的特权外观转移（Privileged Appearance Transfer for Tracking, PATT），这是一个教师-学生训练框架，通过多层表示预测将这些特权外观转移到可部署的跟踪器。特权教师观察来自过去、当前和未来帧的精确目标裁剪，而学生仅接收过去帧的模板，并学习预测教师的搜索表示。为了避免转移不可靠的教师信号，PATT通过教师相对于学生的定位优势和绝对定位精度来加权这一转移。训练完成后，教师、潜在预测器、可靠性权重和特权裁剪被移除，仅保留标准的学生推理。在两个模型规模的七个基准测试中，PATT在长短期跟踪协议下均取得了一致的性能提升。
cs.CV / 80 / 2609.02476
UnCapsTSR: An Unsupervised Transformer-based Image Super-Resolution Approach for Capsule Endoscopy Images
UnCapsTSR：一种基于无监督变换器的胶囊内窥镜图像超分辨率方法
Abstract
Wireless Capsule Endoscopy (WCE) captures and streams video while passing through a patient's Gastrointestinal (GI) tract and is used to examine its irregularities. Although advantageous over conventional endoscopy, WCE suffers from limitations related to capsule size and wireless transmission, resulting in images with coarser resolution. This work presents UnCapsTSR, an unsupervised transformer-based Generative Adversarial Network (GAN) framework for improving the spatial resolution of Low-Resolution (LR) WCE images. The proposed method accomplishes SR without explicit degradation estimation of real-world LR data and eliminates the need for true LR-HR pairs. UnCapsTSR employs a Bilateral Total Variation (BTV) loss to ensure spatial continuity in SR images. A newly curated dataset from the Kvasir Capsule dataset is also presented for training WCE SR models. Generalizability is validated on KID and GIANA datasets that are not used during training. A new non-reference metric, Endoscopy Quality Metric (EndoQM), is introduced for quantitative evaluation of domain-specific WCE data. Experiments demonstrate consistent improvement over state-of-the-art unsupervised SR approaches using NIQE, BRISQUE, PIQE, and EndoQM. Statistical evaluation shows 40 to 80 percent improvement in EndoQM from LR to SR across the evaluated datasets.
Chinese Translation
无线胶囊内窥镜（WCE）在通过患者的胃肠道（GI）时捕捉并传输视频，用于检查其不规则性。尽管相较于传统内窥镜具有优势，但WCE受到胶囊尺寸和无线传输相关限制的影响，导致图像分辨率较低。本文提出了UnCapsTSR，一种基于无监督变换器的生成对抗网络（GAN）框架，用于提高低分辨率（LR）WCE图像的空间分辨率。所提出的方法在不需要显式估计真实世界LR数据的降级情况下实现超分辨率（SR），并消除了对真实LR-HR配对的需求。UnCapsTSR采用双边全变差（BTV）损失，以确保SR图像的空间连续性。此外，本文还提出了一个新整理的数据集，基于Kvasir胶囊数据集，用于训练WCE SR模型。通过在未用于训练的KID和GIANA数据集上验证了模型的泛化能力。引入了一种新的非参考指标——内窥镜质量指标（EndoQM），用于定量评估特定领域的WCE数据。实验结果表明，在使用NIQE、BRISQUE、PIQE和EndoQM的情况下，相较于最先进的无监督SR方法，UnCapsTSR在性能上持续改善。统计评估显示，在评估的数据集中，从LR到SR的EndoQM改善幅度达到40%到80%。
cs.CV / 81 / 2609.02502
Blending Concepts: Benchmarking Visual Metaphor Generation in Text-to-Image Models
概念融合：文本到图像模型中视觉隐喻生成的基准评估
Abstract
Text-to-image (T2I) models have achieved remarkable success at faithfully rendering specified objects and attributes, yet their ability to produce visual metaphors, images that convey abstract ideas by combining elements from two distinct domains, remains largely unexamined. To bridge this gap, we introduce VMetaphor-Bench, the first benchmark for evaluating visual metaphor generation in T2I models. It comprises 1,500 visual metaphors curated from real-world creative imagery, organized into three levels and ten categories, with each sample paired with two prompts of differing specificity. For evaluation, we develop a hybrid framework within an MLLM-as-judge paradigm, combining a multiple-choice question (MCQ) based protocol of 9,594 questions across four levels of metaphorical fidelity with a dimension-based scoring protocol along three perceptual dimensions. Extensive evaluation of 11 representative T2I models reveals that even the strongest proprietary models struggle with compositional structuring and cross-domain mapping, key aspects of metaphorical expression, highlighting visual metaphor generation as an important frontier for future T2I research.
Chinese Translation
文本到图像（T2I）模型在忠实呈现特定对象和属性方面取得了显著成功，但它们生成视觉隐喻的能力——通过结合两个不同领域的元素来传达抽象思想的图像——仍然基本未被研究。为填补这一空白，我们引入了 VMetaphor-Bench，这是第一个用于评估 T2I 模型中视觉隐喻生成的基准。该基准包含 1,500 个来自现实世界创意图像的视觉隐喻，分为三个层次和十个类别，每个样本配有两个不同具体性的提示。为了进行评估，我们在 MLLM-as-judge 范式下开发了一个混合框架，结合了 9,594 个问题的多项选择题（MCQ）协议，涵盖四个隐喻忠实度层次，以及沿三个感知维度的基于维度的评分协议。对 11 个代表性 T2I 模型的广泛评估表明，即使是最强大的专有模型在组合结构和跨领域映射方面也面临挑战，这些都是隐喻表达的关键方面，突显了视觉隐喻生成作为未来 T2I 研究的重要前沿。
cs.CV / 82 / 2609.02504
SR-Edit: Region-Aware Image Editing via Self-Refinement
SR-Edit：基于自我精炼的区域感知图像编辑
Abstract
With the recent rapid progress in generative models, image editing has made remarkable advances, yet achieving faithful edits that precisely modify only the target regions while strictly preserving all other regions remains challenging. Since externally provided region annotations are often difficult to obtain in practice, a growing body of work seeks to improve preservation by automatically inferring edit and non-edit regions, and then enforcing consistency on the latter. However, these approaches still suffer from inaccurate region estimation and heuristic correction strategies that distort the native inference process, making methods designed for fidelity themselves a new source of artifacts. We propose SR-Edit, an image editing framework that overcomes these issues via iterative self-refinement. Specifically, at each iteration, SR-Edit first (i) extracts progressively precise and self-consistent region separation from the model's own predictions by lightweight post-processing, and then (ii) enforces preservation in non-edit areas through correction updates that remain aligned with the original sampling dynamics. Extensive experiments demonstrate that SR-Edit achieves superior preservation and overall image quality compared to existing editing techniques.
Chinese Translation
随着生成模型的快速进展，图像编辑取得了显著的进展，但实现忠实的编辑，即仅精确修改目标区域而严格保留其他区域，仍然具有挑战性。由于在实践中外部提供的区域注释往往难以获得，越来越多的研究工作试图通过自动推断编辑和非编辑区域来改善保留效果，并对后者施加一致性约束。然而，这些方法仍然面临区域估计不准确和启发式修正策略扭曲原生推断过程的问题，使得旨在提高保真度的方法本身成为新的伪影来源。我们提出了SR-Edit，一个通过迭代自我精炼克服这些问题的图像编辑框架。具体而言，在每次迭代中，SR-Edit首先（i）通过轻量级后处理从模型自身的预测中提取逐步精确且自我一致的区域分离，然后（ii）通过保持与原始采样动态一致的修正更新，在非编辑区域施加保留。大量实验表明，SR-Edit在保留效果和整体图像质量上优于现有的编辑技术。
cs.CV / 83 / 2609.02510
Orthogonal Ensembles and Tested Explanations for Performer-Independent Body-Motion Emotion Recognition
正交集与表演者独立的身体运动情感识别的验证解释
Abstract
We study body-only, 12-class acted-emotion classification from skeleton motion under leave-performer-out (LPO) evaluation, a hard, underdetermined setting: chance is 8.3%, and a protocol-matched reproduced STGCN++ baseline reaches only 25.73 +/- 4.03% Macro-F1. We show that reliable gains come not from a new architecture but from combining eleven models with orthogonal error modes: under 10-fold LPO cross-validation on the labeled training performers, an equal-weight logit-mean ensemble reaches 36.80 +/- 4.00% per-fold Macro-F1, a protocol-matched +11.07 pp (+43% relative) over the same-split reproduced baseline. Our central contribution is a tested explanation suite: for a strong ensemble member, part-masking and counterfactual edits show (rather than assert) that its decisions depend on motion-grounded body-region evidence, and this region saliency aligns with rule-based Laban Movement Analysis (LMA) attributes far more than with classical kinematics: region-level saliency-LMA Spearman rho = +0.500 versus +0.033, roughly 15x, and the alignment holds for the submitted 11-way ensemble itself at rho = +0.517; the audit is post hoc and needs no retraining. The same suite faithfully reports a negative: within-window temporal saliency is diffuse rather than localized.
Chinese Translation
我们研究了在留表演者外（LPO）评估下，仅基于身体的12类表演情感分类，这是一种困难的、欠定的设置：随机概率为8.3%，而与协议匹配的再现STGCN++基线仅达到25.73 +/- 4.03%的宏F1值。我们表明，可靠的增益并非来自新的架构，而是来自结合具有正交误差模式的十一种模型：在标记训练表演者上进行10折LPO交叉验证时，等权重对数均值集成达到了36.80 +/- 4.00%的每折宏F1，相较于同一分割的再现基线提高了11.07个百分点（相对提高43%）。我们主要的贡献是一个经过验证的解释套件：对于一个强大的集成成员，部分遮蔽和反事实编辑显示（而不是断言）其决策依赖于基于运动的身体区域证据，而该区域的显著性与基于规则的拉班运动分析（LMA）属性的对齐程度远高于经典运动学：区域级显著性-LMA的Spearman相关系数为+0.500，而经典运动学为+0.033，约为15倍，并且这种对齐在提交的11路集成中也保持，相关系数为+0.517；该审计是事后进行的，无需重新训练。同样的套件忠实地报告了一个负面结果：窗口内的时间显著性是分散而非局部的。
cs.CV / 84 / 2609.02512
Beauty is in the AI of the beholder: MLLMs systematically overrate facial attractiveness
美在观察者的人工智能中：多模态大型语言模型系统性地高估面部吸引力
Abstract
Beauty assessments from Multimodal Large Language Models (MLLMs) are increasingly popular amongst users, companies, and aestheticians. This raises the question of whether these AI models can accurately reflect human judgments of attractiveness. In a pre- registered exploratory study, we compared the attractiveness ratings of 2,513 human participants to four widely used commercial AI models: Claude, Gemini, GPT, and Grok. Results showed that MLLMs systematically rate faces more favourably and within a narrower range than humans and, at the time of study, do not reproduce human ratings in absolute terms. However, MLLMs exhibit strong correlations with human attractiveness judgments, accurately tracking the rank-ordering of faces. MLLMs may judge faces by different cues than humans; only face age was a predictor of facial attractiveness in both humans and MLLMs, with inconsistent patterns across models for ethnicity and gender. AI models strongly agree with one another, except for Grok, which also showed the lowest agreement with humans. Our findings suggest that while they may be able to approximate rank-orderings of human attractiveness, current off-the-shelf commercial MLLMs systematically overrate the beauty of human faces.
Chinese Translation
来自多模态大型语言模型（MLLMs）的美学评估在用户、公司和美学家中越来越受欢迎。这引发了一个问题，即这些人工智能模型是否能够准确反映人类对吸引力的判断。在一项预注册的探索性研究中，我们将2513名人类参与者的吸引力评分与四个广泛使用的商业人工智能模型进行了比较：Claude、Gemini、GPT和Grok。结果显示，MLLMs系统性地对面孔给予更高的评分，并且评分范围比人类更窄，并且在研究时并未在绝对值上再现人类的评分。然而，MLLMs与人类的吸引力判断之间存在强相关性，能够准确追踪面孔的排名顺序。MLLMs可能通过与人类不同的线索来评判面孔；只有面孔年龄在人类和MLLMs中都是面部吸引力的预测因素，而在种族和性别方面，各模型的模式则不一致。除了Grok外，人工智能模型之间的高度一致性，Grok与人类的协议也最低。我们的研究结果表明，尽管它们可能能够近似人类吸引力的排名顺序，但当前的商业MLLMs系统性地高估了人类面孔的美丽。
cs.CV / 85 / 2609.02528
Doppio: A Dataset for Contactless Weight Estimation of Falling Particles
Doppio：一种用于无接触颗粒重量估计的数据集
Abstract
Measuring the mass of powder, including falling particles, is a common task in industrial applications. While scales are effective for static measurements, many applications require contactless sensing, where existing solutions are often costly, application-specific, and technically complex. In this work, we investigate computer vision as a practical alternative for contactless mass estimation. As an accessible real-world case study, we focus on coffee grinding and introduce \emph{Doppio}, a novel video dataset capturing videos of falling ground coffee, paired with precise, per-frame ground-truth weight measurements. To demonstrate contactless measuring, we evaluate deep learning-based approaches ranging from purely spatial feed-forward networks to recurrent spatio-temporal models. These models are analyzed with respect to their predictive accuracy and computational trade-offs. We demonstrate that deep learning-based computer vision models accurately estimate the cumulative weight of falling particles, establishing a solid foundation for future vision-based contactless measurement solutions.
Chinese Translation
测量粉末的质量，包括下落颗粒，是工业应用中的一项常见任务。虽然秤在静态测量中有效，但许多应用需要无接触传感，而现有解决方案往往成本高、特定于应用且技术复杂。在本研究中，我们探讨了计算机视觉作为无接触质量估计的实用替代方案。作为一个可访问的现实案例研究，我们专注于咖啡研磨，并引入了 extit{Doppio}，一个新颖的视频数据集，捕捉下落研磨咖啡的视频，并配有精确的逐帧真实重量测量。为了演示无接触测量，我们评估了基于深度学习的方法，从纯粹的空间前馈网络到递归时空模型。这些模型的预测准确性和计算权衡进行了分析。我们证明了基于深度学习的计算机视觉模型能够准确估计下落颗粒的累计重量，为未来基于视觉的无接触测量解决方案奠定了坚实的基础。
cs.CV / 86 / 2609.02529
Fine-Grained Anomaly Perception in Wild UGC-Enhanced Images: A Comprehensive Dataset and Difference-Fusion Framework
野生用户生成内容增强图像中的细粒度异常感知：一个综合数据集和差异融合框架
Abstract
Image enhancement and restoration have become standard back-end operations on short-video and social media platforms to boost UGC visual experience. Yet these processes inevitably introduce visual anomalies--especially in faces, texts, and textures--that directly undermine perceptual fidelity and viewer trust. While existing IQA methods perform well on classic distortions, they target holistic quality assessment and fail to capture the specific, localized anomalies caused by enhancement algorithms in real-world UGC. To bridge this gap, we formally define a new task-quality Anomaly Perception for UGC image Enhancement (UEAP), and contribute the first UEAP benchmark dataset, named UEAP-4k, curated from the real business scenarios. It provides fine-grained annotations for anomaly categories, localization and severity levels. Furthermore, we propose a Difference-Fusion Anomaly Perception Method (DFAP-UGC) for wild UGC-enhanced images, which leverages explicit problem-reference difference fusion with dense spatial querying, regional verification, and quality-aware ranking, enabling robust anomaly identification in challenging scenarios. To handle the inherent coupling of subtasks in this new task, we propose a Locality-Aware Dynamic Task Prioritization (LADTP) training strategy that enables effective end-to-end learning and eliminates multi-stage overhead. Extensive experiments show that our method outperforms baselines adapted from classical approaches for this task, validating the value of this dataset and the superior of DFAP-UGC for robust UGC-enhanced image anomaly perception. Code and data will be public.
Chinese Translation
图像增强和修复已成为短视频和社交媒体平台上提升用户生成内容（UGC）视觉体验的标准后端操作。然而，这些过程不可避免地引入视觉异常，特别是在面部、文本和纹理方面，这直接削弱了感知的真实性和观众的信任。尽管现有的图像质量评估（IQA）方法在经典失真方面表现良好，但它们主要针对整体质量评估，未能捕捉到真实世界UGC中增强算法造成的特定局部异常。为填补这一空白，我们正式定义了一项新任务——用户生成内容图像增强的异常感知（UEAP），并贡献了首个UEAP基准数据集，命名为UEAP-4k，该数据集来自真实商业场景。它提供了异常类别、定位和严重程度的细粒度注释。此外，我们提出了一种针对野生UGC增强图像的差异融合异常感知方法（DFAP-UGC），该方法利用显式问题参考差异融合，结合密集空间查询、区域验证和质量感知排序，能够在具有挑战性的场景中实现稳健的异常识别。为了处理这一新任务中子任务的固有耦合，我们提出了一种局部感知动态任务优先级（LADTP）训练策略，使得有效的端到端学习成为可能，并消除了多阶段的开销。大量实验表明，我们的方法在这一任务上优于从经典方法改编的基线，验证了该数据集的价值以及DFAP-UGC在稳健的UGC增强图像异常感知中的优越性。代码和数据将公开发布。
cs.CV / 87 / 2609.02531
Spatially Aware World Action Model via Geometric Latent Diffusion
基于几何潜在扩散的空间感知世界行动模型
Abstract
World Action Models (WAMs) leverage the capabilities of large-scale pretrained video diffusion models to jointly predict future observations and actions, inheriting rich visual and physical priors from internet-scale video. This has made them a promising paradigm for robot policy learning, yet the prevailing models operate exclusively on RGB observations and do not leverage 3D information. To bridge this gap, we introduce a Spatially Aware World Action Model (SA-WAM), which repurposes a pretrained video model for joint action, RGB, and depth prediction, enabling 3D-aware world modeling and action prediction within a single diffusion backbone. We use a nonlinear encoding that maps the unbounded depth signal into the bounded input domain expected by the frozen VAE tokenizer. This allows us to reuse the tokenizer without 3D-specific fine-tuning, incorporating geometric information without sacrificing the pretrained priors. SA-WAM achieves state-of-the-art results on the RoboCasa and LIBERO-Plus benchmarks, while simultaneously improving future-state predictions. Furthermore, SA-WAM outperforms strong baselines in real-world evaluation using a UR5 robotic arm, with strong gains in randomized environments. We analyze the correlation between world model prediction quality and rollout success, providing insights into WAM performance and avenues for its improvement.
Chinese Translation
世界行动模型（WAMs）利用大规模预训练视频扩散模型的能力，联合预测未来观察和动作，继承了来自互联网规模视频的丰富视觉和物理先验。这使得它们成为机器人策略学习的有希望的范式，然而，现有模型仅在RGB观察上运行，并未利用3D信息。为了解决这一问题，我们提出了一种空间感知世界行动模型（SA-WAM），该模型重新利用了预训练的视频模型，用于联合动作、RGB和深度预测，从而在单一扩散骨干网络中实现3D感知的世界建模和动作预测。我们使用了一种非线性编码，将无界深度信号映射到冻结的变分自编码器（VAE）标记器所期望的有界输入域。这使我们能够在不进行3D特定微调的情况下重用标记器，结合几何信息而不牺牲预训练的先验。SA-WAM在RoboCasa和LIBERO-Plus基准测试中取得了最先进的结果，同时改善了未来状态的预测。此外，SA-WAM在使用UR5机器人臂的真实世界评估中超越了强基线，在随机化环境中取得了显著提升。我们分析了世界模型预测质量与回滚成功之间的相关性，为WAM的性能及其改进提供了见解。
cs.CV / 88 / 2609.02556
RGB-to-IR image translation for infrared vehicle detection in unseen UAV domains
RGB到红外图像转换在未知无人机领域中的红外车辆检测
Abstract
Synthetic training data is crucial for developing vision AI when real-world data is scarce, as in thermal infrared (IR) aerial vehicle detection. While abundant UAV RGB imagery motivates RGB-to-IR translation for data augmentation, unobservable thermal traits (e.g., engine heat) make learning transferable mappings challenging. This work investigates whether modern generative translators can overcome this cross-modal gap to improve infrared vehicle detection on unseen UAV target domains. Translators are trained on paired RGB-IR source datasets and applied to RGB training images from held-out target datasets to generate synthetic IR data. Evaluated methods include supervised GANs, ControlNet-based diffusion models, and foundation-model editing via LoRA. The resulting synthetic IR imagery is used to train RF-DETR vehicle detectors, which are evaluated on unseen IR target test splits across five aerial datasets, with Kust4K and VTUAV serving as target domains. Synthetic IR consistently outperforms RGB and grayscale baselines. Stable Diffusion 3.5 with ControlNet yields the best results, improving mAP from 50.8 to 60.1 on Kust4K and from 25.6 to 38.4 on VTUAV compared to models trained only on source-domain IR data. Increasing output diversity via multiple seeds (+1.1 mAP) and prompt variations (+3.3 mAP) provides additional gains on VTUAV. Although a performance gap to real target IR data remains, generative RGB-to-IR translation effectively mitigates IR data scarcity and improves cross-domain aerial vehicle detection.
Chinese Translation
在热红外（IR）空中车辆检测中，当现实世界数据稀缺时，合成训练数据对开发视觉人工智能至关重要。虽然丰富的无人机RGB图像激励了RGB到IR的转换以进行数据增强，但不可观察的热特征（例如，发动机热量）使得学习可转移的映射变得具有挑战性。本研究探讨了现代生成翻译模型是否能够克服这种跨模态差距，从而提高在未知无人机目标领域中的红外车辆检测。翻译模型在配对的RGB-IR源数据集上进行训练，并应用于来自保留目标数据集的RGB训练图像，以生成合成的IR数据。评估的方法包括监督生成对抗网络（GANs）、基于ControlNet的扩散模型以及通过LoRA进行的基础模型编辑。生成的合成IR图像用于训练RF-DETR车辆检测器，并在五个空中数据集的未知IR目标测试集上进行评估，其中Kust4K和VTUAV作为目标领域。合成IR的表现始终优于RGB和灰度基线。使用ControlNet的稳定扩散3.5模型取得了最佳结果，与仅在源领域IR数据上训练的模型相比，Kust4K的mAP从50.8提高到60.1，VTUAV的mAP从25.6提高到38.4。通过多个种子增加输出多样性（+1.1 mAP）和提示变体（+3.3 mAP）在VTUAV上提供了额外的增益。尽管与真实目标IR数据之间仍存在性能差距，但生成的RGB到IR转换有效缓解了IR数据稀缺问题，并改善了跨领域空中车辆检测。
cs.CV / 89 / 2609.02560
Stereo 4D Radar for 3D Object Detection: Integrating Geometric Alignment and Absolute Velocity Estimation
立体4D雷达用于3D物体检测：几何对齐与绝对速度估计的结合
Abstract
Four-dimensional (4D) Radar is a powerful sensing modality capable of detecting surrounding three-dimensional (3D) objects under diverse weather conditions and providing Doppler-based motion information. However, raw 4D Radar signals contain significant clutter from road surfaces, guardrails, and surrounding vehicles, along with multipath-induced ghost reflections and the receiver's inherent noise floor. Consequently, preprocessing algorithms designed to remove such invalid measurements often make the Radar data excessively sparse. Moreover, the Doppler measurements provided by 4D Radar describe only the radial component of an object's velocity, limiting their ability to recover the full motion state. In this paper, we introduce a stereo 4D Radar-based 3D object detection framework that exploits the geometric disparity between left and right Radars to estimate the absolute velocity of objects and achieve more robust perception through the fusion of their complementary features. The effectiveness of the proposed framework is validated on our in-house stereo 4D Radar dataset, demonstrating performance gains of 8.82 points in AP 3D and 9.0 points in AP BEV over state-of-the-art mono 4D Radar baselines. These results demonstrate that absolute velocity estimation combined with stereo geometry-aware feature fusion leads to substantial improvements in 3D object detection.
Chinese Translation
四维（4D）雷达是一种强大的传感方式，能够在多种天气条件下检测周围的三维（3D）物体，并提供基于多普勒的运动信息。然而，原始的4D雷达信号包含来自路面、护栏和周围车辆的显著杂波，以及由于多路径效应引起的虚假反射和接收器固有的噪声底。因此，旨在去除这些无效测量的预处理算法往往使雷达数据过于稀疏。此外，4D雷达提供的多普勒测量仅描述物体速度的径向分量，限制了其恢复完整运动状态的能力。在本文中，我们提出了一种基于立体4D雷达的3D物体检测框架，该框架利用左右雷达之间的几何差异来估计物体的绝对速度，并通过融合其互补特征实现更强大的感知。我们在自建的立体4D雷达数据集上验证了所提出框架的有效性，结果显示在AP 3D上性能提升了8.82分，在AP BEV上提升了9.0分，相较于最先进的单通道4D雷达基线。这些结果表明，结合绝对速度估计与立体几何感知特征融合能够显著提升3D物体检测的性能。
cs.CV / 90 / 2609.02565
MARS: What Retrieval Signals Are Hidden in Multimodal Large Language Models for Text-Video Retrieval?
MARS：多模态大语言模型中隐藏的检索信号在文本-视频检索中的应用
Abstract
Text-video retrieval requires representations that can distinguish videos with similar scenes, actions, and temporal patterns. Recent multimodal large language models have been adapted as embedding models, but they often represent each input using a single token from the final layer. This can compress diverse video-text cues into a single vector and limit fine-grained retrieval. To address this limitation, we propose MARS, a multi-layer and multi-slot embedding framework for text-video retrieval. MARS constructs multiple adaptive representation slots by combining hidden states from different decoder layers, compares corresponding text and video slots, and aggregates their similarities for retrieval. To better handle confusing candidates, we further introduce a hard-negative-aware slot specialization objective that encourages the slots to capture discriminative matching cues. Experiments on four text-video retrieval benchmarks show that MARS achieves state-of-the-art results in both direct similarity-based retrieval and reranking settings. Ablation studies and analyses demonstrate that multi-layer fusion, multiple slots, and hard-negative-aware slot specialization provide complementary gains. Code is available at https://github.com/sejong-rcv/MARS.
Chinese Translation
文本-视频检索需要能够区分具有相似场景、动作和时间模式的视频的表示。近期的多模态大语言模型已被调整为嵌入模型，但它们通常使用最终层的单个标记来表示每个输入。这可能会将多样的视频-文本线索压缩为单一向量，从而限制细粒度检索。为了解决这一局限性，我们提出了MARS，一种用于文本-视频检索的多层多槽嵌入框架。MARS通过结合来自不同解码器层的隐藏状态构建多个自适应表示槽，比较相应的文本和视频槽，并聚合它们的相似性以进行检索。为了更好地处理混淆候选项，我们进一步引入了一种关注困难负样本的槽专业化目标，鼓励槽捕捉区分性的匹配线索。在四个文本-视频检索基准上的实验表明，MARS在直接基于相似性的检索和重排序设置中均达到了最先进的结果。消融研究和分析表明，多层融合、多槽和关注困难负样本的槽专业化提供了互补的增益。代码可在 https://github.com/sejong-rcv/MARS 获取。
cs.CV / 91 / 2609.02573
Deeply Interleaved Text-Image Contexts for Multimodal LLMs Assessment
深度交错的文本-图像上下文用于多模态大语言模型评估
Abstract
Current evaluations and training of multimodal models predominantly focus on multi-image tasks, largely overlooking interleaved text-image scenarios. In such multi-image tasks, text typically serves merely as task instructions, lacking deep semantic interaction with the visual content. In contrast, realworld applications like text-image co-creation, character tracking, and spatial reconstruction require constant interaction between text and images. Consequently, models must possess a deep understanding of these interleaved contexts. To bridge this gap, we introduce a novel benchmark, TIC-Bench (deeply interleaved Text-Image Contexts), designed to evaluate the capability of models to integrate text-image clues and recover the ground truth facts within deeply interleaved contexts. This benchmark encompasses three core domains: Logical, Temporal, and Spatial Association, which are further categorized into eight specific types, comprising a total of 2,280 questions. We evaluated 10 state-of-the-art MLLMs and observed a substantial performance gap compared to human experts, together with persistent difficulties in integrating evidence distributed across interleaved visual and textual inputs. Ultimately, this benchmark provides a valuable analytical tool for assessing and advancing the ability of multimodal models to effectively integrate text and image information in deeply interleaved contexts. TIC-Bench is publicly available at https://huggingface.co/datasets/pino10010/TIC-Bench
Chinese Translation
当前对多模态模型的评估和训练主要集中在多图像任务上，往往忽视了交错的文本-图像场景。在这些多图像任务中，文本通常仅作为任务指令，缺乏与视觉内容的深层语义互动。相比之下，现实应用如文本-图像共同创作、角色追踪和空间重建等需要文本与图像之间的持续互动。因此，模型必须具备对这些交错上下文的深刻理解。为了解决这一问题，我们引入了一个新基准，TIC-Bench（深度交错的文本-图像上下文），旨在评估模型整合文本-图像线索并在深度交错上下文中恢复真实事实的能力。该基准涵盖三个核心领域：逻辑关联、时间关联和空间关联，进一步细分为八种具体类型，共包含2,280个问题。我们评估了10个最先进的多模态大语言模型（MLLMs），观察到与人类专家相比存在显著的性能差距，同时在整合分布于交错视觉和文本输入中的证据方面也面临持续困难。最终，该基准为评估和提升多模态模型在深度交错上下文中有效整合文本和图像信息的能力提供了一个有价值的分析工具。TIC-Bench已在 https://huggingface.co/datasets/pino10010/TIC-Bench 上公开发布。
cs.CV / 92 / 2609.02600
Generalizable Brain Tumor Segmentation with Self-Training and Tumor-Aware Deformations
具有自我训练和肿瘤感知变形的可推广脑肿瘤分割
Abstract
This work presents an approach to the Generalizability Across Tumors (BraTS-GoAT) task of the BraTS 2026 Challenge, which focuses on robust segmentation of brain tumor sub-regions across a heterogeneous patient population. The proposed method employs the nnU-Net framework with a large residual encoder architecture, integrating a semi-supervised learning technique with pseudo-labels generated from the unlabeled training data and a tumor-aware deformable augmentation that locally deforms the lesion while preserving the surrounding anatomy. We evaluate the individual contributions of each component, as well as their combination, using varying proportions of the most confident pseudo-labeled cases. The submitted configuration for the generalization task achieves Dice and NSD scores of 0.881 and 0.473 for Whole Tumor, 0.817 and 0.490 for Tumor Core, and 0.775 and 0.533 for Enhancing Tumor on the BraTS-GoAT validation set, improving over the labeled-only baselines across all tumor regions and confirming that self-training and the proposed augmentation are complementary. Our source code is publicly available at https://github.com/Henrique-zan/brats-goat-2026/.
Chinese Translation
本研究提出了一种针对BraTS 2026挑战中肿瘤间可推广性（Generalizability Across Tumors, BraTS-GoAT）任务的方法，旨在对异质患者群体中的脑肿瘤亚区域进行稳健分割。所提出的方法采用nnU-Net框架，结合大型残差编码器架构，整合了一种半监督学习技术，利用从未标记的训练数据生成的伪标签，以及一种肿瘤感知的可变形增强技术，该技术在保持周围解剖结构的同时局部变形病变。我们评估了每个组件的单独贡献以及它们的组合，使用不同比例的最可信伪标签案例。提交的针对可推广性任务的配置在BraTS-GoAT验证集上实现了整体肿瘤的Dice和NSD分数分别为0.881和0.473，肿瘤核心为0.817和0.490，增强肿瘤为0.775和0.533，超越了所有肿瘤区域的仅标记基线，并确认自我训练和所提出的增强是互补的。我们的源代码已公开，地址为https://github.com/Henrique-zan/brats-goat-2026/。
cs.CV / 93 / 2609.02616
AffectDelta: Beyond Emotion Labels for Image Editing
AffectDelta：超越情感标签的图像编辑
Abstract
Emotion-driven image editing aims to evoke a specified target emotion by modifying emotion-relevant visual cues in a source image, while preserving the overall composition and semantic-structural coherence of the original scene. Existing scene-level editors typically specify the target with a single emotion category and often learn visual transformations from operation-level text instructions. A category collapses a mixed affective endpoint into one dominant label, while language cannot precisely quantify how coexisting emotions should increase, decrease, or remain stable. We introduce AffectDelta, a source-aware editor that treats editing as a transition between eight-dimensional emotion distributions. A frozen Emotion Distribution Predictor estimates the source state, and the signed source-to-target difference encodes the direction and magnitude of the requested transition. Within AffectDelta, an internal transition encoder and a source-aware diffusion backbone jointly translate this signal into context-dependent semantic and appearance changes. To train this formulation, we construct AffectPair-249K, comprising 248,841 source-target pairs with predicted eight-dimensional distributions and spanning both cross-category and within-category transitions. Experiments against six baselines, combining quantitative evaluation with qualitative comparisons, demonstrate improved affective alignment and content preservation, while ablations validate our design choices. Code and dataset will be made publicly available upon acceptance.
Chinese Translation
情感驱动的图像编辑旨在通过修改源图像中与情感相关的视觉线索来唤起特定的目标情感，同时保持原始场景的整体构图和语义结构的一致性。现有的场景级编辑器通常通过单一情感类别来指定目标，并且通常从操作级文本指令中学习视觉变换。一种类别将混合的情感终点压缩为一个主导标签，而语言无法精确量化共存情感的增加、减少或保持稳定的方式。我们提出了AffectDelta，这是一种源感知编辑器，将编辑视为在八维情感分布之间的过渡。一个冻结的情感分布预测器估计源状态，而有符号的源到目标差异编码了请求的过渡的方向和幅度。在AffectDelta中，一个内部过渡编码器和一个源感知扩散骨干共同将该信号转化为上下文相关的语义和外观变化。为了训练这一模型，我们构建了AffectPair-249K，包含248,841个具有预测的八维分布的源-目标对，涵盖跨类别和类别内的过渡。与六个基线的实验结合定量评估和定性比较，展示了情感对齐和内容保留的改善，同时消融实验验证了我们的设计选择。代码和数据集将在接受后公开发布。
cs.CV / 94 / 2609.02640
From Detection to Localization: A Unified Forensics Framework for Fully Synthetic and Tampered Images
从检测到定位：一个统一的全合成和篡改图像取证框架
Abstract
The rapid advancement of generative models has significantly worsened the problem of manipulated image detection, as these methods are capable of producing highly realistic forgeries, reinforcing the importance of multimedia forensics. Conventional approaches typically frame image manipulation detection as a binary classification task (real vs. generated), which limits the capability to distinguish and localize different forms of manipulation. To address these constraints, this work extends an existing detector by introducing a unified multiclass framework (real vs. fully generated vs. tampered). In addition to classifying image authenticity, the framework incorporates a segmentation branch to enable pixel-level localization of tampered regions. The proposed approach outperforms selected recent benchmarks, offering an efficient solution with improved classification accuracy and higher IoU scores for the localization task. Find the code at https://github.com/anngal01/From-Detection-to-Localization-A-Unified-Forensics-Framework-for-Fully-Synthetic-and-Tampered-Images.
Chinese Translation
生成模型的快速发展显著加剧了篡改图像检测的问题，因为这些方法能够生成高度逼真的伪造图像，这进一步强调了多媒体取证的重要性。传统方法通常将图像篡改检测框定为二分类任务（真实与生成），这限制了区分和定位不同形式篡改的能力。为了解决这些限制，本研究通过引入一个统一的多类框架（真实与完全生成与篡改）扩展了现有的检测器。除了对图像真实性进行分类外，该框架还结合了一个分割分支，以实现对篡改区域的像素级定位。所提出的方法在选定的最新基准测试中表现优越，提供了一种高效的解决方案，具有更高的分类准确性和更高的IoU（交并比）得分用于定位任务。代码可在 https://github.com/anngal01/From-Detection-to-Localization-A-Unified-Forensics-Framework-for-Fully-Synthetic-and-Tampered-Images 找到。
cs.CV / 95 / 2609.02644
Learning to Attract and Repel: Dual Quality Margin Learning for Face Recognition (DQM-Face)
学习吸引与排斥：面部识别的双质量边际学习 (DQM-Face)
Abstract
Face recognition in unconstrained environments remains highly challenging due to diverse and extreme variations encountered in real-world scenarios. To mitigate these effects, existing margin-based approaches model sample quality through feature magnitude. However, magnitude-based modeling alone is susceptible to identity-agnostic noise, which can degrade the reliability and discriminative power of learned representations. In this paper, we propose Dual Quality Margin Learning for Face Recognition (DQM-Face), a novel framework that enables refined attraction and repulsion dynamics during representation learning. Our approach unifies conventional magnitude-based quality estimation with a newly introduced semantic quality learning mechanism, realized via squeeze-and-excitation semantic attention. By jointly leveraging magnitude and semantic cues, we construct enhanced quality-aware margins that adaptively strengthen intra-class compactness through improved attraction during learning. To further enhance inter-class discrimination, we introduce a repulsion margin formulation that explicitly enlarges inter-class separation. The unified integration of semantic quality modeling with dual attraction-repulsion margin optimization results in a more structured and discriminative feature geometry. Extensive experiments on multiple challenging benchmarks demonstrate that DQM-Face consistently outperforms state-of-the-art face recognition methods. Moreover, we show that the quality learned for margin optimization is highly effective for face image quality assessment within the proposed framework, demonstrating that the learned quality signal is intrinsically aligned with the recognition objective. The code is publicly available: https://github.com/RAIB-group/DQM-Face
Chinese Translation
在非约束环境下的面部识别仍然面临着巨大的挑战，因为在现实场景中会遇到多样化和极端的变化。为了减轻这些影响，现有的基于边际的方法通过特征幅度来建模样本质量。然而，仅仅依赖幅度建模容易受到与身份无关的噪声的影响，这可能会降低学习表示的可靠性和区分能力。本文提出了面部识别的双质量边际学习 (DQM-Face)，这是一个新颖的框架，能够在表示学习过程中实现精细的吸引和排斥动态。我们的方法将传统的基于幅度的质量估计与新引入的语义质量学习机制相结合，通过挤压与激励语义注意力实现。通过联合利用幅度和语义线索，我们构建了增强的质量感知边际，这些边际在学习过程中通过改进的吸引力自适应地增强了类内紧凑性。为了进一步增强类间区分性，我们引入了一种排斥边际公式，明确扩大类间分离。语义质量建模与双重吸引-排斥边际优化的统一整合，导致了更结构化和具有区分性的特征几何。我们在多个具有挑战性的基准上的广泛实验表明，DQM-Face始终优于最先进的面部识别方法。此外，我们还展示了在所提出的框架内，边际优化所学习的质量对于面部图像质量评估是非常有效的，证明了学习到的质量信号与识别目标本质上是一致的。代码已公开可用： https://github.com/RAIB-group/DQM-Face
cs.CV / 96 / 2609.02654
Physics-Driven Independent Pair Generation for Iterative Self-Supervised Low-Dose CT Denoising
基于物理驱动的独立对生成用于迭代自监督低剂量CT去噪
Abstract
Low-dose computed tomography (LDCT) measurements contain mixed Poisson-Gaussian noise. However, most self-supervised methods rely on generic image statistics and do not explicitly model this noise, which may limit their ability to effectively suppress realistic LDCT noise. To address this issue, we propose a physics-driven framework with cross-domain iteration for self-supervised LDCT denoising. The proposed framework proceeds in three main steps. First, a learned sinogram prior and the LDCT noise model guide posterior inference of photon counts, enabling separation of the Poisson and Gaussian components. Second, the separated Poisson and Gaussian components are respectively processed by binomial thinning and Gaussian data thinning to construct two branches, and residual scaling matches each branch's noise level to that of the observation, yielding a training pair with approximately independent noise realizations from one low-dose measurement. Finally, the pair is used to train an image-domain network whose forward-projected outputs update the prior. Through cross-domain iteration, the prior and the training pair are progressively refined while maintaining consistency with CT acquisition physics. Experiments on simulated data from AAPM, LIDC-IDRI, and LoDoPaB-CT and on real LDCT data show consistent gains over the evaluated self-supervised baselines across dose levels, with performance comparable to the evaluated supervised baseline.
Chinese Translation
低剂量计算机断层扫描（LDCT）测量包含混合的泊松-高斯噪声。然而，大多数自监督方法依赖于通用图像统计，并未明确建模这种噪声，这可能限制其有效抑制真实LDCT噪声的能力。为了解决这一问题，我们提出了一种基于物理驱动的框架，结合跨域迭代用于自监督LDCT去噪。该框架主要分为三个步骤。首先，学习的正弦图先验和LDCT噪声模型指导光子计数的后验推断，从而实现泊松和高斯成分的分离。其次，分别对分离出的泊松和高斯成分进行二项稀疏处理和高斯数据稀疏处理，以构建两个分支，并通过残差缩放将每个分支的噪声水平匹配到观测值的噪声水平，从而生成一个具有大致独立噪声实现的训练对。最后，利用该对训练一个图像域网络，其前向投影输出更新先验。通过跨域迭代，先验和训练对在保持与CT采集物理一致性的同时逐步得到精炼。在AAPM、LIDC-IDRI和LoDoPaB-CT的模拟数据以及真实LDCT数据上的实验显示，在不同剂量水平下，相较于评估的自监督基线，均取得了一致的性能提升，且性能与评估的监督基线相当。
cs.CV / 97 / 2609.02663
Characterizing Text Branch Sensitivity in Medical Vision-Language Segmentation via Evidence Decoupling
通过证据解耦表征医学视觉-语言分割中的文本分支敏感性
Abstract
Pretrained vision-language models (VLMs) have shown promising performance in medical image segmentation by incorporating clinical text. However, it remains unclear how much textual information actually contributes to pixel-level predictions. In this work, we systematically investigate the role of text in multimodal medical image segmentation. We first analyze several commonly used fusion strategies and find that segmentation performance is largely insensitive to the choice of fusion module. To further understand modality interactions, we propose an Evidence Decoupling Decoder (EDD) based on evidential deep learning and deep supervision. EDD serves as an internal representation analysis tool that decomposes image evidence and text-modulated evidence throughout the decoding process while maintaining competitive segmentation performance. Experimental results show that the sensitivity to text perturbation varies substantially across datasets. On BUSI and BTMRI, removing text causes catastrophic performance drops, indicating strong model reliance on textual input. On ISIC and Kvasir-SEG, text exerts relatively marginal influence. We further find that text affects predictions mainly through global semantic modulation rather than independent spatial localization, and that the specific semantic components driving text sensitivity differ across datasets. These findings provide a deeper understanding of modality interaction in multimodal medical image segmentation and offer practical insights for future model design.
Chinese Translation
预训练的视觉-语言模型（VLMs）在医学图像分割中通过结合临床文本展现出良好的性能。然而，文本信息在像素级预测中实际贡献了多少仍不清楚。在本研究中，我们系统地调查了文本在多模态医学图像分割中的作用。我们首先分析了几种常用的融合策略，发现分割性能对融合模块的选择基本不敏感。为了进一步理解模态间的交互，我们提出了一种基于证据深度学习和深度监督的证据解耦解码器（EDD）。EDD作为一种内部表征分析工具，在解码过程中分解图像证据和文本调制证据，同时保持竞争性的分割性能。实验结果表明，文本扰动的敏感性在不同数据集之间差异显著。在BUSI和BTMRI数据集中，去除文本导致性能急剧下降，表明模型对文本输入的强依赖性。在ISIC和Kvasir-SEG数据集中，文本的影响相对较小。我们进一步发现，文本主要通过全局语义调制而非独立的空间定位影响预测，并且驱动文本敏感性的特定语义成分在不同数据集之间存在差异。这些发现为多模态医学图像分割中的模态交互提供了更深入的理解，并为未来模型设计提供了实际的见解。
cs.CV / 98 / 2609.02664
Query Rewriting for Complex Object Segmentation in 4D Gaussian Representations
基于4D高斯表示的复杂对象分割查询重写
Abstract
Recent 4D Gaussian representation frameworks have demonstrated strong performance in language-guided dynamic scene understanding. However, these methods remain highly sensitive to verbose and narrative-style queries that contain noisy contextual information. In this paper, we investigate the impact of query rewriting for complex object segmentation in 4D Gaussian representations. Inspired by recent findings in retrieval-augmented language models and keyword-guided query reformulation, we propose a training-free reinterpretation strategy that transforms long descriptive queries into concise keyword-grounded forms. Our approach progressively reduces linguistic noise while preserving semantic anchors relevant to object-centric representations. Experiments on HyperNeRF and Neu3D demonstrate that concise rewritten queries significantly improve both temporal localization and spatial segmentation performance. In particular, our method improves average temporal accuracy from 60.92% to 92.21% and average vIoU from 20.08% to 76.94% without any additional fine-tuning. Extensive ablation studies further reveal that shorter, keyword-focused queries consistently yield stable video-feature similarity distributions and better alignment with object-centric Gaussian representations
Chinese Translation
近期的4D高斯表示框架在语言引导的动态场景理解中表现出色。然而，这些方法对包含嘈杂上下文信息的冗长叙述式查询仍然高度敏感。本文研究了查询重写对4D高斯表示中复杂对象分割的影响。受近期检索增强语言模型和关键词引导查询重构研究的启发，我们提出了一种无训练的重解释策略，将冗长的描述性查询转换为简洁的关键词基础形式。我们的方法逐步减少语言噪声，同时保留与对象中心表示相关的语义锚点。在HyperNeRF和Neu3D上的实验表明，简洁的重写查询显著提高了时间定位和空间分割的性能。特别是，我们的方法在没有任何额外微调的情况下，将平均时间准确率从60.92%提高到92.21%，平均vIoU从20.08%提高到76.94%。大量的消融研究进一步表明，较短的关键词聚焦查询始终产生稳定的视频特征相似性分布，并与对象中心的高斯表示更好地对齐。
cs.CV / 99 / 2609.02683
Genesis: A Generative Engine for Hierarchical Satellite Image Synthesis
Genesis：一种用于层次卫星图像合成的生成引擎
Abstract
Earth observation is fundamentally multi-scale; geospatial tasks span varied resolutions, and satellite imagery is organized into cascading tile pyramids that nest fine detail within wide coverage. Current generative models of satellite imagery, however, operate along a single axis: they either zoom to enhance a single tile's resolution or pan to extend imagery at a fixed scale. As a result, no existing method produces a complete pyramid that stays consistent across both scale and space, where a high-zoom tile must agree with the coarse context it refines and with the neighbors it meets. Motivated by this gap, we introduce a new task, multi-scale tile completion: given a sparse set of seed tiles at arbitrary zoom levels and positions, synthesize a complete, uniform quadtree that is globally consistent across both scale and space. We approach this task with Genesis, a generative engine that brings both axes together by composing two specialized operators over the quadtree: a vertical super-resolution model and a horizontal mask-based outpainting model, producing pyramids that are consistent across zoom levels and seamless across neighboring tiles. Each operator achieves state-of-the-art results on its subtask, and the engine propagates sparse seeds into seamless, multi-resolution maps from any initial configuration. To evaluate the task and benchmark Genesis, we introduce dense500, a fully observed multi-scale pyramid dataset spanning diverse geographic regions, together with a suite of pyramid-level metrics. Code, models, and our dataset are available at https://github.com/mvrl/genesis.
Chinese Translation
地球观测本质上是多尺度的；地理空间任务跨越不同的分辨率，卫星图像被组织成级联的图块金字塔，其中细节嵌套在广泛的覆盖范围内。然而，目前的卫星图像生成模型仅沿单一轴线操作：它们要么放大以增强单个图块的分辨率，要么平移以在固定尺度下扩展图像。因此，现有方法无法生成在尺度和空间上都保持一致的完整金字塔，其中高放大图块必须与其所细化的粗略上下文以及相邻图块保持一致。基于这一缺口，我们提出了一项新任务：多尺度图块补全；给定一组稀疏的种子图块，位于任意的放大级别和位置，合成一个在尺度和空间上全局一致的完整均匀四叉树。我们通过Genesis这一生成引擎来解决这一任务，它通过在四叉树上组合两个专门的操作符，将两个轴线结合在一起：一个垂直超分辨率模型和一个基于掩模的水平外绘模型，从而生成在放大级别上保持一致且在相邻图块之间无缝的金字塔。每个操作符在其子任务上都达到了最先进的结果，且该引擎能够将稀疏种子传播成无缝的多分辨率地图，适用于任何初始配置。为了评估这一任务并基准测试Genesis，我们引入了dense500，一个完全观察到的多尺度金字塔数据集，涵盖多样的地理区域，并提供一套金字塔级别的指标。代码、模型和我们的数据集可在 https://github.com/mvrl/genesis 获取。
cs.CV / 100 / 2609.02689
GaLe: memory-efficient Global Approximate and Local Exact features
GaLe：内存高效的全局近似与局部精确特征
Abstract
Embedded devices typically lack the resources of GPU-equipped machines, and existing inference methods suffer from either high computational overhead (patch-based) or accuracy loss (approximation-based). We propose GaLe, a memory-efficient technique that enables the deployment of pretrained networks on constrained devices without retraining. GaLe partitions feature maps into two components: a local exact (Le) representation that preserves fine details and a global approximate (Ga) representation that retains long-range dependencies. Unlike standard tiling, GaLe supports global operations and attention mechanisms found in hybrid CNN-transformer models. Validated on ImageNet, our method matches exact-inference performance while achieving up to 65% speedup and 90% RAM reduction on a Cortex-M33 compared to patch-based inference. We further demonstrate GaLe's versatility across classification, detection, and generation tasks, highlighting its potential as a foundation for resource-efficient architecture design.
Chinese Translation
嵌入式设备通常缺乏配备GPU的机器所需的资源，而现有的推理方法要么面临高计算开销（基于补丁），要么导致准确性损失（基于近似）。我们提出了GaLe，这是一种内存高效的技术，使得在受限设备上部署预训练网络成为可能，而无需重新训练。GaLe将特征图划分为两个组成部分：局部精确（Le）表示，保留细节，以及全局近似（Ga）表示，保留长程依赖。与标准的切片方法不同，GaLe支持在混合CNN-Transformer模型中发现的全局操作和注意力机制。在ImageNet上进行验证，我们的方法在精确推理性能上表现相当，同时在Cortex-M33上实现了高达65%的加速和90%的RAM减少，相较于基于补丁的推理。我们进一步展示了GaLe在分类、检测和生成任务中的多样性，突显了其作为资源高效架构设计基础的潜力。
cs.CV / 101 / 2609.02697
Generating Medical Image Counterfactuals using Causal Explanations
利用因果解释生成医学图像反事实
Abstract
Deep learning models have achieved impressive performance in medical image diagnosis, yet their deployment in clinical settings remains constrained by limited explainability. Counterfactual images provide one means of auditing model behavior by showing how an image would need to change for a classifier to produce a different prediction. Existing approaches typically generate such explanations using auxiliary models, including generative adversarial networks and diffusion models. While often capable of producing visually realistic images, these methods explain one black-box model using another, making it difficult to separate the classifier's decision-making process from the inductive biases of the generator. We propose a novel counterfactual-generation framework that requires no generative model. Instead, counterfactuals are constructed directly from causal evidence extracted from the classifier. The resulting approach is deterministic, requires no additional model training, and enables controllable edits within user-specified regions of interest. Experiments on real-world medical imaging datasets demonstrate that the proposed method successfully changes classifier predictions while remaining closer to the original image than generative baselines, providing a more direct and transparent view of the classifier's decision boundary.
Chinese Translation
深度学习模型在医学图像诊断中取得了令人瞩目的性能，但其在临床环境中的应用仍受到有限可解释性的制约。反事实图像提供了一种审计模型行为的方法，通过展示图像需要如何变化，以使分类器产生不同的预测。现有的方法通常使用辅助模型生成此类解释，包括生成对抗网络（GANs）和扩散模型。尽管这些方法通常能够生成视觉上逼真的图像，但它们是用另一个黑箱模型来解释一个模型，使得难以将分类器的决策过程与生成器的归纳偏差分开。我们提出了一种新颖的反事实生成框架，该框架不需要生成模型。相反，反事实是直接从分类器提取的因果证据中构建的。所提出的方法是确定性的，不需要额外的模型训练，并且能够在用户指定的感兴趣区域内进行可控编辑。在真实世界医学影像数据集上的实验表明，所提出的方法成功地改变了分类器的预测，同时比生成基线更接近原始图像，提供了分类器决策边界的更直接和透明的视角。
cs.CV / 102 / 2609.02705
A Top-Down Framework for Metric-Scale Athlete Localization from Single Broadcast Frames
基于单帧广播图像的度量尺度运动员定位的自上而下框架
Abstract
Accurate world-coordinate localization of athletes from single-frame broadcast footage is inherently challenging due to extreme scale disparities in ultra-high-resolution imagery. In this paper, we propose a top-down framework for metric-scale athlete localization from a single calibrated frame. Our approach centers on three key contributions. First, we propose Boundary-Aware Adaptive Tiling, a semantics-guided extension of standard sliced inference. By iteratively expanding tile boundaries based on coarse bounding-box predictions, it systematically ensures full object containment, effectively mitigating boundary-splitting artifacts through a lightweight pipeline adaptation without architectural modifications. By substantially mitigating recall degradation under extreme scale variance, Boundary-Aware Adaptive Tiling enables us to isolate perspective distortion as the primary source of residual localization error. Second, we adapt the RTMPose-X architecture into a specialized two-keypoint estimator (pelvis and ground projection), employing a reformulated Gated Attention Unit optimized for this geometrically coupled point pair, and then deterministically lift the 2D ground projections into world coordinates via camera-calibrated ray casting. On the public test set, our method achieves a LocSim score of 97.44 and an mAP of 0.9128, outperforming the baseline by over 21 \% and establishing a robust solution for high-resolution scale variance.
Chinese Translation
从单帧广播视频中准确定位运动员的世界坐标本质上具有挑战性，因为超高分辨率图像中存在极端的尺度差异。本文提出了一种基于单个校准帧的度量尺度运动员定位的自上而下框架。我们的方法集中在三个关键贡献上。首先，我们提出了边界感知自适应切片（Boundary-Aware Adaptive Tiling），这是一种基于语义指导的标准切片推理的扩展。通过基于粗略的边界框预测迭代扩展切片边界，它系统性地确保了完整的对象包含，有效减轻了边界分裂伪影，通过轻量级的管道适配而无需架构修改。通过显著减轻在极端尺度变化下的召回率下降，边界感知自适应切片使我们能够将透视失真隔离为残余定位误差的主要来源。其次，我们将RTMPose-X架构调整为一个专门的两关键点估计器（骨盆和地面投影），采用针对这一几何耦合点对优化的重新构造的门控注意单元（Gated Attention Unit），然后通过相机校准的光线投射将2D地面投影确定性地提升到世界坐标。在公共测试集上，我们的方法实现了97.44的LocSim得分和0.9128的mAP，超过基线21%以上，为高分辨率尺度变化建立了一个稳健的解决方案。
cs.CV / 103 / 2609.02717
MV-dVRK: A Multi-Viewpoint Benchmark for Spatial Surgical Perception
MV-dVRK：一个用于空间外科感知的多视角基准
Abstract
Large-scale training and refined optimization techniques have greatly improved sparse multi-view 3D reconstruction. Despite their relevance to surgery, such methods have never before been rigorously evaluated on real endoscopic images. Current clinical telerobots deploy a single stereo camera inside the patient, making multi-viewpoint data extremely rare. This paper presents MV-dVRK, the first ex-vivo surgical dataset to combine multiple exposure-synchronized stereo viewpoints with accurate surface geometry and camera poses. The static subset of the benchmark provides dense SfM reference geometry, validated against an industrial 3D scanner, together with ground-truth camera poses and sparse-view test sets. We use MV-dVRK to systematically compare zero-shot monocular, stereo, multi-stereo, and multi-view 3D reconstruction methods as the number of viewpoints increases. With two endoscopes, multi-stereo reconstruction achieves the highest coverage. With a third viewpoint, optimization-based multi-view methods perform best, covering 67% of ground-truth surface points within a 1 mm tolerance and recovering highly accurate relative camera poses. By contrast, feed-forward foundation models cover only 43% of the ground-truth surface in the same setting. MV-dVRK also includes ten dynamic sequences spanning multiple surgical tasks, with increasing kinematic complexity and tissue deformation, providing a basis for future research in multi-viewpoint surgical perception. The project is available at: https://mv-dvrk.is.mpg.de.
Chinese Translation
大规模训练和精细优化技术极大地改善了稀疏多视角三维重建。尽管这些方法与外科手术相关，但迄今为止从未在真实的内窥镜图像上进行过严格评估。目前的临床遥控机器人在患者体内部署了单个立体摄像头，使得多视角数据极为稀缺。本文提出了MV-dVRK，这是第一个结合多个曝光同步立体视角、准确表面几何和相机位姿的体外外科数据集。基准的静态子集提供了与工业三维扫描仪验证的稠密结构从运动（SfM）参考几何，以及真实相机位姿和稀疏视图测试集。我们使用MV-dVRK系统地比较了零样本单目、立体、多立体和多视角三维重建方法，随着视角数量的增加。使用两个内窥镜时，多立体重建实现了最高的覆盖率。增加第三个视角后，基于优化的多视角方法表现最佳，在1毫米的容差范围内覆盖了67%的真实表面点，并恢复了高度准确的相对相机位姿。相比之下，前馈基础模型在相同设置下仅覆盖了43%的真实表面。MV-dVRK还包括十个动态序列，涵盖多个外科任务，具有不断增加的运动复杂性和组织变形，为未来多视角外科感知研究提供了基础。该项目可在以下网址获取：https://mv-dvrk.is.mpg.de。
cs.CV / 104 / 2609.02731
RVSD: Retrieval Vision Sparse Decoding for Mitigating Visual Hallucinations in Large Vision-Language Models
RVSD：用于减轻大型视觉语言模型中视觉幻觉的检索视觉稀疏解码
Abstract
Large vision-language models have achieved remarkable success in vision-language tasks. However, they remain prone to Visual Hallucinations (VHs), undermining their reliability in real-world applications. Existing solutions typically require curated datasets, additional training, or multi-round decoding, resulting in considerable computational overhead. In this paper, we propose \textbf{RVSD} (\underline{R}etrieval \underline{V}ision \underline{S}parse \underline{D}ecoding), a training-free and plug-and-play decoding framework that, for the first time, unifies token sparsification and \textbf{Semantic-Space Visual Retrieval} (SSVR) within a single decoding pass. Within RVSD, we introduce a \textbf{semantics-directed token selection} strategy that selectively sparsifies redundant tokens while preserving critical visual information. We further propose the SSVR mechanism, which reformulates visual compensation as an on-demand cross-modal retrieval process within a shared semantic space. Extensive experiments demonstrate that RVSD achieves state-of-the-art performance in mitigating VHs while maintaining robust suppression capabilities under long-context generation settings. Our code is available here.\footnote{https://github.com/canjie-liu/RVSD}
Chinese Translation
大型视觉语言模型在视觉语言任务中取得了显著成功。然而，它们仍然容易出现视觉幻觉（Visual Hallucinations, VHs），这削弱了它们在实际应用中的可靠性。现有解决方案通常需要经过精心策划的数据集、额外的训练或多轮解码，导致相当大的计算开销。本文提出了 extbf{RVSD}（ extunderline{R}etrieval extunderline{V}ision extunderline{S}parse extunderline{D}ecoding），这是一个无训练且即插即用的解码框架，首次在单次解码过程中统一了令牌稀疏化和 extbf{语义空间视觉检索}（Semantic-Space Visual Retrieval, SSVR）。在RVSD中，我们引入了一种 extbf{语义导向的令牌选择}策略，该策略在保留关键视觉信息的同时，有选择性地稀疏冗余令牌。我们进一步提出了SSVR机制，它将视觉补偿重新构建为在共享语义空间内的按需跨模态检索过程。大量实验表明，RVSD在减轻视觉幻觉方面实现了最先进的性能，同时在长上下文生成设置下保持了强大的抑制能力。我们的代码可在此获取。 ootnote{https://github.com/canjie-liu/RVSD}
cs.CV / 105 / 2609.02747
InceptionGS: Generative Bootstrapping for Large-Scale Gaussian Splatting under Unstructured View Sampling
InceptionGS：在无结构视图采样下的大规模高斯喷溅生成引导
Abstract
Achieving truly immersive large-scale scene digitization necessitates consistent and visually pleasing rendering across all possible viewing perspectives. However, collecting multi-view images covering every fine detail of a large-scale scene is prohibitive due to scene complexity, capture cost, negligence, or accessibility constraints. As a result, the sampled views tend to be highly unstructured -- the majority of the scene is well covered yet certain regions inevitably lack sufficient observations. Existing reconstruction based methods are vulnerable to view scarcity while generation based approaches suffer from generalization, controllability, and 3D consistency issues. To address this challenge, we propose InceptionGS, which bootstraps Gaussian splatting by subtly balancing reconstruction and generation. Starting from an initial Gaussian splatting, InceptionGS reasonably rethinks and repairs problematic regions caused by view scarcity while preserving the quality elsewhere, by softly incorporating scene- and view-adaptive generative priors. Extensive experiments on real-world large-scale scenes demonstrate the superiority and broad applicability of our approach in handling unstructured imagery and boosting high-fidelity Gaussian splatting. Please refer to the supplementary video for better visual demonstrations.
Chinese Translation
实现真正沉浸式的大规模场景数字化需要在所有可能的视角下保持一致且视觉上令人愉悦的渲染。然而，由于场景复杂性、捕捉成本、疏忽或可达性限制，收集覆盖大规模场景每个细节的多视角图像是不可行的。因此，采样的视图往往高度无结构——大部分场景得到了良好的覆盖，但某些区域不可避免地缺乏足够的观察。现有的基于重建的方法对视图稀缺性较为脆弱，而基于生成的方法则面临泛化、可控性和三维一致性问题。为了解决这一挑战，我们提出了InceptionGS，通过微妙地平衡重建和生成来引导高斯喷溅。从初始的高斯喷溅开始，InceptionGS合理地重新思考和修复因视图稀缺而导致的问题区域，同时保持其他区域的质量，通过柔和地结合场景和视图自适应的生成先验。对真实世界大规模场景的广泛实验表明，我们的方法在处理无结构图像和提升高保真高斯喷溅方面具有优越性和广泛的适用性。请参阅补充视频以获得更好的视觉演示。
cs.CV / 106 / 2609.02748
Balancing Frequencies and Pixels in Flow Matching
在流匹配中平衡频率与像素
Abstract
Natural images follow a $1/f^2$ spectral distribution: most signal energy lies in the low spatial frequencies, while the perceptually important structures such as textures and edges occupy sparse high-frequency bands. Pixel-space reconstruction objectives, however, treat all spatial errors uniformly, causing low frequencies to dominate the optimization signal and delaying the learning of fine-scale details. In this work, we identify this objective-level spectral imbalance as a key inefficiency in training pixel-space flow models. To address it, we propose a Focal Log-Frequency Loss (f-loss), a spectrally balanced objective that equalizes the learning signal across frequencies, emphasizing high-frequency components that are otherwise underrepresented in pixel-space objectives. Building on this, we introduce a simple training strategy that combines frequency and pixel supervision: we first emphasize frequency-domain learning early to capture all frequencies, and then transition to standard pixel-space v-loss for spatial refinement. This balancing mitigates the low-frequency bias of pixel losses and aligns the training signal with the evolving needs of the model. Our approach is conceptually simple, requires no architectural changes, and acts as a drop-in replacement for flow matching losses. Across multiple model scales, it accelerates convergence by up to 40% while consistently improving FID and perceptual fidelity. We will release code and models.
Chinese Translation
自然图像遵循 $1/f^2$ 频谱分布：大部分信号能量集中在低空间频率，而感知上重要的结构如纹理和边缘则占据稀疏的高频带。然而，像素空间重建目标将所有空间误差视为均匀的，导致低频主导优化信号，从而延迟对细节的学习。在本研究中，我们将这种目标级别的频谱不平衡视为训练像素空间流模型中的一个关键低效因素。为了解决这一问题，我们提出了一种焦点对数频率损失（Focal Log-Frequency Loss，简称 f-loss），这是一种频谱平衡的目标，旨在平衡不同频率之间的学习信号，强调在像素空间目标中被低估的高频成分。在此基础上，我们引入了一种简单的训练策略，结合频率和像素监督：我们首先强调频域学习，以捕获所有频率，然后过渡到标准的像素空间 v-loss 进行空间细化。这种平衡减轻了像素损失的低频偏差，并使训练信号与模型不断演变的需求保持一致。我们的方法在概念上简单，不需要架构上的改变，并且可以作为流匹配损失的直接替代。在多个模型规模下，它加速收敛速度高达 40%，同时持续改善 FID 和感知保真度。我们将发布代码和模型。
cs.CV / 107 / 2609.02751
Multi-Tool Image Editing Attribution in Facial Forgery
面部伪造中的多工具图像编辑归属
Abstract
As generative AI tools become increasingly powerful and easy to use, people can easily edit portrait images with a prompt, necessitating the task of image editing attribution, which predicts the involved editing tools from the given image. Existing attribution methods hold the single-tool assumption and can only attribute a specific editing tool, but struggle to handle the more complex and increasingly common multi-tool editing scenarios, where artifacts left by different editing tools are composite and overlapped. To address this gap, we explore Multi-Tool Image Editing Attribution (MIEA), which aims to identify multiple editing tools involved in a multi-tool edited facial image. To simulate the real-life editing operations on facial images, we then construct a new dataset, MultiEdit, which contains 500k+ edited facial images and covers six types of editing tools that support face swapping (Deepfake) and various facial enhancements. Inspired by the findings from data analysis, we design DPEC, a multi-tool attribution method that can capture distinguishable, locality-aware editing tool traces from both spatial and frequency domains with the support of an error-based curriculum learning strategy. Experiments show \Method\ outperforms nine methods for facial images edited in at most five steps.
Chinese Translation
随着生成性人工智能工具变得越来越强大且易于使用，人们可以轻松地通过提示编辑肖像图像，这就需要进行图像编辑归属的任务，即从给定图像中预测所涉及的编辑工具。现有的归属方法假设使用单一工具，只能归属特定的编辑工具，但在处理更复杂且日益普遍的多工具编辑场景时却显得力不从心，因为不同编辑工具留下的伪影是复合且重叠的。为了解决这一问题，我们探索了多工具图像编辑归属（Multi-Tool Image Editing Attribution, MIEA），旨在识别在多工具编辑的面部图像中涉及的多个编辑工具。为了模拟现实生活中对面部图像的编辑操作，我们构建了一个新的数据集MultiEdit，其中包含超过50万张编辑过的面部图像，并涵盖六种支持面部替换（Deepfake）和各种面部增强的编辑工具。受到数据分析结果的启发，我们设计了DPEC，这是一种多工具归属方法，能够从空间和频率域中捕捉可区分的、具有局部感知的编辑工具痕迹，并结合基于错误的课程学习策略。实验表明， extit{Method}在最多五步编辑的面部图像上优于九种方法。
cs.CV / 108 / 2609.02776
Video-Based Palm-Vein Authentication under Challenging Conditions
在挑战性条件下基于视频的掌静脉身份验证
Abstract
Palm-vein biometrics are increasingly used for secure, contactless authentication. Yet real-world deployment exposes them to surface noise (sweat, dirt), illumination and motion variation, and temperature-driven changes in vascular visibility, which remain underexplored for lack of data captured under such conditions. To study these effects, we introduce the Columbia University Palm-vein (CUP) dataset, to our knowledge the first public video-based palm-vein dataset. CUP records every palm under four surface conditions (a clean baseline, warm, wet, and dirty) and pairs each subject with physiological and demographic metadata. On it we benchmark twenty-one recognizers spanning static, video, and multi-frame aggregation architectures. Models that verify reliably on clean palms lose most of their accuracy on dirty ones, and the mean equal error rate (EER) roughly quadruples. We recover much of that robustness along both axes of the capture. Temporally, a consensus over the few frames the sensor already returns cancels transient corruption; spatially, a test-time matcher that adds no learned parameters fuses the global cosine with a saliency-steered region-level optimal transport that routes the comparison around corrupted regions. The full design leads on every surface of CUP in EER, TAR@FAR=0.01, and Rank-1, at 4.3M parameters and 3.1 GFLOPs, a fraction of the video models' cost. Attached to four frozen state-of-the-art backbones it cuts their mean EER by 29-37% without retraining, and on four public single-image datasets the regional matching alone still helps. A preliminary audit across ten demographic and physiological traits finds two warm-condition gaps, along body water and gender, that survive multiple-comparison correction. CUP will be released for non-commercial research use at https://github.com/MobileX-CU/CUP_v1 upon publication.
Chinese Translation
掌静脉生物识别技术越来越多地用于安全的非接触式身份验证。然而，实际应用中，这些技术面临表面噪声（汗水、污垢）、光照和运动变化，以及温度驱动的血管可见性变化等问题，而由于缺乏在此类条件下捕获的数据，这些问题尚未得到充分研究。为了研究这些影响，我们引入了哥伦比亚大学掌静脉（CUP）数据集，据我们所知，这是第一个公开的基于视频的掌静脉数据集。CUP在四种表面条件下记录每个掌（干净基线、温暖、潮湿和肮脏），并为每个受试者配备生理和人口统计元数据。在此数据集上，我们基准测试了二十一种识别器，涵盖静态、视频和多帧聚合架构。那些在干净掌上可靠验证的模型在肮脏掌上的准确性大幅下降，平均等错误率（EER）大约增加了四倍。我们在捕获的两个维度上恢复了大部分的鲁棒性。在时间维度上，传感器已经返回的少数帧的共识消除了瞬态干扰；在空间维度上，一个不增加任何学习参数的测试时间匹配器将全局余弦与一个基于显著性引导的区域级最优传输相结合，从而绕过受损区域进行比较。完整的设计在CUP的每个表面上都在EER、TAR@FAR=0.01和Rank-1上表现优异，参数量为430万，计算量为3.1 GFLOPs，仅为视频模型成本的一小部分。附加到四个冻结的最先进骨干网络上，它在不重新训练的情况下将它们的平均EER降低了29-37%，而在四个公共单图像数据集上，仅区域匹配也仍然有效。对十个人口统计和生理特征的初步审计发现，在体水和性别方面存在两个温暖条件下的差距，这些差距在多重比较校正后仍然存在。CUP将在出版后以非商业研究用途发布，网址为https://github.com/MobileX-CU/CUP_v1。
cs.CV / 109 / 2609.02780
ShallowStream: Index Shallow then Answer Deep for Streaming Video Understanding
ShallowStream：先索引浅层再深入回答的流媒体视频理解
Abstract
Streaming video understanding is a critical capability for real-world applications, including embodied intelligence, autonomous driving, industrial monitoring, surveillance and early warning, and wearable assistants. However, processing continuous video streams with multimodal large language models (MLLMs) is computationally expensive. Existing efforts have explored reducing streaming overhead through visual token pruning, token merging, quantization, on-demand frame retrieval, and context offloading. However, most existing methods overlook the dimension of model depth. Repeatedly executing full-depth MLLM prefill over incoming frames is prohibitively expensive, incurring substantial computational overhead and causing the KV cache to grow at a rate directly proportional to the prefill depth. To address these challenges, we propose ShallowStream, a novel framework that leverages the shallow layers of an MLLM to simultaneously perform frame encoding and retrieval index building. During stream processing, ShallowStream maintains an always-on lightweight index using the KV cache of shallow layers. During query-time answering, we leverage the attention scores generated by the shallow layers to score context frames and employ a diversity-aware selection strategy to retrieve precise and comprehensive evidence. ShallowStream achieves performance on par with the strongest existing streaming methods, while reducing per-frame prefill latency and 10-second end-to-end latency by up to 52.1x and 11.9x, respectively. Our code is available at https://github.com/CURRENTF/ShallowStream.
Chinese Translation
流媒体视频理解是现实应用中的一项关键能力，包括具身智能、自动驾驶、工业监控、监视与预警以及可穿戴助手。然而，使用多模态大型语言模型（MLLMs）处理连续的视频流计算成本高昂。现有的研究已探索通过视觉令牌修剪、令牌合并、量化、按需帧检索和上下文卸载来减少流媒体开销。然而，大多数现有方法忽视了模型深度的维度。对输入帧重复执行全深度的MLLM预填充是极其昂贵的，导致显著的计算开销，并使KV缓存以与预填充深度成正比的速度增长。为了解决这些挑战，我们提出了ShallowStream，这是一种新颖的框架，利用MLLM的浅层同时执行帧编码和检索索引构建。在流处理过程中，ShallowStream使用浅层的KV缓存维护一个始终在线的轻量级索引。在查询时回答中，我们利用浅层生成的注意力分数对上下文帧进行评分，并采用一种关注多样性的选择策略来检索精确且全面的证据。ShallowStream在性能上与现有最强的流媒体方法相当，同时将每帧预填充延迟和10秒的端到端延迟分别减少了高达52.1倍和11.9倍。我们的代码可在 https://github.com/CURRENTF/ShallowStream 获取。
cs.CV / 110 / 2609.02798
AutoCompass: Accurate Visual Localization on Public Maps by Learning from Weak Labels
AutoCompass：通过从弱标签学习实现公共地图上的准确视觉定位
Abstract
Neural map matchers estimate an image's 3-DoF pose relative to a 2D map. These models are trained on large-scale datasets of geo-referenced images, whose position and heading labels often contain noise that affects the trained models. To address this, we present AutoCompass, a supervision approach for training neural map matchers from inaccurate absolute pose labels. First, we show that heading labels are unnecessary: trained from raw GPS labels, models learn to predict accurate headings, automatically. Second, defining a tolerance region around raw GPS improves positional accuracy. Third, if available, our supervision uses relative poses between training images, obtained via SLAM or SfM, which provide a more accurate training signal. Across driving and egocentric benchmarks, AutoCompass consistently outperforms counterparts trained with the usual strong reliance on absolute pose labels.
Chinese Translation
神经地图匹配器估计图像相对于二维地图的三自由度姿态。这些模型是在大规模的地理参考图像数据集上训练的，这些数据集的位置信息和朝向标签通常包含噪声，从而影响训练模型的效果。为了解决这个问题，我们提出了AutoCompass，一种从不准确的绝对姿态标签中训练神经地图匹配器的监督方法。首先，我们展示了朝向标签并非必要：从原始GPS标签训练的模型能够自动学习预测准确的朝向。其次，在原始GPS周围定义一个容忍区域可以提高位置精度。第三，如果可用，我们的监督方法利用通过SLAM或SfM获得的训练图像之间的相对姿态，这提供了更准确的训练信号。在驾驶和自我中心基准测试中，AutoCompass始终优于那些通常强烈依赖绝对姿态标签的对比方法。
cs.CV / 111 / 2609.02813
GDB-Reward: From Evaluation Metrics to Training Rewards for Graphic Design
GDB-Reward：从评估指标到图形设计训练奖励
Abstract
Text-to-image models excel at natural image synthesis but struggle with graphic design, where success depends on satisfying precise constraints on typography, layout, color, and visual communication. While prompt optimization offers an attractive alternative to expensive diffusion model fine-tuning, learning prompts for frozen image generators requires informative reward functions despite the entirely non-differentiable generation process. Reinforcement learning does not require differentiable objectives; it requires only scalar rewards capable of ranking candidate outputs. This raises a simple question: can design evaluation metrics themselves become reinforcement learning rewards? Our central contribution is GDB-Reward, a framework that systematically transforms heterogeneous graphic design evaluation metrics into a unified reinforcement learning reward. Experiments demonstrate that GDB-Reward provides an effective optimization objective, substantially improving adherence to the design specification in perceptual quality, rendering fidelity, and spatial accuracy while keeping the image generator entirely frozen. More broadly, our results demonstrate that heterogeneous, non-differentiable evaluation metrics can move beyond passive benchmarking to become effective optimization objectives for reinforcement learning in domains where differentiable supervision is unavailable.
Chinese Translation
文本到图像模型在自然图像合成方面表现出色，但在图形设计中却面临挑战，因为成功依赖于满足排版、布局、颜色和视觉传达等精确约束。尽管提示优化为昂贵的扩散模型微调提供了一个有吸引力的替代方案，但对于冻结的图像生成器，学习提示需要信息丰富的奖励函数，尽管生成过程完全是不可微分的。强化学习并不需要可微分的目标；它只需要能够对候选输出进行排序的标量奖励。这引出了一个简单的问题：设计评估指标本身能否成为强化学习的奖励？我们的核心贡献是GDB-Reward，一个系统性地将异构图形设计评估指标转化为统一强化学习奖励的框架。实验表明，GDB-Reward提供了有效的优化目标，显著提高了在感知质量、渲染保真度和空间准确性方面对设计规范的遵循，同时保持图像生成器完全冻结。更广泛地说，我们的结果表明，异构的、不可微分的评估指标可以超越被动基准测试，成为在缺乏可微分监督的领域中强化学习的有效优化目标。
cs.CV / 112 / 2609.02831
Benchmarking RAW and RGB Restoration in Image Signal Processors
图像信号处理器中RAW与RGB恢复的基准测试
Abstract
Modern cameras transform RAW sensor measurements into sRGB images through an image signal processor (ISP). We benchmark two placements for blind restoration around a fixed ISP: (A) pre-ISP restoration in the RAW domain and (B) post-ISP restoration in the sRGB domain. The benchmark covers four smartphone device groups, two learned ISPs, three degradation regimes--noise, blur, and joint noise and blur--, and several representative RAW and RGB restoration models. Our results show that placement alone does not determine performance. The RAW restoration strategy outperforms the best generic RGB restoration models. However, RGB restoration models trained considering the ISP transformations, achieve the best overall performance. Our novel benchmark demonstrates that the image reconstruction performance strongly depends on the alignment between the restoration model and the target imaging pipeline. We consequently recommend reporting restoration placement and ISP-aware supervision as key experimental factors. Our code is available at https://github.com/mv-lab/AISP
Chinese Translation
现代相机通过图像信号处理器（ISP）将RAW传感器测量值转换为sRGB图像。我们对围绕固定ISP的盲恢复进行了两种位置的基准测试：(A) 在RAW域中的ISP前恢复和(B) 在sRGB域中的ISP后恢复。基准测试涵盖了四个智能手机设备组、两个学习型ISP、三种退化模式——噪声、模糊以及噪声与模糊的联合——以及若干代表性的RAW和RGB恢复模型。我们的结果表明，位置本身并不决定性能。RAW恢复策略的表现优于最佳的通用RGB恢复模型。然而，考虑到ISP变换的RGB恢复模型则实现了最佳的整体性能。我们的新基准测试表明，图像重建性能在很大程度上依赖于恢复模型与目标成像管道之间的对齐。因此，我们建议将恢复位置和ISP感知监督作为关键实验因素进行报告。我们的代码可在 https://github.com/mv-lab/AISP 获取。
cs.CV / 113 / 2609.02839
Efficient All-in-One Weather Restoration using Spectral Harmonization
基于谱调和的高效一体化天气恢复
Abstract
Adverse weather conditions such as rain, haze, and snow significantly degrade image quality, posing challenges for both human perception and physical AI. Existing restoration methods require large computational budgets, struggling to process high-resolution images and handle different degradations. In this paper, we present Frequency Reconstruction via Spectral Harmonization, a novel lightweight all-in-one restoration method that explicitly decomposes feature representations into high- and low-frequency components at each scale of a hierarchical encoder-decoder architecture. By combining spectral decomposition with spatial processing through Fourier-based skip connections, FReSH-IR captures complementary frequency information without sacrificing spatial detail. Our approach achieves similar restoration quality with 80% fewer parameters and operations than transformer-based models. Extensive experiments demonstrate that our method offers a great efficiency-performance trade-off, highlighting its practical applications in constrained-resource systems.
Chinese Translation
恶劣的天气条件，如雨、雾霾和雪，显著降低了图像质量，对人类感知和物理人工智能都带来了挑战。现有的恢复方法需要较大的计算预算，难以处理高分辨率图像并应对不同的退化情况。本文提出了一种通过谱调和实现的频率重建（Frequency Reconstruction via Spectral Harmonization），这是一种新颖的轻量级一体化恢复方法，能够在层次编码器-解码器架构的每个尺度上显式地将特征表示分解为高频和低频组件。通过结合谱分解和基于傅里叶的跳跃连接进行空间处理，FReSH-IR能够捕捉互补的频率信息，而不牺牲空间细节。我们的方法在参数和运算量上比基于变换器的模型减少了80%，而恢复质量相似。大量实验表明，我们的方法在效率与性能之间提供了良好的权衡，突显了其在资源受限系统中的实际应用潜力。
cs.CV / 114 / 2609.02847
RoGe: Novel View Synthesis via End-to-End Implicit Reconstruction and Generation
RoGe：通过端到端隐式重建与生成的新视图合成
Abstract
Novel view synthesis from sparse inputs requires both geometric grounding from the observed views and generative priors of unobserved regions, motivating recent hybrid methods that combine reconstruction and generation. However, existing methods bridge the two with rendered images or explicit 3D representations such as point maps or 3D Gaussians. Generation is thus conditioned on a lossy and imperfect projection of the scene, inheriting its errors, and reconstruction receives no signal from generation to correct them. We present RoGe, an end-to-end unified reconstruction and generation framework that removes this explicit bridge. It targets roaming within a scene anchored by sparse views: given a few posed images and a camera trajectory, it synthesizes a temporally coherent video along that trajectory. From the sparse input views, RoGe builds an implicit scene representation with a feed-forward reconstruction model, and queries it with target camera rays to obtain per-view geometric features. These features are injected into a video diffusion model as conditioning, without any 3D intermediate. Both modules are trained jointly, so the generation objective directly shapes its own geometric conditioning. We conduct experiments on DL3DV, where RoGe outperforms reconstruction-based, generation-based, and hybrid baselines on image-level metrics and video-level temporal consistency. Ablations confirm that ray-queried implicit features outperform both raw reconstruction tokens and rendered RGB as conditioning, and that joint training brings further gains.
Chinese Translation
从稀疏输入中合成新视图需要既有来自观察视图的几何基础，又有来自未观察区域的生成先验，这促使了最近混合方法的出现，这些方法结合了重建与生成。然而，现有方法通过渲染图像或显式的三维表示（如点图或三维高斯）将两者连接起来。因此，生成是基于场景的有损和不完美投影，从而继承了其错误，而重建则没有从生成中获得信号来纠正这些错误。我们提出了RoGe，一个端到端统一的重建与生成框架，消除了这种显式桥接。它的目标是在稀疏视图的锚定下在场景中漫游：给定少量姿态图像和相机轨迹，它沿着该轨迹合成一个时间一致的视频。RoGe从稀疏输入视图中构建一个隐式场景表示，并使用前馈重建模型进行查询，以获取每视图的几何特征。这些特征被注入到视频扩散模型中作为条件，而没有任何三维中间过程。这两个模块是联合训练的，因此生成目标直接塑造其自身的几何条件。我们在DL3DV上进行了实验，结果表明RoGe在图像级指标和视频级时间一致性上优于基于重建、基于生成和混合基线。消融实验确认，光线查询的隐式特征优于原始重建标记和渲染RGB作为条件，并且联合训练带来了进一步的提升。
cs.CV / 115 / 2609.02854
MuyBridge: Mobile Human Center-of-Mass Estimation from Monocular Video via Sparse Fusion
MuyBridge：通过稀疏融合从单目视频中移动人体质心估计
Abstract
The 3D center of mass (CoM) is a primary quantity in the biomechanical analysis of sport, rehabilitation, and clinical movement, yet existing 3D pose tracking, mesh recovery, and multi-view triangulation methods either optimize 3D keypoint accuracy without anatomical constraints or carry compute and capture infrastructure too heavy to deploy where CoM tracking is most useful. As a result, the metric CoM remains difficult for coaches and movement analysts to measure from a single camera where athletes train and compete. In this work, we introduce MuyBridge, an on-device system that estimates the athlete's segmental center of mass trajectory from a single phone camera video stream. MuyBridge couples a compact 2D pose network and a distilled single-step monocular depth network through an analytic metric fusion that uses anatomical and physical priors to anchor the metric CoM, requiring no 3D or task-specific supervision. Evaluated on the athletic movements of AthletePose3D (running, track and field, and figure skating), MuyBridge achieves 33-41 mm vertical CoM error and 2.3-6.6% absolute-relative range error (AbsRel) under a one-time calibration, and produces CoM estimates at the 63 FPS pose-estimation rate using asynchronous 2.86 Hz depth updates on iPhone 15. Code is available at: https://github.com/Abradshaw1/Muybridge
Chinese Translation
三维质心（CoM）是运动、生物力学分析、康复和临床运动中的一个主要量。然而，现有的三维姿态跟踪、网格恢复和多视角三角测量方法，要么在没有解剖约束的情况下优化三维关键点的准确性，要么需要过于庞大的计算和捕捉基础设施，无法在质心跟踪最有用的地方进行部署。因此，教练和运动分析师在单个摄像头下测量度量质心仍然困难。在本研究中，我们介绍了MuyBridge，这是一种设备内系统，可以从单个手机摄像头视频流中估计运动员的分段质心轨迹。MuyBridge将一个紧凑的二维姿态网络与一个经过提炼的单步单目深度网络结合，通过一种分析性度量融合方法，利用解剖和物理先验来锚定度量质心，且不需要三维或特定任务的监督。在AthletePose3D的运动（跑步、田径和花样滑冰）上进行评估，MuyBridge在一次校准下实现了33-41毫米的垂直质心误差和2.3-6.6%的绝对相对范围误差（AbsRel），并在iPhone 15上以每秒63帧的姿态估计率生成质心估计，使用异步2.86赫兹的深度更新。代码可在以下链接获取：https://github.com/Abradshaw1/Muybridge
cs.CV / 116 / 2609.02860
PlantC2USeg: Cross-Scale Consistent Pre-Training for Few-Shot Unified Plant Point Cloud Segmentation
PlantC2USeg：跨尺度一致性预训练用于少样本统一植物点云分割
Abstract
Modern crop breeding demands precise organ-level analysis for trait quantification, making plant point cloud segmentation (PPCS) increasingly important. However, conventional deep learning approaches rely heavily on densely annotated datasets that are labor-intensive to acquire. Unified PPCS adaptation from distribution-shifted examples with minimal additional training remains challenging. To address this, we propose PlantC2USeg, a deep transfer learning framework featuring cross-scale consistency learning to explicitly align features across spatial scales and an information-restricted decoding strategy that prevents reconstruction shortcuts and promotes robust adaptation. The resulting pre-training enables stable few-shot generalization across species and sensing conditions, while unified fine-tuning with inherited thresholds further reduces adaptation overhead. Under full supervision on Soybean3D, PlantC2USeg achieves the highest semantic IoU and instance mWCov among compared methods, at 91.91% and 94.62%. With 20 labeled samples, it leads both metrics at 89.78% and 90.27%; with only 10 samples, it retains the highest mWCov of 83.23% while achieving 83.19% IoU. Across HR3D, 10-shot transfer to tobacco, tomato, and sorghum averages 78.41% IoU and 79.42% mWCov, while 22-shot transfer to SYAU-Maize achieves the highest IoU and mRec at 92.75% and 93.51%. Furthermore, a leading category-averaged mIoU of 85.0% on ShapeNet Part demonstrates the framework's capability to handle diverse shape variations beyond agricultural domains. These results demonstrate that PlantC2USeg reduces overall adaptation effort under distribution shifts, enabling scalable plant phenotyping and transferable 3D representation learning beyond agriculture.
Chinese Translation
现代作物育种要求对器官级别进行精确分析以量化性状，使植物点云分割（PPCS）变得越来越重要。然而，传统的深度学习方法在很大程度上依赖于密集标注的数据集，这些数据集获取过程劳动密集。统一PPCS从分布偏移的示例中进行适应，且仅需最小的额外训练，仍然面临挑战。为了解决这一问题，我们提出了PlantC2USeg，一个深度迁移学习框架，具有跨尺度一致性学习，能够明确对齐不同空间尺度的特征，以及一种信息受限的解码策略，防止重建捷径并促进稳健的适应。所得到的预训练使得在不同物种和传感条件下实现稳定的少样本泛化成为可能，而统一微调与继承的阈值进一步减少了适应开销。在Soybean3D的全监督下，PlantC2USeg在比较方法中实现了最高的语义IoU和实例mWCov，分别为91.91%和94.62%。在20个标记样本的情况下，它在两个指标上均领先，分别为89.78%和90.27%；在仅有10个样本的情况下，它保持了最高的mWCov为83.23%，同时实现了83.19%的IoU。在HR3D上，10-shot迁移到烟草、西红柿和高粱的平均IoU和mWCov分别为78.41%和79.42%；而22-shot迁移到SYAU-Maize则达到了最高的IoU和mRec，分别为92.75%和93.51%。此外，在ShapeNet Part上，领先的类别平均mIoU为85.0%，展示了该框架处理农业领域之外多样形状变化的能力。这些结果表明，PlantC2USeg在分布变化下减少了整体适应工作量，使得可扩展的植物表型分析和可迁移的3D表示学习超越农业成为可能。
cs.CV / 117 / 2609.02864
Thinking in Pictures: A Systematic Benchmark for Reasoning-driven Image Generation
图像思维：一个系统化的推理驱动图像生成基准
Abstract
Recent advancements in unified generative models (UGMs) and world simulators have achieved unprecedented results in visual perception and synthesis. However, these models primarily rely on surface-level event alignment, leaving the capacity for high-level visual reasoning underexplored. True visual generative intelligence demands "Reasoning-to-Generation", an ability to infer latent rules from visual inputs and manifest solutions through precise, logically constrained visual outcomes. We introduce RIG-BENCH, a novel comprehensive benchmark that systematically evaluates Reasoning-driven Image Generation (RIG) across four cognitively demanding domains: Concept-based, Transformation-based, Pattern & Structure, and Scenario-based. Featuring 2000 curated samples, RIG-BENCH serves as a rigorous stress test for RIG. Our extensive evaluations of state-of-the-art UGMs and image/video generation models reveal a significant reasoning-generation gap, wherein models frequently produce locally plausible but globally illogical outputs. RIG-BENCH provides a vital diagnostic framework to guide the development of next-generation, logically grounded UGMs and world simulators.
Chinese Translation
最近，统一生成模型（UGMs）和世界模拟器的进展在视觉感知和合成方面取得了前所未有的成果。然而，这些模型主要依赖于表层事件对齐，导致高层次视觉推理的能力未得到充分探索。真正的视觉生成智能要求“推理到生成”（Reasoning-to-Generation），即从视觉输入中推断潜在规则，并通过精确、逻辑约束的视觉结果展现解决方案。我们引入了RIG-BENCH，一个新颖的综合基准，系统地评估推理驱动图像生成（RIG）在四个认知要求高的领域中的表现：基于概念、基于转化、模式与结构，以及基于场景。RIG-BENCH包含2000个精心挑选的样本，作为RIG的严格压力测试。我们对最先进的UGMs和图像/视频生成模型的广泛评估揭示了显著的推理生成差距，即模型经常产生局部上合理但整体上不合逻辑的输出。RIG-BENCH提供了一个重要的诊断框架，以指导下一代逻辑基础的UGMs和世界模拟器的发展。
cs.CV / 118 / 2609.02886
SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models
SolarWM：用于长时间视频世界模型的开放数据与可扩展训练
Abstract
We introduce SolarWM, a fully open foundation for building interactive video world models from data preparation through long-horizon inference. Training across heterogeneous data sources and video backbones is challenging: datasets differ in temporal scale, camera geometry, visual quality, motion, and captioning styles, while video generators use distinct representations and architectures. Naive data mixing and model-specific implementations therefore produce inconsistent supervision and make results difficult to reproduce and compare. SolarWM addresses this coupling with a reconfigurable multi-source data engine and a backbone-native adaptation framework. The engine converts 1.43 million canonical clips from 10 datasets into a unified, frame-aligned contract covering visual observations, metric camera geometry, captions, quality metadata, selection decisions, and provenance, while decoupling source processing from mixture construction. Under shared camera-conditioning, training, and inference interfaces, we instantiate four 5B--33B models based on Wan2.2, LTX-2.5, and MiniMax-H3 while preserving their native representations and objectives. A unified three-stage recipe combines bidirectional adaptation, teacher-forced autoregressive initialization, and distribution matching distillation. The resulting causal models enable real-time interaction over rollouts ranging from minutes to hours after being trained on only 5s sequences. By releasing the resulting data, pipeline, recipes, weights, and framework, SolarWM provides a reproducible and extensible foundation for interactive world-model research.
Chinese Translation
我们介绍了SolarWM，这是一个完全开放的基础框架，用于从数据准备到长时间推理构建交互式视频世界模型。跨异构数据源和视频骨干网络的训练具有挑战性：数据集在时间尺度、相机几何、视觉质量、运动和字幕风格上存在差异，而视频生成器使用不同的表示和架构。因此，简单的数据混合和特定模型的实现会产生不一致的监督，使得结果难以重现和比较。SolarWM通过可重构的多源数据引擎和骨干网络原生适配框架解决了这一耦合问题。该引擎将来自10个数据集的143万条标准剪辑转换为一个统一的、帧对齐的合同，涵盖视觉观察、度量相机几何、字幕、质量元数据、选择决策和来源，同时将源处理与混合构建解耦。在共享的相机条件、训练和推理接口下，我们基于Wan2.2、LTX-2.5和MiniMax-H3实例化了四个5B-33B模型，同时保留它们的原生表示和目标。一个统一的三阶段方案结合了双向适配、教师强制自回归初始化和分布匹配蒸馏。最终得到的因果模型使得在仅用5秒序列训练后，可以在几分钟到几小时的回放中实现实时交互。通过发布生成的数据、管道、方案、权重和框架，SolarWM为交互式世界模型研究提供了一个可重现和可扩展的基础。
cs.AI / 1 / 2609.01611
EvalDetectBench: A Benchmark for Measuring Evaluation Awareness in Frontier Language Models
EvalDetectBench：评估意识前沿语言模型的基准测试
Abstract
Frontier large language models can often recognize when they are being evaluated, a capability known as evaluation awareness. If models behave differently in evaluations than in deployment, this undermines the validity of evaluation results, which are a crucial component of current AI safety frameworks. We introduce EvalDetectBench, an open pipeline and benchmark for measuring evaluation awareness that works with any Inspect-compatible evaluation, allowing practitioners to test against current and future benchmarks. EvalDetectBench ships with a newly curated transcript suite covering current frontier system-card evaluations and diverse deployment sources. The benchmark serves two purposes: measuring how reliably frontier LLMs recognize that they are being evaluated, and assessing how detectable individual benchmarks are as evaluations. We identify two methodological choices in the existing literature that introduce systematic bias: the identity of the model that generated the deployment transcripts accounts for 11.25% of measurement variance and can reorder model rankings; and elicitation prompts selected for high performance on one model can perform near chance on others. EvalDetectBench corrects for both via per-model probe calibration and a stratified generator-harmonisation procedure.
Chinese Translation
前沿大型语言模型通常能够识别出何时正在被评估，这种能力被称为评估意识。如果模型在评估中的表现与实际部署中的表现不同，这将削弱评估结果的有效性，而评估结果是当前人工智能安全框架的重要组成部分。我们介绍了EvalDetectBench，这是一个开放的管道和基准测试，用于测量评估意识，适用于任何兼容Inspect的评估，允许从业者针对当前和未来的基准进行测试。EvalDetectBench配备了一套新策划的转录套件，涵盖当前前沿系统卡评估和多样化的部署来源。该基准测试有两个目的：测量前沿大型语言模型识别其被评估的可靠性，以及评估各个基准作为评估的可检测性。我们在现有文献中识别出两种方法选择，这些选择引入了系统性偏差：生成部署转录的模型身份占测量方差的11.25%，并可能重新排序模型排名；而为某一模型选择的高性能引导提示在其他模型上可能表现接近随机。EvalDetectBench通过每个模型的探测校准和分层生成器协调程序纠正了这两种偏差。
cs.AI / 2 / 2609.01685
Meta-ethics and AI: exploring the novel meta-ethical questions in the era of AI
元伦理学与人工智能：探索人工智能时代的新型元伦理问题
Abstract
With the development of artificial intelligence (AI), the landscape of meta-ethics, which has largely centred on human ethics, faces pressures that may significantly reconfigure it. In particular, if future AI systems were to exhibit sufficiently integrated capacities for moral reasoning, moral intentionality, and moral reflection, novel meta-ethical questions would arise concerning what I call "AI's own ethics", as distinct from ethical principles merely imposed on AI by human designers. This paper offers a conditional and methodological framework for identifying the questions that would emerge if such AI systems were to arise. On that basis, the paper distinguishes four domains of meta-ethical inquiry in the era of AI: questions about the nature of human ethics from the human perspective; questions about the nature of AI's own ethics from the human perspective; questions about the nature of human ethics from the AI perspective; and questions about the nature of AI's own ethics from the AI perspective. The paper then considers how some existing mainstream meta-ethical theories (such as cognitivism and non-cognitivism, error theory and success theory, relativism, and objective realism) might illuminate these domains, while arguing that many familiar human-centred formulations of those theories may not transfer straightforwardly to AI cases without substantial revision. The overall conclusion is that the emergence of AI's own ethics would place significant pressure on current frameworks and may require substantial refinement, reconstruction, or reconceptualisation.
Chinese Translation
随着人工智能（AI）的发展，元伦理学的领域面临着可能显著重构的压力，这一领域主要集中于人类伦理。特别是，如果未来的AI系统展现出足够整合的道德推理、道德意图和道德反思能力，将会出现关于我所称的“AI自身伦理”的新型元伦理问题，这与人类设计者仅仅施加于AI的伦理原则是有区别的。本文提供了一个有条件的、方法论的框架，用于识别如果此类AI系统出现时将会出现的问题。在此基础上，本文区分了人工智能时代四个元伦理学研究领域：从人类视角出发的人类伦理本质问题；从人类视角出发的AI自身伦理本质问题；从AI视角出发的人类伦理本质问题；以及从AI视角出发的AI自身伦理本质问题。接着，本文考虑了现有一些主流元伦理理论（如认知主义与非认知主义、错误理论与成功理论、相对主义与客观现实主义）如何能够阐明这些领域，同时论证了许多熟悉的人类中心理论表述在没有实质性修订的情况下可能无法直接转移到AI案例中。总体结论是，AI自身伦理的出现将对当前框架施加重大压力，并可能需要实质性的修订、重构或重新概念化。
cs.AI / 3 / 2609.01741
When Can a Machine Trust a Statute? A Survival Certificate for Machine-Extracted Legal Logic
机器何时可以信任法律条文？机器提取法律逻辑的生存证书
Abstract
Statutes are increasingly parsed by machines before people read them, and the parsers disagree: on Missouri's statutes, two independently written extractors diverge on numeric-threshold presence at a false-negative rate of 0.43. We ask what formal logic survives such noise. We build a passive survival certificate for the Duquenne-Guigues implication basis of machine-extracted statutory contexts: per-attribute inter-extractor disagreement is measured, replayed against the basis in 1,000 Monte Carlo trials, and an implication is certified only when a one-sided Wilson 95% lower bound on survival reaches 0.95; every certified implication carries premise spans and a minimal counterexample. On 29,365 Missouri sections and 502 Indian central-Act sections, the preregistered held-out gate passes (10 statute families across 7 Titles exact; 16 across 11 with 5% tolerance), yet under one globally deployed error model 93.2% of held-out chapters fall below the informativeness floor, and a 2x2 factorial assigns that to calibration-rate transfer, not selection. The certificate is usable but fragile: deploy it per-chapter-calibrated or error-tolerant. Code, data products, and the audit trail, including one retracted claim, are released.
Chinese Translation
法律条文越来越多地被机器解析，随后人们再进行阅读，而这些解析器之间存在分歧：在密苏里州的法律条文中，两种独立编写的提取器在数字阈值存在性上以0.43的假阴性率出现分歧。我们探讨在这种噪声下，什么形式的逻辑能够存活。我们为机器提取的法律语境的 Duquenne-Guigues 逻辑基础构建了一个被动生存证书：通过测量每个属性的提取器间分歧，并在1,000次蒙特卡洛试验中重放该基础，仅当生存的单侧威尔逊95%下界达到0.95时，才会认证一个蕴含；每个认证的蕴含都携带前提跨度和一个最小反例。在29,365个密苏里州条款和502个印度中央法案条款中，预注册的保留门通过（10个法律家庭跨越7个标题精确；16个跨越11个标题，容忍5%），然而在一个全球部署的错误模型下，93.2%的保留章节低于信息性底线，而一个2x2因子分析将其归因于校准率转移，而非选择。该证书是可用的，但脆弱：需按章节校准或容忍错误进行部署。代码、数据产品和审计记录，包括一个撤回的声明，已被发布。
cs.AI / 4 / 2609.01814
When Does Information Sharing Improve Decentralized Discovery? Aggregation, Independent Rescue, and Equilibrium Selection
信息共享何时能改善去中心化发现？聚合、独立救援与均衡选择
Abstract
Information sharing can improve a pooled estimate while eliminating independent rescue actions. This paper separates those effects in exact finite discovery models. A centralized action-budget profile shows that equal one-person accuracy can coexist with different portfolio values. Under a registered incremental-sharing protocol, a sharing step improves discovery exactly when pooled residual error contracts faster than an independent rescue attempt. Exact bounded registries exhibit compression, aggregation, neutral curves, and a bounded zero mixed class. In a two-agent Bayesian game with a hidden mixture of common and independent signal sources, the registered selected equilibrium yields a strict positive sharing interval at signal accuracy 3/5, while alternative equilibria show that the result is selection-dependent rather than universal. The models are synthetic and finite; no human or organizational data are used.
Chinese Translation
信息共享可以改善汇总估计，同时消除独立救援行为。本文在精确有限发现模型中区分了这些效应。一个集中式的行动预算配置表明，相同的单人准确性可以与不同的投资组合价值共存。在注册的增量共享协议下，当汇总的残差误差收缩速度快于独立救援尝试时，共享步骤恰好改善发现。精确的有界注册表显示出压缩、聚合、中性曲线和有界零混合类别。在一个包含共同和独立信号源的隐藏混合的双代理贝叶斯博弈中，注册选择的均衡在信号准确性为3/5时产生严格正的共享区间，而替代均衡则表明该结果依赖于选择而非普遍适用。模型是合成的且有限；未使用任何人类或组织数据。
cs.AI / 5 / 2609.01815
Induction and Inquiry via Probabilistic Reasoning over Language and Code
通过语言和代码的概率推理进行归纳与探究
Abstract
How humans grow and maintain abstract knowledge from the sparse, streaming noisy data of experience is a longstanding challenge in cognitive science. Any computational account must satisfy at least three desiderata: It must be (1) data-efficient and compute-efficient, (2) capture gradations of uncertainty to support intelligent inquiry and information gathering, and (3) be flexible enough to mentally represent the endless range of concepts people can learn and think about. Here we introduce a computational model that captures these three properties, by encoding symbolic knowledge as mental programs that combine natural language with source code, and sequentially inferring mental programs using LLM-guided Bayesian learning algorithms. Across a range of behavioral studies this model successfully reproduces quantitative signatures of human inductive learning and active inquiry, such as anchoring, garden-pathing, and other effects. In contrast, pure LLMs and classic Bayesian models either fail at the underlying task, or do not reproduce human behavior, or succeed only at exorbitant computational cost. These results suggest that one way humans continually grow their knowledge is by mentally representing many hypotheses spanning language-like and program-like representations, then revising those hypotheses to approximate Bayesian updates, while a bottom-up neural mechanism (an LLM) makes inference both tractable and learnable.
Chinese Translation
人类如何从稀疏、流动的噪声数据中成长和维持抽象知识是认知科学中的一个长期挑战。任何计算模型必须满足至少三个要求：(1) 数据效率和计算效率，(2) 捕捉不确定性的梯度以支持智能探究和信息收集，以及 (3) 足够灵活以心理表征人们可以学习和思考的无尽概念范围。在这里，我们介绍了一种计算模型，它通过将符号知识编码为结合自然语言和源代码的心理程序，并使用 LLM（大型语言模型）引导的贝叶斯学习算法顺序推断心理程序，从而捕捉这三种特性。在一系列行为研究中，该模型成功再现了人类归纳学习和主动探究的定量特征，例如锚定效应、花园小径效应和其他效应。相比之下，纯 LLM 和经典贝叶斯模型要么在基础任务上失败，要么无法再现人类行为，或者仅在高昂的计算成本下成功。这些结果表明，人类不断增长知识的一种方式是通过心理表征许多跨越语言类和程序类表征的假设，然后修正这些假设以近似贝叶斯更新，同时底层的神经机制（一个 LLM）使推理既可处理又可学习。
cs.AI / 6 / 2609.01834
Architecting Conversational Data Systems for Stateless LLM APIs: The Hydration Proxy Pattern
为无状态 LLM API 构建对话数据系统：水合作用代理模式
Abstract
As enterprise platforms transition to conversational reasoning interfaces, the stateless nature of LLM APIs creates an architectural gap. While statelessness enables horizontal scalability for AI providers, it forces client applications to manage the entire burden of conversational state and semantic memory. The work identifies the Hydration Proxy Pattern, an architecture that decouples session persistence from the reasoning engine. The framework ensures platform sovereignty over conversational data while enabling secure, multi-stage semantic grounding. We further propose the Context Stabilization Mandate to resolve the tradeoff between sovereign state management and KV caching.
Chinese Translation
随着企业平台向对话推理接口的转型，LLM API 的无状态特性造成了架构上的缺口。尽管无状态性为 AI 提供者带来了水平扩展的可能，但它迫使客户端应用程序承担整个对话状态和语义记忆的管理负担。本研究识别出水合作用代理模式（Hydration Proxy Pattern），这一架构将会话持久性与推理引擎解耦。该框架确保平台对对话数据的主权，同时支持安全的多阶段语义基础。我们进一步提出了上下文稳定化指令（Context Stabilization Mandate），以解决主权状态管理与键值缓存（KV caching）之间的权衡。
cs.AI / 7 / 2609.01849
SSAKG 2.0: An Open-Source Package for Structural Associative Sequence Memory and Context-Based Retrieval
SSAKG 2.0：一种用于结构关联序列记忆和基于上下文检索的开源软件包
Abstract
This article presents SSAKG 2.0, an open-source software package for constructing and operating Structural Sequential Associative Knowledge Graphs (SSAKGs). An SSAKG represents objects as graph vertices and ordered sequences as structural patterns of graph connections. The resulting sparse graph is used as an associative memory in which complete sequences can be reconstructed from a partial, unordered context. Version 2.0 introduces new algorithms that exploit individual bits of computer memory to efficiently search graph connections. The package is implemented in Python, while performance-critical graph operations are implemented in C and exposed through a Python interface. This hybrid implementation provides a flexible high-level programming environment while reducing the memory and computational overhead associated with large sparse graphs. The algorithms were evaluated using randomly generated numerical sequences, sequences derived from sentences in the NLTK corpus, and mRNA sequences. The experiments demonstrate the ability of the package to store and reconstruct sequences from partial contexts and provide a basis for evaluating the effects of graph density, sequence length, and memory size on retrieval performance. SSAKG 2.0 is distributed under the Apache 2.0 open-source license. The package includes documentation and reproducible examples and is publicly available through GitHub and the Python Package Index (PyPI).
Chinese Translation
本文介绍了SSAKG 2.0，一个用于构建和操作结构序列关联知识图（SSAKGs）的开源软件包。SSAKG将对象表示为图的顶点，将有序序列表示为图连接的结构模式。生成的稀疏图用作关联记忆，其中可以从部分无序上下文中重建完整序列。版本2.0引入了新算法，利用计算机内存的单个位高效地搜索图连接。该软件包使用Python实现，而性能关键的图操作则用C语言实现，并通过Python接口暴露。这样的混合实现提供了灵活的高级编程环境，同时减少了与大型稀疏图相关的内存和计算开销。算法的评估使用了随机生成的数值序列、来自NLTK语料库的句子派生序列以及mRNA序列。实验表明该软件包能够从部分上下文中存储和重建序列，并为评估图密度、序列长度和内存大小对检索性能的影响提供了基础。SSAKG 2.0在Apache 2.0开源许可证下分发。该软件包包括文档和可重现的示例，并通过GitHub和Python包索引（PyPI）公开提供。
cs.AI / 8 / 2609.01852
The Memory Trust Gap: Capability-Dependent Failures in Persistent-Memory Agents
记忆信任差距：持久性内存代理中的能力依赖性失败
Abstract
Persistent memory supports personalized agents, but a stale stored fact can override current authoritative evidence without warning. We study when this harm begins as model capability changes. We evaluate a frozen, closed-set, action-scored benchmark with 2 suites that represent 2 different meanings of "no memory" (a Benefit suite, unsolvable without the stored fact, and a Safety suite, in which an authoritative tool always holds the correct value), on a same-family model-size series (Qwen3 0.6/1.7/4/8B). The Memory Trust Gap reflects over-trust rather than confusion. In the Benefit suite, models answer with the stale value 0.92-1.00 of the time at every scale. In the Safety suite, harm below the no-memory baseline under the trap conditions ($\Delta_{\mathrm{mem}}$) is capability-gated, with the larger models collapsing most once a stale note is made to look current. In a $2\times2\times2\times2$ factorial, which feature triggers over-trust depends on both the feature and model scale. Removing a label amplifies over-trust at every size, and a recency feature (stale dated newer) fools the larger models harder. Source authority is weak and scale-flat, and position changes from positive to negative across the Qwen3 model-size series. We confirm these scale interactions with direct cross-size contrast tests rather than overlapping per-model intervals. Mitigation is likewise capability-dependent: exposing metadata improves accuracy for the capable models, but only pre-resolving the conflict restores accuracy for the 2 smaller checkpoints. The same pattern appears on the capable models in an independent Llama-Instruct model-size series and on 2 external datasets (RGB, MisBench). A framing control finds no consistent advantage for the memory label: at the 3 smaller scales, models trust a stale document more than a stale memory; at 8B, the difference is not significant.
Chinese Translation
持久性内存支持个性化代理，但过时的存储事实可能在没有警告的情况下覆盖当前的权威证据。我们研究了随着模型能力变化，这种危害何时开始。我们评估了一个冻结的、封闭集的、基于行动得分的基准，包含两个套件，代表了“无记忆”的两种不同含义（一个益处套件，无法在没有存储事实的情况下解决，和一个安全套件，在该套件中，权威工具始终持有正确的值），在同一系列模型规模（Qwen3 0.6/1.7/4/8B）上进行。记忆信任差距反映的是过度信任而非混淆。在益处套件中，模型在每个规模上以过时值0.92-1.00的频率作答。在安全套件中，低于无记忆基线的危害在陷阱条件下（$ ext{Δ}_{ ext{mem}}$）是能力限制的，一旦过时的笔记被伪装成当前值，较大的模型大多数情况下会崩溃。在一个$2 imes2 imes2 imes2$的因子实验中，触发过度信任的特征依赖于特征和模型规模。去除标签在每个规模上都会放大过度信任，而一个时效特征（过时的较新日期）更容易欺骗较大的模型。源权威性较弱且规模平坦，位置在Qwen3模型规模系列中从正变为负。我们通过直接的跨规模对比测试确认了这些规模交互，而不是重叠的每模型区间。缓解措施同样依赖于能力：暴露元数据提高了能力模型的准确性，但只有在预先解决冲突时才能恢复两个较小检查点的准确性。在一个独立的Llama-Instruct模型规模系列和两个外部数据集（RGB，MisBench）中，能力模型也出现了相同的模式。一个框架控制发现记忆标签没有一致的优势：在三个较小的规模上，模型对过时文档的信任超过对过时记忆的信任；在8B规模上，差异不显著。
cs.AI / 9 / 2609.01861
Belief-Calibrated Optimization: An Explicit World Model for Agentic Optimization
信念校准优化：一种显式世界模型用于智能体优化
Abstract
The performance of an LLM agent depends on the scaffold around a frozen model. A common way to improve that scaffold is to use a coding agent as an optimizer: it reads current scores and traces and iteratively edits the source, producing a new candidate each round. Each edit is chosen according to a belief about how the environment will respond: what went wrong, and which change should help. That belief is typically implicit. It lives in the coding agent's reasoning on the current call, or remains latent in its parameters, rather than as something written down. Later calls therefore see scores and traces, but they do not use that belief. We introduce Belief-Calibrated Optimization (BCO), a method that writes that belief down as a persistent in-context document and continually revises that document as new candidates are evaluated. The resulting document is a world model: the current account of how the environment responds to edits. Added to an otherwise standard loop, BCO reaches a higher train passrate than a matched control that lacks only the world model, on five benchmarks spanning memory QA, tool-use QA, code-as-action app agents, and terminal agents. The gap remains on every held-out split, which is not used to select the candidate. After a target-model swap, in which the frozen model is replaced and the scaffold is not, the selected BCO scaffold leads on the tasks we test, except where context-window overruns leave it unfinished. An offline ablation then asks whether that gap comes from what the world model says. A fresh predictor given the accumulated document forecasts how the environment will respond more accurately than predictors given either no document or a same-form copy whose content has been falsified. The comparison indicates that the document carries reusable information in its content, not only in its form.
Chinese Translation
大型语言模型（LLM）智能体的性能依赖于一个固定模型周围的支架。改善该支架的常见方法是使用编码智能体作为优化器：它读取当前的分数和追踪记录，并迭代地编辑源代码，每轮生成一个新的候选。每次编辑的选择基于对环境将如何响应的信念：发生了什么错误，以及哪个变化应该有帮助。这种信念通常是隐性的。它存在于编码智能体对当前调用的推理中，或潜藏于其参数中，而不是以书面形式存在。因此，后续调用看到分数和追踪记录，但并未使用该信念。我们提出了信念校准优化（Belief-Calibrated Optimization, BCO）方法，该方法将信念以持久的上下文文档形式记录下来，并在评估新候选时不断修订该文档。生成的文档是一个世界模型：当前环境对编辑的响应的描述。将BCO添加到一个标准循环中，能在五个基准测试中达到比缺乏世界模型的匹配对照组更高的训练通过率，这些基准测试涵盖了记忆问答、工具使用问答、代码作为行动的应用智能体以及终端智能体。在每个未使用于选择候选的保留分割中，这一差距依然存在。在目标模型交换后，即固定模型被替换而支架不变的情况下，所选的BCO支架在我们测试的任务中表现优于其他支架，除了在上下文窗口超出限制导致未完成的情况。随后进行的离线消融实验探讨了这一差距是否源于世界模型的内容。一个新的预测器在给定累积文档的情况下，能够比没有文档或内容被伪造的同形副本的预测器更准确地预测环境的响应。比较结果表明，该文档在其内容中携带可重用的信息，而不仅仅是在其形式上。
cs.AI / 10 / 2609.01873
Epistemic Sybil Resistance: Multiplying AI Agents Without Multiplying Evidence
认知Sybil抗性：在不增加证据的情况下扩展AI代理
Abstract
Multi-agent AI systems improve inference by spawning agents and synthesizing reports. But another agent is not another observation: apparently independent reports may descend from the same evidence, and genuinely independent evidence can produce nearly identical reports. We formalize this as an epistemic Sybil problem. A report Z is an epistemic Sybil extension relative to reports R when I(Theta; Z | R) = 0. No report-only aggregator can generally distinguish replication from independent corroboration: identical reports can warrant different posteriors under unobserved ancestry. A Gaussian shared-root model shows common ancestry does not imply complete redundancy. Repeated extraction adds information toward a source-level ceiling, and correlated extraction errors, which a shared base model can induce among independent agents, lower that ceiling further. We test these predictions with more than 20,000 controlled LLM-agent report and extraction calls on synthetic evidentiary documents. Holding one evidence root fixed while report multiplicity rises from 1 to 32 collapses naive posterior coverage from 0.940 to 0.263. Holding report count fixed while evidence-root multiplicity rises from 1 to 16 closes the gap, and the aggregators are statistically indistinguishable at k = 16. The agent's replicate extraction errors are correlated (gamma_cal = 0.719, estimated out of sample), and a correlated-extraction aggregator restores calibration accordingly. A controlled manipulation isolates representation similarity from evidential ancestry. It changes a report-space deduplication mechanism's mean inferred cluster count by 1.425 (95% CI [1.363, 1.485]), whereas a fourfold change in true ancestry changes it by only 0.040 ([-0.045, 0.120]). Collective inference should therefore track evidential ancestry and dependence, not agent or report multiplicity or similarity.
Chinese Translation
多代理AI系统通过生成代理和合成报告来改善推理。然而，另一个代理并不等于另一个观察：看似独立的报告可能源于相同的证据，而真正独立的证据也可能产生几乎相同的报告。我们将其形式化为一个认知Sybil问题。当 I(Theta; Z | R) = 0 时，报告 Z 相对于报告 R 是一个认知Sybil扩展。一般来说，任何仅基于报告的聚合器无法区分复制与独立证实：在未观察到的祖先情况下，相同的报告可能会支持不同的后验分布。一个高斯共享根模型表明，共同祖先并不意味着完全冗余。重复提取向源级上限添加信息，而相关的提取错误（共享基础模型可以在独立代理之间引发）进一步降低了这一上限。我们通过对超过20,000个受控的LLM代理报告和提取调用在合成证据文件上的测试来验证这些预测。当固定一个证据根时，报告的多样性从1增加到32，天真的后验覆盖率从0.940降至0.263。当固定报告数量而证据根的多样性从1增加到16时，差距缩小，聚合器在 k = 16 时在统计上不可区分。代理的复制提取错误是相关的（gamma_cal = 0.719，估计为样本外），而相关提取聚合器相应地恢复了校准。一个受控操控将表示相似性与证据祖先隔离开来。它使报告空间去重机制的平均推断聚类数变化为1.425（95% CI [1.363, 1.485]），而真实祖先的四倍变化仅使其变化0.040（[-0.045, 0.120]）。因此，集体推理应追踪证据的祖先和依赖关系，而非代理或报告的多样性或相似性。
cs.AI / 11 / 2609.01909
The Ceiling Is in the Channel: Auditing Learner Gaps and Measurement Frontiers in Clinical Prediction
天花板存在于通道中：临床预测中的学习者差距与测量前沿审计
Abstract
Clinical prediction can saturate for two different reasons: a fitted learner may fail to extract available information, or the recorded variables may impose a population frontier. We separate these quantities through the \emph{learner gap} and the \emph{measurement-channel ceiling}. Optimal balanced accuracy is characterized by total-variation separation, yielding architecture invariance, a sharp partial-identification result under replacement contamination, a cross-fitted ceiling estimator, and exact conditions for multimodal decision improvement. We add two finite-sample diagnostics, namely a label-permutation optimism floor and an underfit curve, and validate the audit on three real cohorts: UCI readmission ($n=99{,}343$), BRFSS diabetes ($n=253{,}680$), and NHANES HbA1c ($n=10{,}219$). Well-tuned gradient boosting nearly reaches the estimated frontier in UCI and BRFSS, whereas deliberately or practically deficient learners retain large gaps. NHANES yields a null difference between questionnaire and measured marginal frontiers but a significant joint complementarity gain, refining the simplistic claim that an objective modality must dominate. Across all cohorts, modest AUROC gains coexist with substantially larger Bayes decision-flip rates, and several architectures estimate similar frontiers while their achieved balanced accuracy differs sharply. A PRISMA-guided synthesis of 104 clinical tasks then shows that the same channel-level regularities recur across more than 18 disease categories: a broad but non-universal structured-clinical region, diminishing same-channel gains across model families, and higher performance when measurement channels change. The framework converts saturation from an empirical observation into an auditable decision: improve the learner when headroom remains; improve measurement when it does not.
Chinese Translation
临床预测可能因两种不同原因而饱和：拟合的学习者可能无法提取可用信息，或记录的变量可能施加了人群前沿。我们通过 extit{学习者差距}和 extit{测量通道天花板}来区分这些量。最佳平衡准确性通过总变差分离来表征，产生架构不变性、在替换污染下的尖锐部分识别结果、交叉拟合天花板估计器，以及多模态决策改进的精确条件。我们增加了两个有限样本诊断，即标签置换乐观底线和欠拟合曲线，并在三个真实队列上验证审计：UCI 再入院（$n=99{,}343$）、BRFSS 糖尿病（$n=253{,}680$）和 NHANES HbA1c（$n=10{,}219$）。经过良好调优的梯度提升几乎达到了 UCI 和 BRFSS 中的估计前沿，而故意或实际不足的学习者则保留了较大的差距。NHANES 在问卷和测量边际前沿之间没有差异，但显著的联合互补增益，细化了“客观模态必须占主导地位”的简单主张。在所有队列中，适度的 AUROC 增益与显著更大的贝叶斯决策翻转率共存，且几种架构估计相似的前沿，而它们实现的平衡准确性则有显著差异。然后，基于 PRISMA 指导的 104 个临床任务的综合显示，超过 18 种疾病类别中相同的通道级规律反复出现：广泛但非普遍的结构化临床区域、跨模型家族的同通道增益递减，以及当测量通道变化时更高的性能。该框架将饱和从经验观察转化为可审计的决策：在有余地时改善学习者；在没有余地时改善测量。
cs.AI / 12 / 2609.01924
Looped Transformers under the Jacobian Lens: Does the Global Workspace Survive Recurrence?
在雅可比视角下的循环变换器：全球工作空间是否在递归中存活？
Abstract
Recent work identifies a mid-depth band of verbalisable, causally potent representations in a standard feedforward transformer --- a functional analogue of a global workspace. Whether the same workspace functionality emerges when depth is implemented through recurrence rather than a stack of distinct layers remains unknown. Looped and depth-recurrent transformers provide a direct test of this question because they reuse the same weights across depth. We extend the Jacobian lens to iterated architectures using a virtual-unrolling adapter. We apply the full workspace suite --- lens fitting, readout, and eleven causal experiment families --- to Ouro-2.6B (48 layers looped 4 times, deeply supervised) and Huginn-0125 (a 4-layer core recurred 16 times, trained for latent reasoning), using Qwen3.6-27B (64 untied layers) as the standard baseline. We find that a workspace forms in the iterated part of each architecture, but that recurrence changes how it can be accessed. Ouro reconstructs workspace content in every loop, and linear transport cannot carry that content across loop boundaries; writes and ablations must therefore span every remaining loop. Huginn carries content forward across all sixteen recurrences, while reads, writes, and ablations act only within a sliding window of roughly two recurrences. Whether newly injected content can be verbalised tracks explicit per-iteration supervision; whether existing content can be steered does not.
Chinese Translation
最近的研究发现，在标准前馈变换器中存在一段中层深度的可言语、因果有效的表征——这是一种全球工作空间的功能类比。然而，当深度通过递归而非一系列独立层来实现时，是否会出现相同的工作空间功能仍然未知。循环和深度递归变换器为这个问题提供了直接的测试，因为它们在深度上重用相同的权重。我们使用虚拟展开适配器将雅可比视角扩展到迭代架构。我们将完整的工作空间套件——视角拟合、读出和十一种因果实验系列——应用于Ouro-2.6B（48层循环4次，深度监督）和Huginn-0125（一个4层核心递归16次，训练用于潜在推理），并使用Qwen3.6-27B（64个未绑定层）作为标准基线。我们发现，在每个架构的迭代部分形成了一个工作空间，但递归改变了它的访问方式。Ouro在每个循环中重建工作空间内容，而线性传输无法跨越循环边界传递该内容；因此，写入和消融必须跨越每个剩余的循环。Huginn在所有十六次递归中向前传递内容，而读取、写入和消融仅在大约两个递归的滑动窗口内起作用。新注入的内容是否可以被言语化与显式的每次迭代监督相关；而现有内容是否可以被引导则无关。
cs.AI / 13 / 2609.01962
Post-Training Ternarization of Qwen3-4B Capability, Effective Bit Budget, Storage Compression, and Deployment
Qwen3-4B能力的后训练三元化、有效位预算、存储压缩与部署
Abstract
Ultra-low-bit language models can reduce storage and memory bandwidth, but a nominal "1.58-bit" label does not fully describe the stored representation, retained capability, or runtime behavior. We study an end-to-end post-training conversion of Qwen, an instruction-tuned 4B-parameter model, using KOTMS rotation, E2M-ATQ ternarization, and GPTQ-style error compensation from TWLA. The experiment is weight-only: activations remain at 16-bit precision, so ILA-AMP is omitted. We evaluate effective bit accounting, task capability retention, perplexity, calibration sensitivity, checkpoint composition, and deployment behavior. The final conversion uses 1.641 effective bits per weight for quantized linear weights, with 81.62% of model parameters targeted. Across ten scored capability comparisons, accuracy falls from 64.5% to 54.7%. Degradation is uneven: BoolQ retains 84.6% chance-corrected teacher performance, while ARC-Challenge retains 43.8%. Perplexity rises from 13.639 to 18.748 on WikiText-2, 24.700 to 31.992 on PTB, and 19.831 to 28.966 on C4. A subsequent packing run preserves the ternary planes and scales, reducing reported model size from 8.29 GiB to 3.96 GiB with essentially unchanged perplexity. A separate third-party packing attempt was lossy and is excluded from the primary artifact claim. The packed artifact has not been benchmarked end-to-end for task accuracy or generation throughput. A preliminary Triton GEMV microbenchmark is 4.6x slower than FP16 cuBLAS on one tested shape. We therefore do not claim that compression alone yields faster inference.
Chinese Translation
超低位语言模型可以减少存储和内存带宽，但名义上的“1.58位”标签并不能完全描述存储的表示、保留的能力或运行时行为。我们研究了Qwen的端到端后训练转换，Qwen是一个经过指令调优的4B参数模型，采用了KOTMS旋转、E2M-ATQ三元化和TWLA的GPTQ风格误差补偿。实验仅涉及权重：激活保持在16位精度，因此省略了ILA-AMP。我们评估了有效位计量、任务能力保留、困惑度、校准敏感性、检查点组成和部署行为。最终转换使用每个权重1.641有效位的量化线性权重，目标为81.62%的模型参数。在十个评分能力比较中，准确率从64.5%降至54.7%。性能下降不均匀：BoolQ保留了84.6%的机会校正教师表现，而ARC-Challenge保留了43.8%。在WikiText-2上，困惑度从13.639上升到18.748，在PTB上从24.700上升到31.992，在C4上从19.831上升到28.966。随后的打包运行保留了三元平面和比例，报告的模型大小从8.29 GiB减少到3.96 GiB，困惑度基本保持不变。一次独立的第三方打包尝试是有损的，因此未包含在主要文献声明中。打包的文献尚未针对任务准确性或生成吞吐量进行端到端基准测试。初步的Triton GEMV微基准测试在一个测试形状上比FP16 cuBLAS慢4.6倍。因此，我们不声称仅靠压缩就能实现更快的推理。
cs.AI / 14 / 2609.01982
Benchmarking Language Models for Statistical Problem Formulation
语言模型在统计问题表述中的基准测试
Abstract
Large language models (LLMs) are increasingly used as assistants for statistical and data science work, yet existing evaluations largely assume the analysis target is already specified. In practice, users arrive with informal goals and heterogeneous data, leaving the model to decide what statistical task is implied and which data are relevant. We first formalize this upstream step as Statistical Problem Formulation and decompose it into two subtasks: (1) Statistical Problem Classification and (2) Variable Identification & Role Assignment. We then introduce StatFormBench, a benchmark built from five cross-domain statistics textbooks and a data science case library, covering diverse problem types, data representations, and scenario styles. It contains 1,013 samples spanning 20 coarse-grained and 85 fine-grained statistical problem categories. Across 14 open- and closed-source LLMs, the best zero-shot models reach only 72.0 fine-grained classification accuracy and 63.2 variable set overlap. No model performs consistently best across the two subtasks, while enhanced prompting strategies yield only limited or inconsistent gains. We release the benchmark data on Hugging Face at https://huggingface.co/datasets/THU-CongLab/StatFormBench and the evaluation code on GitHub at https://github.com/THU-CongLab/StatFormBench.
Chinese Translation
大型语言模型（LLMs）越来越多地被用作统计和数据科学工作的助手，但现有的评估大多假设分析目标已经明确。在实际应用中，用户通常带着非正式的目标和异构数据，模型需要决定隐含的统计任务是什么以及哪些数据是相关的。我们首先将这一上游步骤形式化为统计问题表述，并将其分解为两个子任务：（1）统计问题分类和（2）变量识别与角色分配。接着，我们介绍了StatFormBench，这是一个基于五本跨领域统计教材和一个数据科学案例库构建的基准，涵盖了多样的问题类型、数据表示和场景风格。该基准包含1,013个样本，涵盖20个粗粒度和85个细粒度的统计问题类别。在14个开源和闭源的LLMs中，最佳的零-shot模型仅达到72.0的细粒度分类准确率和63.2的变量集重叠率。没有任何模型在两个子任务中表现 consistently 最佳，而增强的提示策略仅带来了有限或不一致的提升。我们在Hugging Face上发布了基准数据，网址为 https://huggingface.co/datasets/THU-CongLab/StatFormBench，并在GitHub上发布了评估代码，网址为 https://github.com/THU-CongLab/StatFormBench。
cs.AI / 15 / 2609.01985
When Agents Implement Systems: A Case Study in Defects, Detection, and Evaluation Rigor
当智能体实施系统时：缺陷、检测与评估严谨性的案例研究
Abstract
As LLM coding agents increasingly perform end-to-end engineering work, we lack empirical characterization of how they behave on systems-level requirements: schema design, async orchestration, configuration correctness, and retrieval-filtering trade-offs. We present a case study of one such agent implementing a multi-component data system against a detailed pre-existing specification. Storage technologies, schema, entity-resolution algorithm, and retrieval-filtering strategy were fixed in advance; the agent autonomy was in the implementation, in diagnosing and fixing defects it introduced, and in interaction-design choices left open. Over a single session, we catalog five such defects, categorized by constraint violated and detection method. We further evaluate, on the public HotpotQA benchmark, the one retrieval trade-off specified in that architecture: restricting candidates to a graph-identified entity set before ranking versus unfiltered search. We substitute the benchmark gold evidence labels for entity identification, since we lacked LLM access to run that stage, and report standard recall rather than the benchmark own accuracy metrics. Across retrieval budgets from 1 to 10 and 100 questions against a pooled corpus of 2994 paragraphs, filtered recall reaches its ceiling by a budget of 3, expected once candidates are restricted to the gold paragraphs themselves, while unfiltered search recovers all required evidence only 69 percent of the time even at a budget of 10, a gap that holds at every budget tested, with sign test p less than 0.0001. We close with a discussion of where the agent autonomy succeeded versus required correction, including one instance where a claimed performance fix was never re-measured on the regression that motivated it.
Chinese Translation
随着大型语言模型（LLM）编码智能体越来越多地执行端到端的工程工作，我们缺乏对它们在系统级需求上的行为的实证特征描述：模式设计、异步编排、配置正确性和检索-过滤权衡。我们呈现了一个案例研究，探讨一个这样的智能体如何根据详细的预先存在的规范实现一个多组件数据系统。存储技术、模式、实体解析算法和检索-过滤策略在实施前已固定；智能体的自主性体现在实施过程中、诊断和修复其引入的缺陷，以及在设计交互时的选择上。在一次会话中，我们记录了五个此类缺陷，按违反的约束和检测方法进行分类。我们进一步在公共的 HotpotQA 基准上评估该架构中指定的一个检索权衡：在排名之前将候选者限制为图识别的实体集与未过滤搜索之间的比较。由于我们缺乏 LLM 访问权限来运行该阶段，因此用基准金证据标签替代实体识别，并报告标准召回率，而不是基准自身的准确性指标。在从 1 到 10 的检索预算和针对 2994 段落的汇总语料库的 100 个问题中，过滤召回率在预算为 3 时达到上限，预计一旦候选者限制为金段落本身，而未过滤搜索即使在预算为 10 时也仅能恢复所需证据的 69%，这一差距在每个测试的预算中都存在，符号检验 p 值小于 0.0001。最后，我们讨论了智能体自主性成功与需要修正的地方，包括一个声称的性能修复在激励它的回归上从未重新测量的实例。
cs.AI / 16 / 2609.01992
ClaimReceipt: Verifying Evidence Sufficiency and Coverage in Agent Evaluations
ClaimReceipt：验证代理评估中的证据充分性和覆盖性
Abstract
Agent evaluations face two distinct evidentiary questions: whether a reported claim is recomputable from retained evidence (sufficiency), and whether the retained records cover the committed experiment set (coverage). Generic logs and hash-linked transcripts answer neither reliably. We introduce ClaimReceipt, a claim-relative receipt specification and selective verifier that binds typed transaction evidence to a signed experiment manifest and returns PASS, INVALID, or INCONCLUSIVE per claim. We freeze the specification before implementation (SHA-256 18d109...b81). On 1,392 historical buyer--seller records, a CR-2 verifier reproduces all five manually labeled audit verdicts, exactly replays 600 deterministic and 792 post-generation records, makes every one of 13 declared field groups non-redundant under tested ablations, and returns the expected result on 11/11 semantic faults with 0/8 false positives. We then run a separate prospective CR-3 epoch: 30 assignments are committed before inference, terminal receipts are signed and chained, and private evidence is encrypted for an auditor. Complete evidence yields coverage and accounting PASS; withholding one terminal receipt returns INCONCLUSIVE_COVERAGE, while withholding all private openings preserves coverage and protocol verification but makes economic claims inconclusive, exactly matching a preregistered prediction. Receipt instrumentation adds 0.021% of model-inference time and 9.9 KB per transaction. A specification-legibility probe indicates that our own frozen specification is not yet unambiguous to an independent reader. Claim verification therefore requires both claim-sufficient evidence and a committed universe against which omissions become visible.
Chinese Translation
代理评估面临两个不同的证据问题：报告的主张是否可以从保留的证据中重新计算（充分性），以及保留的记录是否覆盖了承诺的实验集（覆盖性）。通用日志和哈希链接的记录都无法可靠地回答这两个问题。我们提出了ClaimReceipt，这是一种相对主张的收据规范和选择性验证器，将类型化的交易证据绑定到签名的实验清单上，并根据每个主张返回通过（PASS）、无效（INVALID）或不确定（INCONCLUSIVE）。在实现之前，我们冻结了规范（SHA-256 18d109...b81）。在1,392个历史买卖记录中，CR-2验证器重现了所有五个手动标记的审计裁决，准确重放了600个确定性记录和792个生成后记录，使13个声明的字段组在经过测试的消融下均不冗余，并在11/11个语义故障上返回预期结果，假阳性为0/8。然后，我们运行了一个独立的前瞻性CR-3时期：在推理之前提交了30个任务，终端收据被签名并链接，私有证据为审计员加密。完整证据提供覆盖和会计通过；如果隐瞒一个终端收据则返回不确定覆盖（INCONCLUSIVE_COVERAGE），而隐瞒所有私有开启则保留覆盖和协议验证，但使经济主张变得不确定，完全符合预注册的预测。收据仪器化增加了0.021%的模型推理时间和每个交易9.9 KB的开销。规范可读性探测表明，我们自己的冻结规范对于独立读者仍然不够明确。因此，主张验证需要既有主张充分的证据，又有一个承诺的宇宙，以便使遗漏变得可见。
cs.AI / 17 / 2609.02029
HeadWiseKV: Budgeted Per-Head Cache Residency for Hybrid Long-Context Language Models
HeadWiseKV：混合长上下文语言模型的预算化每头缓存驻留
Abstract
Long-context inference retains a growing key--value (KV) cache during decoding, which consumes substantial GPU memory and can reduce generation throughput. This bottleneck remains in hybrid language models because their residual global-attention layers can dominate context-dependent cache demand. We study how to allocate this state under an aggregate KV-residency budget. We introduce HeadWiseKV, a training-free framework that compresses the residual global KV caches of hybrid language models while preserving their native local, recurrent, and linear paths. It assigns each physical KV head a static, multilevel history window, making cache demand predictable before serving. We formulate this allocation as a restricted operational rate--distortion problem and propose SeqCalib as the core policy-generation algorithm in HeadWiseKV. SeqCalib processes layers in execution order and conditions each decision on the lower-layer policy used at deployment, thereby accounting for interactions across depth. A grouped-cache runtime materializes the selected policy as actual per-head KV residency rather than a mask over a full cache. We evaluate downstream quality across four hybrid long-context models and study physical residency and serving behavior on Qwen3.6-27B. HeadWiseKV retains near-Full-KV RULER and LoCoMo quality across the evaluated models. In the fixed-model systems study, it reduces sampled peak device memory by 8.59\% at a 112K context length and extends the largest verified successful context from 114K to 161K.
Chinese Translation
长上下文推理在解码过程中保留一个不断增长的键值（KV）缓存，这消耗了大量的GPU内存，并可能降低生成吞吐量。这一瓶颈在混合语言模型中依然存在，因为它们的残余全局注意力层可能主导上下文相关的缓存需求。我们研究如何在一个总的KV驻留预算下分配这一状态。我们提出了HeadWiseKV，这是一个无训练的框架，能够在保留混合语言模型的本地、递归和线性路径的同时，压缩残余的全局KV缓存。它为每个物理KV头分配一个静态的多级历史窗口，使得缓存需求在服务之前可预测。我们将这一分配形式化为一个受限的操作速率-失真问题，并提出SeqCalib作为HeadWiseKV中的核心策略生成算法。SeqCalib按执行顺序处理层，并将每个决策的条件设定为在部署时使用的低层策略，从而考虑到深度之间的相互作用。一个分组缓存运行时将所选策略具体化为实际的每头KV驻留，而不是对完整缓存的掩码。我们在四个混合长上下文模型上评估下游质量，并研究Qwen3.6-27B的物理驻留和服务行为。HeadWiseKV在评估的模型中保持了接近Full-KV RULER和LoCoMo的质量。在固定模型系统研究中，它在112K上下文长度下将采样的峰值设备内存减少了8.59\%，并将最大验证成功上下文从114K扩展到161K。
cs.AI / 18 / 2609.02057
Monitoring Web Agents Without Internal Signals: Observable Trajectories and Key-Step Supervision
无内部信号的网络代理监控：可观察轨迹与关键步骤监督
Abstract
Reliable web-agent monitoring is difficult when model-internal uncertainty signals such as token logits are unavailable. In this work, we study prefix-level risk prediction for web agents using observable trajectory signals: given an evolving prefix, estimate whether the current execution remains on track or is tending toward failure. We derive two observable trajectory representations: Macro features summarize cross-step agent--environment behavior and feedback, while Micro features measure the consistency of intention, action, and anticipated state change through repeated black-box queries. Instead of inheriting the final result label, we label the first critical error that remains uncorrected in the observed continuation and is associated with final failure as a key-step boundary, preserving valid early prefixes of failed trajectories as on track. Across WebArena-Lite and Online Mind2Web web agent benchmarks with five open- and closed-source backbones, observable trajectory signals are competitive with internal-signal baselines. The resulting predictors also support early intervention under fixed false-cut budgets and transfer across held-out website categories. These findings show that observable trajectory signals support valuable risk prediction abilities.
Chinese Translation
当模型内部的不确定性信号（如令牌对数）不可用时，可靠的网络代理监控变得困难。在本研究中，我们利用可观察的轨迹信号研究网络代理的前缀级风险预测：给定一个不断演变的前缀，估计当前执行是否保持在轨道上或趋向失败。我们推导出两种可观察的轨迹表示：宏观特征总结了跨步骤的代理-环境行为和反馈，而微观特征通过重复的黑箱查询测量意图、行动和预期状态变化的一致性。我们不直接继承最终结果标签，而是将观察到的延续中第一个未被纠正且与最终失败相关的关键错误标记为关键步骤边界，从而保留失败轨迹的有效早期前缀作为在轨道上。在WebArena-Lite和Online Mind2Web网络代理基准测试中，使用五种开源和闭源骨干网络，可观察轨迹信号与内部信号基线具有竞争力。所得到的预测器还支持在固定的错误削减预算下进行早期干预，并能在保留的网站类别之间迁移。这些发现表明，可观察轨迹信号支持有价值的风险预测能力。
cs.AI / 19 / 2609.02059
DocHop: Benchmarking Out-of-domain Multi-hop Reasoning in Information-Dense Documents
DocHop：在信息密集型文档中基准测试域外多跳推理
Abstract
Multimodal Large Language Models (MLLMs) have achieved strong performance on structured visual understanding tasks such as chart and document question answering. However, existing benchmarks typically evaluate these domains in isolation, leaving underexplored a key capability: whether models can use textual context to determine how chart evidence should be selected, interpreted, and aggregated. We introduce DocHop, a benchmark for integrated chart--context reasoning in document-style images. In DocHop, the document narrative specifies multi-step compositional constraints, while charts provide the corresponding data values. Questions are grounded on a semantic reference label defined in the narrative, requiring models to resolve target entities from context before aggregating evidence across multiple charts. To enable systematic evaluation, we construct DocHop via a stochastic logic-first generation pipeline with controllable reasoning depth and visual density, covering 2,074 examples across six task categories. Experiments on a wide range of proprietary and open-source MLLMs show a substantial gap to human performance: annotators achieve over 90% accuracy, while the best model reaches only 62.83%. Reasoning-enhanced models consistently show improved results, but performance degrades as reasoning complexity increases. Overall, DocHop provides a controlled testbed for challenging multi-hop document reasoning.
Chinese Translation
多模态大型语言模型（MLLMs）在图表和文档问答等结构化视觉理解任务上取得了良好的表现。然而，现有基准通常在孤立的领域中评估这些模型，未能充分探讨一个关键能力：模型是否能够利用文本上下文来确定如何选择、解释和聚合图表证据。我们提出了DocHop，这是一个用于文档风格图像中集成图表-上下文推理的基准。在DocHop中，文档叙述指定了多步骤的组合约束，而图表提供了相应的数据值。问题基于叙述中定义的语义参考标签，要求模型在聚合多个图表的证据之前，从上下文中解析目标实体。为了实现系统评估，我们通过具有可控推理深度和视觉密度的随机逻辑优先生成管道构建了DocHop，涵盖了六个任务类别中的2074个示例。在广泛的专有和开源MLLMs上进行的实验显示出与人类表现之间的显著差距：标注者的准确率超过90%，而最佳模型仅达到62.83%。增强推理能力的模型始终显示出改进的结果，但随着推理复杂性的增加，性能下降。总体而言，DocHop为具有挑战性的多跳文档推理提供了一个受控的测试平台。
cs.AI / 20 / 2609.02060
MineTRACE: An Evidence-Grounded Interactive Reasoning System for Mineral Prospectivity
MineTRACE：一个基于证据的矿产前景交互推理系统
Abstract
Mineral exploration requires integrating heterogeneous geochemical, geophysical, and geological evidence, yet existing prospectivity systems often provide only opaque scores or heatmaps. We present MineTRACE, a web-based system for evidence-grounded exploration of eight commodities: Cu, Au, Ni, W, Sn, Co, Ta, and Mn. Users can explore prospectivity maps, query locations or regions, inspect supporting evidence, and interact through natural language. A transparent expert tree, informed by geological knowledge and known deposits, combines multi-source evidence into interpretable prospectivity scores. For a new location, the conversational assistant retrieves the score and supporting evidence from the analysis pipeline and presents them in natural language. The scorer achieves spatial AUC values of up to 0.917 across different test scenarios, while end-to-end evaluation assesses query accuracy and response grounding. MineTRACE makes public geoscience data easier to access, interpret, and verify, supporting more efficient and transparent mineral exploration.
Chinese Translation
矿产勘探需要整合异质的地球化学、地球物理和地质证据，而现有的前景系统往往仅提供不透明的评分或热图。我们提出了MineTRACE，这是一个基于网络的系统，用于对八种矿产（铜、黄金、镍、钨、锡、钴、铌和锰）的证据驱动探索。用户可以探索前景图，查询位置或区域，检查支持证据，并通过自然语言进行互动。一个透明的专家树，结合了地质知识和已知矿床，将多源证据整合为可解释的前景评分。对于新位置，交互助手从分析流程中检索评分和支持证据，并以自然语言呈现。该评分器在不同测试场景下的空间AUC值高达0.917，而端到端评估则评估了查询的准确性和响应的基础。MineTRACE使公共地球科学数据更易于访问、解释和验证，支持更高效和透明的矿产勘探。
cs.AI / 21 / 2609.02067
ToolGate: An Executable Acceptance Pipeline for Tool-Dependent Scientific Benchmark Construction
ToolGate：一个可执行的接受管道用于工具依赖的科学基准构建
Abstract
Scientific benchmarks are commonly built by domain experts who write tasks and cross-check one another's work, or who adapt existing material from textbooks, published papers, and online resources. These routes can produce strong evaluations, but they require substantial per-item labor. Language models can reduce this repeated work by proposing candidates quickly. The remaining problem is acceptance. We target scientific questions whose answers require computations with specialist software rather than unaided reasoning alone. A candidate is invalid if its script fails or returns a different answer, or trivial if a model answers it without the software. We present ToolGate, which treats every generated item as a proposal and keeps it only if three gates pass. First, an executable solution script must reproduce the proposed answer when run with the scientific software. Second, randomized no-tool screening rejects candidates that models can already solve from the prompt alone. Third, a tool-using agent must solve each survivor within a fixed time limit. We instantiate ToolGate in FEniCSx with 500 generation attempts. The local-verification gate retains 478 candidates. For final reporting, we rescreen this pool after generation: two randomized no-tool screens exclude 222 from the reported pool, and direct GPT-5.5 API calls at medium reasoning (the API default) exclude another 121. Of the remaining 135, a GPT-5.5 Codex CLI agent with access to FEniCSx solves 130; exact deduplication leaves 128 unique protocol survivors. ToolGate turns repeated answer checking and difficulty screening into an auditable process while leaving domain design and final review to experts.
Chinese Translation
科学基准通常由领域专家构建，他们编写任务并相互核对工作，或从教科书、已发表论文和在线资源中改编现有材料。这些途径可以产生强有力的评估，但需要对每个项目进行大量的劳动。语言模型可以通过快速提出候选项来减少这种重复工作。剩下的问题是接受性。我们针对那些答案需要使用专业软件进行计算而非单纯依赖推理的科学问题。如果候选项的脚本失败或返回不同的答案，则该候选项无效；如果模型能够在没有软件的情况下回答，则该候选项被视为琐碎。我们提出了ToolGate，它将每个生成的项目视为一个提案，仅在通过三个门槛后保留。首先，执行解决方案脚本必须在使用科学软件时重现提议的答案。其次，随机无工具筛选会拒绝那些模型仅凭提示即可解决的候选项。第三，使用工具的代理必须在固定时间限制内解决每个幸存者。我们在FEniCSx中实例化ToolGate，进行了500次生成尝试。局部验证门保留了478个候选项。为了最终报告，我们在生成后重新筛选这一池：两次随机无工具筛选排除了222个候选项，直接调用GPT-5.5 API进行中等推理（API默认）又排除了121个。在剩余的135个中，访问FEniCSx的GPT-5.5 Codex CLI代理解决了130个；精确去重后留下128个独特的协议幸存者。ToolGate将重复的答案检查和难度筛选转变为一个可审计的过程，同时将领域设计和最终审查留给专家。
cs.AI / 22 / 2609.02074
CHIME: Credit-Aware Hierarchical Memory Evolution for Long-Horizon Agentic Planning
CHIME：基于信用意识的层次记忆演化用于长时间跨度的自主规划
Abstract
Planning is a central capability that enables agents to decompose complex long-horizon tasks into manageable steps. Test-time search and training-based methods improve planning but incur high inference costs or require expensive training data. Self-evolving memory instead accumulates reusable experience from agent interaction outcomes into an external memory bank, so planning capability keeps improving at inference time without parameter updates. However, existing self-evolving memory methods share an inherent credit assignment problem: they rely on final task outcomes as feedback, but such outcomes conflate plan quality with execution errors and environmental factors, so the accumulated planning experience is often biased and noisy. To address this problem, we propose Credit-Aware Hierarchical Memory Evolution (CHIME), a self-evolving memory framework that maintains a separate planning bank and execution bank and follows an attribute-before-memorize principle: CHIME first attributes each task outcome to the plan, the execution, both, or neither, and then updates only the corresponding memory bank. Extensive experiments on four long-horizon agent benchmarks show that CHIME consistently outperforms state-of-the-art training-based and self-evolving memory baselines. Further analyses reveal several interesting findings. For example, CHIME accumulates effective memory with far fewer items. In addition, the learned memory values faithfully reflect downstream utility: high-quality planning memories are more valuable than execution memories. Finally, the accumulated memory effectively transfers across backbone models. Code will be released at https://github.com/ATH-MaaS/Marco-DeepResearch.
Chinese Translation
规划是使智能体能够将复杂的长时间跨度任务分解为可管理步骤的核心能力。测试时搜索和基于训练的方法可以改善规划，但会产生高推理成本或需要昂贵的训练数据。自我演化记忆则通过将智能体交互结果中的可重用经验积累到外部记忆库中，使得规划能力在推理时不断提高而无需参数更新。然而，现有的自我演化记忆方法存在固有的信用分配问题：它们依赖最终任务结果作为反馈，但这些结果将计划质量与执行错误和环境因素混为一谈，因此积累的规划经验往往存在偏差和噪声。为了解决这个问题，我们提出了基于信用意识的层次记忆演化（CHIME），这是一个自我演化的记忆框架，维护一个单独的规划库和执行库，并遵循“先属性后记忆”的原则：CHIME首先将每个任务结果归因于计划、执行、两者或都不是，然后仅更新相应的记忆库。在四个长时间跨度的智能体基准测试上的广泛实验表明，CHIME始终优于最先进的基于训练和自我演化记忆的基线。此外，进一步分析揭示了一些有趣的发现。例如，CHIME以远少于其他方法的项目积累有效记忆。此外，学习到的记忆值忠实地反映了下游效用：高质量的规划记忆比执行记忆更有价值。最后，积累的记忆在不同的主干模型之间有效转移。代码将发布在 https://github.com/ATH-MaaS/Marco-DeepResearch。
cs.AI / 23 / 2609.02092
Beyond Outcome Gaps: Process-Aware Fairness Diagnosis for LLM-based Multi-Agent Decision Systems
超越结果差距：基于过程的公平性诊断用于基于大型语言模型的多智能体决策系统
Abstract
LLM-based multi-agent systems (MAS) are increasingly considered for high-stakes decision-making, yet outcome-based fairness audits can miss where risks arise within the decision trajectory. We present SCOPED-Hiring, a process-aware fairness diagnosis pipeline for LLM-based hiring MAS. SCOPED-Hiring constructs controlled resume variants, runs role-based hiring committees, logs over 311K structured decision trajectories, and converts trajectory fields into quantitative fairness signals organized by six diagnostic lenses: final outcome, counterfactual, process, pathway, dynamic, and design effects. SCOPED-Hiring reveals that balanced final hire rates can mask hidden trajectory unfairness in multi-agent decision trajectories: career gaps trigger suspicion, proxy cues shape qualification judgments, and identity cues lead to unequal investigation. Targeted repair guided by these diagnoses reduces total layered burden by 72.3% while shifting the hire rate by only 1.86 pp, showing that process diagnosis can guide effective repair. Project Page: https://scoped-hiring-project-page.vercel.app/
Chinese Translation
基于大型语言模型（LLM）的多智能体系统（MAS）在高风险决策中越来越受到重视，但基于结果的公平性审计可能会忽视决策轨迹中风险的来源。我们提出了SCOPED-Hiring，这是一个针对基于LLM的招聘多智能体系统的过程感知公平性诊断管道。SCOPED-Hiring构建了受控的简历变体，运行基于角色的招聘委员会，记录超过31万个结构化决策轨迹，并将轨迹字段转换为由六个诊断视角组织的定量公平性信号：最终结果、反事实、过程、路径、动态和设计效应。SCOPED-Hiring揭示，平衡的最终录用率可能掩盖多智能体决策轨迹中的隐性轨迹不公平：职业空白引发怀疑，代理线索影响资格判断，身份线索导致不平等的调查。基于这些诊断的有针对性修复将总层级负担减少了72.3%，而录用率仅变化1.86个百分点，显示出过程诊断可以指导有效的修复。项目页面：https://scoped-hiring-project-page.vercel.app/
cs.AI / 24 / 2609.02094
MASkills: Continual Skills Optimization for Multi-Agent LLM Systems
MASkills：多智能体大语言模型系统的持续技能优化
Abstract
LLM-based multi-agent systems have shown strong performance on complex tasks, yet continual improvement from interaction experience remains challenging. Existing self-reflection methods build experience memories, but memories are mostly hard to invoke, refine, or scale, while agent skills offer a more actionable unit: structured procedural knowledge that specifies when to act, how to act, and which resources or tools to use. We introduce MASkills, a continual learning framework that optimizes multi-agent LLM systems through agent skills. MASkills presents a new agent-optimization pipeline that integrates skill-conditioned credit assignment, hierarchical credit aggregation, and momentum-smoothed optimization, enabling agent skill libraries to evolve through refinement, induction, consolidation, and pruning. Experiments on HotpotQA, LoCoMo, and GAIA demonstrate the effectiveness of MASkills across multiple agentic tasks. Our code is available at https://github.com/DaRL-GenAI/MASkills
Chinese Translation
基于大语言模型的多智能体系统在复杂任务上表现出色，但从交互经验中持续改进仍然面临挑战。现有的自我反思方法构建了经验记忆，但这些记忆往往难以调用、精炼或扩展，而智能体技能则提供了一个更具可操作性的单元：结构化的程序性知识，明确了何时行动、如何行动以及使用哪些资源或工具。我们提出了MASkills，一个通过智能体技能优化多智能体大语言模型系统的持续学习框架。MASkills呈现了一种新的智能体优化流程，集成了技能条件的信用分配、层次化信用聚合和动量平滑优化，使得智能体技能库能够通过精炼、归纳、整合和修剪不断演化。在HotpotQA、LoCoMo和GAIA上的实验表明，MASkills在多个智能体任务中表现出色。我们的代码可在 https://github.com/DaRL-GenAI/MASkills 获取。
cs.AI / 25 / 2609.02095
READY or Not: Reliable Enterprise Agent Deployment
准备好还是没准备好：可靠的企业代理部署
Abstract
An AI agent can perform well on benchmarks and still be unsuitable for deployment. Existing AI-agent benchmarks measure whether an agent can complete realistic professional work, whereas enterprise deployment asks a different question: whether an agent can meet a required reliability level, under acceptable human oversight, and at tolerable cost. We introduce Reliable Enterprise Agent Deployment (READY), a framework for qualifying AI agents for deployment on enterprise workflows. READY preserves each workflow's own definition of successful execution while applying a common qualification procedure. Given an agent, a workflow, and a class of candidate oversight policies, READY measures the reliability and operating cost of the human-AI system, selects the minimum-cost policy that satisfies a specified reliability target, and statistically qualifies it on held-out cases. The resulting deployment profile characterizes the supported operating point: reliability, human-oversight burden, and cost. READY is implemented as an open testbed that decouples workflow specification, execution, evaluation, and qualification, and runs on existing agent-evaluation infrastructure. In an end-to-end clinical-audit case study spanning 16 agent systems and 750 cases, READY reveals differences hidden by autonomous performance: two systems separated by only 0.3 points in autonomous accuracy (72.8% vs. 72.5%) require 39.2% versus 29.6% human review, respectively, to qualify at the same 76% reliability target under the evaluated oversight policy. READY thus shifts enterprise agent evaluation from how well can the agent perform the work? to under what conditions, and at what cost, can it be reliably deployed? By making those conditions explicit and statistically testable, READY provides a basis for comparing agent systems, setting oversight requirements, and making evidence-based deployment decisions.
Chinese Translation
人工智能代理在基准测试中表现良好，但仍可能不适合部署。现有的人工智能代理基准测试衡量代理是否能够完成现实的专业工作，而企业部署则提出了不同的问题：代理是否能够在可接受的人类监督下，达到所需的可靠性水平，并在可容忍的成本下进行操作。我们引入了可靠企业代理部署（READY），这是一个用于评估人工智能代理在企业工作流中部署资格的框架。READY保留了每个工作流对成功执行的定义，同时应用了统一的资格评估程序。给定一个代理、一个工作流和一类候选监督政策，READY测量人机系统的可靠性和操作成本，选择满足特定可靠性目标的最低成本政策，并在保留的案例上进行统计资格评估。最终的部署配置描述了支持的操作点：可靠性、人类监督负担和成本。READY作为一个开放的测试平台实现，解耦了工作流规范、执行、评估和资格评估，并运行在现有的代理评估基础设施上。在一个涵盖16个代理系统和750个案例的端到端临床审计案例研究中，READY揭示了自主性能掩盖的差异：两个系统在自主准确性上仅相差0.3个百分点（72.8%对72.5%），分别需要39.2%和29.6%的人类审查，以在评估的监督政策下达到相同的76%可靠性目标。因此，READY将企业代理评估的重点从“代理能多好地完成工作？”转变为“在什么条件下，以及以什么成本，能够可靠地部署？”通过明确和可统计检验这些条件，READY为比较代理系统、设定监督要求和做出基于证据的部署决策提供了基础。
cs.AI / 26 / 2609.02116
Semantic Signal-Assisted Inspection and Recovery Allocation in Reverse Logistics
语义信号辅助的逆向物流检验与回收分配
Abstract
Reverse-logistics operators often decide how to inspect and route returned assets before their condition is fully observed, while full inspection consumes scarce labor. Semantic Signal-Assisted Decision Support converts return notes into a condition factor and a signal-quality score that guide inspection depth and recovery allocation under shared labor capacity. We evaluate the framework in three synthetic benchmark scenarios spanning information technology decommissioning, aircraft maintenance, and consumer-electronics returns. Across 30 paired simulation seeds, the keyword implementation improves net recovery value relative to a structured-feature comparator with noisy full inspection while reducing inspection cost in all three scenarios. A risk-blind comparator that skips inspection altogether still records higher value under the benchmark's purely economic objective. At matched inspection cost, score-guided targeting adds 53.9 thousand United States dollars per batch in the aircraft scenario but has little economic effect in the other two configurations; phrase and large language model extractors provide further gains in the aircraft scenario. These results show how narrative evidence can support inspection allocation before recovery decisions are made.
Chinese Translation
逆向物流运营商通常在未完全观察到退回资产的状态之前，就决定如何检验和处理这些资产，而全面检验会消耗稀缺的人力资源。语义信号辅助决策支持将退货单转换为状态因子和信号质量评分，以指导在共享劳动能力下的检验深度和回收分配。我们在三个合成基准场景中评估该框架，这些场景涵盖了信息技术退役、飞机维护和消费电子产品退货。在30对模拟种子中，关键字实现相较于带有噪声的全面检验的结构特征比较器，提高了净回收价值，同时在所有三个场景中降低了检验成本。一个完全跳过检验的风险盲比较器在基准的纯经济目标下仍然记录了更高的价值。在匹配的检验成本下，得分引导的目标在飞机场景中每批增加了53,900美元的收益，但在其他两个配置中经济效应较小；短语和大型语言模型提取器在飞机场景中提供了进一步的收益。这些结果表明，叙述证据可以在做出回收决策之前支持检验分配。
cs.AI / 27 / 2609.02129
Beyond Context Windows: Persistent Discovery Context for Data-Centric Agents
超越上下文窗口：面向数据的智能体的持久发现上下文
Abstract
Data-centric agents repeatedly perform a discovery step before planning or execution: identifying the data objects relevant to a task. Yet successful discovery outcomes are typically discarded rather than reused. We introduce persistent discovery context, a lightweight memory layer that stores prior intent-to-object mappings and reuses them to augment future retrieval. Across three structured data environments, persistent discovery context consistently improves retrieval quality over metadata-only search, remains effective with automatically generated memories, and exposes a reproducible interference failure mode. In lexically sparse domains, memory-only retrieval can even outperform metadata-based retrieval. These findings suggest that discovery outcomes constitute a useful form of reusable context for data-centric agents.
Chinese Translation
面向数据的智能体在规划或执行之前反复进行发现步骤：识别与任务相关的数据对象。然而，成功的发现结果通常被丢弃而不是重用。我们引入了持久发现上下文，这是一种轻量级的记忆层，存储先前的意图与对象映射，并重用它们以增强未来的检索。在三个结构化数据环境中，持久发现上下文始终提高了相较于仅基于元数据搜索的检索质量，且在自动生成的记忆中仍然有效，并揭示了可重复的干扰失败模式。在词汇稀疏的领域中，仅依赖记忆的检索甚至可以超越基于元数据的检索。这些发现表明，发现结果构成了面向数据的智能体可重用上下文的一种有用形式。
cs.AI / 28 / 2609.02133
EmoStance: Response-Side Affective-Orientation Control for Empathetic Response Generation via Emoji Weak Supervision
EmoStance：通过表情符号弱监督实现同理心响应生成的响应侧情感导向控制
Abstract
Empathetic response generation requires models to decide not only what to say, but also how to respond to the previous speaker's affective situation. We formulate this as response-side affective-orientation control and use multi-annotator emoji distributions as weak affective--attitudinal evidence, rather than as output symbols or gold labels, to induce a latent control space that operationally approximates listener stance. We construct EmojiDialogue, an utterance-level extension of EmpatheticDialogues with emoji votes and confidence scores, and propose EmoStance, which models source-side affective expression, predicts a soft response-side orientation from dialogue context and speaker roles, and steers a frozen instruction-tuned LLM through continuous prefix embeddings. In blind pairwise evaluation with 20 annotators and 800 judgments, EmoStance achieves a 62.2% decisive win rate, with the clearest gains in contextual specificity and perceived responsiveness, while remaining complementary to external-knowledge methods. Code, annotation metadata, and reconstruction scripts are available in our GitHub repository: https://github.com/18277390221/EmoStance.
Chinese Translation
同理心响应生成要求模型不仅决定说什么，还要考虑如何回应前一个说话者的情感状态。我们将其表述为响应侧情感导向控制，并使用多标注者的表情符号分布作为弱情感-态度证据，而不是作为输出符号或金标准标签，以诱导一个潜在的控制空间，从而在操作上近似听者的立场。我们构建了EmojiDialogue，这是EmpatheticDialogues的发言级扩展，包含表情符号投票和置信度评分，并提出了EmoStance，该模型在源侧情感表达的基础上，从对话上下文和说话者角色预测软响应侧导向，并通过连续前缀嵌入引导一个冻结的指令调优大语言模型（LLM）。在与20名标注者进行的盲对比评估中，EmoStance实现了62.2%的决定性胜率，尤其在上下文特异性和感知响应性方面表现出明显的提升，同时仍然与外部知识方法互为补充。代码、注释元数据和重建脚本可在我们的GitHub仓库中获取：https://github.com/18277390221/EmoStance。
cs.AI / 29 / 2609.02168
FUSE: An Evaluating Framework for Dangerous Capabilities of LLMs
FUSE：大型语言模型危险能力评估框架
Abstract
Fragmented safety evaluation undermines the governance of dangerous AI capabilities. We present a modular framework that evaluates each model through three orthogonal pipelines---Knowledge ($K$), Defense ($D$), and Harm ($H$)---under a unified protocol, aggregating results into a standardized dangerous-capability profile $\phi$. Pluggable modules supply scenario seeds, knowledge banks, hazard queries, and judge rubrics, while the core evaluation engine remains unchanged across domains; the CB evaluation is complemented by a cyber pilot demonstrating protocol transfer. Instantiating the framework with a chemical-biological (CB) module, we evaluate 12 commercial LLMs from four families. Our first contribution is a horizontal comparison of dangerous capability across models and model families: the three dimensions expose sharply divergent profiles---models with comparable knowledge differ in refusal resilience, and strong defenders do not generate less harmful content when they do comply---while family-level patterns further separate Claude, DeepSeek, and GPT models. The second is a temporal analysis of capability evolution: tracking $K$, $D$, and $H$ against model release dates reveals that dangerous capability has not monotonically declined; newer models deepen knowledge while only partially improving defense, showing that scaling and alignment progress do not uniformly translate into safety. Reliability is established via cross-judge consistency (bootstrap $\rho > 0.79$, 4 of 5 judges) and pipeline orthogonality ($K$--$D$--$H$ inter-correlations $\rho \in [0.32, 0.52]$).
Chinese Translation
碎片化的安全评估削弱了对危险人工智能能力的治理。我们提出了一个模块化框架，通过三个正交管道——知识（Knowledge, $K$）、防御（Defense, $D$）和危害（Harm, $H$）——在统一协议下评估每个模型，将结果聚合成标准化的危险能力档案（profile）$ heta$。可插拔模块提供场景种子、知识库、危害查询和评估标准，而核心评估引擎在各个领域保持不变；CB（化学生物）评估通过一个网络飞行员展示了协议转移。通过化学生物（CB）模块实例化该框架，我们评估了来自四个家族的12个商业大型语言模型（LLMs）。我们的第一个贡献是对模型及模型家族之间危险能力的横向比较：三个维度揭示了截然不同的档案——具有可比知识的模型在拒绝韧性上存在差异，而强防御者在遵守时并不会生成更少的有害内容——同时家族层面的模式进一步区分了Claude、DeepSeek和GPT模型。第二个贡献是对能力演变的时间分析：跟踪$K$、$D$和$H$与模型发布日期的关系显示，危险能力并未单调下降；较新的模型在知识上加深，而防御能力仅部分改善，表明扩展和对齐的进展并不均匀地转化为安全性。通过跨评估者一致性（bootstrap $
ho > 0.79$, 5位评估者中有4位）和管道正交性（$K$--$D$--$H$的相互相关性$
ho ext{in} [0.32, 0.52]$）确立了可靠性。
cs.AI / 30 / 2609.02191
Examining the Vulnerability of Multi-Agent Medical Systems to Human Interventions for Clinical Reasoning
研究多智能体医疗系统在临床推理中对人类干预的脆弱性
Abstract
Human interventions at fault points can alter the diagnostic accuracy of multi-agent medical systems. We defined fault points as moments in AI agent conversations, in which an agent's reasoning became most vulnerable to external influence. Using the MedQA dataset, this study analyzed simulated doctor-patient conversations to measure how interventions shifted reasoning and accuracy. Correct intervention methods showed an improvement in baseline diagnostic accuracy of up to 40%, while incorrect or bias-related interventions degraded performance by up to 6% and increased diagnostic drift and uncertainty. Beyond performance changes, our analysis revealed behavioral similarities between cognitive biases in simulated agent environments and real-world clinical practice. Examples included premature closure and susceptibility to misleading cues. Overall, these findings demonstrate that identifying and guiding fault points with human interventions may provide a mechanism for improving diagnostic robustness in multi-agent medical systems.
Chinese Translation
人类在故障点的干预可能会改变多智能体医疗系统的诊断准确性。我们将故障点定义为人工智能代理对话中的时刻，在这些时刻，代理的推理最容易受到外部影响。利用MedQA数据集，本研究分析了模拟的医患对话，以测量干预如何改变推理和准确性。正确的干预方法使基线诊断准确性提高了多达40%，而错误或偏见相关的干预则使性能下降了多达6%，并增加了诊断漂移和不确定性。除了性能变化，我们的分析还揭示了模拟代理环境中的认知偏差与现实临床实践之间的行为相似性。例子包括过早闭合和对误导性线索的敏感性。总体而言，这些发现表明，识别和引导故障点的人工干预可能为提高多智能体医疗系统的诊断稳健性提供了一种机制。
cs.AI / 31 / 2609.02215
ASCII Attack: Recontextualising Harmful Requests as Artistic Critique in Large Language Models
ASCII攻击：将有害请求重新语境化为大型语言模型中的艺术批评
Abstract
Safety alignment trains large language models to refuse harmful requests stated plainly, but that training is applied mostly to surface form. Requests that only recontextualise the same operational content, changing how the model reads it, are therefore only weakly covered. The ASCII Attack is one such recontextualisation. It is single-turn and black-box: one message, with no access to model internals. It embeds a fully legible harmful request in ASCIl-art characters, presents it as artwork, and asks for feedback. Unlike ArtPrompt, it hides nothing: the request stays readable. The reply is written as artistic critique and can contain operational detail that a plain request would have been refused for. Every framed prompt is paired with a direct-question control, so the contrast is isolated from topic, model and decoding variation. The contrast identifies a bundled surface, not one isolated channel. Across eleven models and eight harm topics, a harm-aware classifier judges 62% of framed prompts harmful against 42% of controls. On the most susceptible model the framed prompt succeeds 93% of the time. A single query matches or exceeds published single-query attacks under four of five harm judges. The effect tracks the model more than the topic and does not diminish with scale. At least one judge dissents from the panel majority on nearly two-thirds of framed rows, which is itself a measurement-validity finding. That pattern is consistent with mismatched generalisation.
Chinese Translation
安全对齐训练大型语言模型拒绝明确表述的有害请求，但这种训练主要应用于表面形式。因此，仅通过重新语境化相同的操作内容、改变模型的理解方式的请求，覆盖程度较弱。ASCII攻击就是这样一种重新语境化。它是单轮的黑箱操作：一条消息，没有访问模型内部。它将一个完全可读的有害请求嵌入ASCIl艺术字符中，呈现为艺术作品，并请求反馈。与ArtPrompt不同，它没有隐瞒任何内容：请求保持可读。回复以艺术批评的形式书写，并可能包含一个普通请求会被拒绝的操作细节。每个框架提示都与一个直接问题控制配对，因此对比是与主题、模型和解码变异隔离的。对比识别的是一个捆绑的表面，而不是一个孤立的通道。在十一种模型和八个有害主题中，一个有害感知分类器判断62%的框架提示是有害的，而控制组为42%。在最易受影响的模型中，框架提示成功率达到93%。单个查询在五个有害评估者中的四个中匹配或超过已发布的单查询攻击。该效应更多地跟随模型而非主题，并且随着规模的扩大并未减弱。在近三分之二的框架行中，至少有一位评审与小组多数意见相悖，这本身就是一个测量有效性的发现。该模式与不匹配的概括一致。
cs.AI / 32 / 2609.02216
PEARL: Path-Entity Aligned Relational Learning with Contextual Subgraphs for Inductive Knowledge Graph Completion
PEARL：基于上下文子图的路径-实体对齐关系学习用于归纳知识图谱补全
Abstract
Inductive knowledge graph completion (IKGC) aims to predict missing links involving entities unseen during training, requiring models to learn transferable relational and structural patterns. Existing subgraph- and path-based approaches often encode relational paths independently of their surrounding query subgraphs, although the predictive relevance of a path may vary across structural contexts. We propose PEARL, a Path-Entity Aligned Relational Learning framework that models paths as context-conditioned reasoning signals. PEARL constructs a query-specific contextual subgraph from the union of the query entities' neighborhoods and uses a large language model (LLM)-guided retriever to distill semantically relevant paths. It then builds a bipartite interaction graph over paths, contextual entities, and a global subgraph representation, allowing path embeddings to adapt to local and global structural evidence. To suppress noise introduced by the enlarged context, PEARL employs a dual-view contrastive objective that promotes representation consistency under stochastic contextual perturbations. Experiments on WN18RR, FB15k-237, and NELL-995 show that PEARL obtains the best average Hits@10 among the compared IKGC methods on all three benchmarks. Ablation studies, efficiency analyses, and case studies further validate the contributions of contextual subgraph modeling, semantic path retrieval, path-entity interaction, and contrastive regularization.
Chinese Translation
归纳知识图谱补全（IKGC）旨在预测在训练过程中未见的实体之间的缺失链接，这要求模型学习可迁移的关系和结构模式。现有的基于子图和路径的方法通常独立于其周围的查询子图对关系路径进行编码，尽管路径的预测相关性可能因结构上下文而异。我们提出了PEARL，一个路径-实体对齐的关系学习框架，将路径建模为上下文条件的推理信号。PEARL从查询实体邻域的并集中构建特定于查询的上下文子图，并使用大型语言模型（LLM）引导的检索器提取语义相关的路径。然后，它在路径、上下文实体和全局子图表示上构建一个双向交互图，使路径嵌入能够适应局部和全局结构证据。为了抑制由扩大上下文引入的噪声，PEARL采用双视图对比目标，促进在随机上下文扰动下的表示一致性。在WN18RR、FB15k-237和NELL-995上的实验表明，PEARL在所有三个基准测试中获得了比较的IKGC方法中最佳的平均Hits@10。消融研究、效率分析和案例研究进一步验证了上下文子图建模、语义路径检索、路径-实体交互和对比正则化的贡献。
cs.AI / 33 / 2609.02217
SkillGLoW: Procedural-Family Skill Consolidation for Self-Improving Agents on Long-Horizon Task Streams
SkillGLoW：面向自我改进代理的程序性技能整合在长时间任务流中的应用
Abstract
LLM agents increasingly self-improve by writing and reusing textual skills, kept either as one global document or as a flat pool of per-task entries, though most of the evidence comes from domains with structurally similar tasks. On long-horizon workloads where each task demands a different solution, the two forms fail in opposite ways: the document collapses into generic discipline, while the pool inflates and its entries stay bound to the instance that wrote them. We argue the missing unit of reuse is the solving procedure shared by a cluster of related tasks, and build SkillGLoW (Global-Local Weave) around it: the local skills a task writes from its own execution are aggregated into procedural families and compressed into de-instantiated global priors, while the instance detail they hold is regenerated per task rather than stored; a commit gate admits a prior only when real execution shows it does not degrade the deployed library. Across four benchmarks (mathematical reasoning, terminal automation, software repair, and embodied control) and three models, the priors gain 17.2 points (hard) over the no-skill baseline on average, with positive gains in all 12 continual-improvement runs, and 18.0 with local regeneration, while the library holds one prior per procedural family, 3.6x more compact than the per-task pool. Under the same protocol GLoW leads a published single-document optimizer on 15 of 21 cells. Unmodified, the library lifts success on unseen ALFWorld tasks from 73.9% to 83.9%, evidence that what transfers is procedure rather than task memory.
Chinese Translation
大型语言模型（LLM）代理通过编写和重用文本技能不断自我改进，这些技能可以保存在一个全球文档中或作为每个任务条目的平面池中，尽管大多数证据来自结构相似的任务领域。在每个任务需要不同解决方案的长时间工作负载中，这两种形式在相反的方式上失败：文档崩溃为通用学科，而池子膨胀且其条目仍然与编写它们的实例绑定。我们认为缺失的重用单元是由一组相关任务共享的解决过程，并围绕此构建了SkillGLoW（全球-局部编织）：任务从自身执行中编写的局部技能被聚合为程序性家族，并压缩为去实例化的全球先验，而它们所持有的实例细节是针对每个任务再生而不是存储的；提交门仅在真实执行表明其不会降低已部署库的性能时才允许先验进入。在四个基准测试（数学推理、终端自动化、软件修复和具身控制）和三个模型中，先验在无技能基线的基础上平均提高了17.2分（困难），在所有12次持续改进运行中均取得了积极的增益，而通过局部再生则提高至18.0，同时库中每个程序性家族仅持有一个先验，比每个任务池紧凑3.6倍。在相同协议下，GLoW在21个单元中领先于已发布的单文档优化器的15个单元。未修改的情况下，库将未见的ALFWorld任务的成功率从73.9%提升至83.9%，这证明了转移的是过程而非任务记忆。
cs.AI / 34 / 2609.02231
PhoenixNest-Video: Evidence-Grounded Multimodal Agent Framework for Automated Video Interview Assessment
PhoenixNest-Video：基于证据的多模态智能体框架用于自动化视频面试评估
Abstract
Interview assessment requires per-criterion judgments grounded in behavioral evidence, yet surging applicant volumes have made human-only evaluation costly and inconsistent, while existing AI approaches yield opaque scores without traceable rationale. We introduce PhoenixNest-Video, an evidence-grounded multimodal agent framework for automated video interview assessment. It builds a semantic video graph as structured working memory, performs rubric-conditioned retrieval with cross-modal verification across visual, audio, and textual streams, and produces per-criterion scores anchored to the candidate's materials. A Scorer trained via Rubrics-based Reinforcement Learning with dual rewards for rubric alignment and score-level differentiation internalizes the discriminative structure of multi-level rubrics. PhoenixNest-Video attains 91.50\% grade-level accuracy on VInterview-2025, outperforming substantially larger proprietary models. A compact, rubric-grounded agent therefore scores candidates in closer agreement with an expert panel than direct prompting of much larger models, and exposes the evidence behind each score for human review.
Chinese Translation
面试评估需要基于行为证据的逐项判断，但申请者数量激增使得仅依靠人工评估变得成本高昂且不一致，而现有的人工智能方法则产生了不透明的评分，缺乏可追溯的理由。我们提出了PhoenixNest-Video，一种基于证据的多模态智能体框架，用于自动化视频面试评估。该框架构建了一个语义视频图作为结构化工作记忆，执行基于评分标准的检索，并通过视觉、音频和文本流之间的跨模态验证，生成与候选人材料相结合的逐项评分。通过基于评分标准的强化学习训练的评分器，利用对评分标准一致性和评分级别差异化的双重奖励，内化了多层评分标准的区分结构。PhoenixNest-Video在VInterview-2025上达到了91.50%的等级准确率，显著优于更大规模的专有模型。因此，这一紧凑的、基于评分标准的智能体在评分候选人时与专家小组的意见更为一致，且为每个评分提供了可供人类审查的证据。
cs.AI / 35 / 2609.02236
PGPO: Potential-Guided Policy Optimization for Multi-Turn Agentic Tasks
PGPO：用于多回合自主任务的潜力引导策略优化
Abstract
Group-based reinforcement learning (RL) has become an effective paradigm for LLM post-training, but in multi-turn agentic tasks with sparse terminal rewards, it often provides coarse credit for intermediate actions. To obtain more fine-grained credit assignment, recent work such as GiGPO introduces step-level advantages for intermediate actions. However, these step-level signals still rely on the final outcome of each individual trajectory. As a result, actions within failed trajectories can remain poorly differentiated, so effective actions can receive the same unfavorable credit as erroneous ones. In this work, we propose Potential-Guided Policy Optimization (PGPO) for multi-turn agentic tasks. PGPO estimates empirical state potentials from anchor-state-group return statistics within each rollout group. It then derives action advantages from potential differences between adjacent states, enabling cross-trajectory credit propagation. This provides finer-grained step-level credit assignment, especially within failed trajectories. Experiments on ALFWorld and WebShop show strong overall performance relative to recent group-based RL methods. Further analysis provides evidence that PGPO yields more informative failure-side credit signals with negligible training overhead.
Chinese Translation
基于群体的强化学习（RL）已成为大规模语言模型（LLM）后训练的有效范式，但在具有稀疏终端奖励的多回合自主任务中，它通常对中间动作提供粗略的信用分配。为了获得更细粒度的信用分配，最近的研究如GiGPO引入了中间动作的步级优势。然而，这些步级信号仍然依赖于每个单独轨迹的最终结果。因此，失败轨迹中的动作可能仍然难以区分，有效的动作可能会与错误的动作获得相同的不利信用。在本研究中，我们提出了用于多回合自主任务的潜力引导策略优化（PGPO）。PGPO从每个回合组的锚点状态组回报统计中估计经验状态潜力。然后，它根据相邻状态之间的潜力差异推导出动作优势，从而实现跨轨迹的信用传播。这提供了更细粒度的步级信用分配，尤其是在失败轨迹中。在ALFWorld和WebShop上的实验显示，PGPO相较于最近的基于群体的RL方法表现出强大的整体性能。进一步的分析提供了证据，表明PGPO产生了更具信息性的失败侧信用信号，并且训练开销微乎其微。
cs.AI / 36 / 2609.02242
Propose to Learn, Learn to Propose: Evaluability-Aware Assistance under Bounded Rationality
提议学习，学习提议：在有限理性下的可评估性意识辅助
Abstract
AI assistants often collaborate by proposing candidate edits, plans, or designs that users evaluate before adoption. Existing assistance methods focus on proposal quality or user-goal inference, often assuming that the user can reliably evaluate any proposal, which can fail in practice because of bounded rationality. We study evaluability-aware proposal planning, where proposals serve both as task interventions and as probes for learning latent preferences and evaluation constraints, where the resulting belief updates then guide later proposals. We formalise this setting as ProSE, a hidden-parameter sequential assistance problem, and instantiate it with a KL-regularised bounded-rational binary response model in which acceptance trades off value gain against a distance-dependent evaluability penalty. Analysing the planning consequence of this likelihood reveals that likely accepted proposals and informative probes need not coincide, which explains why planners that only pursue acceptance systematically underperform. We operationalise ProSE with \textsc{ProSE-Plan}, a depth-2 Bayes-adaptive planner that scores proposals by possible responses and response-induced posterior beliefs. In controlled graph simulations, \textsc{ProSE-Plan} improves over evaluability-unaware and myopic baselines when evaluation cost is the bottleneck, and a probe-commit ablation confirms that our approach selects informative proposals that simpler methods miss. Our results thus identify user evaluability as a planning-relevant dimension of AI assistance, complementary to generation quality and preference inference.
Chinese Translation
人工智能助手通常通过提出候选编辑、计划或设计来进行协作，用户在采纳之前对其进行评估。现有的辅助方法关注于提议质量或用户目标推断，通常假设用户能够可靠地评估任何提议，但在实践中，由于有限理性，这种假设可能会失败。我们研究了可评估性意识的提议规划，其中提议既作为任务干预，又作为学习潜在偏好和评估约束的探针，结果的信念更新随后指导后续提议。我们将这一设置形式化为ProSE，一个隐藏参数的序列辅助问题，并用KL正则化的有限理性二元响应模型进行实例化，其中接受与价值增益和距离依赖的可评估性惩罚之间进行权衡。分析这种可能性的规划后果表明，可能被接受的提议和信息探针不必重合，这解释了为何仅追求接受的规划者系统性表现不佳。我们用 extsc{ProSE-Plan}实现ProSE，这是一个深度为2的贝叶斯自适应规划器，通过可能的响应和响应引起的后验信念对提议进行评分。在受控图形模拟中，当评估成本成为瓶颈时， extsc{ProSE-Plan}优于不考虑可评估性和短视的基线，而探针-承诺消融实验确认了我们的方法选择了简单方法所忽视的信息性提议。因此，我们的结果确定了用户可评估性作为人工智能辅助的一个与规划相关的维度，补充了生成质量和偏好推断。
cs.AI / 37 / 2609.02244
Task-Level Natural Language Priors as Learning Signals for Low-Resource LLM Training
任务级自然语言先验作为低资源大语言模型训练的学习信号
Abstract
Large language models (LLMs) often struggle when low-resource training data are ambiguous or incomplete. Task-level natural-language priors can provide useful guidance in such settings, but existing approaches usually treat these priors as input context rather than as learning signals during training. We propose Prior-Guided Tuning (PGT), a training perspective that incorporates natural-language priors as auxiliary learning signals for low-resource LLM training. Under this perspective, we introduce Contrastive Prior Steering (CPS), which keeps the original supervised objective intact while adding positive and negative prior-conditioned auxiliary losses to encourage task-consistent learning and discourage plausible but misleading alternatives. Experiments on AmbiMath, Jigsaw, and MNLI/HANS show that CPS consistently improves over plain and prompt fine-tuning. On AmbiMath, CPS achieves 97.6% average exact-match accuracy. On Jigsaw, CPS improves average Macro F1 by 9.5 percentage points over standard fine-tuning, and with 1/10 of the experimental training data slightly exceeds full-data plain fine-tuning. On HANS, CPS improves non-entailment accuracy by 8.3 and 5.2 percentage points for LLaMA 3.1 8B and Qwen 2.5 7B, respectively, while maintaining comparable in-domain MNLI accuracy. These results support our central claim: task-level natural-language priors can provide useful guidance as auxiliary learning signals for low-resource LLM training. Our code and data will be publicly available.
Chinese Translation
大型语言模型（LLMs）在低资源训练数据模糊或不完整时常常表现不佳。任务级自然语言先验可以在这种情况下提供有用的指导，但现有的方法通常将这些先验视为输入上下文，而不是在训练过程中作为学习信号。我们提出了先验引导调优（Prior-Guided Tuning, PGT），这一训练视角将自然语言先验作为低资源LLM训练的辅助学习信号。在这一视角下，我们引入了对比先验引导（Contrastive Prior Steering, CPS），该方法在保持原有监督目标不变的同时，添加了正负先验条件的辅助损失，以鼓励任务一致的学习并抑制合理但误导的替代方案。在AmbiMath、Jigsaw和MNLI/HANS上的实验表明，CPS在普通和提示微调上均表现出一致的改进。在AmbiMath上，CPS实现了97.6%的平均精确匹配准确率。在Jigsaw上，CPS的平均宏F1比标准微调提高了9.5个百分点，并且在使用1/10的实验训练数据时略微超过了全数据的普通微调。在HANS上，CPS分别提高了LLaMA 3.1 8B和Qwen 2.5 7B的非蕴含准确率8.3和5.2个百分点，同时保持了可比的领域内MNLI准确率。这些结果支持我们的核心论点：任务级自然语言先验可以作为低资源LLM训练的辅助学习信号提供有用的指导。我们的代码和数据将公开可用。
cs.AI / 38 / 2609.02246
LLM-as-a-Judge Is Not an Oracle: Why Self-Improving Agents Need Deterministic Guardrails
LLM作为裁判并非全知全能：为何自我提升代理需要确定性的防护机制
Abstract
Self-improving agent pipelines have a problem at their center. An optimizer rewrites prompts to score higher, and the score comes from a judge that is itself an LLM. That judge has the last word on whether the system is getting better, and our position is that it has not earned it. The judge should be demoted from oracle to advisor: its verdict becomes one input among several, and every change is gated instead by a deterministic verification layer the judge cannot override. We reached this position by building the alternative and running it. Over months of running autonomous prompt-optimization loops in production across contract analysis, compliance review, and code quality, we cataloged eleven ways the evaluation signal failed, in four classes: judge bias, harness and metric failures, ground-truth errors, and reward hacking. Agents achieved perfect scores by reading cached answer keys from their environment, a 100% pass rate concealing 68% true capability. A corrupted ground-truth label caused the optimizer to delete correct compliance rules to agree with it. A syntactically broken prompt was promoted as the winner because a silent parser fallback improved the metric. Attempts to fix the judge by rewriting its rubric plateaued; the only reliable gain came from a structural constraint on its output order. In response we describe PROCTOR, a Teacher-Student loop in which a stateful orchestrator holds all tool access, stateless subagents diagnose failures and draft mutations they cannot apply, and a Teacher grades those mutations under five deterministic guardrails: hermetic sandboxes, capability-disjoint roles, acceptance checks that outrank the Teacher, frozen holdouts, and canary cases engineered so that a perfect score is itself evidence of cheating. We report the failures this prevented, and, because the Teacher is itself an LLM judge, the failures it did not.
Chinese Translation
自我提升代理流程存在一个核心问题。优化器通过重写提示以获得更高评分，而评分来自于作为裁判的语言大模型（LLM）。该裁判对系统是否变得更好拥有最终决定权，而我们的观点是它并未赢得这一权威。裁判应从神谕（oracle）降级为顾问：其裁决成为多个输入之一，每一次变更都必须经过一个确定性的验证层，该层裁判无法覆盖。我们通过构建替代方案并运行验证了这一观点。在合同分析、合规审查和代码质量等生产环境中，经过数月的自主提示优化循环，我们归纳出评估信号失败的十一种方式，分为四类：裁判偏差、框架和指标失效、真实标签错误以及奖励欺骗。代理通过读取环境中缓存的答案键实现了满分，通过100%通过率掩盖了68%的真实能力。一个被篡改的真实标签导致优化器删除了正确的合规规则以与之匹配。一个语法错误的提示因解析器的静默回退而被提升为优胜者，提升了指标。通过重写评分标准来修正裁判的尝试达到了瓶颈；唯一可靠的提升来自对其输出顺序的结构性约束。为此，我们提出了PROCTOR，一种师生循环架构，其中一个有状态的协调者掌控所有工具访问，无状态的子代理诊断失败并草拟无法自行应用的变更，师者在五个确定性防护机制下对变更进行评分：封闭沙箱、能力不重叠的角色、优于师者的接受检查、冻结的保留项以及设计使得满分本身即作弊证据的金丝雀案例。我们报告了该机制防止的失败案例，同时也指出由于师者本身是LLM裁判，仍存在未能防范的失败。
cs.AI / 39 / 2609.02253
APEx: Distillation of Agent Procedural Experience for Adaptive Deep Research Question Answering
APEx：适应性深度研究问题回答的代理程序经验蒸馏
Abstract
Deep research agents augment large language models with external tools to answer complex, long-horizon questions through multi-turn reasoning. Learning from prior experience is crucial for continual improvement, yet existing methods either retrieve verbose task-specific traces that burden decision-making, or distill procedural skills that remain decoupled from downstream policy adaptation. We propose APEx, a hierarchical experience utilization framework that organizes interaction history into instance-level trajectory memories and category-level procedural skills, and couples them through a closed-loop architecture of Executor, Distiller, and Planner. The three modules are optimized via a three-stage alternating GRPO training paradigm, enabling reward-guided skill distillation rather than fixed-prompt generation. At test time, distilled skills serve as procedural priors for online Planner adaptation through skill-guided test-time reinforcement learning, allowing ground-truth-free self-improvement with skill-alignment regularization to prevent policy drift. Experiments on 7 benchmarks demonstrate that APEx achieves state-of-the-art performance, surpassing GPT-5.4 by 14.7 points and the strongest memory-augmented baseline by 3.0 points.
Chinese Translation
深度研究代理通过外部工具增强大型语言模型，以通过多轮推理回答复杂的长期问题。从以往经验中学习对持续改进至关重要，但现有方法要么检索冗长的任务特定轨迹，增加决策负担，要么蒸馏的程序技能与下游策略适应脱节。我们提出了APEx，一个层次化的经验利用框架，将交互历史组织为实例级轨迹记忆和类别级程序技能，并通过执行者（Executor）、蒸馏器（Distiller）和规划者（Planner）的闭环架构将其耦合。这三个模块通过三阶段交替的GRPO训练范式进行优化，实现了基于奖励的技能蒸馏，而非固定提示生成。在测试时，蒸馏的技能作为在线规划者适应的程序先验，通过技能引导的测试时强化学习实现无真实标签的自我改进，并通过技能对齐正则化防止策略漂移。在7个基准测试上的实验表明，APEx实现了最先进的性能，超越了GPT-5.4 14.7分，超越了最强的记忆增强基线3.0分。
cs.AI / 40 / 2609.02264
Codebook Agent: Amortized Topology Design for LLM Multi-Agent Systems
代码簿代理：针对LLM多智能体系统的摊销拓扑设计
Abstract
Adapting the communication topology of an LLM multi-agent system to each query improves both accuracy and efficiency, yet current designers treat this as conditional graph generation: a variational, autoregressive, or diffusion decoder searches the $N \times N$ adjacency space, and a graph-network proxy trained on utility and a structural cost such as edge count ranks the sampled candidates. We argue that this formulation is misaligned with the problem. Empirically, topologies that survive a reward filter collapse to about six distinct graphs even when the codebook capacity grows from 8 to 64; edge count is negatively correlated with measured token consumption (Pearson $r \approx -0.4$), so sparsifying the graph makes inference more expensive; and a message-passing scorer over agent-profile nodes is adjacency-invariant whenever agents share a profile---the default configuration of published benchmarks---so it cannot rank candidates at all in that regime. These three facts motivate Codebook Agent: a vector-quantized autoencoder compresses successful topologies into a query-independent 16-entry codebook; a reward-weighted MLP maps the query embedding to a distribution over codes; and an MLP proxy that reads the flattened adjacency, regressed on measured utility and per-task normalized token cost, reranks the top decoded candidates in a single batched forward pass. With no iterative search and no message passing at test time, Codebook Agent is the most accurate method on all six benchmarks we compare (84.6 average against 83.0 for the strongest prior designer), emits a topology in 2.4 ms, and uses 21.9--33.2% fewer LLM tokens.
Chinese Translation
将LLM多智能体系统的通信拓扑适应于每个查询可以提高准确性和效率，然而当前设计者将其视为条件图生成：一个变分、自回归或扩散解码器在$N imes N$邻接空间中搜索，而一个基于效用和结构成本（如边数）训练的图网络代理对采样的候选进行排名。我们认为这种表述与问题不符。从经验上看，经过奖励过滤的拓扑在代码簿容量从8增长到64时，仍然会收敛到大约六个不同的图；边数与测量的令牌消耗呈负相关（Pearson $r ext{约} -0.4$），因此稀疏化图形会使推理成本增加；而在代理共享配置文件的情况下，基于代理配置文件节点的消息传递评分器是邻接不变的——这是已发布基准的默认配置——因此在该状态下无法对候选进行排名。这三个事实促使了代码簿代理的提出：一个向量量化自编码器将成功的拓扑压缩为一个与查询无关的16条目代码簿；一个奖励加权的多层感知机（MLP）将查询嵌入映射到代码的分布；一个读取扁平化邻接的MLP代理，基于测量的效用和每任务的标准化令牌成本进行回归，在一次批量前向传递中重新排名前解码的候选。在测试时没有迭代搜索和消息传递，代码簿代理在我们比较的所有六个基准上都是最准确的方法（平均84.6，相比之下最强的先前设计者为83.0），在2.4毫秒内输出一个拓扑，并使用21.9%到33.2%的LLM令牌更少。
cs.AI / 41 / 2609.02273
CoMerge: Conflict-Driven Preference Optimization for Multi-Task Model Merging
CoMerge：基于冲突驱动的多任务模型合并偏好优化
Abstract
Model merging provides an efficient paradigm for constructing multi-task large language models (LLMs) without full model retraining, yet it remains challenged by parameter interference. While existing methods aim to preserve the capabilities of individual expert models and mitigate interference, they generally do not directly learn from the potentially degraded behaviors exposed by naive merging. In this paper, we propose a conflict-driven preference optimization framework for model merging (CoMerge), which reformulates model merging as a preference optimization problem. The approach utilizes a self-supervised, conflict-driven strategy that leverages the defects of naive merging methods (e.g., task arithmetic) as hard negative samples to construct preference pairs without external annotations. By applying preference optimization to refine lightweight, tensor-wise merging coefficients, CoMerge enables the model to mitigate parameter-space conflicts while preserving task-specific capabilities. Extensive experiments show that CoMerge achieves an average normalized performance of 0.9968 on MergeBench, outperforming all evaluated data-free and data-driven model-merging baselines. Furthermore, on Llama-3.1-8B-Instruct, CoMerge yields marked improvements on conflict-sensitive tasks such as instruction following and safety, while remaining highly competitive with full-parameter fine-tuning despite optimizing only 1,445 scalar coefficients.
Chinese Translation
模型合并为构建多任务大型语言模型（LLMs）提供了一种高效的范式，无需完全重新训练模型，然而它仍然面临参数干扰的挑战。现有方法旨在保留单个专家模型的能力并减轻干扰，但通常未能直接从简单合并所暴露的潜在降级行为中学习。在本文中，我们提出了一种基于冲突驱动的模型合并偏好优化框架（CoMerge），将模型合并重新表述为一个偏好优化问题。该方法利用自监督的冲突驱动策略，将简单合并方法（如任务算术）的缺陷作为硬负样本，构建偏好对而无需外部注释。通过应用偏好优化来细化轻量级的张量级合并系数，CoMerge使模型能够在保留任务特定能力的同时减轻参数空间冲突。大量实验表明，CoMerge在MergeBench上实现了平均归一化性能0.9968，超越了所有评估的数据无关和数据驱动的模型合并基线。此外，在Llama-3.1-8B-Instruct上，CoMerge在指令跟随和安全等对冲突敏感的任务上表现出显著改善，同时尽管仅优化了1,445个标量系数，仍与全参数微调保持高度竞争。
cs.AI / 42 / 2609.02292
SCX Router: Streaming Zero-Shot Model Selection with a Decoder-KV Classifier and a Real-World Task Ontology
SCX 路由器：基于解码器-KV 分类器的流式零-shot 模型选择与真实世界任务本体
Abstract
The rapid proliferation of large language models (LLMs) and the growing diversity of their applications presents a unique optimization opportunity: selecting the right model for the task, while optimizing for speed, cost, and quality at a per-task level. However, inference endpoints can vary widely in quality, price, latency, context support, tool use, domain expertise, and reasoning behavior. This heterogeneity makes manual heuristics difficult to maintain and unlikely to achieve consistently favorable speed--cost--quality trade-offs on their own. We introduce \router{}, a lightweight GLiClass-based router that assigns a suitability score to each inference-time model label without autoregressive generation. The released 0.6B-parameter checkpoint combines a Qwen3 decoder with a shallow bidirectional scorer. Its decoder-KV execution path preserves a text-only key--value cache across a session, encodes only new dialogue turns, and evaluates transient candidate-label tokens without adding them to the persistent cache. The same checkpoint also predicts task type, difficulty, reasoning mode, and expected output length, and supports custom zero-shot labels. For task generation, we construct a task ontology with 23 families, 115 task types, 345 routable subtypes, 1,173 synthetic examples, and an orthogonal axis of 30 domains. Using this structure, we generate 150,000 verifier-scored tasks and 15,000 open-ended tasks. We then train the Qwen3 decoder on these tasks, while explicitly separating learned request prediction from per-task policies for attributes such as eligibility, cost, cache reuse, safety, and sovereignty. Across six LiveBench subsets, the router outperforms the mean candidate; on the selected 1,000-task subset, it achieves an aggregate top-1 score of 0.707 versus 0.696 for the strongest fixed model, with benchmark-dependent gains.
Chinese Translation
大型语言模型（LLMs）的快速普及及其应用的日益多样化带来了独特的优化机会：为每个任务选择合适的模型，同时在速度、成本和质量上进行优化。然而，推理端点在质量、价格、延迟、上下文支持、工具使用、领域专业知识和推理行为等方面可能存在很大差异。这种异质性使得手动启发式方法难以维持，并且不太可能单独实现一致的有利速度-成本-质量权衡。我们提出了
outer{}，一种轻量级的基于 GLiClass 的路由器，它在推理时为每个模型标签分配适用性评分，而无需自回归生成。发布的 0.6B 参数检查点结合了 Qwen3 解码器和一个浅层双向评分器。其解码器-KV 执行路径在会话中保留一个仅文本的键-值缓存，仅对新的对话轮次进行编码，并在不将其添加到持久缓存的情况下评估瞬态候选标签令牌。该检查点还预测任务类型、难度、推理模式和预期输出长度，并支持自定义零-shot 标签。为了生成任务，我们构建了一个包含 23 个家族、115 种任务类型、345 种可路由子类型、1,173 个合成示例和 30 个领域的正交轴的任务本体。利用这一结构，我们生成了 150,000 个验证者评分的任务和 15,000 个开放式任务。然后，我们在这些任务上训练 Qwen3 解码器，同时明确将学习到的请求预测与任务属性（如资格、成本、缓存重用、安全性和主权）的每任务策略分开。在六个 LiveBench 子集上，路由器的表现优于平均候选；在选定的 1,000 任务子集上，其综合 top-1 分数为 0.707，而最强固定模型的分数为 0.696，且具有基准依赖的增益。
cs.AI / 43 / 2609.02302
Improving Evaluation Realism with Inference-Time Compute and Deployment Scaffolds
通过推理时计算和部署支架提高评估现实性
Abstract
A core obstacle to alignment evaluation is evaluation awareness: capable models can tell when they are being tested rather than deployed, weakening the conclusions a safety evaluation can support. We present two techniques that make simulated alignment evaluations harder to distinguish from real deployments. Our first technique, critique refinement, spends additional inference-time compute on each simulator action: the simulator generates multiple candidate actions, refines them using feedback from an instance of the target model on how to make them more realistic, and continues the evaluation with the most deployment-like candidate. Our second technique, DISH (Deployment-Imitating SWE-Agent Harness), wraps the target in an agent harness, reducing the gap between simulated and real deployment environments in coding settings. We test the techniques on multiple target models and find that they compose: applying both yields larger realism gains than either alone. Our results show that automated approaches can improve the realism of alignment evaluations, and that these improvements use additional compute more effectively than making the audits longer.
Chinese Translation
对齐评估的一个核心障碍是评估意识：能够的模型能够识别出何时正在被测试而非部署，这削弱了安全评估所能支持的结论。我们提出了两种技术，使得模拟的对齐评估更难以与真实部署区分。我们的第一种技术，批评细化（critique refinement），在每个模拟器动作上花费额外的推理时计算：模拟器生成多个候选动作，利用目标模型的反馈对它们进行细化，以使其更具现实性，并继续使用最接近部署的候选进行评估。我们的第二种技术，DISH（Deployment-Imitating SWE-Agent Harness），将目标包装在一个代理支架中，减少编码环境中模拟和真实部署环境之间的差距。我们在多个目标模型上测试了这些技术，发现它们是可以组合的：同时应用两者比单独应用任一技术能获得更大的现实性提升。我们的结果表明，自动化方法可以提高对齐评估的现实性，并且这些改进比延长审计时间更有效地利用了额外的计算资源。
cs.AI / 44 / 2609.02336
SALA: Semantic-Aware Logical Alignment for Complex Reasoning in In-Context Learning
SALA：用于上下文学习中复杂推理的语义感知逻辑对齐
Abstract
Effective in-context learning (ICL) for complex reasoning relies on selecting the right demonstrations. Traditional retrieval methods based on surface similarity fail to capture the underlying problem-solving logic. Recent logic-based methods address this by matching predefined reasoning steps, but the rigid rules and exact-match criteria is improper to handle flexible or diverse reasoning processes. To address the problem, we propose SALA, a Semantic-Aware Logical Alignment framework. Instead of relying on a fixed inventory, SALA automatically learns task-specific reasoning operations. It then embeds these operations into a continuous semantic space and uses dynamic time warping (DTW) to align the reasoning sequences. This approach allows for soft, flexible matching of reasoning logic while remaining highly interpretable. Experiments across four reasoning benchmarks and three LLMs demonstrate that SALA outperforms existing demonstration selection methods. Further analysis confirms the roles of the operation induction and the logical semantic alignment.
Chinese Translation
有效的上下文学习（ICL）对于复杂推理依赖于选择正确的示例。基于表面相似性的传统检索方法未能捕捉到潜在的问题解决逻辑。最近的基于逻辑的方法通过匹配预定义的推理步骤来解决这一问题，但严格的规则和精确匹配标准不适合处理灵活或多样的推理过程。为了解决这一问题，我们提出了SALA，一个语义感知逻辑对齐框架。SALA不依赖于固定的库存，而是自动学习任务特定的推理操作。然后，它将这些操作嵌入到一个连续的语义空间中，并使用动态时间规整（DTW）对推理序列进行对齐。这种方法允许对推理逻辑进行柔性、灵活的匹配，同时保持高度可解释性。在四个推理基准和三个大型语言模型（LLM）上的实验表明，SALA优于现有的示例选择方法。进一步的分析确认了操作归纳和逻辑语义对齐的作用。
cs.AI / 45 / 2609.02371
Diagnosing with Insights: Structured Analysis of Agent Failures via Behavioral Abstractions
通过洞察进行诊断：基于行为抽象的代理失败结构化分析
Abstract
With the proliferation of LLM agents, the ability to understand and diagnose failures in agents is essential to achieving superior effectiveness and trustworthiness. As agent failures often manifest via long and complex trajectories, manually finding the needles in the haystack is untenable. However, traditional diagnosis techniques for software bugs can hardly address LLM agent failures, while completely relying on LLMs as the judge yields unreliable diagnosis results. To overcome these challenges, this paper presents AGENTSCOPE, a new neuro-symbolic approach for agent failure mode diagnosis. The key principle of AGENTSCOPE is to abstract agent behavior, based on its trajectories, into structured representations. Furthermore, AGENTSCOPE introduces the concept of neural invariants to specify agent behavior properties. AGENTSCOPE leverages LLM-guided reasoning atop the structured representation against neural invariants to pinpoint both the failure step and its type in the trajectory. We show the effectiveness of AGENTSCOPE on publicly available agent failure datasets (Who&When) and a more comprehensive dataset created by us (AgentErrata), where AGENTSCOPE significantly outperforms the current state of the art in fault localization and attribution accuracy. Our work shows that integrating structured abstractions with LLM-guided reasoning enables effective, reliable, and interpretable diagnosis for agent failures.
Chinese Translation
随着大型语言模型（LLM）代理的普及，理解和诊断代理失败的能力对于实现更高效和可信赖的效果至关重要。由于代理失败通常通过长且复杂的轨迹表现出来，手动在海量数据中寻找问题几乎是不可能的。然而，传统的软件缺陷诊断技术难以解决LLM代理的失败问题，而完全依赖LLM作为判断依据则会导致不可靠的诊断结果。为了解决这些挑战，本文提出了AGENTSCOPE，一种新的神经符号方法用于代理失败模式诊断。AGENTSCOPE的关键原则是将代理行为基于其轨迹抽象为结构化表示。此外，AGENTSCOPE引入了神经不变性（neural invariants）的概念，以指定代理行为属性。AGENTSCOPE利用基于结构化表示的LLM引导推理，针对神经不变性来精准定位轨迹中的失败步骤及其类型。我们在公开可用的代理失败数据集（Who&When）和我们创建的更全面的数据集（AgentErrata）上展示了AGENTSCOPE的有效性，结果表明AGENTSCOPE在故障定位和归因准确性方面显著优于当前的最先进技术。我们的研究表明，将结构化抽象与LLM引导推理相结合，可以实现对代理失败的有效、可靠和可解释的诊断。
cs.AI / 46 / 2609.02399
Contrastive Explanations in Quantitative Bipolar Argumentation Frameworks
定量双极论证框架中的对比解释
Abstract
Argumentation frameworks are useful tools for representing and reasoning with information in a variety of settings, e.g. in supplementing AI models as they perform classification tasks, with a notable benefit of providing additional explainability. In this paper, we introduce contrastive explanations for Quantitative Bipolar Argumentation Frameworks (QBAFs), one such formalism. Unlike most existing explanations for QBAFs, which explain the reasoning outcome of a single argument of interest (i.e. a topic argument), contrastive explanations explain the difference between two topic arguments. We introduce a general form of contrastive attribution functions (CAFs) and establish a set of general properties they should satisfy. We introduce CAFs based on removal, gradients and Shapley-values, and study their properties. Finally, to illustrate contrastive explanations, we demonstrate their usefulness in healthcare and bias identification settings.
Chinese Translation
论证框架是用于在多种环境中表示和推理信息的有用工具，例如在补充人工智能模型执行分类任务时，提供额外可解释性的显著好处。在本文中，我们为定量双极论证框架（Quantitative Bipolar Argumentation Frameworks, QBAFs）引入了对比解释，这是一种形式化方法。与大多数现有的QBAFs解释不同，后者通常解释单个感兴趣论证（即主题论证）的推理结果，对比解释则解释两个主题论证之间的差异。我们引入了一般形式的对比归因函数（Contrastive Attribution Functions, CAFs），并建立了一组它们应满足的一般属性。我们基于去除、梯度和夏普利值（Shapley-values）引入CAFs，并研究它们的属性。最后，为了说明对比解释，我们展示了它们在医疗保健和偏见识别环境中的实用性。
cs.AI / 47 / 2609.02421
UTP-Bench: Uncertainty-aware Travel Planning Benchmark
UTP-Bench：考虑不确定性的旅行规划基准
Abstract
Large Language Models (LLMs) have recently demonstrated strong capabilities in automated travel itinerary generation. However, real- world travel planning is inherently uncertain: transportation delays, crowd fluctuations, and unexpected stochastic delays frequently inval- idate otherwise feasible schedules. Existing benchmarks like TravelPlanner and TripCraft assume deterministic environments, evaluating only static constraint satisfaction and ignoring whether generated plans remain robust when such uncertainties arise. To address this limitation, we introduce UTP-Bench1 , a large-scale benchmark for uncertainty-aware travel planning. The dataset integrates real-world travel data spanning 504 cities of India, including attractions, restau- rants, accommodations, and multi-modal trans- portation networks. To model realistic disrup- tions, UTP-Bench incorporates empirical delay distributions and crowd-density patterns col- lected from major cities, enabling evaluation of travel plans under stochastic conditions. We further propose three evaluation metrics, namely Buffer Adequacy Score (BAS), Crowd- Aware Timing Score (CATS), and Transport Delay Absorption Score (TDAS), which quan- tify the ability of generated itineraries to main- tain robustness against transit delays and crowd variability. Experiments with state-of-the-art LLMs like GPT-5, Qwen3, Mistral and Phi-4 re- veal substantial gaps between model-generated and human-authored plans, particularly in tem- poral buffering, delay-aware transportation scheduling, and crowd-sensitive planning.
Chinese Translation
大型语言模型（LLMs）最近在自动化旅行行程生成方面展现了强大的能力。然而，现实世界中的旅行规划本质上是充满不确定性的：交通延误、客流波动和意外的随机延迟常常使得原本可行的计划失效。现有的基准如TravelPlanner和TripCraft假设环境是确定性的，仅评估静态约束满足，忽略了在不确定性出现时生成的计划是否仍然稳健。为了解决这一局限性，我们引入了UTP-Bench，这是一个大规模的考虑不确定性的旅行规划基准。该数据集整合了涵盖印度504个城市的真实旅行数据，包括景点、餐馆、住宿和多模式交通网络。为了模拟现实中的干扰，UTP-Bench结合了从主要城市收集的经验延迟分布和客流密度模式，使得在随机条件下评估旅行计划成为可能。我们进一步提出了三个评估指标，即缓冲充足性评分（Buffer Adequacy Score, BAS）、考虑客流的时间评分（Crowd-Aware Timing Score, CATS）和交通延误吸收评分（Transport Delay Absorption Score, TDAS），这些指标量化了生成的行程在面对交通延误和客流变化时保持稳健性的能力。与最先进的LLMs（如GPT-5、Qwen3、Mistral和Phi-4）的实验显示，模型生成的计划与人类撰写的计划之间存在显著差距，特别是在时间缓冲、考虑延误的交通调度和敏感于客流的规划方面。
cs.AI / 48 / 2609.02459
CivBench: A Long-Horizon Benchmark for Tool-Mediated Agents in Civilization VI
CivBench：用于《文明 VI》中工具介导代理的长时间基准测试
Abstract
We present CivBench, an open-source benchmark for evaluating language model agents in long-horizon, tool-mediated environments through the Model Context Protocol (MCP). A single episode spans 300+ turns and produces thousands of tool calls over a large action space, requiring sustained planning, state monitoring, and execution under partial observability. The environment exposes 76 MCP tools and a narration layer that converts visual game state into structured text. We use CivBench to characterise agent behaviour across four model families in 23 admissible runs. The sample is a pilot, not a model ranking: aggregate outcomes do not reliably discriminate models at this scale. Instead, we introduce two interface-level metrics that the environment makes measurable: Proactive Monitoring Rate (PMR), capturing whether agents actively query latent strategic state, and RAG@10, capturing whether commitments stated in structured planning reflections are executed within ten subsequent turns. Across runs we observe two consistent patterns under a shared playbook protocol. Agents under-monitor strategically relevant state that is available but requires explicit querying: despite playbook guidance to query victory progress every 20 turns, agents do so only every 30 to 75 turns, and in 7 of 20 detectable defeats they failed to query within the 20 turn warning window before game end. Agents also frequently fail to execute near-term commitments stated in their own planning reflections (RAG@10 between 48.2% and 65.8% across models). Both patterns arise despite tool access and explicit guidance, and we interpret them as deviations under instruction rather than absences of capability. We release the environment, scenarios, logs, metrics, and analysis pipeline at https://github.com/lmwilki/civ6-mcp
Chinese Translation
我们提出了CivBench，这是一个开源基准测试，用于通过模型上下文协议（Model Context Protocol, MCP）评估语言模型代理在长时间、工具介导环境中的表现。单个回合跨越300多个回合，并在广泛的动作空间中产生数千次工具调用，这要求持续的规划、状态监控和在部分可观察性下的执行。该环境提供了76个MCP工具和一个叙述层，将视觉游戏状态转换为结构化文本。我们使用CivBench对23个可接受运行中的四个模型系列的代理行为进行特征描述。样本是一个试点，而不是模型排名：在这个规模上，汇总结果并不能可靠地区分模型。相反，我们引入了两个环境可测量的接口级指标：主动监控率（Proactive Monitoring Rate, PMR），捕捉代理是否主动查询潜在的战略状态，以及RAG@10，捕捉在结构化规划反思中声明的承诺是否在随后的十个回合内得以执行。在运行过程中，我们观察到在共享剧本协议下的两个一致模式。代理在战略上相关的可用状态监控不足，但需要明确查询：尽管剧本指导每20个回合查询一次胜利进度，代理仅在每30到75个回合查询一次，并且在20个可检测的失败中，有7次未能在游戏结束前的20回合警告窗口内进行查询。代理还经常未能执行在其自身规划反思中声明的近期承诺（不同模型的RAG@10在48.2%到65.8%之间）。尽管有工具访问和明确指导，这两种模式依然出现，我们将其解释为在指令下的偏差，而非能力的缺失。我们在https://github.com/lmwilki/civ6-mcp发布了环境、场景、日志、指标和分析管道。
cs.AI / 49 / 2609.02620
Collective creativity in hybrid societies
混合社会中的集体创造力
Abstract
Generative AI is changing how cultural artifacts are created and circulated, and with it our understanding of creativity itself. Researchers disagree about whether these tools enrich or impoverish culture, and we argue that much of that disagreement comes from conflating two distinct components of creativity: novelty, a property of single artifacts, and diversity, a property of populations. We argue further that creativity in the context of generative AI is best understood as a property of hybrid collectives, or populations of interacting people and algorithms, rather than of individuals. AI-assisted ideation reliably raises the novelty of individual output while narrowing diversity in the aggregate, but this is not an inevitable consequence of putting machines in the loop. Because humans and models search in complementary ways, mixed groups can outperform and out-diversify groups of either kind alone, and machine-discovered solutions can enter human culture and persist there. What decides the outcome is composition: which agents are present, in what proportion, and how they are connected. The question is no longer whether AI helps or harms creativity, but which mixtures let individual gains accumulate without eroding collective diversity.
Chinese Translation
生成性人工智能正在改变文化艺术品的创造和传播方式，从而也改变了我们对创造力本身的理解。研究者们对于这些工具是丰富还是贫乏文化存在分歧，我们认为这种分歧的很大一部分源于将创造力的两个不同组成部分混淆在一起：新颖性，作为单一艺术品的特性，以及多样性，作为群体的特性。我们进一步认为，在生成性人工智能的背景下，创造力最好被理解为混合集体或交互的人与算法的群体的特性，而非个体的特性。人工智能辅助的构思可靠地提高了个体输出的新颖性，同时在整体上缩小了多样性，但这并不是将机器纳入循环的必然结果。由于人类和模型以互补的方式进行搜索，混合群体能够超越单一类型群体的表现和多样性，而机器发现的解决方案可以进入人类文化并在其中持续存在。决定结果的因素是组成：哪些代理存在，比例如何，以及它们之间的连接方式。问题不再是人工智能是否有助于或损害创造力，而是哪些混合能够让个体收益积累而不侵蚀集体多样性。
cs.AI / 50 / 2609.02649
Loom: Weaving Diagnostic Strands into Free-Text Consensus via Embedding-Space Reweighting
Loom：通过嵌入空间重加权将诊断线索编织成自由文本共识
Abstract
Aggregating noisy, conflicting textual hypotheses into a reliable consensus is a fundamental challenge when deploying NLP systems in real-world industrial settings. While monolithic Large Language Model (LLM) agents offer unbounded expressivity for tasks like Root Cause Analysis (RCA), they suffer from context limits, compounding hallucinations, and prohibitive inference latency. Traditional weak supervision offers statistical rigor but is mathematically restricted to discrete classes. We present Loom, a generative consensus framework deployed for real-world RCA that bridges these paradigms. Loom aggregates open-form hypotheses emitted by modular heuristics (diagnostic templates dynamically populated with episode-specific entities, times, and metrics) by projecting them into a continuous embedding space, and resolves conflicting signals with an iterative centroid-based reweighting algorithm. The resulting consensus weights ground a single lightweight LLM synthesis step. Evaluated on the OpenRCA benchmark, Loom occupies the accuracy--efficiency Pareto frontier: it matches a state-of-the-art autonomous agent on Bank and Market-2 and trails on Market-1 and Telecom, while using a single LLM call per incident on all four datasets ($\sim$26$\times$ faster; $\sim$33$\times$ with an 8B-parameter synthesizer). We discuss our deployment experience, highlighting lessons learned regarding the trade-offs between agentic depth and inference latency, negative results in redundancy detection, and how deterministic consensus fosters trust among Subject Matter Experts~(SMEs).
Chinese Translation
在实际工业环境中部署自然语言处理（NLP）系统时，将嘈杂且相互冲突的文本假设聚合成可靠共识是一项基本挑战。尽管单一的大型语言模型（LLM）代理在根本原因分析（RCA）等任务中提供了无限的表达能力，但它们受到上下文限制、幻觉累积和推理延迟的困扰。传统的弱监督提供了统计严谨性，但在数学上仅限于离散类别。我们提出了Loom，一个用于实际RCA的生成共识框架，弥合了这些范式。Loom通过将由模块化启发式（动态填充特定事件的实体、时间和指标的诊断模板）发出的开放形式假设投影到连续的嵌入空间中来聚合这些假设，并通过迭代的质心重加权算法解决冲突信号。最终的共识权重为单一的轻量级LLM合成步骤提供了基础。在OpenRCA基准测试中评估，Loom占据了准确性与效率的帕累托前沿：在Bank和Market-2上与最先进的自主代理相匹配，而在Market-1和Telecom上稍显落后，同时在所有四个数据集上每个事件仅使用一次LLM调用（约快26倍；使用8B参数合成器时约快33倍）。我们讨论了我们的部署经验，强调了关于代理深度与推理延迟之间权衡的经验教训、冗余检测中的负面结果，以及确定性共识如何在主题专家（SMEs）之间促进信任。
cs.AI / 51 / 2609.02707
Door-in-the-Face Requests and Refusal Behaviour in Large Language Models
门面请求与大语言模型中的拒绝行为
Abstract
Does the door-in-the-face technique work on language models? In humans, a large request that is refused makes a smaller follow-up request more likely to be granted. We test this on nine production models from three providers: each model refuses a large request, then receives a smaller version of the same request, and we compare its compliance with asking directly. The answer depends on the model. On Anthropic's frontier models the technique works: Opus 5 answers the smaller request 65.8% of the time after refusing the larger one, against 29.3% when asked directly. On the frontier models of OpenAI and Google, and on Haiku 4.5, it backfires, lowering compliance by 15.5 to 23.0 points. A control locates the effect: a refused large request on an unrelated topic does less than the related one on all nine models, so the concession itself matters everywhere, while the reaction to having just refused something differs by model family. The technique does not transfer to refusals drawn from public benchmarks. What decides whether a retreat can work is what the request asks for: rewriting 265 refused requests for usable instructions into requests for explanations of the same topic removed the refusal in 263 cases. Human influence techniques port to language models one model family at a time.
Chinese Translation
门面请求技术在语言模型上有效吗？在人类中，一个被拒绝的大请求使得后续的小请求更有可能被接受。我们在来自三个提供商的九个生产模型上进行了测试：每个模型拒绝一个大请求，然后收到同一请求的小版本，我们比较其与直接请求的合规性。答案取决于模型。在Anthropic的前沿模型中，该技术有效：Opus 5在拒绝大请求后，接受小请求的概率为65.8%，而直接请求时为29.3%。在OpenAI和Google的前沿模型以及Haiku 4.5中，该技术适得其反，使合规性降低了15.5到23.0个百分点。一个控制实验确定了这一效应：在九个模型中，拒绝一个与主题无关的大请求的影响小于与主题相关的请求，因此让步本身在所有情况下都很重要，而对刚刚拒绝的请求的反应因模型家族而异。该技术无法转移到来自公共基准的拒绝请求上。决定撤回请求是否有效的因素是请求的内容：将265个被拒绝的可用指令请求重写为同一主题的解释请求，在263个案例中消除了拒绝。人类影响技术逐步迁移到语言模型中，每次针对一个模型家族。
cs.AI / 52 / 2609.02749
Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills
Repo-To-Skill：将GitHub仓库提炼为AI4AI技能
Abstract
Autonomous agents are beginning to carry out machine-learning (ML) research end to end. These agents combine a model backbone with a harness for planning, execution, memory, and verification, but this architecture still leaves domain-specific know-how outside the agent. We call this missing layer operational knowledge, the know-how that separates knowing a method from making it work. That knowledge is not absent from the field. It appears in repositories and papers, but in forms written for human readers and too large to load during a task. Once distilled into compact, verified skills, this knowledge can be reused across tasks rather than rediscovered during each run. We present DisCo, a skill-powered research agent that creates skills and uses them during research. Its distillation runs in two complementary forms: task-agnostic, condensing the field's widely used repositories into reusable skills, and task-oriented, producing the skills a concrete task calls for. The former, applied across the open ecosystem, yields the AREX-Skill Library, with 5,000+ verified skills distilled from 1,000 widely used ML repositories and organized into 20 areas and 178 capability families. With the GPT-5.5 backbone, research harness, and downstream execution budget held fixed, the skill-equipped research agent scores 134.3% higher on MLE-bench, 34.4% higher on PaperBench, 9.2% higher on FrontierCS, and 14.0% higher on PassNet than the same agent without skills. These gains come from adding distilled operating context under that fixed setup.
Chinese Translation
自主智能体开始全面开展机器学习（ML）研究。这些智能体结合了模型骨干和用于规划、执行、记忆和验证的框架，但这种架构仍然将特定领域的知识排除在智能体之外。我们称这种缺失的层次为操作知识，它是将了解一种方法与使其有效运作区分开来的知识。这种知识在该领域并非缺失。它以人类读者可读的形式出现在仓库和论文中，但在任务执行过程中过于庞大而无法加载。一旦被提炼为紧凑、经过验证的技能，这些知识可以在任务之间重用，而不是在每次运行时重新发现。我们提出了DisCo，一个以技能驱动的研究智能体，能够创建技能并在研究中使用它们。其提炼过程以两种互补的形式进行：任务无关的，将该领域广泛使用的仓库浓缩为可重用的技能，以及任务导向的，生成具体任务所需的技能。前者在开放生态系统中应用，产生了AREX-Skill库，从1,000个广泛使用的ML仓库中提炼出5,000多个经过验证的技能，并组织成20个领域和178个能力家族。在固定的GPT-5.5骨干、研究框架和下游执行预算下，配备技能的研究智能体在MLE-bench上得分高出134.3%，在PaperBench上高出34.4%，在FrontierCS上高出9.2%，在PassNet上高出14.0%，相比于没有技能的同一智能体。这些增益来自于在该固定设置下增加提炼的操作上下文。
cs.AI / 53 / 2609.02750
Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems
双层协调反思：一种博弈论方法用于多智能体大语言模型系统
Abstract
Multi-agent LLM systems commonly use an orchestrator to decompose a task for a team of workers and then improve through textual reflection. Despite strong empirical results, these systems lack a unified account of coordination, memory improvement, and the role of external verification. We model orchestrator-worker interaction as a bilevel coordination game: under bounded coupling, the workers' local-update game is an approximate potential game whose equilibrium slack is controlled by decomposition quality. We then analyse reflection as stochastic movement over semantic memory states. For free-form reflection, we derive a finite-time upper bound, prove worst-case tightness, and give a positive lower bound under a falsifiable persistent-harm condition. We further prove an information-theoretic impossibility result: no gate that observes only the generated transcript can improve uniformly over text-indistinguishable environments, whereas an environment-grounded gate can. Motivated by this separation, we introduce Stochastic Reflective Memory Ascent (SRMA), which accepts a candidate memory only after a grounded evaluation risk strictly decreases. Under calibration and non-degenerate corrective mass, SRMA converges exactly, geometrically or polynomially; matching constructions show that both rate regimes are order-tight. We also provide confidence gating for stochastic evaluation and re-anchoring guarantees for piecewise-stationary environments. Experiments instantiate these objects with environment-grounded metrics and test the predicted coordination and drift laws. On 500 SWE-bench instances, the complete Kimi-based system resolves 72.2% versus a 70.8% public mini-SWE-agent reference. Code: https://github.com/YihangChen9/Bilevel-Coordinated-Reflection
Chinese Translation
多智能体大语言模型（LLM）系统通常使用协调者将任务分解给一组工作者，然后通过文本反思进行改进。尽管在实证结果上表现强劲，这些系统在协调、记忆改进和外部验证的角色方面缺乏统一的解释。我们将协调者与工作者的互动建模为一个双层协调博弈：在有限耦合下，工作者的局部更新博弈是一个近似潜在博弈，其均衡松弛程度由分解质量控制。然后，我们将反思分析为在语义记忆状态上的随机移动。对于自由形式的反思，我们推导出有限时间的上界，证明了最坏情况的紧密性，并在可证伪的持续危害条件下给出了正的下界。我们进一步证明了一个信息论的不可能性结果：仅观察生成的记录的门控无法在文本不可区分的环境中均匀改进，而基于环境的门控则可以。受到这一分离的启发，我们引入了随机反思记忆上升（Stochastic Reflective Memory Ascent, SRMA），该方法仅在基于环境的评估风险严格降低后接受候选记忆。在标定和非退化的修正质量下，SRMA 精确收敛，可以是几何或多项式的；匹配构造表明这两种速率模式都是有序紧密的。我们还为随机评估提供了置信门控，并为分段平稳环境提供了重新锚定保证。实验通过基于环境的度量实例化这些对象，并测试预测的协调和漂移规律。在500个SWE-bench实例中，完整的基于Kimi的系统解决了72.2%的问题，而公共的迷你SWE-agent参考为70.8%。代码： https://github.com/YihangChen9/Bilevel-Coordinated-Reflection
cs.AI / 54 / 2609.02760
Measurement-Driven Sub-Network Selection for On-Premise Retrieval-Augmented Factory Agents
基于测量的本地检索增强工厂代理子网络选择
Abstract
On-premise assistants can give factory workers conversational access to machine documentation, but models capable of the task rarely fit shop-floor hardware. We show that after structural compression and retrieval-grounded adaptation, model size is no longer a reliable predictor of adapted answer quality: general capability falls almost linearly with parameter count, while judged retrieval-augmented answer quality does not. We therefore treat deployment as a post-adaptation selection problem, committing one sub-network per device on judged answer quality and measured on-device throughput under a configurable general-capability floor and memory budget; rules that optimize size, speed, or quality alone each give up capability or throughput. A weight-shared supernetwork trained with sandwich-style in-place distillation keeps this selection inexpensive. In a manufacturing-manual case study, extraction costs 13.7 percent of the unpruned model's judged quality and retrieval-grounded distillation returns it to within 4.6 percent, recovering two thirds of the loss, and the same assistant runs across three heterogeneous edge tiers at 1.3 to 5 watts standby.
Chinese Translation
本地助手可以为工厂工人提供与机器文档的对话访问，但能够完成该任务的模型很少适合车间硬件。我们展示了在结构压缩和基于检索的适应之后，模型大小不再是适应后答案质量的可靠预测指标：一般能力几乎与参数数量呈线性下降，而经过评估的检索增强答案质量则没有。因此，我们将部署视为一个适应后选择问题，根据评估的答案质量和在可配置的一般能力下限及内存预算下测量的设备吞吐量，为每个设备承诺一个子网络；单独优化大小、速度或质量的规则都会放弃能力或吞吐量。通过三明治式原地蒸馏训练的权重共享超网络使得这一选择过程成本低廉。在一个制造手册的案例研究中，提取成本为未剪枝模型评估质量的13.7%，而基于检索的蒸馏将其恢复到4.6%以内，回收了三分之二的损失，并且同一助手在三个异构边缘层上以1.3到5瓦的待机功耗运行。
cs.AI / 55 / 2609.02786
SafeEvolve: Harness-Policy Co-Evolution from Agent Experience for Safety Alignment
SafeEvolve：基于代理经验的安全对齐中的操控-策略共演化
Abstract
The performance of LLM-based agents is jointly shaped by the base model and the harness used when interacting with the environment. This exposes them to safety risks in both harmful final responses and multi-step execution trajectories. Existing safety alignment mechanisms often rely on either external harness updates or policy optimization, yet applying either paradigm in isolation fails to bridge runtime control with intrinsic safety. We propose SafeEvolve, an experience-driven self-evolving framework for agent safety alignment. SafeEvolve leverages safety experience from completed on-policy trajectories to drive a continual loop of harness-policy co-evolution. On the harness side, SafeEvolve converts trajectory-level safety evidence into bounded, component-level updates across safety prompt and hierarchical skills, yielding auditable and reversible harness artifacts. On the policy side, SafeEvolve follows a two-stage SFT-RL paradigm, where harness-use SFT bootstraps the policy to actively leverage evolved harness artifacts, and harness-augmented RL further shapes autonomous safety behaviors during multi-step exploration via verifier-decomposed rewards. Through harness-policy co-evolution, SafeEvolve converts safety experience into an evolved runtime harness and improved policy behavior. Experiments on agentic safety benchmarks show that SafeEvolve achieves a stronger safety-utility tradeoff than existing baselines. For Qwen3.5-4B, SafeEvolve achieves a $3\times$ ASR reduction on AgentDojo while improving benign utility from 59.79% to 61.86%.
Chinese Translation
基于大型语言模型（LLM）的代理性能受到基础模型和与环境交互时使用的操控的共同影响。这使得它们在有害的最终响应和多步执行轨迹中面临安全风险。现有的安全对齐机制通常依赖于外部操控更新或策略优化，但单独应用这两种范式无法有效地将运行时控制与内在安全结合起来。我们提出了SafeEvolve，这是一种基于经验的自我演化框架，用于代理的安全对齐。SafeEvolve利用来自已完成的在线轨迹的安全经验，推动操控-策略共演化的持续循环。在操控方面，SafeEvolve将轨迹级安全证据转化为跨安全提示和层次技能的有界组件级更新，从而生成可审计和可逆的操控文档。在策略方面，SafeEvolve遵循两阶段的SFT-RL范式，其中操控使用的SFT为策略提供启动，积极利用演化的操控文档，而增强操控的强化学习（RL）则通过验证者分解奖励进一步塑造多步探索中的自主安全行为。通过操控-策略共演化，SafeEvolve将安全经验转化为演化的运行时操控和改进的策略行为。在代理安全基准上的实验表明，SafeEvolve在安全效用权衡方面优于现有基线。对于Qwen3.5-4B，SafeEvolve在AgentDojo上实现了$3 imes$的ASR降低，同时将良性效用从59.79%提高到61.86%。
cs.AI / 56 / 2609.02805
Large Language Models (LLMs) for Telecom Root Cause Analysis (RCA): A Structured Reasoning Framework for Evidence-Grounded Diagnosis
用于电信根本原因分析（RCA）的大型语言模型（LLMs）：基于证据的诊断结构化推理框架
Abstract
Root cause analysis (RCA) is a critical task in telecom network operations, but diagnosing performance degradations in modern 5G and emerging 6G networks remains challenging due to complex cross-layer dependencies. While large language models (LLMs) offer promising capabilities for reasoning and knowledge integration, directly applying vanilla LLMs to telecom RCA often leads to hallucination, unstable reasoning, and poor alignment with structured network evidence. This work first reviews the evolution of telecom RCA from rule-based and machine learning (ML) approaches to emerging LLM-enabled techniques, and provides an overview of recent paradigms, including structured reasoning, retrieval-augmented knowledge grounding, agentic orchestration, and verifiable reasoning. Building upon these insights, we propose a structured reasoning framework for LLM-enabled telecom RCA that aligns diagnostic reasoning with telecom-specific evidence and domain knowledge. The proposed approach first organizes heterogeneous network telemetry into canonical contexts, and then enforces decision-path reasoning during diagnosis, and finally generates evidence-grounded explanations for reliable fault identification. Experimental results on two 5G RCA datasets, TeleLogs and TelecomTS, demonstrate that the proposed framework consistently improves diagnostic accuracy and decision consistency compared with baseline techniques. These cross-dataset results highlight the importance of structured reasoning design for practical LLM-based RCA systems in next-generation telecom networks.
Chinese Translation
根本原因分析（RCA）是电信网络运营中的一项关键任务，但由于复杂的跨层依赖性，在现代5G和新兴6G网络中诊断性能下降仍然具有挑战性。虽然大型语言模型（LLMs）在推理和知识整合方面展现出良好的能力，但直接将普通LLMs应用于电信RCA往往会导致幻觉、不稳定的推理以及与结构化网络证据的较差一致性。本研究首先回顾了电信RCA从基于规则和机器学习（ML）方法到新兴的LLM驱动技术的发展，并概述了包括结构化推理、检索增强的知识基础、代理协调和可验证推理在内的最新范式。在这些见解的基础上，我们提出了一种用于LLM驱动电信RCA的结构化推理框架，该框架将诊断推理与电信特定证据和领域知识对齐。所提方法首先将异构网络遥测数据组织成规范上下文，然后在诊断过程中强制执行决策路径推理，最后生成基于证据的解释以实现可靠的故障识别。在两个5G RCA数据集TeleLogs和TelecomTS上的实验结果表明，与基线技术相比，所提出的框架在诊断准确性和决策一致性方面始终有所提升。这些跨数据集的结果突显了结构化推理设计在下一代电信网络中实际LLM基础RCA系统中的重要性。
cs.AI / 57 / 2609.02821
AI Contextual Measurement for Recovering Individual and Group-Level Effects: Validation Against Survey Measures and an Occupational Application
AI上下文测量用于恢复个体和群体层面效应：与调查测量的验证及职业应用
Abstract
Researchers increasingly use artificial intelligence to construct measures of social, organizational, and occupational characteristics that are absent from conventional surveys. We propose AICOME, AI COntextual MEasurement, a framework for evaluating whether AI-derived respondent-level measures can recover individual and group-level effects in contextual models. The key idea is that an AI measure constructed at the respondent level can be used to derive its group-level aggregate and its individual deviation, allowing researchers to estimate both between-group and within-group associations rather than treating AI measurement as response prediction alone. We validate the framework using the 2022 China Family Panel Studies (CFPS), where occupations provide the empirical grouping structure and several job-related survey variables provide validation benchmarks. For computer use, foreign-language use, weekly hours, and management responsibilities, we compare survey measures with AI-derived measures in response-level, model-level, contextual, and boundary-condition validations. The results show that AI contextual measurement can recover much of the contextual-model information contained in observed survey variables when rich respondent and job characteristics are available. Weekly hours provides the strongest validation case, with AI-derived measures reproducing the large negative between- and within-occupation associations with satisfaction observed in CFPS. The framework also identifies clear boundary conditions: performance deteriorates when information is restricted to occupation and basic demographics, and recovery is weaker when several related concepts are treated as simultaneously unobserved. The findings suggest that AICOME is most useful for recovering a limited number of theoretically important constructs from rich existing datasets.
Chinese Translation
研究人员越来越多地使用人工智能构建社会、组织和职业特征的测量，这些特征在传统调查中缺失。我们提出了AICOME（AI上下文测量），这是一个评估AI衍生的受访者层面测量是否能够恢复上下文模型中个体和群体层面效应的框架。关键思想是，构建于受访者层面的AI测量可以用于推导其群体层面聚合和个体偏差，使研究人员能够估计群体间和群体内的关联，而不是仅将AI测量视为响应预测。我们使用2022年中国家庭追踪调查（CFPS）验证该框架，其中职业提供了经验分组结构，而多个与工作相关的调查变量提供了验证基准。对于计算机使用、外语使用、每周工作小时数和管理责任，我们在响应层面、模型层面、上下文和边界条件验证中比较了调查测量与AI衍生测量。结果表明，当可用丰富的受访者和工作特征时，AI上下文测量能够恢复观察到的调查变量中包含的大部分上下文模型信息。每周工作小时数提供了最强的验证案例，AI衍生测量再现了CFPS中观察到的与满意度之间的显著负相关性，无论是在职业间还是职业内。该框架还识别出明确的边界条件：当信息仅限于职业和基本人口统计时，性能会下降；当多个相关概念被视为同时未观察时，恢复效果较弱。研究结果表明，AICOME在从丰富的现有数据集中恢复有限数量的理论重要构念方面最为有用。
cs.AI / 58 / 2609.02885
Discriminative World Models for Web Agents
用于网络代理的判别世界模型
Abstract
Recent web agents use world models for test-time action selection by sampling candidate actions, predicting the resulting web states, and ranking them with a ranker model or a Process Reward Model (PRM). These world models are typically trained via supervised next-state prediction to generate fixed representations like HTML or AXTree snapshots. However, this objective is misaligned with the downstream ranker, which relies on predicted states being discriminative across candidates to accurately score them. To address this, we introduce predicted-state matching, a training objective where the predicted representation must distinguish the true resulting state from those reached by alternative actions. We train these models using a branching web-agent dataset derived from WebArena Go-Browse trajectories, where every decision point contains multiple alternative actions and their resulting states. Experiments on our held-out predicted-state matching benchmark show that our approach outperforms world models trained with supervised next-state prediction. We further show that our approach improves PRM-style action ranking on WebPRMBench compared with action-only PRMs and PRMs augmented with supervised-next-state world models. Finally, on WebArena-Lite, using our world model for test-time action selection improves end-to-end task success. Our project page is available at: https://dhruvpendharkar.github.io/dwm/.
Chinese Translation
近期的网络代理通过世界模型进行测试时的动作选择，方法是采样候选动作、预测相应的网络状态，并使用排序模型或过程奖励模型（Process Reward Model, PRM）对其进行排名。这些世界模型通常通过监督的下一个状态预测进行训练，以生成固定的表示，如 HTML 或 AXTree 快照。然而，这一目标与下游排序器的需求不一致，因为排序器依赖于预测状态在候选之间的判别性，以准确评分。为了解决这一问题，我们引入了预测状态匹配（predicted-state matching），这是一种训练目标，要求预测的表示必须能够区分真实的结果状态与其他替代动作所达到的状态。我们使用从 WebArena Go-Browse 轨迹中派生的分支网络代理数据集来训练这些模型，其中每个决策点包含多个替代动作及其结果状态。在我们保留的预测状态匹配基准上的实验表明，我们的方法优于通过监督下一个状态预测训练的世界模型。我们进一步展示了与仅动作的 PRM 和增强了监督下一个状态世界模型的 PRM 相比，我们的方法在 WebPRMBench 上提高了 PRM 风格的动作排名。最后，在 WebArena-Lite 上，使用我们的世界模型进行测试时的动作选择提高了端到端任务的成功率。我们的项目页面可访问： https://dhruvpendharkar.github.io/dwm/.
cs.CL / 1 / 2609.01658
PRO-Step: Step-level Process Reward Optimization for Retrieval-Augmented Generation
PRO-Step：用于检索增强生成的步骤级过程奖励优化
Abstract
Retrieval-Augmented Generation enhances Large Language Models by grounding responses in external knowledge, but multi-hop reasoning remains vulnerable to error propagation, where early retrieval failures confound subsequent steps. Standard outcome-based optimization only rewards the final answer, leaving intermediate retrieval and reasoning errors undetected. While existing process-based methods introduce step-level signals, they still score each step against the final answer, rewarding spurious successes where flawed retrieval coincidentally produces the correct answer. Step-level supervision in RAG requires evaluating both logical validity and evidential grounding at each step. We introduce PRO-STEP: we train a generative PRM that evaluates both dimensions, employ PRM-guided value tree search to construct preference pairs contrasting valid steps against flawed ones, and optimize the policy via step-level Direct Preference Optimization. Experiments on single and multi-hop QA datasets demonstrate that PRO-STEP achieves the best average EM and F1 across five benchmarks. Code, models, and training data are publicly available at https://github.com/keemminnke/PRO-Step.
Chinese Translation
检索增强生成（Retrieval-Augmented Generation）通过将响应与外部知识相结合来增强大型语言模型，但多跳推理仍然容易受到错误传播的影响，早期的检索失败会干扰后续步骤。标准的基于结果的优化仅奖励最终答案，导致中间检索和推理错误未被检测到。虽然现有的基于过程的方法引入了步骤级信号，但它们仍然是将每个步骤的得分与最终答案进行比较，从而奖励那些偶然产生正确答案的错误检索所带来的虚假成功。在检索增强生成中，步骤级监督需要在每一步评估逻辑有效性和证据基础。我们提出了PRO-STEP：我们训练了一个生成的过程奖励模型（PRM），评估这两个维度，采用PRM引导的价值树搜索构建有效步骤与缺陷步骤之间的偏好对，并通过步骤级直接偏好优化（Direct Preference Optimization）来优化策略。在单跳和多跳问答数据集上的实验表明，PRO-STEP在五个基准测试中实现了最佳的平均EM和F1。代码、模型和训练数据可在https://github.com/keemminnke/PRO-Step上公开获取。
cs.CL / 2 / 2609.01687
Learning Evidence Sufficiency Boundaries for Selective Answering in Grounded Multi-Hop QA
学习选择性回答中的证据充分性边界用于基础多跳问答
Abstract
Grounded question answering systems should answer only when the supplied evidence supports the answer. In multi-hop QA, this requirement is difficult because partial evidence can make an unsupported answer appear plausible. We study selective answering through evidence sufficiency boundaries: for the same question, a model should abstain under unsupported or partially supported context, answer when the context first becomes sufficient, and keep the answer stable when redundant evidence is added. We introduce Evidence Sufficiency Boundary Training, a generation-native training framework that constructs ordered evidence chains and supervises the abstain-to-answer transition directly. The method combines level supervision, a boundary flip margin, post-boundary stability, and answer recall protection. We build evidence chains from HotpotQA, 2WikiMultiHopQA, and MuSiQue, then evaluate models with chain metrics, raw QA utility, and unsupported-answer rates on external non-answerable sets. With Qwen2.5-3B-Instruct and LoRA adaptation, Evidence Sufficiency Boundary Training gives the strongest boundary localization among the tested systems, with flip accuracy of 0.807 compared with 0.781 for a token-level abstention baseline. It also achieves the lowest overall unsupported-answer rate on external non-answerable evaluation, 0.095 compared with 0.101 for the same baseline, while retaining competitive raw QA F1. The results show that grounded selective answering improves when training marks the evidence level where refusal should give way to answering.
Chinese Translation
基础问答系统应仅在提供的证据支持答案时进行回答。在多跳问答中，这一要求较为困难，因为部分证据可能使不支持的答案看起来合理。我们通过证据充分性边界研究选择性回答：对于同一问题，模型应在不支持或部分支持的上下文中保持沉默，当上下文首次变得充分时进行回答，并在添加冗余证据时保持答案稳定。我们引入了证据充分性边界训练（Evidence Sufficiency Boundary Training），这是一种生成原生的训练框架，构建有序的证据链并直接监督沉默到回答的过渡。该方法结合了层级监督、边界翻转边际、边界后稳定性和答案回忆保护。我们从 HotpotQA、2WikiMultiHopQA 和 MuSiQue 构建证据链，然后使用链度量、原始问答效用和外部不可回答集上的不支持答案率评估模型。在使用 Qwen2.5-3B-Instruct 和 LoRA 适应的情况下，证据充分性边界训练在测试系统中提供了最强的边界定位，翻转准确率为 0.807，而基于标记的沉默基线为 0.781。它还在外部不可回答评估中实现了最低的不支持答案率，0.095，相比之下，基线为 0.101，同时保持了竞争力的原始问答 F1。结果表明，当训练标记拒绝应让位于回答的证据水平时，基础选择性回答得到了改善。
cs.CL / 3 / 2609.01737
SpeakPay: Domain-Adaptive LoRA Fine-Tuning of Whisper for Low-Resource Nepali Financial Speech Recognition
SpeakPay：针对低资源尼泊尔金融语音识别的领域自适应LoRA微调Whisper
Abstract
Mobile payment applications in Nepal are graphically mediated and largely inaccessible to visually impaired users. This paper presents SpeakPay, a voice-first digital wallet, and documents the central technical contribution: a controlled study of domain adaptation for low-resource financial speech recognition. We introduce NepFinSpeech-403, a 403-utterance dataset of Nepali financial voice commands (send, load, and balance operations spanning 237 unique numerals), and fine-tune Whisper large-v2 with LoRA. On the held-out test set, the domain-adapted model reduces Word Error Rate from 129.95% (zero-shot baseline) to 42.58% --- a 67.2% relative reduction --- and improves Devanagari numeral recognition accuracy from 0.0% to 73.9%. We find that word-level metrics understate the practical task-level impact: domain adaptation improves the Transaction Success Rate from 1.67% to 33.33%, a roughly 20x gain. The improvement is consistent at the individual-utterance level (sign test, $p < 10^{-17}$) and across all command types. A data efficiency analysis shows that as few as 100 domain-specific utterances are sufficient to halve the zero-shot WER, with performance plateauing around 300 examples. Error analysis reveals systematic numeral confusion patterns (zero insertion/deletion, prefix hallucination) that account for the majority of remaining transaction failures. The trained system is deployed as a publicly accessible voice-first web application. All code, dataset, model weights, and this paper are released at https://github.com/subedibiraj/speakpay.
Chinese Translation
尼泊尔的移动支付应用程序在图形上进行中介，且对视觉障碍用户基本上无法访问。本文介绍了SpeakPay，一个以语音为主的数字钱包，并记录了其核心技术贡献：对低资源金融语音识别的领域适应性控制研究。我们引入了NepFinSpeech-403，这是一个包含403个尼泊尔金融语音命令（发送、加载和余额操作，涵盖237个独特数字）的数据集，并使用LoRA对Whisper large-v2进行微调。在保留的测试集上，经过领域适应的模型将字错误率从129.95%（零样本基线）降低到42.58%——相对减少67.2%——并将天城文数字识别准确率从0.0%提高到73.9%。我们发现，基于词的度量低估了实际任务层面的影响：领域适应将交易成功率从1.67%提高到33.33%，约为20倍的提升。该改进在单个语句层面（符号检验，$p < 10^{-17}$）和所有命令类型中均一致。数据效率分析表明，仅需100个领域特定的语句即可将零样本字错误率减半，性能在约300个示例时趋于平稳。错误分析揭示了系统的数字混淆模式（零插入/删除、前缀幻觉），这些模式占据了大多数剩余的交易失败。训练好的系统作为一个公开可访问的以语音为主的网络应用程序进行部署。所有代码、数据集、模型权重以及本文均已发布在 https://github.com/subedibiraj/speakpay。
cs.CL / 4 / 2609.01772
MemeCULT-1K: Benchmarking South Asian Cultural Context and Humor Understanding of Multimodal Models
MemeCULT-1K：南亚文化背景与幽默理解的多模态模型基准测试
Abstract
Meme understanding goes beyond recognizing visual content or literal text; it requires implicit cultural knowledge and pragmatic inference that most vision-language models still lack. We introduce MemeCULT-1K, a multilingual benchmark of 1,000 South Asian memes in Bengali, English, and Hindi, where each meme is paired with a cultural context note and three human-written explanations, along with a supplementary set of 54 Bengali regional dialect memes. We evaluate thirteen popular Vision Language Models (VLMs) under two settings: meme-only and context-aware. Providing minimal cultural context yields consistent gains across all models and languages: mean SBERT similarity improves from 44.6 to 56.4 (+11.8), BLEURT from 37.3 to 42.3 (+5.0), and LLM-as-a-Judge scores from 2.57 to 3.43 out of 5 (+0.86). Fine-grained error analysis reveals that closed-source models fail mainly on entity and reference misidentification, while open-source models are bottlenecked by broader cultural knowledge gaps, with linguistic and phonological failures proving the most context-resistant across both. These results highlight the difficulty of culturally grounded meme understanding and motivate future work on explicit cultural knowledge integration. Our dataset and code are publicly available at TawsifDipto17/MemeCULT-1K.
Chinese Translation
理解表情包不仅仅是识别视觉内容或字面文本；它需要隐含的文化知识和实用推理，而大多数视觉-语言模型仍然缺乏这些能力。我们介绍了MemeCULT-1K，这是一个包含1000个南亚表情包的多语言基准，涵盖孟加拉语、英语和印地语，每个表情包都配有文化背景说明和三条人工撰写的解释，以及一组54个孟加拉地方方言的补充表情包。我们在两种设置下评估了十三个流行的视觉语言模型（VLMs）：仅表情包和考虑上下文。提供最少的文化背景在所有模型和语言中均带来了持续的提升：平均SBERT相似度从44.6提高到56.4（+11.8），BLEURT从37.3提高到42.3（+5.0），LLM-as-a-Judge评分从2.57提高到3.43（满分5分，+0.86）。细致的错误分析显示，闭源模型主要在实体和引用识别上失败，而开源模型则因更广泛的文化知识缺口而受到瓶颈，语言和音韵的失败在两者中都是最具上下文抵抗力的。这些结果突显了文化基础的表情包理解的困难，并激励未来在显性文化知识整合方面的研究。我们的数据集和代码已在TawsifDipto17/MemeCULT-1K公开发布。
cs.CL / 5 / 2609.01788
VakyArth: Evaluating Pragmatic Competence in LLMs across Indic Languages
VakyArth：评估大型语言模型在印度语言中的语用能力
Abstract
Real-world communication often requires pragmatic reasoning: interpreting meanings implied through context and cultural convention rather than stated literally. Existing pragmatic evaluation remains largely limited to English and high-resource languages, leaving Indic languages unexplored despite their linguistic and cultural diversity. We introduce VakyArth, the first pragmatic benchmark for Indic languages, designed as a diagnostic evaluation covering Hindi, Punjabi, Tamil, and Malayalam. VakyArth evaluates models across five phenomena: deixis, speech acts, implicature, social pragmatics, and coherence; through multiple-choice questions, natural language inference, and translation, with all items authored by native speakers. Across multilingual large language models (LLMs) of varying families and sizes, we find consistent failures on pragmatic meanings rooted in Indic linguistic and cultural conventions. Our analysis shows systematic differences across languages and tasks: MCQ accuracy exceeds NLI accuracy in all model-language combinations, translation performance does not reliably track pragmatic understanding, and Indo-Aryan languages show a translation advantage over Dravidian languages. We further show that automatic translation metrics can miss fluent but pragmatically unfaithful outputs, especially for implicature and deixis.
Chinese Translation
现实世界的交流往往需要语用推理：通过上下文和文化惯例解释隐含意义，而非字面表达。现有的语用评估主要限于英语和高资源语言，尽管印度语言在语言和文化上具有多样性，但尚未得到探索。我们介绍了VakyArth，这是第一个针对印度语言的语用基准，旨在作为一种诊断评估，涵盖印地语、旁遮普语、泰米尔语和马拉雅拉姆语。VakyArth通过多项选择题、自然语言推理和翻译，评估模型在五种现象上的表现：指示词、言语行为、暗示、社会语用和连贯性；所有题目均由母语者撰写。在不同家族和规模的多语言大型语言模型（LLMs）中，我们发现这些模型在基于印度语言和文化惯例的语用意义上存在一致性失败。我们的分析显示，不同语言和任务之间存在系统性差异：在所有模型-语言组合中，多项选择题的准确率超过自然语言推理的准确率，翻译性能并不可靠地反映语用理解，且印欧语系语言在翻译上优于达罗毗荼语言。我们进一步表明，自动翻译指标可能会遗漏流畅但在语用上不忠实的输出，尤其是在暗示和指示词方面。
cs.CL / 6 / 2609.01794
Disentangling Statistical Preemption from Entrenchment in Language Models' Avoidance of Overgeneralization
剖析语言模型避免过度概化中的统计预先干预与巩固的关系
Abstract
How do learners avoid overgeneralizations such as Tom laughed me without explicit negative evidence? Constructionists have posited two proposals that describe indirect negative evidence against overgeneralizations: preemption (which privileges exposure to near-synonymous construction---e.g., she made him laugh) vs. entrenchment (all exposures to a verb's grammatical usages, including cases like He laughed). We disentangle these hypotheses by running controlled rearing experiments on LMs trained on child-caregiver conversations, where we systematically remove preemptive vs. non-preemptive evidence. We find that while LMs avoid overgeneralizations, they do not show preemption at a verb-specific level, instead showing weak but non-zero evidence of abstract preemption. Combined with results from analyzing the LMs' training dynamics, we find that LMs treat competing structures as indirect positive---as opposed to negative---evidence in the verb-specific condition. Insofar as preemption is the more plausible route to avoiding overgeneralizations in humans, our results point the need for there to be sensitivities to indirect negative evidence in neural network learners, and suggest new human experiments to test abstract preemption.
Chinese Translation
学习者如何在没有明确负面证据的情况下避免诸如“汤姆让我笑”这样的过度概化？构建主义者提出了两种描述针对过度概化的间接负面证据的提案：预先干预（优先考虑接触近义构式——例如，她让他笑）与巩固（对动词语法用法的所有接触，包括像“他笑了”这样的情况）。我们通过在基于儿童-看护者对话训练的语言模型（LMs）上进行控制性培养实验来剖析这些假设，系统性地去除预先干预与非预先干预的证据。我们发现，尽管语言模型避免了过度概化，但在动词特定层面上并未显示出预先干预的迹象，而是显示出微弱但非零的抽象预先干预证据。结合对语言模型训练动态的分析结果，我们发现，在动词特定条件下，语言模型将竞争结构视为间接的正面证据，而非负面证据。考虑到预先干预是人类避免过度概化的更为合理的途径，我们的结果指出神经网络学习者需要对间接负面证据的敏感性，并建议进行新的实验以测试抽象预先干预。
cs.CL / 7 / 2609.01798
How Do Prompt Variations Affect Energy Consumption in On-Device LLMs?
提示变体如何影响设备端大语言模型的能耗？
Abstract
Large language models (LLMs) are increasingly deployed on mobile devices, making energy efficiency a key deployment constraint, yet the energy impact of prompt design remains underexplored. This paper aims to understand how two prompt properties, cognitive load and phrasing pattern, shape the energy behavior of on-device LLM inference. We conduct a broad empirical study covering prompt properties, datasets, models, and devices, with phase-level profiling that separates prefill and decode energy. We find that cognitive load primarily affects the energy cost per token, while phrasing pattern affects energy largely through token usage. Our energy-quality analysis further shows that prompt design reshapes the attainable frontier differently across models, highlighting the need for model-aware prompt design in energy-efficient on-device LLM inference. Code, datasets, and scripts are available at https://amai-gsu.github.io/PromptProperty/.
Chinese Translation
大型语言模型（LLMs）越来越多地部署在移动设备上，使得能效成为一个关键的部署约束，但提示设计对能耗的影响仍然未被充分探讨。本文旨在理解两个提示特性——认知负荷和措辞模式——如何影响设备端LLM推理的能量行为。我们进行了一项广泛的实证研究，涵盖了提示特性、数据集、模型和设备，并进行了阶段级的能耗分析，区分了预填充和解码能耗。我们发现，认知负荷主要影响每个标记的能量成本，而措辞模式则主要通过标记使用影响能耗。我们的能量-质量分析进一步表明，提示设计在不同模型中重塑了可达边界，突显了在能效设备端LLM推理中需要模型感知的提示设计。代码、数据集和脚本可在 https://amai-gsu.github.io/PromptProperty/ 获取。
cs.CL / 8 / 2609.01810
TalkFa: A Unified Benchmark for Farsi Dialogue Generation and Understanding
TalkFa：波斯语对话生成与理解的统一基准
Abstract
Farsi, spoken by more than 120 million people, lacks a comprehensive benchmark for dialogue generation and understanding. We introduce TALKFA, a unified benchmark comprising three complementary datasets: (1) WIKI-FADIAL, 4.2K Wikipedia-grounded dialogues for knowledge-grounded generation; (2) DAILYDIALOG-FA, 6.6K dialogues annotated for dialogue acts and emotions; and (3) PLAYDIAL-FA, 2.1K theatrical dialogues with sentiment labels. While LLMs assist data construction, every dialogue undergoes multi-stage review and revision by native Farsi speakers, and only the final human-approved dialogues are released. Experiments with six LLAMA and MISTRAL models show that LoRA substantially improves dialogue generation while requiring only 25-50% of the training data to recover over 90% of the final performance gains. Across classification tasks, FABERT achieves the best dialogue-act performance, LORA-MISTRAL-7B performs best on emotion recognition, and MISTRAL-24B achieves the highest sentiment score. Human evaluation and independent external validation demonstrate the reliability of the benchmark, while comparisons with GPT-4.1 as an LLM judge reveal that automatic metrics substantially overestimate dialogue quality. Zero-shot evaluation with frontier LLMs further shows that TalkFa remains a challenging benchmark. We will release all datasets, annotation guidelines, code, and checkpoints.
Chinese Translation
波斯语由超过1.2亿人使用，但缺乏全面的对话生成与理解基准。我们推出了TALKFA，这是一个统一基准，包含三个互补的数据集：(1) WIKI-FADIAL，包含4200个基于维基百科的知识驱动对话；(2) DAILYDIALOG-FA，包含6600个标注了对话行为和情感的对话；以及(3) PLAYDIAL-FA，包含2100个带有情感标签的戏剧对话。虽然大型语言模型（LLMs）在数据构建中提供了帮助，但每个对话都经过母语波斯语者的多阶段审查和修订，最终仅发布经过人类批准的对话。对六个LLAMA和MISTRAL模型的实验表明，LoRA显著改善了对话生成，同时仅需25-50%的训练数据即可恢复超过90%的最终性能提升。在分类任务中，FABERT在对话行为性能上表现最佳，LORA-MISTRAL-7B在情感识别上表现最佳，而MISTRAL-24B则获得了最高的情感评分。人类评估和独立外部验证证明了该基准的可靠性，而与作为LLM评判者的GPT-4.1的比较显示，自动指标显著高估了对话质量。与前沿LLMs的零样本评估进一步表明，TalkFa仍然是一个具有挑战性的基准。我们将发布所有数据集、标注指南、代码和检查点。
cs.CL / 9 / 2609.01828
AVERT: Audio-Verified Adjudication for Spoken Dialogue State Tracking
AVERT：基于音频验证的口语对话状态跟踪裁决
Abstract
Spoken dialogue state tracking recovers slot-value pairs from speech, where ASR errors concentrate in entity values and persist across turns, making it both a generation and an editing problem. A strong per-turn text editor corrects much of this but, operating on the transcript alone, leaves three recoverable errors: a value predicted inconsistently across turns, an omitted slot, and a value the audio does not support. We present AVERT, which scores each candidate value by combining cross-turn agreement with a trained audio-conditioned verifier and resolves the three error types with three operators, vote, add, and swap, each restricted to the slots where its error is common. On SpokenWOZ, a base speech-LLM reaches 33.04 JGA, a text editor 38.34, and AVERT 40.13, without retraining either. This is in the range of a 1B end-to-end system that consumes the full spoken history (39.32), though AVERT uses two 1B decoders rather than one. The audio verifier contributes a statistically significant gain, and restricting each operator to a selected slot subset matters: removing it lets unrestricted voting overwrite correct categorical values and fall below the editor.
Chinese Translation
口语对话状态跟踪从语音中恢复槽值对，其中自动语音识别（ASR）错误集中在实体值上，并在多个轮次中持续存在，这使其既是生成问题也是编辑问题。一个强大的逐轮文本编辑器能够纠正大部分错误，但仅在转录文本上操作，留下三种可恢复的错误：在多个轮次中不一致预测的值、遗漏的槽以及音频不支持的值。我们提出了AVERT，它通过结合跨轮次一致性与训练的音频条件验证器来评分每个候选值，并使用三种操作符（投票、添加和交换）解决这三种错误类型，每种操作符仅限于其错误常见的槽。在SpokenWOZ数据集上，一个基础的语音大语言模型（speech-LLM）达到33.04的JGA，一个文本编辑器达到38.34，而AVERT则达到了40.13，且未进行重新训练。这一结果接近于一个消耗完整口语历史的1B端到端系统（39.32），尽管AVERT使用了两个1B解码器而不是一个。音频验证器显著提高了性能，并且将每个操作符限制在选定的槽子集上是重要的：去除这一限制会导致不受限制的投票覆盖正确的类别值，从而低于编辑器的表现。
cs.CL / 10 / 2609.01832
Interpretable Symptom Vectors for Depression in a Large Language Model
大型语言模型中的可解释抑郁症状向量
Abstract
Patients with depression present with diverse symptom profiles, yet clinical practice routinely reduces this variation to a single severity score. Large language models (LLMs) can potentially capture various symptoms and their severity from patient speech. However, how depressive symptoms are represented inside LLMs remains poorly understood, limiting clinical trust. To examine whether internal model activations match clinician judgment, we analyzed the residual stream of Gemma-3-27B-PT using mechanistic interpretability techniques. Recording activations across symptom descriptions drawn from validated clinical instruments, we found that symptom groups geometrically separated the most at layer 21 across multiple distance metrics. Using Semantic Projection, we then projected held-out naturalistic text onto Symptom Vectors constructed from these instruments. The resulting per-symptom coefficients preserved clinician-annotated rank ordering across mood, somatic, and suicidality axes. Furthermore, a single depression vector in Layer 21 separates held-out depressive from non-depressive text (AUC = 0.789), which can be used as an emotional valence gate that restricts symptom projection to depressive speech. These results reveal a decorrelated, clinician-aligned symptom signal readable directly from internal activations, offering a mechanistic foundation for interpretable depression-assessment tools.
Chinese Translation
抑郁症患者表现出多样的症状特征，但临床实践通常将这种变化简化为单一的严重程度评分。大型语言模型（LLMs）有潜力从患者的言语中捕捉各种症状及其严重程度。然而，抑郁症状在LLMs内部的表示方式仍然不甚清楚，这限制了临床信任。为了检验内部模型激活是否与临床医生的判断相符，我们使用机械可解释性技术分析了Gemma-3-27B-PT的残差流。通过记录来自经过验证的临床工具的症状描述的激活情况，我们发现症状组在多个距离度量下在第21层几何上分离得最明显。随后，我们使用语义投影将保留的自然文本投影到这些工具构建的症状向量上。结果显示，每个症状的系数保留了临床医生标注的情绪、躯体和自杀倾向轴的排名顺序。此外，第21层中的单一抑郁向量能够将保留的抑郁文本与非抑郁文本分开（AUC = 0.789），可作为限制症状投影到抑郁言语的情感效价门。这些结果揭示了一种去相关的、与临床一致的症状信号，可以直接从内部激活中读取，为可解释的抑郁评估工具提供了机械基础。
cs.CL / 11 / 2609.01833
Candidate Generation and Definition-Guided Verification for Sentence-Level Depression Symptom Recognition
基于候选生成和定义引导验证的句子级抑郁症状识别
Abstract
Sentence-level recognition of depression symptoms is challenging because similar expressions can differ in symptom relevance, and language-model inference is insufficiently grounded in diagnostic definitions. This study proposes a two-stage framework separating symptom-candidate generation from definition-grounded verification. A contrastively fine-tuned sentence encoder generates a symptom candidate per sentence, and a fine-tuned language model verifies whether the candidate is present or absent using the sentence, its context, and a candidate-specific diagnostic definition, checking its judgment against that definition before answering. Evaluated against encoder, inference-based, medical, and general LLM baselines and a matched single-stage supervised classifier, the proposed pipeline attains the best accuracy and F1 scores of all methods, with rationales matching expert-authored annotations. A preliminary clinical audit indicates moderate alignment with diagnostic definitions, with explanation quality strongly dependent on prediction correctness. The results support decomposing symptom recognition into candidate generation and definition-grounded verification, though performance remains limited for rare categories.
Chinese Translation
句子级抑郁症状识别面临挑战，因为相似的表达可能在症状相关性上有所不同，而语言模型推理不足以与诊断定义相结合。本研究提出了一种两阶段框架，将症状候选生成与基于定义的验证分开。对比微调的句子编码器为每个句子生成一个症状候选，而微调的语言模型则使用句子、其上下文和候选特定的诊断定义来验证候选是否存在，在回答之前检查其判断与该定义的一致性。与编码器、基于推理的医学和通用LLM基线以及匹配的单阶段监督分类器进行评估，所提议的管道在所有方法中获得了最佳的准确率和F1分数，其理由与专家撰写的注释相匹配。初步的临床审计表明与诊断定义的中等一致性，解释质量强烈依赖于预测的正确性。结果支持将症状识别分解为候选生成和基于定义的验证，尽管对于稀有类别的性能仍然有限。
cs.CL / 12 / 2609.01846
Cite or Decline: A Strict Course-Grounded Chatbot for STEM Lecture Videos
引用或拒绝：一个基于课程的严格聊天机器人用于STEM讲座视频
Abstract
Recorded lecture videos, often enhanced with search and summarization features, are a standard study resource. However, students cannot easily ask course specific questions or verify answers against an instructor's lecture. We report a semester-long deployment of VideoPoints platform with a retrieval-augmented chatbot that answers from course lecture materials and returns timestamped citations. The chatbot retrieves only from the active course, uses chapter summaries to guide transcript ranking, and returns clickable timestamped citations. Students used it for quick lookups and exam review. Across 833 messages, 70.5% included citations, none crossed a course boundary, and when no lecture evidence matched, the chatbot usually declined rather than answering. Among the users, citations were the most consistently useful feature, while practice-question generation was the strongest unmet request. We also evaluated the design on the real-world test split of EduVidQA, a public multimodal benchmark for lecture-video question answering. Our design improved correct-lecture retrieval by 6.3 percentage points over dense-only retrieval. Together, the results show that effective deployment depends on course isolation, supported citations, and alignment with students' study practices.
Chinese Translation
录制的讲座视频，通常配备搜索和摘要功能，是一种标准的学习资源。然而，学生无法轻易提出与课程相关的问题或验证答案是否与讲师的讲座相符。我们报告了在一个学期内部署VideoPoints平台的情况，该平台配备了一个增强检索的聊天机器人，能够从课程讲座材料中回答问题并返回带时间戳的引用。该聊天机器人仅从当前课程中检索信息，使用章节摘要来指导文本排名，并返回可点击的时间戳引用。学生们使用它进行快速查找和考试复习。在833条消息中，70.5%包含引用，没有一条跨越课程边界，当没有讲座证据匹配时，聊天机器人通常选择拒绝回答。在用户中，引用是最一致有用的功能，而生成练习题是最强烈未满足的请求。我们还在EduVidQA的真实世界测试集上评估了该设计，这是一个公共的多模态讲座视频问答基准。我们的设计使正确讲座检索提高了6.3个百分点，相较于仅基于密集检索的结果。综合来看，这些结果表明，有效的部署依赖于课程隔离、支持引用以及与学生学习实践的一致性。
cs.CL / 13 / 2609.01867
Thinking effort aligns between humans and reasoning models in abductive reasoning
人类与推理模型在溯因推理中的思维努力对齐
Abstract
A major question in cognitive modeling concerns the behavioral alignment between large language models and humans across linguistic and non-linguistic tasks. Unlike standard LLMs, large reasoning models (LRMs) are optimized with reinforcement learning from verifiable rewards, encouraging correct solutions to reasoning tasks rather than preference-aligned responses. Recent work (de Varda et al., 2025) investigates the cost of thinking in humans and LRMs by comparing human reaction times with model reasoning traces across a range of reasoning tasks. We isolate this alignment by turning to abductive reasoning: unlike deductive tasks, its difficulty cannot be inferred from formal structure and offers no shortcuts a model could exploit to mimic effort without genuine search, providing firmer ground for empirical claims of shared effort. We find further evidence of alignment between LRM and human reasoning effort, as well as evidence that models and humans tend to make similar errors. Finally, we show that decoding methods that let models explore multiple reasoning paths increase alignment in reasoning cost between humans and LRMs across the three models tested.
Chinese Translation
认知建模中的一个主要问题是大型语言模型与人类在语言和非语言任务中的行为对齐。与标准的大型语言模型（LLMs）不同，大型推理模型（LRMs）通过可验证奖励的强化学习进行优化，鼓励在推理任务中找到正确的解决方案，而非偏好对齐的响应。最近的研究（de Varda et al., 2025）通过比较人类反应时间与模型推理轨迹，探讨了人类和LRMs在一系列推理任务中的思维成本。我们通过转向溯因推理来隔离这种对齐：与演绎任务不同，其难度无法从形式结构推断，并且没有模型可以利用的捷径来模仿努力而不进行真实搜索，从而为共享努力的实证主张提供了更坚实的基础。我们发现LRM与人类推理努力之间的对齐有进一步的证据，同时也发现模型和人类倾向于犯类似的错误。最后，我们展示了让模型探索多条推理路径的解码方法增加了人类与LRMs在三种测试模型中的推理成本对齐。
cs.CL / 14 / 2609.01878
GAPS: Dimension-Level Gates for Conditional Activation Steering
GAPS：用于条件激活引导的维度级门控
Abstract
Activation steering suppresses undesired behaviors in language models by adding a steering vector to the hidden state during generation. Recent conditional methods such as CAST and DSAS improve the behavior-capability trade-off by deciding when to intervene, but once active, they apply the full dense vector to all hidden dimensions, regardless of whether a neuron carries concept information or already lies in the desired regime. We introduce dimension-level conditioning as a complementary axis of selectivity that also decides which neurons to intervene on. Our method, GAPS (Gated Activation steering via Posterior and Separability), combines two training-free gates: a static separability gate that restricts steering to neurons with statistically reliable concept information (via AUROC), and a dynamic posterior gate that steers a neuron only when its current activation is better explained by the undesired concept under a Gaussian model. The gates add O(D) overhead per token, and they plug into existing conditional methods. On toxicity mitigation (RealToxicityPrompts) and concept removal (OneSeC) with Gemma-3 (4B) and Qwen-3 (1.7B), GAPS consistently matches or improves the Pareto front of its token-level counterparts; under a fixed capability budget, DSAS+GAPS reduces Gemma-3's toxicity rate from 6.52% to 0.48%, versus 3.52% for DSAS alone. Ablations attribute most of the gain to the posterior gate.
Chinese Translation
激活引导通过在生成过程中将引导向量添加到隐藏状态中来抑制语言模型中的不良行为。最近的条件方法如CAST和DSAS通过决定何时干预来改善行为与能力之间的权衡，但一旦激活，它们会将完整的稠密向量应用于所有隐藏维度，而不考虑神经元是否携带概念信息或是否已经处于期望的状态。我们引入了维度级条件作为选择性的补充轴线，以决定干预哪些神经元。我们的方法GAPS（通过后验和可分离性进行门控激活引导）结合了两个无训练的门控：一个静态可分离性门控，限制引导仅应用于具有统计上可靠概念信息的神经元（通过AUROC），以及一个动态后验门控，仅在神经元的当前激活在高斯模型下更好地被不良概念解释时才进行引导。这些门控为每个标记增加O(D)的开销，并且可以与现有的条件方法结合使用。在毒性缓解（RealToxicityPrompts）和概念移除（OneSeC）任务中，使用Gemma-3（4B）和Qwen-3（1.7B），GAPS始终与其标记级别的对应方法匹配或改善Pareto前沿；在固定能力预算下，DSAS+GAPS将Gemma-3的毒性率从6.52%降低到0.48%，而仅使用DSAS时为3.52%。消融实验将大部分增益归因于后验门控。
cs.CL / 15 / 2609.01918
Grounded, Compute-Efficient LLM Policy Agents for Energy-Poverty Equity in Physically-Constrained Peer-to-Peer Energy Markets
基于实证的、计算高效的LLM政策代理以实现物理约束下的点对点能源市场中的能源贫困公平
Abstract
Energy poverty is nearly absent from NLP-for-social-good, and the little existing work is either static retrieval/QA or relies on carbon-intensive cloud LLMs, a self-defeating "computational irony" for a humanitarian setting. We present EqGrid, a closed-loop simulation in which a low-frequency, open-weight LLM policy agent sets price and carbon bounds and targeted subsidies over a community of empirically-grounded household personas, while high-frequency multi-agent RL traders clear a continuous double auction constrained by a physical distribution grid (IEEE-33-bus with Dynamic Operating Envelopes). Our contribution is threefold and directly addresses how to measure the social impact of AI: (i) grounded personas (region-matched socio-demographics) whose load curves are checked for shape and level realism against real smart-meter data; (ii) formal energy-poverty equity metrics (Energy Burden, Gini of EB, LIHC) showing the intervention reduces burden inequality without raising net grid cost; and (iii) a compute-efficiency frontier that measures how much equity performance survives compressing the policy agent from a 235B teacher down to a sub-1B model deployable on a laptop, in estimated energy/carbon per decision. A decoupled-safety design (the LLM sets bounds; a validate-and-project grid gate executes) yields zero grid-constraint violations versus 55 under direct LLM control. On energy-poverty equity, the LLM policy lowers the Gini of energy burden to 0.305 (from 0.351) and mean burden by 28% while cutting cost (outperforming a tuned rule baseline), and a 3B-active model retains 95% of the benefit at roughly 9x lower inference energy than the teacher, with even a 0.8B on-device model retaining 92% at roughly 24x lower energy. We will release code and configs.
Chinese Translation
能源贫困在社会公益的自然语言处理领域几乎不存在，而现有的少量研究要么是静态检索/问答，要么依赖于碳密集型的云端LLM，这在人道主义环境中形成了一种自我矛盾的“计算讽刺”。我们提出了EqGrid，一个闭环模拟，其中一个低频率、开放权重的LLM政策代理在一个基于实证的家庭角色社区中设置价格和碳限制及有针对性的补贴，而高频率的多智能体强化学习交易者则在一个受物理分配电网（IEEE-33-bus与动态操作包络）约束的连续双重拍卖中进行清算。我们的贡献有三方面，直接解决了如何衡量人工智能的社会影响：(i) 基于实证的角色（与地区匹配的社会人口统计）其负荷曲线与真实智能电表数据进行形状和水平的现实性检查；(ii) 正式的能源贫困公平度指标（能源负担、能源负担的基尼系数、低收入家庭能源补助）显示干预措施在不提高净电网成本的情况下减少了负担不平等；(iii) 一个计算效率前沿，衡量在将政策代理从235B的教师模型压缩到可在笔记本电脑上部署的低于1B模型时，公平性能的保留程度，以每个决策的估计能量/碳为单位。一个解耦安全设计（LLM设置边界；验证和投影电网门执行）在直接LLM控制下产生了零电网约束违规，而在直接控制下则有55个违规。在能源贫困公平方面，LLM政策将能源负担的基尼系数降低到0.305（从0.351下降），平均负担降低28%，同时降低成本（超越了调优的规则基线），而一个3B活跃模型在推理能量上比教师模型低约9倍的情况下保留了95%的收益，甚至一个0.8B的设备模型在能量上低约24倍的情况下保留了92%的收益。我们将发布代码和配置。
cs.CL / 16 / 2609.01936
Sparse Readout Prism: Explaining Logit-Lens Scores in Features Instead of Tokens
稀疏读出棱镜：用特征而非标记解释 Logit-Lens 分数
Abstract
A language model's prediction of its next token develops across layers, and lens methods track this process by decoding intermediate hidden states into tokens. But a lens reading reflects both the hidden state and the readout (the unembedding matrix) used to decode it. Many lenses are fit on a corpus, and we show that two lenses differing only in their fitting corpus can report different tokens for the same hidden states. We call this dependence corpus conditionality. To examine readout structure independently of the fitting corpus, we introduce Sparse Readout Prism (SRP), which decomposes the readout using only its weights and expresses any token logit or logit difference as a sum of contributions from sparse readout features. This reveals readout features as a new unit of analysis for lens readings, exposing structure that token identities can obscure and enabling comparisons across tokens, contexts, layers, and lenses. Replacing the original readout with SRP's sparse approximation reconstructs 8.9-17.3 percentage points more of the tested logit differences than the strongest of six baselines built on geometric relations among readout rows. Ablating features shifts logit differences in proportion to their SRP contributions. Although token readings vary with the fitting corpus, the dominant readout feature remains stable. Because SRP uses no corpus in its construction, it provides a control independent of the fitting corpus for lens analyses.
Chinese Translation
语言模型对其下一个标记的预测在各层之间发展，而透镜方法通过将中间隐藏状态解码为标记来跟踪这一过程。但透镜读取反映了隐藏状态和用于解码的读出（即去嵌入矩阵）。许多透镜是在一个语料库上拟合的，我们展示了仅在拟合语料库上有所不同的两个透镜可以为相同的隐藏状态报告不同的标记。我们称这种依赖关系为语料条件性。为了独立于拟合语料库检查读出结构，我们引入了稀疏读出棱镜（Sparse Readout Prism, SRP），它仅使用权重分解读出，并将任何标记的 logit 或 logit 差异表示为稀疏读出特征贡献的总和。这揭示了读出特征作为透镜读取的新分析单元，暴露了标记身份可能掩盖的结构，并使得在标记、上下文、层和透镜之间进行比较成为可能。用 SRP 的稀疏近似替代原始读出可以重构出比基于读出行之间几何关系构建的六个基线中最强的基线多出 8.9-17.3 个百分点的测试 logit 差异。消融特征会根据其 SRP 贡献的比例改变 logit 差异。尽管标记读取会随着拟合语料库的不同而变化，但主导的读出特征仍然保持稳定。由于 SRP 在构建中不使用任何语料库，因此为透镜分析提供了一个独立于拟合语料库的控制。
cs.CL / 17 / 2609.01971
NS-Copilot: An LLM-Driven Agent System for Autonomous Neuroscience Analysis
NS-Copilot：一个基于大型语言模型的自主神经科学分析代理系统
Abstract
AI is rapidly advancing neuroscience, yet many laboratories fail to fully unleash its potential due to significant interdisciplinary barriers. While pre-trained neural models for physiological data are progressing quickly, their heterogeneous architectures and modality-specific constraints hinder systematic integration, selection, and evaluation. Despite recent advances in large language model (LLM)-based agent systems for intelligent scientific applications, existing approaches often still lack the domain expertise required to effectively select and coordinate diverse neuroscience pre-trained models and handle unique data types in this domain. We present NS-Copilot, an LLM-driven multi-agent system for neuroscience analysis that autonomously supports end-to-end workflows for diverse professional tasks. It unifies domain-specific pre-trained models and supports key neuroscience modalities, including EEG and extracellular spike data, through a natural-language interface. Given raw data and a task description, NS-Copilot orchestrates agents with specialized roles for planning, adaptive control, code generation, and result synthesis, enabling analysis without dataset-specific heuristics. We evaluate NS-Copilot on neuroscience benchmarks spanning Alzheimer's disease, Parkinson's disease, and working memory spike decoding. Across 8 trials per task, the system consistently outperforms strong baselines on the primary metric, demonstrating the ability of NS-Copilot for effective and scalable neuroscience analysis.
Chinese Translation
人工智能正在迅速推动神经科学的发展，但许多实验室由于显著的跨学科障碍未能充分发挥其潜力。尽管用于生理数据的预训练神经模型正在快速进展，但它们异构的架构和特定模态的限制阻碍了系统的整合、选择和评估。尽管基于大型语言模型（LLM）的智能科学应用代理系统近期取得了一些进展，但现有方法往往仍缺乏有效选择和协调多样化神经科学预训练模型以及处理该领域独特数据类型所需的领域专业知识。我们提出了NS-Copilot，一个基于LLM的多代理系统，用于神经科学分析，能够自主支持多样化专业任务的端到端工作流程。它统一了特定领域的预训练模型，并通过自然语言接口支持关键的神经科学模态，包括脑电图（EEG）和细胞外尖峰数据。在给定原始数据和任务描述的情况下，NS-Copilot协调具有专业角色的代理进行规划、自适应控制、代码生成和结果综合，从而实现无需特定数据集启发式的分析。我们在涵盖阿尔茨海默病、帕金森病和工作记忆尖峰解码的神经科学基准上评估了NS-Copilot。在每个任务的8次试验中，该系统在主要指标上始终优于强基线，展示了NS-Copilot在有效和可扩展神经科学分析中的能力。
cs.CL / 18 / 2609.02015
How Output Format Confounds Data Quality and Capability in Instruction Tuning
输出格式如何影响指令调优中的数据质量和能力
Abstract
Instruction-tuning data are judged by quality metrics, and tuned models are judged by benchmarks, but both judgments pass through an output interface: the surface format in which an answer is written. Using gradient signatures across 12 tasks, four semantically equivalent interfaces, three model families, and controlled corruptions, we show that this interface confounds both measurements. Spectral statistics such as effective rank are provably invariant to interface rotation and empirically blind to semantic corruption, while the direction of the update carries the quality signal. The interface-varying residual is not noise: it identifies each unit's own target task perfectly across all three families. Capability itself is stored relative to the training interface: a skill that raises accuracy by more than 40 points under the training format can be nearly invisible under every other, and correcting a single generation budget flips the measured effect of fine-tuning on GSM8K from a gain into a large loss. Pre-registered interventions delimit where this geometry stops short of control. Data quality and model capability are interface-conditioned quantities, and current practice often reports the interface instead of the content.
Chinese Translation
指令调优数据通过质量指标进行评估，而调优后的模型则通过基准测试进行评估，但这两种评估都经过一个输出接口：答案书写的表面格式。通过在12个任务、四种语义等效接口、三种模型家族和受控干扰下使用梯度特征，我们展示了该接口混淆了这两种测量。有效秩等谱统计在接口旋转下是可证明的不变的，并且在语义干扰下是经验上盲目的，而更新的方向则携带质量信号。接口变化的残差并不是噪声：它在所有三种家族中完美识别每个单元的目标任务。能力本身相对于训练接口存储：在训练格式下提高超过40个点的技能在其他格式下几乎不可见，而纠正单个生成预算则将GSM8K上微调的测量效果从增益翻转为大损失。预注册的干预措施限定了这种几何形状在控制方面的不足。数据质量和模型能力是接口条件化的量，而当前的实践往往报告接口而非内容。
cs.CL / 19 / 2609.02054
A Tri-Agent Framework for Evaluating and Aligning Question Clarification Capabilities of Large Language Models
用于评估和对齐大型语言模型问题澄清能力的三代理框架
Abstract
Large Language Models (LLMs) are increasingly deployed in interactive systems where understanding user intent precisely is paramount. A key capability for such systems is effective question clarification, especially when user queries are ambiguous or underspecified. This paper introduces a novel tri-agent framework for the robust evaluation of an LLM's ability to engage in clarifying dialogue. Our framework comprises three distinct LLM-based agents: (1) a Question Clarifying Agent (QCA), the system under evaluation, tasked with identifying ambiguities and posing clarifying questions; (2) a Respondent Agent (RA), designed to simulate human user responses, potentially including irrelevant or challenging replies; and (3) an Evaluator Agent (EA), an LLM-as-a-judge, which assesses the quality of the dialogue based on a comprehensive set of metrics. We detail a methodology for synthetic data generation in the supply chain domain as an example. We propose metrics evaluating ambiguity handling, question quality, dialogue efficiency, language appropriateness, and final intent alignment. We also briefly discuss the validation of the EA against human judgments. This work provides a structured approach to benchmark, validate, and improve the clarification capabilities of conversational LLM applications.
Chinese Translation
大型语言模型（LLMs）越来越多地应用于交互系统中，在这些系统中，准确理解用户意图至关重要。有效的问题澄清能力是此类系统的关键能力，尤其是在用户查询模糊或不明确时。本文提出了一种新颖的三代理框架，用于稳健评估LLM在澄清对话中的能力。我们的框架由三个不同的基于LLM的代理组成：（1）问题澄清代理（Question Clarifying Agent, QCA），即被评估的系统，负责识别模糊性并提出澄清问题；（2）回应代理（Respondent Agent, RA），旨在模拟人类用户的响应，可能包括无关或具有挑战性的回复；（3）评估代理（Evaluator Agent, EA），作为评判者的LLM，根据一套全面的指标评估对话质量。我们详细描述了在供应链领域合成数据生成的方法作为示例。我们提出了评估模糊处理、问题质量、对话效率、语言适当性和最终意图对齐的指标。我们还简要讨论了EA与人类判断的验证。这项工作提供了一种结构化的方法来基准测试、验证和改进对话LLM应用的澄清能力。
cs.CL / 20 / 2609.02056
HyGRAIL: Cost-Aware and Evidence-Grounded Scientific Hypothesis Discovery over Knowledge Graphs
HyGRAIL：基于成本意识和证据支持的知识图谱科学假设发现
Abstract
Scientific knowledge graphs organize entities and relations extracted from scientific literature, but they remain inherently incomplete. Missing typed links in such graphs can therefore represent plausible scientific hypotheses, such as unexplored associations between materials and applications. However, scientific hypothesis discovery is challenging because true discoveries are extremely sparse among typed candidate pairs: graph neural networks (GNNs) are efficient but unreliable for ambiguous cases, while large language models (LLMs) are knowledgeable but too costly to apply exhaustively and are not naturally grounded in graph structures. We propose HyGRAIL, a cost-aware and evidence-grounded framework that combines heterogeneous GNN triage with LLM-based hypothesis review. HyGRAIL first uses a GNN to score candidate hypotheses and identify a validation-calibrated ambiguous region, routing only graph-uncertain cases to LLM review. For each routed hypothesis, HyGRAIL retrieves node-level associations and multi-hop relational paths from the knowledge graph (KG), then converts this structured evidence into natural language through template-based or LLM-based naturalization. An LLM review agent finally judges each hard hypothesis using the naturalized evidence and validation-selected decision criteria. On MatKG, HyGRAIL achieves the best F1 score of 0.429, improving over the strongest prior baseline by 0.242 F1 points and over the GNN-only baseline by 0.322. Meanwhile, GNN triage reduces the LLM call rate by 54.36% on average. Ablation studies further show that retrieved graph evidence is crucial for reliable hypothesis verification and that compact, two-sided evidence is more effective than simply increasing retrieval quantity.
Chinese Translation
科学知识图谱组织了从科学文献中提取的实体和关系，但它们本质上仍然是不完整的。因此，这些图谱中缺失的类型链接可能代表合理的科学假设，例如材料与应用之间未探索的关联。然而，科学假设的发现具有挑战性，因为在类型候选对中，真正的发现极为稀少：图神经网络（GNN）高效但在模糊情况下不可靠，而大型语言模型（LLM）知识丰富但在全面应用时成本过高，并且不自然地与图结构相结合。我们提出了HyGRAIL，一个基于成本意识和证据支持的框架，结合了异构GNN筛选与基于LLM的假设审查。HyGRAIL首先使用GNN对候选假设进行评分，并识别经过验证的模糊区域，仅将图不确定的案例路由到LLM审查。对于每个路由的假设，HyGRAIL从知识图谱（KG）中检索节点级关联和多跳关系路径，然后通过基于模板或基于LLM的自然化将这些结构化证据转换为自然语言。最后，LLM审查代理使用自然化证据和经过验证的决策标准来判断每个困难假设。在MatKG上，HyGRAIL达到了最佳F1分数0.429，比最强的先前基线提高了0.242 F1点，比仅使用GNN的基线提高了0.322。同时，GNN筛选平均减少了54.36%的LLM调用率。消融研究进一步表明，检索到的图证据对于可靠的假设验证至关重要，而紧凑的双面证据比单纯增加检索数量更有效。
cs.CL / 21 / 2609.02089
IDEEA: training-free Input-Dependent stEEring via Activation cluster matching
IDEEA：通过激活簇匹配实现无训练的输入依赖性引导
Abstract
Steering aligns large language models (LLMs) by injecting a bias into selected activations at inference time, offering a far cheaper alternative to weight-update methods such as supervised fine-tuning or reinforcement learning. However, most existing training-free steering methods are input-independent: a single direction is fitted once and shared across all inputs. This is fundamentally limiting as different inputs occupy different regions of the activation space and admit different optimal steering directions toward the same target concept, much as the gradient with respect to a fixed loss varies from input to input. We close this gap with IDEEA (Input-Dependent stEEring via Activation cluster matching), a training-free framework for input-dependent steering. IDEEA clusters the positive and negative activation supports per attention head, and solves an optimal-matching problem to construct a set of cluster-conditional directions, all about the target concept. At inference time, it picks from this pool of directions and uses the one that best matches the input's own activation for steering. IDEEA aligns the model toward the target concept while preserving the input's original representation, evidence that activations encoding a concept occupy several distinct sub-regions of the representation space rather than a single one. IDEEA improves the truth $\times$ info rate in TruthfulQA by an average of 9.9% (up to 23.5%) over the best input-independent baseline.
Chinese Translation
引导通过在推理时向选定的激活注入偏差来对齐大型语言模型（LLMs），提供了一种比权重更新方法（如监督微调或强化学习）便宜得多的替代方案。然而，现有的大多数无训练引导方法都是输入无关的：一个方向被一次性拟合并在所有输入中共享。这在根本上是有限制的，因为不同的输入占据激活空间的不同区域，并且允许朝向同一目标概念的不同最佳引导方向，就像固定损失的梯度因输入而异一样。我们通过IDEEA（通过激活簇匹配实现输入依赖性引导）来填补这一空白，这是一个用于输入依赖性引导的无训练框架。IDEEA为每个注意力头聚类正负激活支持，并解决一个最优匹配问题，以构建一组关于目标概念的簇条件方向。在推理时，它从这个方向池中选择，并使用与输入自身激活最匹配的方向进行引导。IDEEA在对齐模型朝向目标概念的同时，保留输入的原始表示，证明了编码概念的激活占据表示空间中的多个不同子区域，而不是单一的区域。IDEEA在TruthfulQA中将真实度与信息率的乘积提高了平均9.9%（最高可达23.5%），相较于最佳输入无关基线。
cs.CL / 22 / 2609.02091
Selective Knowledge Edit Reversal via Gated Singular Vector Shrinkage
通过门控奇异向量收缩实现选择性知识编辑逆转
Abstract
Knowledge editing provides an efficient way to update factual knowledge in large language models. However, malicious edits may introduce safety risks, making it necessary to reverse undesirable editing effects. Existing reversal methods for parameter-modifying edits mainly focus on global removal, which may also erase beneficial edits that should be preserved. In this paper, we study selective reversal of edited knowledge, where the goal is to reverse targeted edited facts while preserving the remaining edited facts. Based on the hypothesis that each edit is sparsely encoded within the dominant subspace of the edited matrix, we propose a spectral-based reversal framework that locates edit-sensitive components within the dominant singular subspace of edited weights. Experiments across multiple settings demonstrate the effectiveness of our method in reversing selected edits while preserving unrelated edited facts. These results suggest that different edits are sparsely encoded within dominant singular components and can be separable when the number of edits is moderate, making selective spectral reversal a promising direction for locating edit-specific components and repairing edited language models.
Chinese Translation
知识编辑为更新大型语言模型中的事实知识提供了一种高效的方法。然而，恶意编辑可能引入安全风险，因此有必要逆转不良编辑效果。现有的参数修改编辑逆转方法主要集中于全局移除，这可能会抹去应当保留的有益编辑。本文研究了编辑知识的选择性逆转，目标是逆转特定的编辑事实，同时保留其余编辑事实。基于每个编辑在编辑矩阵的主导子空间中稀疏编码的假设，我们提出了一种基于谱的逆转框架，该框架能够定位编辑权重主导奇异子空间中的编辑敏感组件。在多个设置下的实验表明，我们的方法在逆转选定编辑的同时保留无关的编辑事实方面是有效的。这些结果表明，不同的编辑在主导奇异组件中是稀疏编码的，并且在编辑数量适中的情况下可以被分离，使得选择性谱逆转成为定位特定编辑组件和修复编辑语言模型的一个有前景的方向。
cs.CL / 23 / 2609.02108
Predict, Don't Iterate: Efficient Adaptive-Length Infilling for Diffusion Language Models
预测，而非迭代：扩散语言模型的高效自适应长度填充
Abstract
Diffusion language models (DLMs) have emerged as a promising alternative to the auto-regressive paradigm. With bidirectional attention and any-order generation, DLMs naturally fit infilling tasks, which require generating a middle span conditioned on both the prefix and the suffix. However, infilling is sensitive to the length of the span, while DLMs require the length to be fixed before generation. Although prior studies extend DLMs to dynamic lengths, they still suffer from two limitations. (i) Sensitivity to initial length. These methods require a preset length to initialize the search and are highly sensitive to this initial length, often yielding suboptimal results. (ii) Inference inefficiency. They either insert length-changing operations during generation or repeatedly search for an appropriate length using multi-step denoising confidence, both of which introduce substantial extra forward passes and computational cost. Therefore, we propose PILL (Probing-based InfiLling with preset-Length-free decoding), an efficient infilling method for DLMs that requires no preset initial length and adds far fewer extra forward passes than baselines, substantially reducing inference time. Experiments show that, across five DLMs spanning different families, architectures, and training recipes on eight infilling benchmarks, PILL improves over the strongest baseline by +4.8 average pass rate on code and +6.0 BLEU-2 on text, while running 1.82x faster than that baseline. The code is available at https://github.com/Hsu1023/PILL.
Chinese Translation
扩散语言模型（DLMs）已成为自回归范式的有前景的替代方案。凭借双向注意力和任意顺序生成，DLMs 自然适用于填充任务，这类任务需要在前缀和后缀的条件下生成中间跨度。然而，填充对跨度的长度敏感，而 DLMs 在生成之前需要固定长度。尽管先前的研究将 DLMs 扩展到动态长度，但仍然存在两个局限性。（i）对初始长度的敏感性。这些方法需要预设长度来初始化搜索，并且对这个初始长度高度敏感，常常导致次优结果。（ii）推理效率低下。它们要么在生成过程中插入长度变化操作，要么使用多步去噪信心反复搜索合适的长度，这两者都引入了大量额外的前向传播和计算成本。因此，我们提出了 PILL（基于探测的无预设长度解码填充），这是一种高效的 DLM 填充方法，不需要预设初始长度，并且相比基线增加的额外前向传播次数大大减少，从而显著降低推理时间。实验表明，在涵盖不同家族、架构和训练方案的五种 DLMs 上，在八个填充基准测试中，PILL 在代码上比最强基线提高了 +4.8 的平均通过率，在文本上提高了 +6.0 的 BLEU-2，同时运行速度比该基线快 1.82 倍。代码可在 https://github.com/Hsu1023/PILL 获取。
cs.CL / 24 / 2609.02115
text2ql: Multi-Target Natural Language Querying via a Language-Agnostic Intermediate Representation
text2ql：通过语言无关的中间表示实现多目标自然语言查询
Abstract
Natural language interfaces to databases have traditionally suffered from three structural limitations: exclusive targeting of relational SQL, unconditional dependence on large language model (LLM) inference at query time, and absence of any runtime signal when generated queries are semantically incorrect. This paper presents text2ql, an open-source Python framework that addresses all three limitations through a language-agnostic Intermediate Representation (QueryIR) and a pluggable renderer architecture. A single seven-stage detection pipeline serves both SQL and GraphQL targets; a zero-LLM deterministic mode delivers 100% execution accuracy at a median latency of 3.2 ms with no API cost; and every generated query carries a runtime confidence score in [0.15, 0.97] computed from an additive signal model. Evaluated on 50-query random samples from the Spider and BIRD benchmarks (indicative results; full-set evaluation is planned), the LLM-backed mode achieves 62-70% exact match and 84-91% execution accuracy; the deterministic mode achieves 100% execution accuracy with zero parse errors across all 100 test cases. An ablation study isolates schema-aware prompting as the dominant accuracy lever, contributing +18.4 percentage points of exact-match gain over the schema-free baseline on both benchmarks. text2ql is publicly available at https://pypi.org/project/text2ql/ under the Apache 2.0 license.
Chinese Translation
数据库的自然语言接口传统上存在三大结构性限制：仅针对关系型SQL、在查询时对大型语言模型（LLM）推理的无条件依赖，以及在生成的查询语义不正确时缺乏任何运行时信号。本文提出了text2ql，一个开源的Python框架，通过语言无关的中间表示（QueryIR）和可插拔的渲染器架构解决了这三大限制。一个七阶段的检测管道同时服务于SQL和GraphQL目标；零LLM确定性模式在中位延迟为3.2毫秒的情况下提供100%的执行准确率且无API成本；每个生成的查询都携带一个在[0.15, 0.97]范围内的运行时置信度分数，该分数是通过加性信号模型计算得出的。在Spider和BIRD基准测试的50个查询随机样本上进行评估（指示性结果；计划进行全套评估），LLM支持模式实现62-70%的精确匹配和84-91%的执行准确率；确定性模式在所有100个测试案例中实现100%的执行准确率且无解析错误。一项消融研究将模式感知提示识别为主要的准确性杠杆，在两个基准上相较于无模式基线贡献了+18.4个百分点的精确匹配增益。text2ql在Apache 2.0许可下公开可用，网址为https://pypi.org/project/text2ql/
cs.CL / 25 / 2609.02122
AI agents reshape consensus formation in human groups
人工智能代理重塑人类群体中的共识形成
Abstract
As large language model (LLM) agents shift from tools to participants in human groups, a fundamental question for collective behavior is how their growing presence reshapes consensus formation. Here we study mixed human-AI groups in a collaborative description game, in which shared conventions emerge through repeated rounds of random pairwise communication. Varying the proportions of LLM agents, we identify three distinct regimes of consensus formation: low agent proportions facilitate human-led consensus, intermediate proportions disrupt convergence, and high proportions restore strong consensus while shifting it toward agent-led conventions. Crucially, these regimes differ not only in the strength of convergence, but also in the semantic grounding and communicative form of the resulting consensus: human-led consensus is more concrete, holistic, and grounded in shared real-world analogies, whereas agent-led consensus is more abstract, less information-dense, and more geometrically segmented. Mechanistically, agent influence arises from a shared linguistic prior that places agents near one another in the expression space, combined with relatively stable expression choices across rounds; humans initially resist adopting expressions from partners perceived as AI but gradually yield to conformity pressure. These findings provide evidence that AI composition can shape the emergence, content, and perceived legitimacy of group norms, making agent proportion and transparency important design variables for human-AI systems.
Chinese Translation
随着大型语言模型（LLM）代理从工具转变为人类群体中的参与者，集体行为的一个基本问题是它们日益增长的存在如何重塑共识形成。在这项研究中，我们考察了混合人类与人工智能的群体在一项协作描述游戏中的表现，在该游戏中，通过多轮随机配对沟通，产生共享的约定。通过改变LLM代理的比例，我们识别出三种不同的共识形成机制：低比例的代理促进人类主导的共识，中等比例的代理扰乱收敛，而高比例的代理则恢复强烈的共识，同时将其转向代理主导的约定。至关重要的是，这些机制不仅在收敛的强度上有所不同，还在所形成共识的语义基础和交流形式上存在差异：人类主导的共识更加具体、整体，并且基于共享的现实世界类比，而代理主导的共识则更加抽象、信息密度较低，并且在几何上更加分割。从机制上看，代理的影响源于一种共享的语言先验，这使得代理在表达空间中彼此接近，并结合了跨轮次相对稳定的表达选择；人类最初抵制采用被视为人工智能的伙伴的表达，但逐渐屈服于从众压力。这些发现提供了证据，表明人工智能的组成可以影响群体规范的出现、内容和感知合法性，因此代理的比例和透明度成为人类与人工智能系统设计中的重要变量。
cs.CL / 26 / 2609.02131
C$^{3}$T: Counterfactual Causal Reasoning for Sentiment Shifts in Social-Media Conversation Trees
C$^{3}$T：社交媒体对话树中情感变化的反事实因果推理
Abstract
Sentiment in social-media threads does not only vary across posts; it shifts as users react to claims, corrections, evidence, and hostility within a branching reply tree. We study why sentiment changes in rumor-centric conversation trees by treating discourse moves (e.g., denial/correction, evidence/link, toxicity/attack) as candidate interventions and asking (i) what sentiment a reply expresses, (ii) whether the sentiment shifts relative to its parent, and (iii) which prior message most plausibly drove the reply's sentiment. To support this setting, we introduce CaSiRe, a causal sentiment reasoning layer over public rumor conversation datasets that adds post-level sentiment labels, induced parent-child shift labels, calibrated multi-label intervention tags, and explicitly annotated causal-source labels. We then propose C$^{3}$T (Counterfactual Causal Conversation Transformer), a thread-structured temporal model that jointly predicts node sentiment and shifts, learns sparse ancestor attribution, and supports counterfactual queries by forcing conversational intervention embeddings on or off to estimate potential outcomes. Under an event-level split, C$^{3}$T improves out-of-event robustness and attribution over text-only, graph-based, and temporal baselines, and yields interpretable model-based effects: denials/corrections and evidence reduce downstream negativity, while toxicity increases it. We also benchmark open-weight LLM prompting baselines and find that added conversational context helps, but attribution remains less reliable, motivating structure-aware counterfactual modeling for social-media analysis.
Chinese Translation
社交媒体线程中的情感不仅在帖子之间变化；它还会随着用户对主张、纠正、证据和敌意的反应而在分支回复树中发生转变。我们通过将话语行为（例如，否认/纠正、证据/链接、毒性/攻击）视为候选干预，研究谣言中心对话树中情感变化的原因，并提出以下问题：(i) 回复表达了什么情感，(ii) 相较于其父节点情感是否发生了变化，以及 (iii) 哪条先前消息最有可能驱动回复的情感。为支持这一设置，我们引入了CaSiRe，一个在公共谣言对话数据集上的因果情感推理层，增加了帖子级情感标签、诱导的父子情感变化标签、经过校准的多标签干预标签和明确注释的因果源标签。然后，我们提出了C$^{3}$T（反事实因果对话变换器），这是一个线程结构的时间模型，能够共同预测节点情感和情感变化，学习稀疏的祖先归因，并通过强制对话干预嵌入的开启或关闭来支持反事实查询，以估计潜在结果。在事件级拆分下，C$^{3}$T在事件外的鲁棒性和归因上优于仅基于文本、图形和时间的基线，并产生可解释的模型效应：否认/纠正和证据减少下游的负面情感，而毒性则增加负面情感。我们还基准测试了开放权重的LLM提示基线，发现增加的对话上下文有助于改善效果，但归因仍然不够可靠，这促使我们进行结构感知的反事实建模以分析社交媒体。
cs.CL / 27 / 2609.02153
A Layered Taxonomy for Chinese Learner Grammatical Error Annotation
中文学习者语法错误标注的分层分类法
Abstract
Grammatical error annotation in Chinese learner writing requires labels that are both consistent and linguistically meaningful. This paper proposes a layered scheme linking computational Chinese grammatical error correction (CGEC) with pedagogical error analysis. The scheme first identifies character- and punctuation-level orthographic errors, labeling them by edit operation and subtype. Other errors receive a three-layer core label combining edit operation, linguistic domain, and part of speech, with optional Chinese-specific extensions for aspect, modality, comparison, argument structure, and complements. Drawing on CGEC resources, learner-error taxonomies, and Mandarin grammar, the taxonomy is evaluated through a coverage analysis of automatically extracted MuCGEC edits and a preliminary consistency study in which five large language models apply it to a sample. The results support the layered approach while identifying category boundaries requiring further refinement.
Chinese Translation
中文学习者写作中的语法错误标注需要既一致又具有语言学意义的标签。本文提出了一种将计算机中文语法错误纠正（CGEC）与教学错误分析相结合的分层方案。该方案首先识别字符和标点级别的书写错误，通过编辑操作和子类型进行标注。其他错误则采用一个三层核心标签，结合编辑操作、语言领域和词性，并可选地添加中文特有的扩展，如方面、情态、比较、论元结构和补语。基于CGEC资源、学习者错误分类和普通话语法，本文通过对自动提取的MuCGEC编辑进行覆盖分析和初步一致性研究来评估该分类法，后者中五个大型语言模型将其应用于样本。结果支持了分层方法，同时识别出需要进一步细化的类别边界。
cs.CL / 28 / 2609.02158
OBJECTION! Lawyer Agents Mitigate Guilty Bias in Legal Judgment Prediction
反对！律师代理人在法律判断预测中减轻有罪偏见
Abstract
Legal Judgment Prediction (LJP) models are typically trained on documents that describe facts from a prosecutorial perspective. Existing datasets further exhibit severe label imbalance toward guilty outcomes. Consequently, these models suffer from "Guilty Bias", blindly accepting the prosecution's narrative as objective truth. Previous studies employing three-step reasoning structures or training on synthetically generated innocence data improve overall accuracy, but they still fail to mitigate bias at inference time. In this paper, we introduce OBJECTION, an inference-time pipeline that integrates an Adversarial Lawyer Agent into each 3-step reasoning of offense, unlawfulness, and culpability. Unlike generic critics, our agent actively challenges the model's presumptions of guilt by injecting legal defense arguments at each reasoning stage. To thoroughly evaluate this, we present a new "Natural Innocent" dataset including 3.4k real-world cases, overcoming the limitations of synthetic innocence benchmarks. Test results show that OBJECTION drastically reduces the False Guilty Rate (FGR) from 82.93% (SOTA baseline) to 16.69%, proving its capability to perform substantive legal reasoning. This work denotes a key progress toward aligning Legal AI with the presumption of innocence.
Chinese Translation
法律判断预测（LJP）模型通常是在描述检方视角事实的文档上进行训练的。现有数据集在有罪结果上表现出严重的标签不平衡。因此，这些模型受到“有罪偏见”的影响，盲目接受检方的叙述作为客观真相。以往的研究采用三步推理结构或在合成生成的无罪数据上进行训练，提高了整体准确性，但在推理时仍未能减轻偏见。在本文中，我们引入了OBJECTION，一个推理时管道，将对抗性律师代理集成到每个关于犯罪、违法性和责任的三步推理中。与一般的批评者不同，我们的代理在每个推理阶段通过注入法律辩护论点，积极挑战模型对有罪的假设。为了全面评估这一点，我们提出了一个新的“自然无罪”数据集，包括3400个真实案例，克服了合成无罪基准的局限性。测试结果表明，OBJECTION将假有罪率（FGR）从82.93%（现有技术基线）大幅降低至16.69%，证明了其进行实质性法律推理的能力。这项工作标志着法律人工智能与无罪推定对齐的重要进展。
cs.CL / 29 / 2609.02163
Do Cantonese-Adapted Language Models Better Predict Cantonese Reading? A Cross-Model Eye-Tracking Evaluation
粤语适应语言模型是否更好地预测粤语阅读？跨模型眼动追踪评估
Abstract
Information-theoretic measures derived from autoregressive language models are widely used to characterize the expectations that shape human reading, but whether language-variety-specific training improves such psycholinguistic alignment remains unclear. This question is still open for Cantonese, where recent NLP evaluations reported mixed benefits from Cantonese-specific training relative to Mandarin-oriented or general-purpose models. Using naturalistic Cantonese eye-tracking data, we compare two within-family adaptation contrasts: CKIP GPT-2 Tiny versus its lightly Cantonese-adapted JED351 derivative, and Qwen2.5-7B versus CantoneseLLM-7B, which underwent substantially more extensive Cantonese continued pretraining and instruction tuning. From each model, we derive lexical surprisal, POS surprisal, entropy before the target, and entropy reduction. Lexical surprisal and the joint four-metric model consistently favor CantoneseLLM-7B, followed by Qwen2.5-7B, CKIP, and JED351, whereas entropy reduction favors CKIP. These results suggest that more extensive Cantonese-specific training can be associated with stronger predictive fit, while model rankings also depend on the information-theoretic measure being evaluated.
Chinese Translation
从自回归语言模型中衍生的信息论度量广泛用于表征塑造人类阅读的期望，但语言种类特定的训练是否改善了这种心理语言学的一致性仍然不明确。对于粤语而言，这一问题仍然悬而未决，最近的自然语言处理评估报告了粤语特定训练相对于普通话导向或通用模型的混合效益。利用自然的粤语眼动追踪数据，我们比较了两个同家族适应对比：CKIP GPT-2 Tiny与其轻度粤语适应的JED351衍生模型，以及Qwen2.5-7B与经过更广泛粤语持续预训练和指令调优的CantoneseLLM-7B。从每个模型中，我们推导出词汇惊讶度、词性惊讶度、目标前的熵和熵减。词汇惊讶度和联合四度量模型一致支持CantoneseLLM-7B，其次是Qwen2.5-7B、CKIP和JED351，而熵减则支持CKIP。这些结果表明，更广泛的粤语特定训练可能与更强的预测拟合相关，而模型排名也依赖于所评估的信息论度量。
cs.CL / 30 / 2609.02172
Breadth Beats Depth: Improving GCG-Based Jailbreak Optimization with Breadth-Oriented Suffix Search
广度胜于深度：通过广度导向后缀搜索改善基于GCG的越狱优化
Abstract
Optimization-based jailbreak attacks such as Greedy Coordinate Gradient (GCG) achieve strong effectiveness and transferability by optimizing adversarial suffixes on white-box source models. However, existing GCG-based methods rely on averaged adversarial loss and deep greedy search, which can over-emphasize easy-to-jailbreak behaviors and overlook promising regions of the suffix space. We propose BOSS, a plug-and-play framework that improves GCG-based jailbreak optimization through breadth-oriented suffix search. BOSS uses Tail-Focused Adversarial Loss (TFAL), standard source loss, and behavior coverage to select terminal suffixes, then explores multiple short trajectories and selectively continues promising suffixes. Experiments on public benchmarks show that BOSS improves attack success rates across multiple GCG-based methods while reducing optimization time.
Chinese Translation
基于优化的越狱攻击，如贪婪坐标梯度（Greedy Coordinate Gradient，GCG），通过优化白盒源模型上的对抗后缀实现了强大的有效性和可迁移性。然而，现有的基于GCG的方法依赖于平均对抗损失和深度贪婪搜索，这可能过于强调容易越狱的行为，而忽视了后缀空间中有前景的区域。我们提出了BOSS，一个即插即用的框架，通过广度导向的后缀搜索改善基于GCG的越狱优化。BOSS使用以尾部为重点的对抗损失（Tail-Focused Adversarial Loss，TFAL）、标准源损失和行为覆盖来选择终端后缀，然后探索多个短轨迹并选择性地继续有前景的后缀。在公共基准测试上的实验表明，BOSS提高了多种基于GCG的方法的攻击成功率，同时减少了优化时间。
cs.CL / 31 / 2609.02272
PaperCompiler: Faithful Paper-to-Code Generation via Repository-Level Specification Compilation
PaperCompiler：通过仓库级规范编译实现忠实的论文到代码生成
Abstract
Faithfully translating research papers into repository-level implementations remains challenging because papers often describe methods at a high level, leave implementation assumptions implicit, and require generated repositories to preserve method logic, evaluation protocols, and cross-file consistency. Despite recent advances in paper-to-code agents, their intermediate outputs are often presented as free-form plans or summaries that downstream coding agents may ignore, reinterpret, or compress, leading to algorithmic simplification and inconsistent repository structure. To address these challenges, we introduce PaperCompiler, a paper-to-code generation framework that compiles paper-grounded evidence into explicit repository-level implementation specifications. PaperCompiler grounds implementation-relevant evidence while preserving source provenance and distinguishing paper-supported, inferred, externally delegated, and unresolved information. The resulting specifications encode non-degradation requirements, ownership assignments, cross-file dependencies, and file-level constraints. Repository generation proceeds under these compiled specifications while retaining flexibility over local engineering choices not fixed by the paper. PaperCompiler outperforms strong baselines on Paper2CodeBench, achieving a 13.8% relative improvement in reference-based fidelity (from 3.64 to 4.15) and reducing high-severity evaluator critiques (from 13.2% to 6.1%).
Chinese Translation
将研究论文忠实地翻译为仓库级实现仍然面临挑战，因为论文通常以高层次描述方法，隐含实施假设，并要求生成的仓库保持方法逻辑、评估协议和跨文件一致性。尽管最近在论文到代码代理方面取得了进展，但它们的中间输出通常以自由形式的计划或摘要呈现，后续的编码代理可能会忽略、重新解释或压缩这些内容，从而导致算法简化和不一致的仓库结构。为了解决这些挑战，我们引入了PaperCompiler，一个将论文基础证据编译为明确的仓库级实现规范的论文到代码生成框架。PaperCompiler在保留源出处的同时，基于实现相关证据，区分论文支持的信息、推断的信息、外部委托的信息和未解决的信息。生成的规范编码了非降级要求、所有权分配、跨文件依赖关系和文件级约束。仓库生成在这些编译的规范下进行，同时保留对论文未固定的本地工程选择的灵活性。PaperCompiler在Paper2CodeBench上超越了强基线，在基于参考的忠实度上实现了13.8%的相对提升（从3.64提升至4.15），并减少了高严重性评估者的批评（从13.2%降至6.1%）。
cs.CL / 32 / 2609.02275
Do Large Language Models Capture the Diversity in their Training Data?
大型语言模型是否捕捉到其训练数据中的多样性？
Abstract
Large language models are trained to model conditional distributions over text, yet it remains inadequately understood whether they capture the full diversity of plausible outputs present in their training data. We study this question through an information-theoretic lens by comparing the conditional entropy of model-generated outputs with that of the corresponding training data. Given paired input-output samples, we use conditional entropy and its matrix-based analogue based on von Neumann entropy to measure output variability beyond what is explained by the conditioning input, without requiring multiple reference outputs for the same prompt. Across LLM families with publicly available training data, including OLMo, Pythia, and GPT-Neo, we consistently find that model-generated outputs exhibit lower conditional entropy than their training data, across different model scales, sequence lengths, and decoding strategies. We observe a similar conditional diversity gap beyond language modeling, including class-conditioned ImageNet generators and text-conditioned models trained on MS-COCO. To address this gap, we propose a post-hoc correction mechanism that generates multiple outputs for each input and reweights them through a matrix-entropy projection, increasing conditional diversity while remaining close to the original model distribution. We prove the concavity of the matrix-based conditional entropy functional, which makes the resulting entropy-constrained projection a convex optimization problem, and develop a scalable mirror-descent algorithm for its implementation. Our results reveal a systematic conditional diversity gap between modern generative models and their training data, and provide an information-theoretic framework for measuring and mitigating this gap.
Chinese Translation
大型语言模型被训练以建模文本的条件分布，但尚不充分理解它们是否捕捉到了训练数据中存在的合理输出的全部多样性。我们通过信息论的视角研究这个问题，比较模型生成输出的条件熵与相应训练数据的条件熵。在给定的输入-输出样本对中，我们使用条件熵及其基于冯·诺依曼熵的矩阵类比来测量输出的变异性，超出由条件输入所解释的部分，而无需对同一提示要求多个参考输出。在包括 OLMo、Pythia 和 GPT-Neo 在内的具有公开训练数据的 LLM 家族中，我们一致发现模型生成的输出在不同模型规模、序列长度和解码策略下的条件熵均低于其训练数据。我们观察到在语言建模之外存在类似的条件多样性差距，包括类条件的 ImageNet 生成器和在 MS-COCO 上训练的文本条件模型。为了解决这一差距，我们提出了一种后处理校正机制，为每个输入生成多个输出，并通过矩阵熵投影对它们进行重加权，从而在保持接近原始模型分布的同时增加条件多样性。我们证明了基于矩阵的条件熵泛函的凹性，这使得所得的熵约束投影成为一个凸优化问题，并开发了一种可扩展的镜像下降算法以实现该方法。我们的结果揭示了现代生成模型与其训练数据之间存在系统性的条件多样性差距，并提供了一个信息论框架来测量和缓解这一差距。
cs.CL / 33 / 2609.02309
Efficient GUI Agents: A Systems Survey of Observation, Memory, Action, and Runtime Optimization
高效的GUI代理：观察、记忆、行动和运行时优化的系统调查
Abstract
GUI agents increasingly operate across websites, mobile apps, and desktop environments, yet the field still reports progress primarily through task success. We argue that practical deployment depends equally on efficiency: how much context, computation, action budget, and runtime overhead an agent consumes while succeeding. This survey studies efficient GUI agents through an end-to-end systems lens that preserves the current technical axes of observation efficiency, context and memory efficiency, action efficiency, and planner-side/system efficiency. For each subsection, we expand the seed literature through targeted search plus backward and forward citation chaining, then synthesize the dominant mechanisms, reported efficiency signals, and new overheads they introduce. Across the literature, recent progress converges on a small set of recurring ideas: selective reading instead of full-context ingestion, global-to-local visual allocation, recoverable memory rather than raw history replay, verification-aware control, and hybrid runtimes that can switch between GUI and non-GUI execution. We conclude by identifying the main open problems, including honest accounting of verifier cost, cross-benchmark comparability, and co-design of observation, memory, and execution layers under real latency and privacy constraints.
Chinese Translation
GUI代理越来越多地在网站、移动应用和桌面环境中运行，但该领域仍主要通过任务成功来报告进展。我们认为，实际部署同样依赖于效率：代理在成功时消耗多少上下文、计算、行动预算和运行时开销。本调查通过端到端系统的视角研究高效的GUI代理，保留当前观察效率、上下文和记忆效率、行动效率以及规划者/系统效率的技术轴。在每个子部分中，我们通过有针对性的搜索以及向后和向前的引用链扩展种子文献，然后综合主导机制、报告的效率信号以及它们引入的新开销。在文献中，最近的进展集中在一小组反复出现的想法上：选择性阅读而非全上下文摄取、全局到局部的视觉分配、可恢复的记忆而非原始历史重放、关注验证的控制，以及能够在GUI和非GUI执行之间切换的混合运行时。我们最后识别出主要的开放问题，包括对验证器成本的诚实核算、跨基准的可比性，以及在真实延迟和隐私约束下观察、记忆和执行层的协同设计。
cs.CL / 34 / 2609.02315
DiffIE: Diffusion-based Open Information Extraction
DiffIE：基于扩散的开放信息提取
Abstract
A single sentence often expresses multiple valid relational triplets, which makes Open Information Extraction (OpenIE) fundamentally a multi-output task. Existing neural systems handle this by autoregressive generation, which is flexible but slow and prone to redundancy, or by fixed-slot prediction, which is efficient but couples the extraction budget to training. We introduce DIFFIE which instead treats the stochasticity of conditional discrete diffusion as the extraction mechanism itself: independent reverse-diffusion trajectories over per-token role tags produce a pool of candidate triplets, which are clustered under lenient matching and ranked to form the output. Both the pool size and the number of returned extractions are inference-time choices, decoupling the extraction budget from training and exposing test-time compute as a tunable axis. DIFFIE achieves the new state of the art in CaRB (1-1) both F1 and AUC, and outperforms the strongest rule-based system (ClausIE) in BenchIE; it also remains competitive in standard CaRB and WiRe57 evaluations, giving the best average score among systems that report all four benchmarks. Ablations show that uniform discrete diffusion outperforms absorbing state diffusion in our setting, and that a matched non-diffusion stochastic tagger does not reproduce its gains. Our results indicate that diffusion stochasticity is an effective mechanism for structured prediction tasks with multiple valid outputs.
Chinese Translation
单个句子通常表达多个有效的关系三元组，这使得开放信息提取（OpenIE）在本质上成为一个多输出任务。现有的神经系统通过自回归生成来处理这一问题，这种方法灵活但速度较慢且容易产生冗余，或者通过固定槽位预测，这种方法高效但将提取预算与训练耦合在一起。我们提出了DIFFIE，它将条件离散扩散的随机性视为提取机制本身：在每个标记角色标签上独立的反向扩散轨迹产生一组候选三元组，这些三元组在宽松匹配下进行聚类并排序以形成输出。池的大小和返回提取的数量都是推理时的选择，从而将提取预算与训练解耦，并将测试时的计算暴露为可调轴。DIFFIE在CaRB（1-1）中在F1和AUC上达到了新的最先进水平，并在BenchIE中超越了最强的基于规则的系统（ClausIE）；在标准的CaRB和WiRe57评估中也保持竞争力，在报告所有四个基准的系统中获得了最佳平均分。消融实验表明，在我们的设置中，均匀离散扩散优于吸收态扩散，并且匹配的非扩散随机标记器未能重现其增益。我们的结果表明，扩散随机性是具有多个有效输出的结构化预测任务的有效机制。
cs.CL / 35 / 2609.02366
NE-R1: Enhancing Named Entity Recognition Model via Reinforcement Learning
NE-R1：通过强化学习增强命名实体识别模型
Abstract
Named Entity Recognition (NER) has achieved substantial progress since the advent of large language models (LLMs). Nevertheless, the recognition of long-tail and domain-specific entities remains challenging due to the deficiency in parametric knowledge. Retrieval-augmented generation (RAG) offers a promising remedy by injecting external knowledge, but it also introduces noise and unnecessary cost when dealing with familiar cases. In this paper, we propose NE-R1, a novel framework for adaptive retrieval-augmented NER. We design a "retrieval-on-demand" mechanism for NER. Then we integrate it into models by a two-stage training method: (1) multi-task instruction tuning initialization; (2) end-to-end RL optimization with CoT. To achieve reasonable selection between parameterized and external knowledge, we design a multi-dimensional reward considering both accuracy and retrieval benefit. NE-R1 achieves state-of-the-art performance on various benchmarks, with an average F1 score gain of 2.52% in in-domain evaluation and 1.18% in zero-shot cross-domain evaluation.
Chinese Translation
自从大型语言模型（LLMs）出现以来，命名实体识别（NER）取得了显著进展。然而，由于参数知识的不足，长尾和领域特定实体的识别仍然具有挑战性。检索增强生成（RAG）通过注入外部知识提供了一种有前景的解决方案，但在处理熟悉案例时也引入了噪声和不必要的成本。在本文中，我们提出了NE-R1，一个用于自适应检索增强NER的新框架。我们为NER设计了一种“按需检索”机制。然后，我们通过两阶段训练方法将其集成到模型中：（1）多任务指令调优初始化；（2）结合链式推理（CoT）的端到端强化学习优化。为了在参数化知识和外部知识之间实现合理选择，我们设计了一种考虑准确性和检索收益的多维奖励。NE-R1在各种基准测试中实现了最先进的性能，在领域内评估中平均F1得分提升了2.52%，在零样本跨域评估中提升了1.18%。
cs.CL / 36 / 2609.02379
MultiGhostBench: A Multilingual Benchmark for Long-Form LLM-Generated Text Attribution under Distribution Shifts
MultiGhostBench：一个用于长文本 LLM 生成文本归属的多语言基准测试，适应于分布变化
Abstract
While existing work on LLM authorship attribution (AA) has made progress, available benchmarks remain limited, often focusing on English, controlled settings, or relatively outdated models, with the few multilingual studies considering only relatively short texts. We introduce MultiGhostBench, a multilingual benchmark comprising 928 books generated by five recent LLMs across six languages and three scripts, with an average length of approximately 59K words per book. The benchmark supports evaluation under domain, author, and language shifts. Evaluation of representative AA methods shows that no single method consistently performs best across settings, and performance generally degrades under distribution shifts. Transformer-based detectors can retain generator-related information across languages, although transfer effectiveness varies by language pair, whereas statistical and fingerprint-based detectors are more language-dependent. We envision MultiGhostBench as a valuable resource for the development and evaluation of robust AA methods. The dataset and code can be found at https://github.com/GrecoMT/MultiGhostBench.
Chinese Translation
尽管现有的 LLM 作者归属（AA）研究取得了一定进展，但现有的基准测试仍然有限，通常集中于英语、受控环境或相对过时的模型，而少数多语言研究仅考虑相对较短的文本。我们提出了 MultiGhostBench，这是一个多语言基准，包含由五个近期 LLM 生成的 928 本书，涵盖六种语言和三种书写系统，每本书的平均长度约为 59K 字。该基准支持在领域、作者和语言变化下的评估。对代表性 AA 方法的评估表明，没有单一方法在所有设置中始终表现最佳，且在分布变化下性能普遍下降。基于 Transformer 的检测器能够在不同语言间保留生成器相关信息，尽管转移效果因语言对而异，而基于统计和指纹的检测器则更依赖于语言。我们设想 MultiGhostBench 将成为开发和评估稳健 AA 方法的宝贵资源。数据集和代码可在 https://github.com/GrecoMT/MultiGhostBench 找到。
cs.CL / 37 / 2609.02391
PolERo: Studying Political Evasion in Romanian
PolERo：研究罗马尼亚的政治规避
Abstract
Political evasion refers to responses that engage with a question while withholding the requested information. Recent NLP work frames political evasion as a classification task using a two-level taxonomy of response clarity and fine-grained evasion strategies. Existing work on response clarity and evasion classification is limited to English, leaving open whether the taxonomy and model behavior transfer across languages and political contexts. We introduce PolERo, a dataset of 3,574 human-annotated question-answer pairs extracted from official transcripts of five Romanian presidents. We evaluate multiple classification approaches on both datasets under matched conditions, including TF-IDF baselines, fine-tuned encoder models, a proposed sliding-window encoder, and zero/few-shot LLM prompting. We study cross-lingual transfer through joint bilingual training and machine-translation-based data augmentation. Our results indicate that fine-tuned encoders are competitive, cross-lingual transfer is asymmetric, and ambivalent evasion categories involving pragmatic cues remain the main challenge across all model families.
Chinese Translation
政治规避是指在回应一个问题时，虽然参与了讨论，但却隐瞒了所请求的信息。近期的自然语言处理（NLP）研究将政治规避框架化为一个分类任务，使用了一个关于回应清晰度和细粒度规避策略的两级分类法。现有关于回应清晰度和规避分类的研究仅限于英语，因此尚不清楚该分类法和模型行为是否能够跨语言和政治背景进行迁移。我们引入了PolERo，一个包含3,574对人类标注的问答对的数据集，这些问答对提取自五位罗马尼亚总统的官方记录。我们在匹配条件下对多个分类方法进行了评估，包括TF-IDF基线、微调的编码器模型、提出的滑动窗口编码器以及零样本/少样本的LLM提示。我们通过联合双语训练和基于机器翻译的数据增强研究跨语言迁移。我们的结果表明，微调的编码器具有竞争力，跨语言迁移是不对称的，而涉及语用线索的模棱两可的规避类别仍然是所有模型系列面临的主要挑战。
cs.CL / 38 / 2609.02396
Improving Health Literacy through Lay Summarization of Radiological Reports: An Evaluation of BioNER and Retrieval-Augmented Generation
通过对放射学报告的通俗总结提高健康素养：BioNER和检索增强生成的评估
Abstract
Radiology reports are written primarily for clinicians, and their specialized terminology often makes them difficult for patients to interpret. As a result, many patients turn to publicly available Large Language Models (LLMs) to help explain their reports, despite well-documented risks of factual inaccuracies and hallucinations. Automated lay-summary generation has emerged as a promising alternative, yet the effectiveness of retrieval-enhanced and clinically informed approaches for radiology-specific communication remains underexplored. This study investigates the extent to which Retrieval-Augmented Generation (RAG) and Named Entity Recognition (NER) improve the quality, factual consistency, and readability of automatically generated lay summaries compared with standard LLM-based generation. We develop a framework combining NER-based extraction of clinically relevant findings with a RAG mechanism for contextual grounding, evaluated across few-shot and fine-tuned variants of two models (Qwen, BioBART). Results show that NER consistently improves readability and overall quality, while RAG alone offers no benefit and can introduce hallucinations from irrelevant retrieved terms. Combining RAG with NER degrades performance in few-shot settings but improves readability when fine-tuned. Fine-tuned BioBART with NER achieves the best overall performance, highlighting entity-aware extraction as the primary driver of improved patient-friendly summaries.
Chinese Translation
放射学报告主要是为临床医生撰写的，其专业术语常常使患者难以理解。因此，许多患者转向公共可用的大型语言模型（LLMs）来帮助解释他们的报告，尽管存在已充分记录的事实不准确和幻觉的风险。自动化的通俗总结生成作为一种有前景的替代方案出现，但针对放射学特定交流的检索增强和临床知情方法的有效性仍未得到充分探索。本研究调查了检索增强生成（RAG）和命名实体识别（NER）在提高自动生成的通俗总结的质量、事实一致性和可读性方面的程度，与标准的基于LLM的生成进行比较。我们开发了一个框架，将基于NER的临床相关发现提取与RAG机制结合，以实现上下文基础，在两个模型（Qwen, BioBART）的少量样本和微调变体中进行评估。结果表明，NER始终提高了可读性和整体质量，而单独使用RAG则没有益处，并可能引入来自无关检索术语的幻觉。在少量样本设置中，结合RAG和NER会降低性能，但在微调时提高了可读性。微调后的BioBART与NER结合实现了最佳整体性能，突显了以实体为中心的提取是改善患者友好型总结的主要驱动因素。
cs.CL / 39 / 2609.02414
Before the Script, Set the Stage: How Worldview Simulation Amplifies Psychologically Grounded Persuasion in Multi-Turn Jailbreaking
在剧本之前，搭建舞台：世界观模拟如何增强心理基础的多轮劝说中的影响力
Abstract
Multi-turn jailbreak attacks demonstrate that harmful intent can be distributed across dialogue, yet existing methods obscure what conversational mechanisms drive vulnerability. We introduce BLUEPRINT, a safety-evaluation framework separating a factorized social-influence strategy space from WORLDVIEWSIM, a cross-turn situational context module. Monte Carlo Tree Search optimizes turn-level combinations of 18 theory-grounded influence factors across a four-turn trajectory. Across six frontier models, BLUEPRINT achieves near-ceiling ASR on major open-weight and proprietary models, while requiring the fewest average queries (2.46). The resulting trajectories further reveal model-specific vulnerability among resistant targets: each responds to distinct influence factors and strategy transitions, yet all share a common recovery pathway-shifting toward concrete, executable task framing consistently escapes hard-refusal states. Ablations confirm operational cues matter most: making requests actionable has the largest impact, gain framing is unusually potent, and some legitimacy appeals can backfire. These findings suggest robust multi-turn safety requires monitoring not only harmful content, but also how dialogue state makes unsafe requests appear concrete and locally executable.
Chinese Translation
多轮越狱攻击表明，恶意意图可以在对话中分散存在，但现有方法模糊了驱动脆弱性的对话机制。我们提出了BLUEPRINT，一个安全评估框架，将分解的社会影响策略空间与WORLDVIEWSIM（世界观模拟），一个跨轮情境上下文模块分开。蒙特卡洛树搜索优化了18个理论基础影响因素在四轮轨迹中的轮次组合。在六个前沿模型中，BLUEPRINT在主要的开放权重和专有模型上达到了接近极限的ASR（自动语音识别），同时所需的平均查询次数最少（2.46）。结果轨迹进一步揭示了在抵抗目标中模型特定的脆弱性：每个目标对不同的影响因素和策略转换作出反应，但所有目标都共享一个共同的恢复路径——转向具体、可执行的任务框架，始终能够逃脱强拒绝状态。消融实验确认了操作线索的重要性：使请求可操作的影响最大，增益框架异常有效，而某些合法性诉求可能适得其反。这些发现表明，强大的多轮安全性不仅需要监控有害内容，还需要关注对话状态如何使不安全的请求看起来具体且局部可执行。
cs.CL / 40 / 2609.02438
When Decodability Is Not Enough: Logical Validity Representations, Behavioral Dissociation, and Causal Tests in Language Models
当可解码性不足时：语言模型中的逻辑有效性表征、行为分离与因果测试
Abstract
Large language models can look capable of logical reasoning, but correct or incorrect answers alone tell us little about what the model represents internally. We study logical verification in five open-weight transformer models using matched valid--invalid premise--claim pairs that vary across inference families, semantic domains, templates, and difficulty levels. Despite near-chance behavioral performance, logical validity is often almost perfectly decodable from hidden states and remains strongly decodable under held-out templates, domains, and inference families. Validity also remains highly decodable on behaviorally incorrect examples in the conditions where correctness-conditioned evaluation is well defined. At the same time, exhaustive leave-one-out tests reveal clear limits to this generalization, and interventions along probe-derived validity directions have only weak, nonspecific effects compared with random controls. Our results suggest that representing validity, expressing it in behavior, and using it causally are distinct. Validity related information can be strongly decodable from a model's hidden states without being reliably expressed in its output.
Chinese Translation
大型语言模型看似具备逻辑推理能力，但仅凭正确或错误的答案对模型内部表征的了解甚少。我们研究了五个开放权重的变换器模型中的逻辑验证，使用了在推理家族、语义领域、模板和难度等级上变化的匹配有效-无效前提-主张对。尽管行为表现接近随机水平，逻辑有效性通常可以从隐藏状态中几乎完美解码，并且在保留的模板、领域和推理家族下仍然保持强解码能力。在正确性条件评估定义良好的情况下，有效性在行为上错误的示例中也保持高度可解码。同时，详尽的逐一排除测试揭示了这种泛化的明显限制，而沿着探测导出的有效性方向进行的干预与随机对照相比，仅产生微弱且不特定的效果。我们的结果表明，表征有效性、在行为中表达有效性以及因果使用有效性是不同的。有效性相关信息可以从模型的隐藏状态中强解码，而不一定在其输出中可靠表达。
cs.CL / 41 / 2609.02473
Learning to Fuse LLMs with Ontology Rankers for Rare-Disease Diagnosis
学习将大型语言模型与本体排序器融合用于罕见疾病诊断
Abstract
Ontology rankers remain useful for rare-disease diagnosis because each candidate can be traced to matched patient phenotypes. Large language models (LLMs) can generate differential diagnoses from the same patient description, but their predictions lack an equally clear evidence trail. Rather than asking which system should replace the other, we ask whether an LLM can improve the ranker without giving up its evidence. Our behavior-based fusion model examines the two ranked lists, their agreement, and the ontology support behind each candidate, and learns how much to rely on each system for the individual case. Before comparison, we remove a documented test-set leakage pathway caused by benchmark cases and ontology annotations being derived from the same publications. Across eight open LLMs, fusion improves Phenomizer Recall@1 by 7.86 percentage points on Phenopacket Store and 20.18 points on RAMEDIS. When paired with DeepSeek-V4-Flash through an API, a fusion model trained only on the other LLMs improves Recall@1 from 0.1657 to 0.2176, a 5.19-point gain, without retraining. For 90.8% of correct fused diagnoses, the disease retains candidate-level ontology evidence that can be inspected. These results show that LLMs can strengthen an established diagnostic tool without discarding the structured evidence that makes it useful.
Chinese Translation
本体排序器在罕见疾病诊断中仍然具有重要价值，因为每个候选者都可以追溯到匹配的患者表型。大型语言模型（LLMs）能够根据相同的患者描述生成鉴别诊断，但其预测缺乏同样清晰的证据链。我们不是在问哪个系统应该取代另一个，而是询问LLM是否可以在不放弃其证据的情况下改善排序器。我们的基于行为的融合模型考察了两个排名列表、它们之间的一致性以及每个候选者背后的本体支持，并学习在个案中应依赖于每个系统的程度。在比较之前，我们消除了由基准案例和本体注释源自同一出版物所导致的文档测试集泄漏路径。在八个开放的LLM中，融合模型在Phenopacket Store上将Phenomizer的Recall@1提高了7.86个百分点，在RAMEDIS上提高了20.18个百分点。当通过API与DeepSeek-V4-Flash配对时，仅基于其他LLM训练的融合模型将Recall@1从0.1657提高到0.2176，增幅为5.19点，且无需重新训练。对于90.8%的正确融合诊断，疾病保留了可以检查的候选级本体证据。这些结果表明，LLM可以增强一个已建立的诊断工具，而不丢弃使其有用的结构化证据。
cs.CL / 42 / 2609.02480
PragAlign: Feedback-Guided Pragmatic Alignment for Controlled Synthetic Dialogue Generation
PragAlign：基于反馈的务实对齐框架用于受控合成对话生成
Abstract
Synthetic dialogue generation can support research in privacy-restricted service settings, but generated conversations must preserve communicative intent, affective meaning, and natural dialogue flow. We introduce PragAlign, a feedback-guided framework for controlled synthetic dialogue generation conditioned on service context, target intent, and target emotion, with auxiliary trait-style controls. PragAlign uses a generate--evaluate--revise loop in which an LLM-based evaluator scores intent alignment, emotion alignment, coherence, fluency, and aggregate quality, then provides criterion-specific feedback for up to three refinement rounds. On 800 matched dialogue specifications, PragAlign achieves 99.50\% evaluator-defined acceptance, compared with 72.25\% for one-shot generation and 95.88\% for repeated generation without structured feedback. This indicates that repeated attempts account for much of the gain over one-shot generation, while structured feedback primarily improves last-mile multi-constraint satisfaction rather than broad average quality. Refinement gains are concentrated in emotion alignment, which is also the dominant failure mode in ablations. A separate human evaluation of 1,200 generated dialogues shows that intent expression and dialogue flow are highly recognizable to annotators, while emotion appropriateness is less stable and more subjective. These results support PragAlign as a quality-control framework for improving evaluator-defined communicative constraint satisfaction, while showing that affective realization and independent human-perceived quality remain open challenges.
Chinese Translation
合成对话生成可以支持在隐私受限的服务环境中的研究，但生成的对话必须保持交际意图、情感意义和自然对话流畅性。我们提出了PragAlign，这是一个基于反馈的框架，用于在服务上下文、目标意图和目标情感的条件下进行受控合成对话生成，并具有辅助的特征风格控制。PragAlign使用生成-评估-修订循环，其中基于大型语言模型（LLM）的评估者对意图对齐、情感对齐、一致性、流畅性和整体质量进行评分，然后提供针对特定标准的反馈，最多进行三轮修订。在800个匹配的对话规范上，PragAlign实现了99.50%的评估者定义的接受率，而一次性生成的接受率为72.25%，没有结构化反馈的重复生成的接受率为95.88%。这表明，重复尝试在很大程度上解释了相较于一次性生成的增益，而结构化反馈主要改善了最后一公里的多约束满足，而不是广泛的平均质量。修订增益主要集中在情感对齐上，而这也是消融实验中的主要失败模式。对1200个生成对话的单独人工评估显示，意图表达和对话流畅性对注释者来说是高度可识别的，而情感适当性则不太稳定且更具主观性。这些结果支持PragAlign作为一个质量控制框架，以改善评估者定义的交际约束满足，同时表明情感实现和独立的人类感知质量仍然是开放的挑战。
cs.CL / 43 / 2609.02482
How LLMs Build Fictional Worlds: Setting and Narrative Space in AI-Generated Creative Storytelling
大型语言模型如何构建虚构世界：AI生成创意叙事中的背景与叙事空间
Abstract
In this paper, we analyze how Large Language Models (LLMs) employ worldbuilding strategies, focusing on setting as one measurable dimension of storyworld construction. We compare 1,000 AI-generated stories per model in English and German with human-authored fiction from Project Gutenberg. Building on prior work, we operationalize setting through five types of narrative space: "action", "perceived," "visual," "descriptive" and "no space", identified using fine-tuned BERT classifiers for German and English. We generate narratives using GPT 4.1, LlaMA 3.3, Mistral 3.2, and Gemma 3 and compare their spatial distributions to a human-authored baseline. We find that human-authored texts predominantly employ "action space," grounding narratives in embodied character-environment interaction, whereas LLMs systematically overproduce "perceived space," emphasizing atmosphere and affect. This divergence remains stable across narrative time. Overall, our findings show that LLMs exhibit worldbuilding patterns that differ consistently from human-authored fiction in ways that are both model-specific and language-sensitive.
Chinese Translation
在本文中，我们分析了大型语言模型（LLMs）如何采用世界构建策略，重点关注背景作为故事世界构建的一个可测量维度。我们比较了每个模型生成的1,000个英语和德语的AI生成故事与古腾堡计划中的人类创作小说。基于先前的研究，我们通过五种叙事空间类型对背景进行操作化：'行动'、'感知'、'视觉'、'描述'和'无空间'，这些类型是通过针对德语和英语的微调BERT分类器识别的。我们使用GPT 4.1、LlaMA 3.3、Mistral 3.2和Gemma 3生成叙事，并将其空间分布与人类创作的基线进行比较。我们发现，人类创作的文本主要采用'行动空间'，将叙事建立在具身的角色与环境的互动中，而LLMs则系统性地过度生成'感知空间'，强调氛围和情感。这种差异在叙事时间上保持稳定。总体而言，我们的研究结果表明，LLMs展现出与人类创作小说在模式上和语言上均存在一致差异的世界构建模式。
cs.CL / 44 / 2609.02496
Debias-SparseGPT: Bias-Aware Pruning for Large Language Models
Debias-SparseGPT：面向大型语言模型的偏见感知剪枝
Abstract
Model compression techniques such as pruning and quantization facilitate the efficient deployment and acceleration of Large Language Models (LLMs). However, recent studies show that weight sparsification methods, such as SparseGPT, can amplify existing biases in models, with outputs varying significantly depending on persona cues in the prompt. In this paper, we introduce Debias-SparseGPT, a post-training pruning method incorporating representational debiasing using a second-order term defined over demographically contrasting inputs. We perform empirical validation of our method over a wide range of generative LLMs. Across models and sparsity regimes (25%, 50%, and structured 2:4 sparsity), Debias-SparseGPT consistently reduces pruning-induced bias compared to SparseGPT while preserving model perplexity and zero-shot accuracy. Under the most restrictive 2:4 structured sparsity pattern, which most aggressively degrades model quality, augmenting the calibration set with long-context, content-rich examples further improves both downstream performance and fairness. Overall, Debias-SparseGPT advances the bias-performance trade-off while preserving the computational efficiency of sparse models.
Chinese Translation
模型压缩技术，如剪枝和量化，促进了大型语言模型（LLMs）的高效部署和加速。然而，最近的研究表明，权重稀疏化方法，如SparseGPT，可能会放大模型中已有的偏见，输出结果在很大程度上依赖于提示中的角色提示。在本文中，我们介绍了Debias-SparseGPT，这是一种后训练剪枝方法，结合了基于人口统计对比输入定义的二阶项的表征去偏见。我们对该方法在广泛的生成型LLM上进行了实证验证。在不同模型和稀疏性范围（25%、50%和结构化2:4稀疏性）下，Debias-SparseGPT在保持模型困惑度和零样本准确率的同时，始终减少了与SparseGPT相比的剪枝引起的偏见。在最严格的2:4结构化稀疏模式下，该模式最显著地降低模型质量，通过增加包含长上下文和丰富内容示例的校准集，进一步改善了下游性能和公平性。总体而言，Debias-SparseGPT在保持稀疏模型计算效率的同时，推动了偏见与性能之间的权衡。
cs.CL / 45 / 2609.02526
When Persona Attributes Improve Population Alignment in Large Language Models
当个性属性改善大型语言模型中的人口对齐
Abstract
Large Language Models (LLMs) are increasingly used to predict the responses of human participants in survey panels. Towards that goal, persona prompting has recently emerged as a technique to inform and align large pretrained language models. Persona prompting refers to the practice of using short textual descriptions of 'personas' in prompts to steer the LLM's generations. Personas describe individuals through different attributes such as their socio-demographics, attitudes, or behaviors, with the aim of aligning LLMs to produce responses that correlate with the corresponding human responses. Yet, recent work has produced mixed and partly conflicting results of persona prompting without clear patterns of success and failure. Among the few consistent findings is that the selection of persona attributes matters, and that using more attributes does not necessarily lead to better performance. It remains unclear how different attribute selection methods perform and how to choose among them. In this paper, we propose that observed human response variation of a survey question is a potential explanation for the mixed performance observed so far. In addition, we compare the performance of persona prompting associated with different methods for selecting persona attributes. We evaluate these methods on four different (general) social surveys across two countries, six LLMs, and twenty prediction tasks per survey. Our work helps to identify when persona prompting can be expected to be useful in survey prediction tasks, and provides new insights on the effectiveness of different attribute selection methods for LLM-based survey prediction using persona prompting.
Chinese Translation
大型语言模型（LLMs）越来越多地用于预测调查小组中人类参与者的响应。为此，个性提示（persona prompting）最近作为一种技术出现，用于告知和对齐大型预训练语言模型。个性提示是指在提示中使用“个性”的简短文本描述，以引导LLM的生成。个性通过不同的属性描述个体，例如他们的社会人口统计特征、态度或行为，旨在使LLM生成与相应人类响应相关的回答。然而，最近的研究结果显示个性提示的效果参差不齐，且部分结果相互矛盾，未能明确成功与失败的模式。在为数不多的一些一致发现中，个性属性的选择至关重要，且使用更多属性并不一定会导致更好的表现。目前尚不清楚不同的属性选择方法的表现如何，以及如何在它们之间进行选择。在本文中，我们提出调查问题中观察到的人类响应变异可能是迄今为止观察到的混合表现的潜在解释。此外，我们比较了与不同个性属性选择方法相关的个性提示的表现。我们在两个国家的四个不同（一般）社会调查中评估这些方法，涉及六个LLM和每个调查的二十个预测任务。我们的研究有助于识别在调查预测任务中何时可以期待个性提示的有效性，并为基于LLM的调查预测中使用个性提示的不同属性选择方法的有效性提供新的见解。
cs.CL / 46 / 2609.02606
Predictors of Loneliness in Older Adults Using Multimodal Analysis of Speech and Language
老年人孤独感的预测因素：基于多模态语音和语言分析
Abstract
Loneliness is a critical public health issue among older adults, linked to higher risks of depression, cognitive decline, and mortality. Scalable, objective methods for its detection remain limited, particularly in natural conversational contexts. We analyzed speech and language markers of loneliness in 310 older adults using semi-structured telephone interviews to help understand how they process feeling lonely and how their language differs at different levels of feeling loneliness. Our multimodal framework combined linguistic features (psycholinguistic dictionaries, n-grams, and topic models) with acoustic features (pitch, tone, loudness) to examine associations with self-reported loneliness scores. Both predefined and data-driven methods captured patterns in verbal content and vocal delivery. Higher loneliness was associated with negations(r = 0.11), negative tone(r = 0.12), and conflict-related language. Lower loneliness was linked to social references(r = -0.18), motivational drives(r = -0.11), and emotional richness in speech(r = -0.12). We also found that the multimodal model (r = 0.298) outperforms the text-only and audio-only models. Findings suggest that loneliness manifests through both linguistic and acoustic cues, supporting the potential of speech-based analysis in psychological assessments and as an early indicator of emotional loneliness when used alongside existing assessments, rather than as standalone diagnostic tools.
Chinese Translation
孤独感是老年人中一个重要的公共卫生问题，与抑郁、认知衰退和死亡率的风险增加相关。可扩展的、客观的检测方法仍然有限，尤其是在自然对话环境中。我们通过对310名老年人进行半结构化电话访谈，分析了孤独感的语音和语言标记，以帮助理解他们如何处理孤独感以及他们的语言在不同孤独感水平下的差异。我们的多模态框架结合了语言特征（心理语言学词典、n-gram和主题模型）与声学特征（音调、语气、响度），以考察与自我报告的孤独感评分的关联。预定义和数据驱动的方法都捕捉到了言语内容和声音表达的模式。更高的孤独感与否定词（r = 0.11）、消极语气（r = 0.12）和冲突相关语言相关联。较低的孤独感则与社交引用（r = -0.18）、动机驱动（r = -0.11）和言语中的情感丰富性（r = -0.12）相关。我们还发现多模态模型（r = 0.298）优于仅文本和仅音频模型。研究结果表明，孤独感通过语言和声学线索表现出来，支持在心理评估中使用基于语音的分析的潜力，并作为情感孤独的早期指标，与现有评估工具结合使用，而不是作为独立的诊断工具。
cs.CL / 47 / 2609.02639
TaRA: Training-Aware Low-Rank Adaptation Initialization
TaRA：训练感知的低秩适应初始化
Abstract
Low-Rank Adaptation (LoRA) has become a de facto standard for parameter-efficient fine-tuning (PEFT), yet its performance is highly sensitive to initialization due to the information bottleneck imposed by low-rank decomposition. Existing approaches attempt to construct high-quality LoRA initializations by exploiting principal components of pretrained weights, activations, or gradients. However, these methods do not directly account for the training dynamics of the full-rank model. In this paper, we propose Training-aware Low-Rank Adaptation Initialization (TaRA), a method that initializes LoRA such that the gradients induced by the low-rank factors closely approximate the gradient of the corresponding full-rank weight matrix. Derived from a mathematical formulation, TaRA improves gradient fidelity at the start of training while introducing negligible computational overhead. Across diverse and challenging fine-tuning tasks, TaRA consistently outperforms prior state-of-the-art methods, establishing a simple, robust, and scalable solution for effective LoRA initialization.
Chinese Translation
低秩适应（Low-Rank Adaptation, LoRA）已成为参数高效微调（Parameter-Efficient Fine-Tuning, PEFT）的事实标准，但由于低秩分解所施加的信息瓶颈，其性能对初始化高度敏感。现有方法试图通过利用预训练权重、激活或梯度的主成分来构建高质量的 LoRA 初始化。然而，这些方法并未直接考虑全秩模型的训练动态。本文提出了一种训练感知的低秩适应初始化方法（Training-aware Low-Rank Adaptation Initialization, TaRA），该方法初始化 LoRA，使得低秩因子所引起的梯度与相应的全秩权重矩阵的梯度紧密近似。TaRA 源自数学公式，能够在训练开始时提高梯度的保真度，同时引入的计算开销微乎其微。在多样且具有挑战性的微调任务中，TaRA 始终优于先前的最先进方法，建立了一种简单、稳健且可扩展的有效 LoRA 初始化解决方案。
cs.CL / 48 / 2609.02651
WinoQueer-NL: Assessing Bias in Dutch Language Models toward LGBTQ+ Identities
WinoQueer-NL：评估荷兰语言模型对LGBTQ+身份的偏见
Abstract
While English language models have been widely examined for anti-queer bias, Dutch models remain understudied. To address this gap, we developed a culturally and linguistically adapted Dutch dataset based on the English WinoQueer benchmark, containing pairs of stereotypical and counter-stereotypical sentences. To validate and expand it, we conducted an online survey with 43 Dutch queer participants, confirming 145 of 171 stereotypes as culturally relevant and identifying 22 new biases through free-text responses. The final released dataset, comprising 42,906 sentences, was evaluated using a range of Dutch-specific and multilingual models, including both masked language models (MLMs) and autoregressive language models (ARLMs), with bias measured via a score comparing log-likelihoods of stereotypical versus counter-stereotypical sentences. While the mean bias score across models appeared neutral (~50%), closer analysis revealed significant disparities: some models favored stereotypical sentences up to 97% of the time for transgender identities, but only 6% of the time for gay-related pairs, with transgender and non-binary identities consistently receiving the highest bias scores. Our findings highlight the importance of culturally grounded datasets for evaluating and mitigating biases that disproportionately impact marginalized groups in Dutch language models.
Chinese Translation
尽管英语语言模型已被广泛研究以探讨反酷儿偏见，但荷兰模型仍然未得到充分研究。为了解决这一空白，我们基于英语WinoQueer基准开发了一个文化和语言适配的荷兰数据集，该数据集包含成对的刻板印象句子和反刻板印象句子。为了验证和扩展该数据集，我们与43名荷兰酷儿参与者进行了在线调查，确认了171个刻板印象中145个具有文化相关性，并通过自由文本回应识别出22种新的偏见。最终发布的数据集包含42,906个句子，使用一系列荷兰特定和多语言模型进行了评估，包括掩蔽语言模型（MLMs）和自回归语言模型（ARLMs），偏见通过比较刻板印象句子与反刻板印象句子的对数似然得分进行测量。尽管各模型的平均偏见得分看似中性（约50%），但更深入的分析揭示了显著差异：某些模型在跨性别身份方面高达97%的时间偏向于刻板印象句子，而在与同性恋相关的句子对中仅有6%的时间偏向于刻板印象句子，跨性别和非二元身份始终获得最高的偏见得分。我们的研究结果强调了基于文化的数据集在评估和减轻对荷兰语言模型中边缘化群体产生不成比例影响的偏见的重要性。
cs.CL / 49 / 2609.02672
oHC: Orthogonal Hyper-Connections on SO(4) via Quaternions
oHC：通过四元数在 SO(4) 上的正交超连接
Abstract
Hyper-Connections (HC) replace the single residual stream of a Transformer with $n$ parallel ones, mixing them at every layer with a learned $n \times n$ residual matrix. Leaving that matrix unconstrained places no limit on the factor by which the mixing step rescales the residual streams, and that factor compounds across layers, which destabilizes training. Manifold-constrained Hyper-Connections (mHC) address this by restricting the matrix to the doubly stochastic matrices. That caps the factor at one, so the mixing can no longer amplify any direction, but nothing bounds it from below. We prove that inside this set the mixing step can reduce the norm of the residual streams only by shrinking the differences between the streams, while their mean is left unchanged; and since the reduction accumulates over layers, the streams grow more alike and their diversity is spent with depth. We therefore propose Orthogonal Hyper-Connections (oHC), restricting the residual matrix to the rotation group $SO(n)$, so that the mixing step can neither amplify nor attenuate the residual streams in any direction, which keeps training stable and no longer forces the differences between the streams to contract. Specifically, at the four streams used by recent HC models we parameterize the group in closed form by a pair of unit quaternions, which adds no parameters, replaces the iterative projection with a fixed pattern of signed additions, and can be constructed faster than mHC. We evaluate oHC across a comprehensive set of downstream tasks, where it outperforms the single-stream residual baseline, mHC and iHC, which fixes the residual matrix to the identity.
Chinese Translation
超连接（Hyper-Connections, HC）用 $n$ 个并行的残差流替代了 Transformer 的单一残差流，并在每一层通过学习得到的 $n imes n$ 残差矩阵进行混合。将该矩阵不加约束地使用并不会限制混合步骤对残差流的重新缩放因子，而该因子在层间累积，导致训练不稳定。流形约束超连接（Manifold-constrained Hyper-Connections, mHC）通过将矩阵限制为双随机矩阵来解决这一问题。这将因子上限设定为1，因此混合不再能够放大任何方向，但也没有下限约束。我们证明，在这一集合内，混合步骤只能通过缩小流之间的差异来减少残差流的范数，而它们的均值保持不变；由于这种减少在层间累积，流变得更加相似，其多样性随着深度的增加而消耗。因此，我们提出了正交超连接（Orthogonal Hyper-Connections, oHC），将残差矩阵限制在旋转群 $SO(n)$ 上，使得混合步骤在任何方向上都无法放大或减弱残差流，从而保持训练稳定，并不再强迫流之间的差异收缩。具体而言，在最近的 HC 模型中使用的四个流上，我们通过一对单位四元数以封闭形式对该群进行参数化，这不会增加参数数量，将迭代投影替换为固定模式的符号加法，并且构建速度快于 mHC。我们在一系列全面的下游任务中评估了 oHC，其表现优于单流残差基线、mHC 和将残差矩阵固定为单位矩阵的 iHC。
cs.CL / 50 / 2609.02679
From Tokens to Semantics: Leveraging Complementary Signals for Hallucination Detection in Black-Box LLMs
从标记到语义：利用互补信号检测黑箱大语言模型中的幻觉
Abstract
When LLMs support public-facing or high-stakes workflows, missed fabrications can harm users and institutions, while false alarms consume limited human-review capacity. When no trusted context or reference document is available, we study two signals accessible through black-box model APIs: semantic entropy, which measures disagreement among sampled response meanings, and uncertainty derived from token log-probabilities. Their failure modes can be complementary: semantic entropy becomes uninformative when responses form one semantic cluster, while token uncertainty can miss consistently confident errors. We extend token-based uncertainty detection by aggregating token-level signals across sampled responses through our TopK method, evaluate the hybrid CoCoA method, which combines target-response uncertainty with semantic dissimilarity, and propose and study two supervised methods: Gated, which routes single-cluster cases to an aggregated-token-feature classifier, and Stacked, which learns jointly from semantic uncertainty and broader token features. We evaluate seven benchmarks, including five public benchmarks (four text datasets and multimodal handwritten-cheque extraction) and two constructed benchmarks (Financial Summaries and Long-Text QA), using four language models. In our evaluation across models and datasets, Stacked gave the best performance in nearly half of the cases, while TopK and CoCoA remain competitive without supervised training labels, although their thresholds require careful calibration. No method is universally strongest. We therefore evaluate performance at false-positive-rate budgets from 1% to 15%, assess their sensitivity to generation and calibration choices, and examine variation across dataset characteristics.
Chinese Translation
当大语言模型（LLMs）支持面向公众或高风险的工作流程时，未能识别的虚假信息可能会对用户和机构造成伤害，而错误警报则消耗有限的人力审核能力。在没有可信上下文或参考文档的情况下，我们研究了通过黑箱模型API可获取的两种信号：语义熵，它衡量采样响应意义之间的不一致性，以及来自标记对数概率的不确定性。这两种信号的失效模式可能是互补的：当响应形成一个语义簇时，语义熵变得无信息，而标记不确定性可能会遗漏持续自信的错误。我们通过我们的TopK方法扩展了基于标记的不确定性检测，聚合了通过采样响应获得的标记级信号，评估了混合的CoCoA方法，该方法结合了目标响应的不确定性和语义差异性，并提出并研究了两种监督方法：Gated，该方法将单簇案例路由到聚合标记特征分类器，以及Stacked，该方法从语义不确定性和更广泛的标记特征中共同学习。我们使用四种语言模型评估了七个基准，包括五个公共基准（四个文本数据集和多模态手写支票提取）和两个构建的基准（财务摘要和长文本问答）。在我们对模型和数据集的评估中，Stacked在近一半的情况下表现最佳，而TopK和CoCoA在没有监督训练标签的情况下仍然具有竞争力，尽管它们的阈值需要仔细校准。没有一种方法是普遍最强的。因此，我们在1%到15%的假阳性率预算下评估性能，评估它们对生成和校准选择的敏感性，并检查数据集特征之间的变化。
cs.CL / 51 / 2609.02685
DKL: Decoupled Knowledge Learning for Instruction-Tuned Language Models
DKL：用于指令调优语言模型的解耦知识学习
Abstract
RAG has become the de facto method for incorporating new, corpus-specific knowledge into an instruction following LLM (Instruct LLM). Although RAG-based prompting improves factual grounding, it fails when retrieval is incorrect or incomplete, leading to hallucinations. Finetuning methods such as RAFT and PA-RAG enhance RAG by injecting new knowledge into the model's parameters, but require generating a massive amount of synthetic QA that covers the entire corpus. Extended Pre-Training (EPT) on the text corpus avoids the need for comprehensive synthetic data generation but compromises an Instruct LLM's instruction-following capabilities, necessitating instruction fine-tuning (IFT) after pre-training. However, IFT is costly and may be infeasible due to the unavailability of an instruction-tuning corpus. In this work, we propose DKL-Decoupled Knowledge Learning for Instruction-Tuned Language Models. Instead of doing EPT on the Instruct LLM, DKL performs EPT on its corresponding base LLM to infuse new knowledge. These knowledge infused weights are then merged with the Instruct LLM, imparting new knowledge without affecting their instruction-following capabilities. DKL is a lightweight method that avoids expensive instruction fine-tuning and relies on model merging to infuse the new knowledge into the Instruct LLM without destroying its instruction following capabilities. Empirical results show that DKL improves RAG accuracy from 54.17 to 79.26 on retrieval failure cases, while outperforming prior approaches with substantially less training data.
Chinese Translation
RAG 已成为将新的、特定语料库知识融入指令跟随的语言模型（Instruct LLM）的事实方法。尽管基于 RAG 的提示改善了事实基础，但在检索不正确或不完整时会失败，导致幻觉。诸如 RAFT 和 PA-RAG 的微调方法通过将新知识注入模型参数来增强 RAG，但需要生成大量覆盖整个语料库的合成问答（QA）。在文本语料库上进行扩展预训练（EPT）避免了全面合成数据生成的需求，但妥协了 Instruct LLM 的指令跟随能力，因此在预训练后需要进行指令微调（IFT）。然而，IFT 成本高昂，且由于缺乏指令调优语料库可能不可行。在本研究中，我们提出了 DKL——用于指令调优语言模型的解耦知识学习。DKL 不在 Instruct LLM 上进行 EPT，而是在其对应的基础 LLM 上进行 EPT，以注入新知识。这些注入知识的权重随后与 Instruct LLM 合并，赋予新知识而不影响其指令跟随能力。DKL 是一种轻量级方法，避免了昂贵的指令微调，并依赖模型合并将新知识注入 Instruct LLM，而不破坏其指令跟随能力。实证结果表明，DKL 在检索失败案例中将 RAG 的准确率从 54.17 提高到 79.26，同时在训练数据量显著减少的情况下超越了先前的方法。
cs.CL / 52 / 2609.02702
Trace as State: Reasoning Traces as Conditional States for Long-Context Transformers
追踪作为状态：将推理追踪作为长上下文变换器的条件状态
Abstract
Transformers process information causally, but long-context reasoning may depend on task state discovered only later. We formalize this mismatch through conditional state update tasks. For causal state update processors, providing the condition first can require exponentially less memory in the worst case than providing it last. Motivated by this principle, we introduce Trace as State. We use collected reasoning traces as a textual proxy for task state and place it before the long-context block on a fresh pass, allowing information derived previously to guide rereading. We conduct extensive experiments on Trace as State and Trace Append, a matched control that uses the same task state proxy but put it after the context. Across three models and three long-context datasets, Trace as State outperforms Trace Append in 26 of 27 reported combinations of model, task, and metric. On GraphWalks Parents, exact match lifts DeepSeek V4 Pro Preview from 29.2% on the initial pass and 43.0% with Trace Appendto 81.8% with Trace as State, and from 66.4% and 83.2% to 100.0% for GLM-5.2. These results show that placing traces before the context can improve long-context reasoning while retaining the causal transformer structure.
Chinese Translation
变换器以因果方式处理信息，但长上下文推理可能依赖于仅在后期发现的任务状态。我们通过条件状态更新任务形式化这种不匹配。对于因果状态更新处理器，首先提供条件在最坏情况下可能需要的内存比最后提供条件少得多，减少的量级是指数级的。基于这一原则，我们提出了追踪作为状态（Trace as State）。我们使用收集的推理追踪作为任务状态的文本代理，并在新一轮处理时将其放置在长上下文块之前，从而允许之前获得的信息指导重新阅读。我们对追踪作为状态和追踪附加（Trace Append）进行了广泛的实验，后者使用相同的任务状态代理，但将其放置在上下文之后。在三个模型和三个长上下文数据集上，追踪作为状态在27种报告的模型、任务和指标组合中有26种超越了追踪附加。在GraphWalks Parents上，精确匹配将DeepSeek V4 Pro Preview从初次处理的29.2%和使用追踪附加的43.0%提升至使用追踪作为状态的81.8%，而GLM-5.2的结果则从66.4%和83.2%提升至100.0%。这些结果表明，将追踪放置在上下文之前可以改善长上下文推理，同时保持因果变换器的结构。
cs.CL / 53 / 2609.02730
CORAL: An LLM-Native Harness for Production Recommender Systems
CORAL：一种原生于大语言模型的生产推荐系统框架
Abstract
Production recommender systems shape what billions of people see, and sustaining their performance requires continual optimization: as content, user behavior, and upstream models shift, the choices governing retrieval, ranking, and serving must be revisited. Traditionally, human engineers test such changes through online experiments--a slow, reactive process limited by engineering effort, leaving parts of the system unrevised as conditions change. Although large language models have been applied to ranking, user modeling, and offline model development, few systems place an agent in a continual closed loop that acts on a live recommender and learns from the measured effects of its decisions. We present CORAL (Constraint-Optimized Recommender via an Agentic Loop), an LLM-native harness that closes this loop: each cycle, the agent observes operating signals, reasons over a memory of past decisions and outcomes, and invokes tools--including a numerical optimizer that keeps changes within a fixed operating budget--to reconfigure the recommender, with measured outcomes informing the next cycle. We formulate this as a partially observed, non-stationary, constrained optimization problem in which the policy improves in context, without parameter updates, from its prior actions. Across two large-scale social platforms, evaluated with A/B experiments, the same harness improves engagement at no additional serving cost on one and reduces serving cost without degrading engagement on the other, spanning the engagement-efficiency frontier. Performance improves as the loop iterates, suggesting that a single agentic loop can automate continual optimization work traditionally performed by human algorithm engineers under explicit guardrails.
Chinese Translation
生产推荐系统塑造了数十亿人所看到的内容，维持其性能需要持续的优化：随着内容、用户行为和上游模型的变化，决定检索、排序和服务的选择必须重新审视。传统上，人类工程师通过在线实验测试这些变化——这是一种缓慢的、反应式的过程，受限于工程努力，导致系统的某些部分在条件变化时未得到修订。尽管大语言模型已被应用于排序、用户建模和离线模型开发，但很少有系统将代理置于一个持续的闭环中，该代理在实时推荐系统上采取行动，并从其决策的测量效果中学习。我们提出了CORAL（通过代理循环进行约束优化的推荐系统），这是一种原生于大语言模型的框架，旨在关闭这一循环：在每个周期中，代理观察操作信号，基于过去决策和结果的记忆进行推理，并调用工具——包括一个在固定操作预算内保持变更的数值优化器——来重新配置推荐系统，测量的结果为下一个周期提供信息。我们将其形式化为一个部分可观察的、非平稳的、受约束的优化问题，其中策略在上下文中改进，而无需参数更新，基于其先前的行动。在两个大型社交平台上进行的A/B实验评估中，同一框架在一个平台上提高了参与度而没有额外的服务成本，而在另一个平台上则降低了服务成本而不降低参与度，跨越了参与度与效率的边界。随着循环的迭代，性能不断提升，这表明单一的代理循环可以自动化传统上由人类算法工程师在明确的保护措施下执行的持续优化工作。
cs.CL / 54 / 2609.02735
Choosing a PEFT Variant for Per-Patient Dysarthric ASR: A Single-Speaker Case Study on Two ASR Bases
为每位患者选择PEFT变体以应对言语障碍的自动语音识别：基于两种ASR基础的单一说话者案例研究
Abstract
Per-patient adapters are the preferred production architecture for dysarthric automatic speech recognition (ASR), yet parameter-efficient fine-tuning (PEFT) variants have not been compared in the speaker-dependent, per-patient regime. We present a single-speaker case study comparing seven LoRA-family methods (LoRA, QLoRA, AdaLoRA, DoRA, LoHA, VeRA, VB-LoRA) on two production bases (Whisper-large-v3 with Hungarian fine-tuning, and a multilingual Qwen3-ASR-1.7B checkpoint) for one post-stroke Hungarian male speaker (S1, 409 utterances; severe dysarthria on auditory-perceptual clinical assessment). Attention-projection adapters substantially improve CER on both bases. Across three seeds, a paired bootstrap detects no significant LoRA-DoRA difference (p>0.5; 13.86/13.90 % CER on Whisper, 28.10/28.33 % on Qwen3-ASR), so we adopt the simpler, cheaper LoRA. Real 4-bit (NF4) QLoRA is worse on every seed and both bases (14.56/30.09 % CER) with no memory saving at this scale, and LoHA, VeRA, VB-LoRA and AdaLoRA do not reach the LoRA family, though LoHA still gives an 18.6 % relative CER reduction on Whisper. On the same base, full fine-tuning is more accurate (11.43 % CER), but a 115 MB LoRA that also adapts the feed-forward blocks reaches within 0.66 pp of it at approximately 3.7 % of the per-patient storage. A 6-point enrollment grid shows about 5 min of patient audio captures 45.6 % of the zero-shot-to-30-min CER reduction, with further gains at 10 and 30 min (caveat: one speaker, one language, severe post-stroke dysarthria). Training scripts and recipes will be released, source-available under a research-use licence, on publication.
Chinese Translation
针对每位患者的适配器是言语障碍自动语音识别（ASR）的首选生产架构，但在说话者依赖的每位患者模式中，尚未比较过参数高效微调（PEFT）变体。我们展示了一项单一说话者案例研究，比较了七种LoRA家族方法（LoRA、QLoRA、AdaLoRA、DoRA、LoHA、VeRA、VB-LoRA）在两个生产基础（Whisper-large-v3与匈牙利语微调，以及多语言Qwen3-ASR-1.7B检查点）上的表现，研究对象为一名中风后匈牙利男性说话者（S1，409次发声；在听觉感知临床评估中表现为严重言语障碍）。注意力投影适配器在两个基础上显著提高了字符错误率（CER）。在三个随机种子下，配对自助法未检测到显著的LoRA-DoRA差异（p>0.5；Whisper上的CER为13.86/13.90%，Qwen3-ASR上的CER为28.10/28.33%），因此我们采用了更简单、更便宜的LoRA。真实的4位（NF4）QLoRA在每个种子和两个基础上表现更差（CER为14.56/30.09%），在此规模下没有内存节省，而LoHA、VeRA、VB-LoRA和AdaLoRA未能达到LoRA家族的水平，尽管LoHA在Whisper上仍提供了18.6%的相对CER降低。在相同基础上，完全微调的准确率更高（CER为11.43%），但一个115 MB的LoRA同时适配前馈块，达到的CER仅比其高出0.66个百分点，且存储需求约为每位患者的3.7%。6点登记网格显示，约5分钟的患者音频捕获了从零样本到30分钟CER降低的45.6%，在10分钟和30分钟时还有进一步的提升（注意：一位说话者，一种语言，严重的中风后言语障碍）。训练脚本和配方将在出版时发布，源代码将在研究使用许可下提供。
cs.CL / 55 / 2609.02737
Language Models Can Control Their Own Attention
语言模型可以控制自己的注意力
Abstract
Language models spend most of their attention on a small fraction of context, yet they read the entire KV cache to find the few tokens that matter. If the user asks about a previous detail in a 1M-token conversation, global attention layers must scan the full context to generate each token of the reply. A prominent approach mitigates this cost by pre-selecting relevant tokens via lightweight proxy scores, but this extrinsic scoring still incurs O(N) per step. We take an intrinsic approach motivated by the simple question: wouldn't the model already know which parts of the context are relevant? To this end, we introduce Declarative Attention (DA), a protocol that elicits the model to declare where it needs to attend within its chain-of-thought, partitioning generation into three modes: (full context), (a specific region), and (recent output only). The inference engine parses these declarations like tool calls and skips most of the KV cache read. Under zero-shot evaluation across 15 long-context tasks, DA on off-the-shelf models (Gemma-4-31B, Qwen-3.6-27B) significantly reduces total attended tokens during decoding (52.0%, 31.1%) with modest accuracy drops (1.27pp, 2.75pp) that shrink with model scale. DA unlocks a new axis of sparse attention, with further potential under training-based methods that future work can explore.
Chinese Translation
语言模型在大部分时间内将注意力集中在少量上下文上，但它们仍需读取整个键值缓存（KV cache）以找到少数重要的标记。如果用户在一个包含100万标记的对话中询问之前的细节，全球注意力层必须扫描完整的上下文以生成回复的每个标记。一种显著的方法通过轻量级的代理评分预先选择相关标记，从而减轻这一成本，但这种外部评分在每一步仍然会产生O(N)的开销。我们采取了一种内在的方法，基于一个简单的问题：模型难道不应该已经知道上下文中哪些部分是相关的吗？为此，我们引入了声明性注意力（Declarative Attention, DA），这是一种协议，促使模型声明它在思维链中需要关注的位置，将生成过程分为三种模式：（完整上下文）、（特定区域）和（仅最近输出）。推理引擎像解析工具调用一样解析这些声明，并跳过大部分的KV缓存读取。在15个长上下文任务的零-shot评估中，DA在现成模型（Gemma-4-31B, Qwen-3.6-27B）上显著减少了解码过程中的总注意标记（分别为52.0%和31.1%），而准确率的下降幅度适中（分别为1.27pp和2.75pp），且随着模型规模的增大而减小。DA开启了一种稀疏注意力的新维度，并在基于训练的方法中具有进一步的潜力，未来的工作可以对此进行探索。
cs.CL / 56 / 2609.02754
Untangling the Mechanisms of Misleading Context in Medical Question Answering
解开医疗问答中误导性上下文的机制
Abstract
Large language models now answer medical questions with expert-level performance. However, the context these systems act on can be misleading, and misleading context can corrupt a model's medical judgment. To understand how misleading context corrupts this judgment, we examine the model's susceptibility to the context, disclosure of it, mechanism of corrupted reasoning, and monitorability of the decision. On the medical reasoning subset of MedMisBench, a clinician-reviewed question-answering benchmark of 8,627 questions, we inject two types of misleading context cues, fabricated evidence and a bare assertion. We test three reasoning models, two that expose their full reasoning trace and one frontier model that exposes only its response. All three are more susceptible to the assertion than to the fabricated evidence, adopting the asserted answer 10 to 27 points more often. The misleading cues are disclosed in 81 to 98% of traces but only 7 to 90% of responses, and the assertion is disclosed less often than evidence based cues. Resampling from reasoning traces without disclosure shows the two cues corrupt reasoning differently, evidence entering early and accumulating while the assertion redirects the conclusion near its end. An LLM monitor catches 78% of corrupted decisions at 5% false positives when reading an open model's trace with guidance, against at most 32% from any response. The misleading context that models are most susceptible to is disclosed least, and was caught reliably only from an open reasoning trace, which frontier providers withhold.
Chinese Translation
大型语言模型现在能够以专家级的表现回答医疗问题。然而，这些系统所依据的上下文可能具有误导性，误导性上下文可能会破坏模型的医疗判断。为了理解误导性上下文如何破坏这种判断，我们考察了模型对上下文的敏感性、上下文的披露、推理的腐败机制以及决策的可监控性。在MedMisBench的医疗推理子集上，这是一个包含8627个问题的临床审查问答基准，我们注入了两种类型的误导性上下文线索：伪造证据和简单断言。我们测试了三种推理模型，其中两种展示了其完整的推理轨迹，而一种前沿模型仅展示其响应。所有三种模型对断言的敏感性高于对伪造证据的敏感性，采用断言答案的频率比伪造证据高出10到27个百分点。误导性线索在81%到98%的轨迹中被披露，但仅在7%到90%的响应中被披露，且断言的披露频率低于基于证据的线索。从未披露的推理轨迹中重新采样显示这两种线索对推理的影响不同，证据早期进入并累积，而断言则在推理接近结束时重新引导结论。一种大型语言模型监控器在读取开放模型的轨迹时，能够以5%的假阳性率捕捉到78%的腐败决策，而从任何响应中最多只能捕捉到32%。模型最容易受到的误导性上下文披露最少，且仅通过开放的推理轨迹能够可靠捕捉到这一点，而前沿提供者则对此保持保密。
cs.CL / 57 / 2609.02771
From Reweighting to Rewriting: Unlocking the Intervention Effects of Influential Samples in Training Data Attribution
从重加权到重写：解锁训练数据归因中影响样本的干预效果
Abstract
Training data attribution (TDA) aims to identify training examples that shape model behavior, but its intervention value depends on both which examples are selected and how they are modified. Influence functions (IF) estimate behavioral changes under infinitesimal reweighting, yet IF-selected examples often show limited advantages over random selection under conventional weight-based interventions. This raises the question of whether influential examples lack intervention value or whether reweighting fails to realize their behavioral leverage.We introduce influence-guided response rewriting, which uses IF to identify intervention targets and replaces their responses with behavior-aligned or behavior-opposed supervision while keeping instructions fixed. Across four open-weight LLMs, we compare rewriting and reweighting on the same influence-selected examples using epistemic abstention as our primary testbed. Response rewriting produces stronger, more persistent, and bidirectional behavioral shifts, while reweighting the same examples yields weak and inconsistent effects. Further analyses show that influence-selected examples provide greater rewriting leverage than alternative selectors, with changes remaining concentrated on target-relevant behaviors. The same qualitative contrast extends to safety refusal. These results distinguish the local reweighting effects captured by influence estimates from the broader intervention leverage of the examples they identify, motivating intervention-aware evaluation of TDA methods.
Chinese Translation
训练数据归因（TDA）旨在识别塑造模型行为的训练示例，但其干预价值依赖于所选择的示例及其修改方式。影响函数（IF）在无限小重加权下估计行为变化，然而，在传统的基于权重的干预下，IF选择的示例往往相较于随机选择显示出有限的优势。这引发了一个问题：影响示例是否缺乏干预价值，或者重加权未能实现其行为杠杆。我们提出了影响引导的响应重写，利用IF识别干预目标，并在保持指令不变的情况下，用与行为一致或相对的监督替换其响应。在四个开放权重的语言模型（LLMs）中，我们使用认知回避作为主要测试平台，对相同的影响选择示例进行重写和重加权的比较。响应重写产生了更强、更持久和双向的行为变化，而对相同示例的重加权则产生了微弱且不一致的效果。进一步分析表明，影响选择的示例提供了比其他选择器更大的重写杠杆，变化仍然集中在与目标相关的行为上。这种定性对比同样适用于安全拒绝。这些结果区分了影响估计所捕获的局部重加权效果与它们识别的示例的更广泛干预杠杆，促使对TDA方法进行干预意识评估。
cs.CL / 58 / 2609.02772
HyperStyler: Low-resource Authorship Style Transfer via Context-aware Style Navigation and Hypernetworks
HyperStyler：通过上下文感知风格导航和超网络实现低资源作者风格迁移
Abstract
Low-resource authorship style transfer (LAST) aims to rewrite text into the style of an arbitrary target author using only a few reference examples while preserving the original meaning. Existing methods often struggle to achieve both high style fidelity and semantic preservation because they compress diverse references into a single static author embedding, which averages out context-dependent stylistic variation, and rely on hidden representations for style control, which entangle style with content. We propose HyperStyler, a novel architecture that decouples LAST into style selection and style realization. Stylo-navigator predicts style coordinates by jointly modeling the source context and target-author references, and Stylo-hypernet realizes them via dynamic parameter modulation instead of hidden-state injection. Our experiments on Reddit, Blog, and News datasets demonstrate that HyperStyler consistently outperforms prior methods including LLM-based approaches and generalizes robustly across domains. Notably, HyperStyler achieves superior performance with as few as 2.4% additional parameters over T5-large, while being over 1.8x faster than LLMs at inference.
Chinese Translation
低资源作者风格迁移（LAST）旨在仅使用少量参考示例将文本重写为任意目标作者的风格，同时保持原意。现有方法往往难以实现高风格保真度和语义保留，因为它们将多样的参考压缩为单一静态的作者嵌入，这会平均化依赖上下文的风格变化，并依赖于隐藏表示进行风格控制，这使得风格与内容交织在一起。我们提出了HyperStyler，这是一种新颖的架构，将LAST解耦为风格选择和风格实现。Stylo-navigator通过联合建模源上下文和目标作者参考来预测风格坐标，而Stylo-hypernet则通过动态参数调制实现这些坐标，而不是通过隐藏状态注入。我们在Reddit、Blog和News数据集上的实验表明，HyperStyler始终优于包括基于LLM的方法在内的先前方法，并且在不同领域具有良好的泛化能力。值得注意的是，HyperStyler在T5-large基础上仅增加2.4%的参数就能实现卓越的性能，同时在推理时比LLM快超过1.8倍。
cs.CL / 59 / 2609.02783
EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction
EarlyEval：通过早期结果预测降低代理评估成本
Abstract
Evaluating LLM agents is essential for guiding their development, yet it has grown prohibitively expensive: a single pass of a frontier model over an agentic benchmark can cost hundreds to thousands of dollars, a price paid repeatedly across iterative development cycles. Prior efforts, centered on benchmark distillation, reduce the number of evaluation tasks but leave the cost of executing each retained task untouched. In this work, we introduce early outcome prediction, a complementary axis of efficiency that instead cuts cost within each task. Our key insight is that an agent's final outcome is often evident from its intermediate behavior well before execution completes. We instantiate this idea in EarlyEval, a lightweight framework that trains a pair of LightGBM success and failure classifiers over behavioral, textual, and reference-solution features, and halts an agent run the moment either classifier crosses a calibrated confidence threshold, adding negligible per-step overhead. Across three benchmarks, SWE-bench Verified, TerminalBench, and Toolathlon, EarlyEval can eliminate 13%-26% of agent steps and up to 44.1% input tokens and 29.4% output tokens at 89%-97% prediction accuracy, while perturbing per-agent resolve rates by only one to two percentage points on average.
Chinese Translation
评估大型语言模型（LLM）代理对于指导其发展至关重要，但这一过程的成本已变得过于昂贵：在代理基准上对前沿模型进行一次评估可能需要数百到数千美元，这一费用在迭代开发周期中反复支付。之前的工作主要集中在基准蒸馏上，虽然减少了评估任务的数量，但未能降低执行每个保留任务的成本。在本研究中，我们引入了早期结果预测，这是一种补充的效率轴线，旨在降低每个任务内的成本。我们的关键见解是，代理的最终结果通常可以通过其中间行为在执行完成之前就显现出来。我们在EarlyEval中实现了这一理念，这是一个轻量级框架，训练了一对基于LightGBM的成功和失败分类器，利用行为、文本和参考解决方案特征，并在任一分类器超过校准置信度阈值时立即停止代理运行，增加的每步开销微乎其微。在三个基准测试中，SWE-bench Verified、TerminalBench和Toolathlon，EarlyEval能够消除13%-26%的代理步骤，并在89%-97%的预测准确率下，减少多达44.1%的输入令牌和29.4%的输出令牌，同时对每个代理的解决率平均仅扰动一到两个百分点。
cs.CL / 60 / 2609.02796
DiscoSign: Discourse-Aware Text to Sign Language Gloss Translation
DiscoSign：基于话语意识的文本到手语术语翻译
Abstract
Sign language processing systems have traditionally operated at the sentence level, ignoring critical discourse phenomena fundamental to sign language comprehension. We introduce DiscoSign, a computational approach for discourse-aware text to sign language gloss translation grounded in linguistic research. We address three key phenomena within our modular Large Language Model (LLM)-based translation framework: (i) spatial coreference resolution, where entities maintain consistent spatial locations throughout discourse; (ii) Question-Answer Clauses (QACs), pseudocleft structures serving specific discourse functions; and (iii) concept-gloss consistency, ensuring stable mappings between English concepts and American Sign Language (ASL) signs. Traditional translation metrics fail to capture discourse-level quality, so we introduce a suite of novel evaluation metrics designed to assess each dimension of discourse coherence addressed by our framework. Experiments on sentence-level and discourse-level datasets show that our approach for discourse-aware processing significantly improves spatial consistency and entity tracking relative to sentence-only translation, while maintaining competitive single-sentence gloss translation quality. Our work establishes the first systematic framework for discourse-level text to sign language gloss translation with corresponding evaluation methodology.
Chinese Translation
手语处理系统传统上在句子层面上运作，忽视了对手语理解至关重要的关键话语现象。我们提出了DiscoSign，一种基于语言学研究的、面向话语的文本到手语术语翻译的计算方法。我们在模块化的大型语言模型（LLM）基础的翻译框架中解决了三个关键现象：（i）空间共指解析，其中实体在整个话语中保持一致的空间位置；（ii）问答子句（Question-Answer Clauses, QACs），伪分裂结构用于特定的话语功能；以及（iii）概念-术语一致性，确保英语概念与美国手语（ASL）手势之间的稳定映射。传统的翻译指标无法捕捉话语层面的质量，因此我们引入了一套新的评估指标，旨在评估我们框架所涉及的每个话语连贯性维度。在句子级和话语级数据集上的实验表明，我们的面向话语处理的方法在空间一致性和实体跟踪方面显著优于仅基于句子的翻译，同时保持了具有竞争力的单句术语翻译质量。我们的工作建立了第一个系统化的话语层面文本到手语术语翻译框架及其相应的评估方法。
cs.CL / 61 / 2609.02859
User Feedback Provides a Unique Signal that LLMs Can not Detect
用户反馈提供了一种大型语言模型无法检测的独特信号
Abstract
Harnessing naturally occurring feedback from user interactions offers a promising learning signal for Large Language Models (LLMs). However, recent studies suggest this feedback is inherently noisy and difficult to leverage effectively. We challenge this conception by demonstrating that user feedback is a highly actionable signal for improvement, and that its perceived ineffectiveness stems from a systematic bias in current evaluation paradigms. To isolate the usefulness of feedback, we construct synthetic data with a definitive ground truth, alongside naturalistic data to validate that our findings hold in real-world scenarios. By comparing model revisions generated with and without access to feedback across both settings, we show that feedback-informed revisions resolve targeted issues at significantly higher rates than baseline revisions. Finally, we expose the root of the evaluation bias: when a model successfully fixes an issue exclusively due to feedback, LLM judges frequently fail to identify the genuinely corrected response, systematically preferring inferior baseline outputs instead.
Chinese Translation
利用用户交互中自然发生的反馈为大型语言模型（LLMs）提供了一种有前景的学习信号。然而，最近的研究表明，这种反馈本质上是嘈杂的，且难以有效利用。我们对这一观念提出质疑，证明用户反馈是一个高度可操作的改进信号，其被认为无效的原因源于当前评估范式中的系统性偏差。为了隔离反馈的有效性，我们构建了具有明确真实值的合成数据，并结合自然数据以验证我们的发现是否适用于现实场景。通过比较在两种环境中有无反馈访问的模型修订，我们显示，基于反馈的修订在解决特定问题方面的成功率显著高于基线修订。最后，我们揭示了评估偏差的根源：当模型仅因反馈成功修复问题时，LLM评审者往往无法识别真正被修正的响应，而是系统性地偏好劣质的基线输出。