cs.RO / 1 / 2608.18140
Scheduling and Routing with Degradation-Triggered Job Arrivals: An Application to Forest Firefighting with an Unmanned Aerial Vehicle Fleet
基于退化触发作业到达的调度与路径规划：无人机机队在森林灭火中的应用
Abstract
We define an intertwined scheduling and routing problem where new jobs appear due to the degradation of the existing jobs. Specifically, once a job arrives at a potential job location, a time window begins during which the demand of the job can be fulfilled. The demand degrades within the time window, and once it surpasses a particular threshold, it triggers the arrival of new jobs. Each job location inherently possesses an initial default reward, and the presence of an unprocessed job at a location gradually reduces this default value. The overall objective is to maximize the total remaining reward. The underlying motivation of this problem aligns with the proverb ``a stitch in time saves nine," and the problem itself carries practical implications. We focus on the problem in the context of aerial forest firefighting. Each ignited area has a designated action window; delaying intervention causes the fire to grow, diminishing the area's value and causing it to spread to adjacent areas. We develop a mixed-integer programming model that maximizes value retention in wildfire-threatened regions, and a hybrid model based on dynamic constraint generation to enhance the scalability of the model. We evaluate the performance and practicality of our models through computational experiments and a case study. Additionally, we ensure the study's reproducibility and encourage further research by providing open access to the codebase of our model.
Chinese Translation
我们定义了一个交织的调度与路径规划问题，其中由于现有作业的退化而出现新的作业。具体而言，一旦作业到达潜在作业地点，便会开始一个时间窗口，在此期间可以满足该作业的需求。需求在时间窗口内逐渐退化，一旦超过特定阈值，就会触发新作业的到达。每个作业地点本质上具有一个初始默认奖励，而该地点上未处理作业的存在会逐渐降低这一默认值。总体目标是最大化剩余奖励的总和。该问题的根本动机与“及时行事，省却九次”的谚语相符，并且该问题本身具有实际意义。我们将重点放在空中森林灭火的背景下。每个着火区域都有一个指定的行动窗口；延迟干预会导致火势蔓延，降低该区域的价值，并使其扩散到邻近区域。我们开发了一个混合整数规划模型，以最大化在野火威胁区域的价值保留，并基于动态约束生成的混合模型来增强模型的可扩展性。通过计算实验和案例研究，我们评估了模型的性能和实用性。此外，我们确保研究的可重复性，并通过提供模型代码库的开放访问来鼓励进一步研究。
cs.RO / 2 / 2608.18178
Trust as a Field: A Macroscopic Representation for Vehicular Networks
信任作为一个领域：车辆网络的宏观表示
Abstract
Trust assessment is a fundamental component of cooperative and connected vehicle systems. However, existing approaches operate primarily at the level of individual vehicles, making it difficult to reason about trust evolution across road segments. In this paper, we propose a spatio-temporal trust-field framework that aggregates microscopic vehicle-level trust into a continuous representation over space and time. The trust field is formally defined on road segments. We conducted simulation-based experiments using synthetic trajectories generated under controlled conditions, enabling analysis of trust-field behavior in simple road scenarios. Beyond theoretical modeling, we study an implication of the trust-field concept: reconstructing the full trust field from sparse roadside-unit (RSU) measurements. We compare (i) a coordinate-based deep learning baseline that learns a generic trust field from sparse samples and (ii) a field-informed deep learning method that treats trust as a latent quantity carried by vehicles and enforces measurement consistency through the aggregation mechanism. The field-informed approach more accurately recovers trajectory-aligned low-trust patterns and yields improved reconstruction error.
Chinese Translation
信任评估是合作与连接车辆系统的基本组成部分。然而，现有的方法主要在个别车辆层面上运作，这使得在道路段上推理信任演变变得困难。本文提出了一种时空信任场框架，该框架将微观车辆级别的信任聚合为一个在空间和时间上连续的表示。信任场在道路段上被正式定义。我们进行了基于仿真的实验，使用在受控条件下生成的合成轨迹，从而能够分析简单道路场景中的信任场行为。除了理论建模外，我们还研究了信任场概念的一个应用：从稀疏的路边单元（RSU）测量中重建完整的信任场。我们比较了（i）一个基于坐标的深度学习基线，该基线从稀疏样本中学习通用信任场，以及（ii）一个场信息深度学习方法，该方法将信任视为由车辆携带的潜在量，并通过聚合机制强制测量一致性。场信息方法更准确地恢复了与轨迹对齐的低信任模式，并且降低了重建误差。
cs.RO / 3 / 2608.18227
Revisiting the "Push-T" Robot Manipulation Task with Agentic Robotics
重新审视“Push-T”机器人操作任务与自主机器人技术
Abstract
Push-T is an iconic benchmark for learning manipulation policies from human demonstrations. The robot must use a single point of contact to push a T-shaped block into a target pose. In this short paper, we revisit the Push-T task in the context of emerging advances in Agentic Robotics where an LLM coding agent -- Claude Code with Fable 5 -- is prompted to create an algorithmic solution that does not require any demonstration data. We study how effective the agentic coding loop can solve the Push-T task, and compare the resulting code as policy with the visuomotor imitation learning policy. Results suggest that the agent found the 2D gym simulation online, and used sim experiments to learn push mechanics, iteratively optimizing to achieve 100% success rate using 46% fewer steps than the best diffusion policy trained with 200 human demonstrations. The coding agent also solve extensions from T to the full alphabet (Push-A to Push-Z) using a self generated curriculum and generated simulation code for the Franka and UR5 robot arms in 3D cross-embodiment simulations with visual feedback. Videos, policies and details will be posted online.
Chinese Translation
Push-T是一个标志性的基准，用于从人类示范中学习操作策略。机器人必须利用单一点接触将一个T形块推入目标姿态。在本文中，我们在自主机器人技术的最新进展背景下重新审视Push-T任务，其中一个大型语言模型（LLM）编码代理——Claude Code with Fable 5——被提示创建一个不需要任何示范数据的算法解决方案。我们研究了自主编码循环在解决Push-T任务中的有效性，并将生成的代码作为策略与视觉运动模仿学习策略进行比较。结果表明，该代理在线找到2D健身房模拟，并利用模拟实验学习推力机制，经过迭代优化以实现100%的成功率，所需步骤比使用200个人类示范训练的最佳扩散策略少46%。该编码代理还通过自生成的课程解决了从T到完整字母表（Push-A到Push-Z）的扩展，并为Franka和UR5机器人臂生成了3D跨体现模拟的模拟代码，提供了视觉反馈。视频、策略和详细信息将在线发布。
cs.RO / 4 / 2608.18234
GigaBrain-WBC-0.5: A Behavior World Model for Robust Whole-Body Control with Environment Interaction
GigaBrain-WBC-0.5：一种用于环境交互的鲁棒全身控制行为世界模型
Abstract
Whole-body motion tracking policies turn a humanoid into a robust control interface: the teleoperator---or an upstream model---only supplies a coarse movement intent, while the low-level policy keeps the robot balanced and physically feasible. Existing trackers deliver this interface only on flat ground: trained in empty scenes, they never learn how contact with terrain and objects reshapes their dynamics, and they attempt to teach the policy to balance under any command by continually enlarging the reference-motion corpus, which stops working once feasible behaviors become environment-dependent. We present GigaBrain-WBC-0.5, the first Behavior World Model (BWM) for humanoid whole-body control. Rather than a purely reactive tracker, we train a causal Transformer to jointly predict its next action, next state, and the distribution over its next latent behavior command, so the network that acts also models how the environment shapes what it can do next. An automatic terrain-annotation pipeline recovers full 3D contact geometry from retargeted motion, enabling terrain annotation at the scale of existing motion datasets. The predicted distribution is reused at deployment to detect implausible commands online and retract them onto learned behaviors, so the robot attempts tasks in a "best-effort" manner. The result is a unified policy that takes real-time command, interacts with environment, and stays robust to implausible commands, falls, and disturbances. GigaBrain-WBC-0.5 achieves the highest success rate across all four regimes among three large-scale tracker baselines: 81.3% on terrain interaction (4.3x the strongest baseline), 83.1% under implausible commands, and 99.3% fall recovery (16.8x the strongest baseline). Hardware trials show robust interaction under missing supports and disturbances; the Unitree G1 checkpoint transfers to the Maker L01 robot with simple fine-tuning.
Chinese Translation
全身运动跟踪策略将类人机器人转变为一个鲁棒的控制接口：遥控操作员或上游模型仅提供粗略的运动意图，而低级策略则保持机器人平衡和物理可行。现有的跟踪器仅在平坦地面上提供这一接口：在空旷场景中训练，它们从未学习到与地形和物体接触如何重塑其动态，并试图通过不断扩大参考运动语料库来教导策略在任何指令下保持平衡，但一旦可行行为变得依赖环境，这种方法便失效。我们提出了GigaBrain-WBC-0.5，这是第一个用于类人全身控制的行为世界模型（Behavior World Model, BWM）。我们训练一个因果Transformer，而不是一个纯粹的反应式跟踪器，以共同预测其下一个动作、下一个状态以及其下一个潜在行为指令的分布，因此执行动作的网络也建模了环境如何塑造它接下来能做的事情。一个自动地形标注管道从重新定向的运动中恢复完整的3D接触几何，能够在现有运动数据集的规模上进行地形标注。预测的分布在部署时被重用，以在线检测不合理的指令并将其收回到学习的行为上，因此机器人以“尽力而为”的方式尝试任务。最终结果是一个统一的策略，能够实时接收指令，与环境互动，并对不合理的指令、跌倒和干扰保持鲁棒性。GigaBrain-WBC-0.5在三个大规模跟踪器基线中，在所有四个模式下实现了最高的成功率：在地形交互中为81.3%（是最强基线的4.3倍），在不合理指令下为83.1%，以及99.3%的跌倒恢复（是最强基线的16.8倍）。硬件试验表明在缺失支撑和干扰下的鲁棒交互；Unitree G1检查点经过简单微调后可转移至Maker L01机器人。
cs.RO / 5 / 2608.18254
GAPL: Grounded Action-effect Policy Learning for LLM-Based Trajectory Planning
GAPL：基于LLM的轨迹规划中的基础行动效果策略学习
Abstract
Trajectory planning for autonomous driving requires both high-level reasoning and precise low-level control. Large Language Models (LLMs) offer semantic-rich planning capabilities, however, their application is limited by hallucinated reasoning, poor grounding in environment dynamics, and limited numerical precision in control. We propose GAPL (Grounded Action-effect Policy Learning), a unified framework that integrates LLM-based effect estimation, simulation-based effect grounding, and policy optimization into a closed-loop system. GAPL consists of three modules: (1) an LLM-based Effect Evaluator for structured multi-dimensional action-effect estimation; (2) a Simulation-based Effect Grounder that predicts dynamics-consistent effects from simulator rollouts; and (3) an Effect-Aware Decision Maker that grounds LLM effect estimates against simulation via a distiller to guide Proximal Policy Optimization (PPO)-based policy learning. Experiments on four Highway-env scenarios demonstrate that GAPL consistently outperforms baselines, achieving average reductions of {0.76, 0.86, 2.00} in collision rate, average displacement error (ADE), and final displacement error (FDE), and an average reward gain of 1.44.
Chinese Translation
自主驾驶的轨迹规划需要高层次的推理和精确的低层次控制。大型语言模型（LLMs）提供了丰富的语义规划能力，然而，它们的应用受到虚幻推理、对环境动态的较差基础和控制中的有限数值精度的限制。我们提出了GAPL（基础行动效果策略学习），这是一个统一框架，将基于LLM的效果估计、基于仿真的效果基础和策略优化整合为一个闭环系统。GAPL由三个模块组成：（1）基于LLM的效果评估器，用于结构化的多维行动效果估计；（2）基于仿真的效果基础器，从仿真器的滚动预测动态一致的效果；（3）效果感知决策者，通过蒸馏器将LLM效果估计与仿真相结合，以指导基于近端策略优化（PPO）的策略学习。在四个高速公路环境场景中的实验表明，GAPL始终优于基线，分别在碰撞率、平均位移误差（ADE）和最终位移误差（FDE）上实现了平均减少{0.76, 0.86, 2.00}，并获得了1.44的平均奖励增益。
cs.RO / 6 / 2608.18258
VERAGMIL: Virtual Environment for Scooping Granular Foods with Imitation Learning Models
VERAGMIL：用于模仿学习模型的颗粒食品舀取虚拟环境
Abstract
Robot-Assisted Feeding (RAF) systems are essential for assisting individuals with disabilities or motor impairments in eating tasks. Manipulating granular food items, such as rice and beans, poses significant challenges due to their dynamic physical properties. Learning from human demonstrations offers a promising solution, but acquiring high-quality demonstrations is complex. To address this, we present VERAGMIL, a framework that combines a high-fidelity simulator with an intuitive Virtual Reality (VR) interface for recording demonstrations and supporting different imitation learning methods. VERAGMIL provides a realistic environment for training RAF systems to handle granular materials, including robots, sensors, and various food items with distinct physical characteristics. We evaluate VERAGMIL by training three imitation learning models, BC, BC-RNN, and BCQ, on granular scooping and transporting tasks using both VR interface and 3D space mouse demonstrations, comparing them with a human-expert baseline. The models are assessed on success rate, spillage, generalization to unseen food items, and task completion time. Results show that VR-based demonstrations significantly outperform 3D space mouse data, with BCQ achieving the best overall performance, particularly in reducing spillage and approaching human performance. These findings underscore the effectiveness of our framework for training RAF systems in granular material handling. The code for our framework is publicly available at: https://github.com/AmanuelErgogo/VERAGMIL.git.
Chinese Translation
机器人辅助喂养（RAF）系统对于帮助残疾人士或运动障碍者完成进食任务至关重要。然而，操控颗粒状食品（如米饭和豆类）由于其动态物理特性，面临重大挑战。从人类示范中学习提供了一种有前景的解决方案，但获取高质量的示范却相当复杂。为了解决这一问题，我们提出了VERAGMIL，一个将高保真模拟器与直观的虚拟现实（VR）界面相结合的框架，用于记录示范并支持不同的模仿学习方法。VERAGMIL为训练RAF系统处理颗粒材料提供了一个真实的环境，包括机器人、传感器以及具有不同物理特性的各种食品。我们通过在颗粒舀取和运输任务上训练三种模仿学习模型（BC、BC-RNN和BCQ），使用VR界面和3D空间鼠标示范进行评估，并与人类专家基线进行比较。模型的评估指标包括成功率、溢出、对未见食品的泛化能力以及任务完成时间。结果表明，基于VR的示范显著优于3D空间鼠标数据，其中BCQ在减少溢出和接近人类表现方面取得了最佳整体表现。这些发现强调了我们框架在颗粒材料处理方面训练RAF系统的有效性。我们的框架代码已公开发布，地址为：https://github.com/AmanuelErgogo/VERAGMIL.git。
cs.RO / 7 / 2608.18270
Transferable Tool-Tissue Contact Detection from Stereo Depth in Robot-Assisted Surgery
可转移的工具-组织接触检测：基于立体深度在机器人辅助手术中的应用
Abstract
Reliable tool--tissue contact detection can support interaction-aware control and downstream force estimation in robot-assisted surgery. Most existing methods learn a contact classifier from RGB appearance, which is hard to generalize. In this work, we use the depth image generated from a stereo pair to give more information about tool--tissue contact. For each depth frame, we localize a spatially supported minimum-distance patch around the tool boundary and reduce it to a single scalar, $-\log_{10}|d|$; this signal rises and falls in step with ground-truth contact. We formalize this observation with a fully supervised two-state hidden Markov model. We fit this model as a six-fold leave-one-session-out (LOSO) ensemble on six palpation sessions against a single silicone cup-like phantom, with the decision threshold selected from the pooled out-of-fold predictions. It is evaluated on four held-out sessions of three categories: 1. same task on same phantom; 2. same task on different phantom; 3. different task on different phantom. This model reaches held-out macro F1 $0.927$ and AUPRC $0.980$. We further compare against a reproduction of an RGB-based contact classifier from prior work. This RGB-based model achieves high performance on the first category (F1 $0.965$), but substantially lower performance on the other two, resulting in macro F1 $0.320$ across all four sessions. These results indicate that the tool--tissue distance is a strong, transferable cue for contact detection in robot-assisted surgery.
Chinese Translation
可靠的工具-组织接触检测可以支持机器人辅助手术中的交互感知控制和下游力估计。现有的大多数方法从RGB外观中学习接触分类器，这在泛化上存在困难。在本研究中，我们利用立体对生成的深度图像提供关于工具-组织接触的更多信息。对于每个深度帧，我们在工具边界周围定位一个空间支持的最小距离补丁，并将其简化为一个标量 $- ext{log}_{10}|d|$；该信号随着真实接触的变化而上升和下降。我们用一个完全监督的双状态隐马尔可夫模型形式化这一观察。我们将该模型作为六折留一会话外（LOSO）集成拟合于六个触诊会话，针对一个单一的硅胶杯状幻影，决策阈值从汇总的会话外预测中选择。该模型在三个类别的四个保留会话中进行评估：1. 在同一幻影上执行相同任务；2. 在不同幻影上执行相同任务；3. 在不同幻影上执行不同任务。该模型在保留会话中达到了宏观 F1 值 $0.927$ 和 AUPRC 值 $0.980$。我们进一步与先前工作的RGB基础接触分类器的重现结果进行比较。该RGB基础模型在第一类上表现优异（F1 值 $0.965$），但在其他两个类别上的表现明显较低，导致四个会话的宏观 F1 值为 $0.320$。这些结果表明，工具-组织距离是机器人辅助手术中接触检测的一个强大且可转移的线索。
cs.RO / 8 / 2608.18292
GuideFetch: A Task Coordination Framework for Concurrent Navigation and Object Retrieval in Assistive Robot Dogs
GuideFetch：用于辅助机器人犬的并发导航与物体检索的任务协调框架
Abstract
Consider a robot guide dog escorting a blind user to an available seat while a second assistive robot dog concurrently retrieves a cup of coffee and delivers it to the same seat. This setting motivates concurrent execution because navigation and object retrieval can overlap. A syntactically valid Large Language Model (LLM) plan may still violate embodiment constraints, and successful-looking controller motion does not by itself establish task completion. We introduce \textsc{GuideFetch}, a coordination framework for concurrent navigation and object retrieval by a heterogeneous guider and fetcher team. An LLM instantiates a schedule-conditioned four-action schema from a natural-language instruction. Before execution, robot, skill, and target aliases are normalized, and proposed actions are validated against registered targets, robot capabilities, and the selected schedule. Robot and object states then govern sequential and parallel execution. In a matched $2\times2$ study across 90 combinations of scene and seed (360 executions), all 180 online LLM responses validate without fallback or replay and match the corresponding scripted plans. For each planner source, sequential and parallel execution achieve $72/90$ and $71/90$ operational successes, respectively. Among the 56 cases completed by both schedules, parallel execution reduces mean makespan by 41.3\%. Within this controlled setting, role specialization and action overlap shorten completed missions, while state checks distinguish plan validity from verified mission completion. Source code will be available.
Chinese Translation
设想一只机器人导盲犬在为盲人用户引导到一个空座位的同时，另一只辅助机器人犬正在并行地取回一杯咖啡并将其送到同一座位。这种场景激励了并发执行，因为导航和物体检索可以重叠。一个语法上有效的大型语言模型（LLM）计划可能仍然违反具体实施约束，而看似成功的控制器运动本身并不能证明任务的完成。我们引入了 extsc{GuideFetch}，这是一个用于异构引导者和取物者团队的并发导航与物体检索的协调框架。LLM 从自然语言指令中实例化一个调度条件的四动作模式。在执行之前，机器人、技能和目标别名被标准化，建议的动作经过对已注册目标、机器人能力和所选调度的验证。然后，机器人和物体的状态决定顺序和并行执行。在一个匹配的 $2 imes2$ 研究中，涵盖了90种场景和种子组合（360次执行），所有180个在线LLM响应都在没有回退或重放的情况下得到验证，并与相应的脚本计划相匹配。对于每个规划源，顺序和并行执行分别实现了 $72/90$ 和 $71/90$ 的操作成功率。在56个由两个调度完成的案例中，并行执行将平均完成时间缩短了41.3。在这个受控环境中，角色专业化和动作重叠缩短了完成任务的时间，而状态检查则将计划的有效性与验证的任务完成区分开来。源代码将会公开。
cs.RO / 9 / 2608.18364
A Task-Agnostic Control Strategy for Dynamic Assistance with Pneumatically Actuated Soft Exosuits
一种任务无关的控制策略用于气动驱动软外骨骼的动态辅助
Abstract
Pneumatic artificial muscles have provided new opportunities to develop upper-extremity soft exosuits for reha- bilitation, augmentation, and assisted daily living. However, the complex dynamics and limited bandwidth of these actuators has made providing responsive assistance based on user intention a longstanding challenge. In this work, we present an inverse-plant control strategy for pneumatically actuated soft exosuits that only relies on kinematic sensing for task-agnostic and dynamic assistance during daily living. We model the human-robot system using a Hammerstein dynamic model, consisting of a Preisach hysteresis model and a linear time-invariant filter, to capture the static and dynamic behavior of the system. We personalize our model to each user using 140 s of data and approximate an inverse to integrate into our control loop. When evaluated on a test rig that emulated a soft assistive exosuit for the wrist, our controller reduced the interaction torque by up to 73% and the activation of key flexor and extensor muscles by up to 47% relative to the condition with no assistance for speeds ranging from 8{\deg}/s to 120{\deg}/s. Overall, this work presents a control strategy that can provide task-agnostic, dynamic assistance with pneumatically actuated soft exosuits without the need for physiological or force sensors to interpret user intention.
Chinese Translation
气动人工肌肉为开发用于康复、增强和日常生活辅助的上肢软外骨骼提供了新的机会。然而，这些驱动器的复杂动态和有限带宽使得根据用户意图提供响应性辅助成为一个长期挑战。在本研究中，我们提出了一种仅依赖运动学传感器的气动驱动软外骨骼的逆植物控制策略，以实现日常生活中的任务无关和动态辅助。我们使用Hammerstein动态模型对人机系统进行建模，该模型由Preisach滞后模型和线性时不变滤波器组成，以捕捉系统的静态和动态行为。我们使用140秒的数据为每个用户个性化我们的模型，并近似计算逆模型以集成到我们的控制回路中。在一个模拟手腕软辅助外骨骼的测试装置上进行评估时，我们的控制器将交互扭矩降低了高达73%，关键屈肌和伸肌的激活降低了高达47%，相对于无辅助条件，适用的速度范围为8°/s到120°/s。总体而言，本研究提出了一种控制策略，可以在不需要生理或力传感器来解读用户意图的情况下，为气动驱动软外骨骼提供任务无关的动态辅助。
cs.RO / 10 / 2608.18433
The Embodiment Gap in Robot Foundation Models
机器人基础模型中的体现差距
Abstract
Robot foundation models (RFMs), including vision-language-action (VLA) policies, are often discussed through a scaling view: more data, larger models, and broader benchmarks should improve generalization. In robotics, however, a model can generalize while work still remains before it can run on a robot with a particular body. The work required differs across methods and target robots, and those differences affect practical deployment. We call the gap between reusable models, representations, or data and their use in execution on the target robot the embodiment gap. This survey examines what can be reused across robot embodiments and what must still be implemented on a new robot. We place existing methods on a two-axis map that shows the type of shared structure and the stage at which adaptation is needed for execution on the target robot. We then examine recent work through three overlapping research directions: sharing semantics and perception, sharing robot data and interfaces, and learning correspondence across embodiments. We also propose a reporting framework for adaptation work that success rate alone does not reveal. The framework identifies the work that should be checked when comparing cross-embodiment learning and highlights work that remains on a new robot and questions for future study.
Chinese Translation
机器人基础模型（Robot Foundation Models, RFMs），包括视觉-语言-动作（Vision-Language-Action, VLA）策略，通常通过扩展视角进行讨论：更多的数据、更大的模型和更广泛的基准应该改善泛化能力。然而，在机器人领域，一个模型可以实现泛化，但在其能够在特定机体的机器人上运行之前，仍需进行一些工作。所需的工作因方法和目标机器人而异，这些差异影响实际部署。我们将可重用模型、表示或数据与其在目标机器人上执行之间的差距称为体现差距（embodiment gap）。本调查研究了在机器人体现之间可以重用的内容以及在新机器人上仍需实施的内容。我们将现有方法置于一个双轴图中，展示了共享结构的类型以及在目标机器人上执行时需要适应的阶段。接着，我们通过三个重叠的研究方向来审视近期的工作：共享语义和感知、共享机器人数据和接口，以及跨体现学习对应关系。我们还提出了一个适应性工作的报告框架，仅靠成功率无法揭示该框架。该框架识别在比较跨体现学习时应检查的工作，并强调在新机器人上仍需完成的工作以及未来研究的问题。
cs.RO / 11 / 2608.18446
HarvestPoint-ACT: Explicit Target Selection and Harvest-Point Conditioning for Robotic Fruit Harvesting under Occlusion
HarvestPoint-ACT：在遮挡条件下进行机器人水果采摘的显式目标选择与采摘点条件化
Abstract
End-to-end imitation learning avoids hand-made robot motion for approaching and grasping, but the policy must still decide which fruit to pick and where to close the gripper. Occlusion can make the policy lose the selected fruit during harvesting, and the correct closing point is difficult to infer from pixels alone. This paper presents HarvestPoint-ACT, which makes both decisions explicit in perception and provides them to the policy. An instance segmentation front end with a keypoint branch predicts a mask and a harvest point for each visible fruit, where the harvest point specifies the location to close the gripper. A scheduler ranks detected candidates by occlusion and travel distance and selects one target. After each attempt, it redetects and reranks the candidates because the canopy may have changed. The selected fruit is encoded for an action chunking transformer as an eight-dimensional state, containing the absolute harvest point, the vector from the gripper to that point, a validity flag, and a confidence score. When the selected fruit is temporarily undetected, the system retains the last harvest point estimate in the robot base frame and marks it as stale, and aborts the attempt if the loss persists. On a canopy mock-up, HarvestPoint-ACT achieves a success rate of 88%, and of 75% under heavy occlusion.
Chinese Translation
端到端的模仿学习避免了手工制作的机器人运动来接近和抓取，但策略仍然必须决定选择哪个水果以及在哪里闭合夹爪。遮挡可能导致策略在采摘过程中失去选定的水果，而仅凭像素很难推断出正确的闭合点。本文提出了HarvestPoint-ACT，它在感知中明确这两个决策并将其提供给策略。一个实例分割前端与关键点分支预测每个可见水果的掩膜和采摘点，其中采摘点指定了闭合夹爪的位置。调度器根据遮挡和行驶距离对检测到的候选者进行排名并选择一个目标。在每次尝试后，它重新检测并重新排名候选者，因为树冠可能发生变化。选定的水果被编码为一个八维状态，以供动作分块变换器使用，包含绝对采摘点、从夹爪到该点的向量、有效性标志和置信度分数。当选定的水果暂时未被检测到时，系统在机器人基座坐标系中保留最后的采摘点估计并将其标记为过时，如果损失持续，则中止尝试。在一个树冠模型上，HarvestPoint-ACT实现了88%的成功率，在严重遮挡情况下为75%。
cs.RO / 12 / 2608.18454
Backward Layout Search for Sequence-Constrained Robotic Assembly
基于逆向布局搜索的序列约束机器人装配
Abstract
Robotic assembly layout planning must determine the assembly site and the initial pose of each part while ensuring collision-free execution of a prescribed assembly sequence. This problem is challenging because the obstacle environment changes after each assembly step, and unassembled parts re maining in the workspace may block robot motions. We observe that the feasibility of each assembly step depends only on the initial poses of the current and later-assembled parts. Based on this dependency, we propose Backward Layout Search (BLS), which assigns initial part poses in reverse assembly order. Each expansion performs geometric, kinematic, grasp, and prescribed-motion checks, while collision masks and candidate set filtering remove infeasible initial part pose candidates. Promising partial layouts are retained through beam selection, and complete layouts are validated by full motion planning in forward assembly order. Experiments on five assembly models show that BLS produces collision-free executable layouts and reduces step evaluations and search time compared with a matched forward search.
Chinese Translation
机器人装配布局规划必须确定装配地点和每个部件的初始姿态，同时确保按照规定的装配顺序无碰撞地执行。该问题具有挑战性，因为在每个装配步骤后，障碍环境会发生变化，工作空间中未装配的部件可能会阻碍机器人的运动。我们观察到，每个装配步骤的可行性仅依赖于当前和后续装配部件的初始姿态。基于这一依赖关系，我们提出了逆向布局搜索（Backward Layout Search, BLS），该方法按逆装配顺序分配初始部件姿态。每次扩展都执行几何、运动学、抓取和规定运动检查，同时碰撞掩码和候选集过滤去除不可行的初始部件姿态候选。通过束选择保留有前景的部分布局，并通过正向装配顺序的完整运动规划验证完整布局。在五个装配模型上的实验表明，BLS生成了无碰撞的可执行布局，并且与匹配的正向搜索相比，减少了步骤评估和搜索时间。
cs.RO / 13 / 2608.18470
DevGRU: Depth-guided Visual Navigation using a Collision-aware Recurrent Model
DevGRU：基于深度引导的视觉导航使用碰撞感知递归模型
Abstract
Existing visual navigation models often aim to develop foundation models that can generalize robot navigation across diverse platforms. However, many of these models are prone to collisions when deployed in complex indoor environments, particularly in structured layouts and narrow passages. To address this problem, we propose a depth image- and point-goal-conditioned navigation system, DevGRU. The proposed system employs an action predictor (AP) that generates collision-aware future trajectories, enabling effective avoidance of immediate obstacles. In conjunction with a collision predictor, the AP further compensates for errors accumulated in the goal pose estimation and proactively mitigates future deviations. To evaluate our method, we conducted experiments across nine different scenes and three state-of-the-art approaches - ViNT, NoMaD, and NavDP - as well as four additional variants of ViNT and NoMaD. In terms of navigation performance, DevGRU significantly outperforms ViNT and NoMaD by a large margin. In addition, the proposed model has a relatively small number of trainable parameters, resulting in the fastest inference time among the baselines, particularly outperforming NavDP by 7x in model size and 17x in inference time.
Chinese Translation
现有的视觉导航模型通常旨在开发能够在多种平台上推广机器人导航的基础模型。然而，这些模型在复杂的室内环境中部署时，尤其是在结构化布局和狭窄通道中，容易发生碰撞。为了解决这一问题，我们提出了一种基于深度图像和点目标条件的导航系统DevGRU。该系统采用了一个动作预测器（Action Predictor, AP），生成碰撞感知的未来轨迹，从而有效避免即时障碍物。结合碰撞预测器，AP进一步补偿在目标姿态估计中积累的误差，并主动减轻未来的偏差。为了评估我们的方法，我们在九个不同场景中进行了实验，并与三种最先进的方法——ViNT、NoMaD和NavDP——以及ViNT和NoMaD的四个额外变体进行了比较。在导航性能方面，DevGRU显著优于ViNT和NoMaD，差距较大。此外，所提出的模型具有相对较少的可训练参数，在基线中实现了最快的推理时间，特别是在模型大小上比NavDP快7倍，在推理时间上快17倍。
cs.RO / 14 / 2608.18507
An Experimental Study of Downwash Effects on a Continuum Manipulator Integrated with a Multirotor UAV
多旋翼无人机集成的连续操纵器下洗效应的实验研究
Abstract
Continuum arm aerial manipulation systems leverage soft-manipulator compliance and dexterity for tasks in confined or hazardous environments, but propeller downwash can degrade performance, particularly near walls and the ground. This effect remains uncharacterized for continuum manipulators. This letter experimentally studies downwash-induced kinematic deviations of a tendon-driven continuum manipulator integrated with a multirotor platform. Under still-air conditions, the CM is compared with a constant-curvature (CC) model. Downwash- induced end-effector pose deviations are then quantified relative to the mean still-air experimental baseline at four propeller throttle levels in free space, and at maximum throttle near a wall, and near the ground. Vertical position and yaw show the largest deviations and are amplified by ground effect. A CC-guided Gaussian process regression (GPR) residual model is learned from experimental data that improves forward pose prediction RMSE (position by 89-95%, orientation by 47-79%), and support compensation-oriented, downwash-aware modeling of continuum arm aerial manipulation systems.
Chinese Translation
连续臂空中操作系统利用软操纵器的柔性和灵活性在狭小或危险环境中执行任务，但螺旋桨的下洗效应可能会降低性能，特别是在靠近墙壁和地面时。这种效应在连续操纵器中尚未被表征。本文通过实验研究了与多旋翼平台集成的由腱驱动的连续操纵器在下洗作用下的运动学偏差。在静止空气条件下，将连续操纵器（CM）与恒曲率（CC）模型进行比较。然后，相对于四个螺旋桨油门水平下的静止空气实验基线，量化了下洗引起的末端执行器姿态偏差，包括在自由空间和靠近墙壁及地面的最大油门下。垂直位置和偏航角显示出最大的偏差，并受到地面效应的放大。通过实验数据学习的CC引导高斯过程回归（GPR）残差模型改善了前向姿态预测的均方根误差（RMSE）（位置改善89-95%，方向改善47-79%），并支持针对下洗效应的补偿建模，提升连续臂空中操作系统的性能。
cs.RO / 15 / 2608.18552
Real-Time Control-Constrained DDP for Underactuated Balancing of Legged Robots
实时控制约束下的差分动态规划（DDP）用于欠驱动腿式机器人的平衡
Abstract
This paper presents a real-time control-constrained Differential Dynamic Programming (DDP) framework for underactuated legged robots. To address the limitation of classical DDP in handling control constraints, we propose an Accelerated Projected Gradient (APG)-based control-constrained DDP (ABC-DDP), which efficiently computes constrained solutions and identifies active sets without repeated Karush-Kuhn-Tucker (KKT) inversions. A virtual constraint is introduced to integrate control constraints within a feasibility-driven multiple-shooting framework, enabling stable optimization even from dynamically infeasible initializations. The proposed method supports real-time model predictive control (MPC) with short horizons under strong underactuation. Simulation results demonstrate static two-leg standing under external disturbances, along with diverse dynamic motions including slow catwalk, upright walking, and high-speed running within a unified MPC framework. To the best of our knowledge, this is the first demonstration of static two-leg standing of a quadruped robot achieved using real-time finite-horizon MPC.
Chinese Translation
本文提出了一种实时控制约束的差分动态规划（DDP）框架，旨在解决欠驱动腿式机器人控制约束的问题。为了解决经典DDP在处理控制约束时的局限性，我们提出了一种基于加速投影梯度（APG）的控制约束DDP（ABC-DDP），该方法能够高效计算约束解并识别活动集，而无需重复进行Karush-Kuhn-Tucker（KKT）逆运算。引入了一种虚拟约束，以在可行性驱动的多重发射框架内整合控制约束，从而即使在动态不可行的初始化条件下也能实现稳定优化。所提出的方法支持在强欠驱动条件下的短时间范围内的实时模型预测控制（MPC）。仿真结果展示了在外部干扰下的静态双腿站立，以及包括缓慢走秀、直立行走和高速奔跑等多种动态动作，均在统一的MPC框架内实现。据我们所知，这是首次通过实时有限时间范围的MPC实现四足机器人静态双腿站立的演示。
cs.RO / 16 / 2608.18618
LabDex: A Hierarchical Benchmark for Dexterous Manipulation in Laboratories
LabDex：实验室灵巧操作的分层基准
Abstract
Autonomous laboratories hold great promise for accelerating scientific discovery. To achieve this vision, robots are supposed to dexterously manipulate diverse labware and instruments and execute long-horizon, state-dependent experimental procedures. Yet existing benchmarks do not jointly capture dexterous hand use, real-world laboratory interactions, and multi-stage experimental procedures, limiting systematic training and evaluation. To bridge this gap, we introduce LabDex, a large-scale real-world dataset and benchmark for dexterous manipulation in chemistry laboratories, organized around a hierarchical task taxonomy spanning atomic skills, compositional tasks, and long-horizon experiments. First, LabDex is cross-platform and, for the first time, unifies real-world and simulation platforms under a common framework, providing standardized task definitions, demonstrations, and evaluation protocols. Second, LabDex is large-scale and systematically organizes chemistry laboratory operations into three interconnected levels: Atomic Skills, which characterize fundamental dexterous manipulation capabilities; Compositional Skills; and Long-Horizon Laboratory Workflows. This hierarchical design not only supports the evaluation of end-task performance, but also enables the analysis of how fundamental dexterous skills compose and influence more complex laboratory operations. We conduct cross-level evaluations of representative robot learning methods in both real-world and simulation environments. The experimental results validate the effectiveness of the LabDex task design and demonstration data, and show that the benchmark supports the training and systematic evaluation of existing robotic policies across laboratory dexterous manipulation tasks at different levels, providing a foundation for further research and development of autonomous laboratory robots.
Chinese Translation
自主实验室在加速科学发现方面具有巨大的潜力。为了实现这一愿景，机器人需要灵巧地操作各种实验器具和仪器，并执行长期、依赖状态的实验程序。然而，现有的基准测试并未共同捕捉灵巧手部使用、现实世界的实验室交互以及多阶段实验程序，从而限制了系统的训练和评估。为了解决这一问题，我们引入了LabDex，这是一个大规模的现实世界数据集和化学实验室灵巧操作的基准，围绕着一个分层任务分类法组织，涵盖原子技能、组合任务和长期实验。首先，LabDex是跨平台的，并首次在一个共同框架下统一了现实世界和仿真平台，提供标准化的任务定义、演示和评估协议。其次，LabDex是大规模的，并系统地将化学实验室操作组织为三个相互关联的层次：原子技能，表征基本的灵巧操作能力；组合技能；以及长期实验室工作流程。这种分层设计不仅支持最终任务性能的评估，还使得分析基本灵巧技能如何组合并影响更复杂的实验室操作成为可能。我们在现实世界和仿真环境中对代表性的机器人学习方法进行了跨层次评估。实验结果验证了LabDex任务设计和演示数据的有效性，并表明该基准支持对不同层次的实验室灵巧操作任务中现有机器人策略的训练和系统评估，为自主实验室机器人的进一步研究和开发提供了基础。
cs.RO / 17 / 2608.18632
Evaluation of Monocular SLAM Systems on High-Altitude Nadir UAV Footage
高空垂直无人机影像中单目SLAM系统的评估
Abstract
Aerial nadir video combines weak geometric constraints with severe perceptual aliasing, making it a difficult regime for monocular SLAM. We benchmark five monocular SLAM systems on local UAV flights, synthetic city-scale imagery, and long-range aerial sequences. To isolate visual performance, we provide no inertial or GNSS aiding. Performance varies strongly with environment and trajectory scale: MASt3R-SLAM achieves the lowest mean horizontal MAE on the five DJI flights (0.53\% of reference path length), whereas no system consistently preserves global trajectory shape on the long GES and ALTO sequences. Overall, DROID-SLAM performs best, averaging 2.88\% of reference path length across completed runs. Vertical position remains poor, and large-area trajectories remain highly distorted despite loop-closure capability. Current monocular SLAM methods are by themselves therefore insufficient for reliable visual-only aerial navigation.
Chinese Translation
航拍垂直视频结合了弱几何约束和严重的感知混淆，使得单目SLAM成为一个困难的领域。我们在本地无人机飞行、合成城市规模影像和长距离航拍序列上对五个单目SLAM系统进行了基准测试。为了隔离视觉性能，我们没有提供任何惯性或GNSS辅助。性能在环境和轨迹规模上差异显著：MASt3R-SLAM在五次DJI飞行中实现了最低的平均水平绝对误差（MAE），为参考路径长度的0.53\%；而在长距离的GES和ALTO序列中，没有任何系统能够持续保持全球轨迹形状。总体而言，DROID-SLAM表现最佳，在完成的运行中平均为参考路径长度的2.88\%。垂直位置表现仍然较差，尽管具备闭环能力，但大面积轨迹仍然高度失真。因此，目前的单目SLAM方法本身不足以实现可靠的仅基于视觉的航拍导航。
cs.RO / 18 / 2608.18647
Progressive Experience Fusion for Multi-Task World Model Control in Endovascular Navigation
用于血管内导航的多任务世界模型控制的渐进经验融合
Abstract
Autonomous endovascular navigation could support the delivery of mechanical thrombectomy to underserved areas, but controllers must navigate long, multi-stage paths across varying vascular anatomies. This study investigates Progressive Experience Fusion (PEF) to train a multi-task TD-MPC2 controller. We additionally evaluate a heuristic that changes the Model Predictive Path Integral planning horizon using residual action-sequence dispersion, and fine-tuning in a patient-specific simulation. Across five subtasks in ten known training anatomies with held-out targets, PEF achieved a mean success rate of 74%, compared with 37% for Soft Actor-Critic (p < 0.001) and 65% for base TD-MPC2 (p = 0.053). A PEF controller with adaptive-horizon planning trained on 30 vasculatures achieved a mean success rate of 90% in ten held-out vasculatures. The PEF agent successfully transferred to an unseen in vitro stroke patient vasculature under fluoroscopy, achieving a mean path ratio improvement from 63% to 80% with fine-tuning (p < 0.001), following 40x103 fine-tuning steps (corresponding to approximately 107 min of clinical inter-hospital transfer time). This work represents a proof of concept for multi-vasculature training and patient-specific adaptation, while further validation is required before clinical deployment.
Chinese Translation
自主血管内导航可以支持将机械取栓技术送达服务不足的地区，但控制器必须在不同的血管解剖结构中导航长且多阶段的路径。本研究探讨了渐进经验融合（Progressive Experience Fusion, PEF）以训练多任务TD-MPC2控制器。我们还评估了一种启发式方法，该方法利用残余动作序列分散性改变模型预测路径积分（Model Predictive Path Integral, MPPI）规划的时间范围，并在特定患者的模拟中进行微调。在十个已知训练解剖结构的五个子任务中，PEF的平均成功率为74%，而软演员-评论家（Soft Actor-Critic）的成功率为37%（p < 0.001），基础TD-MPC2的成功率为65%（p = 0.053）。在30个血管上训练的具有自适应时间范围规划的PEF控制器在十个保留血管中达到了90%的平均成功率。PEF代理成功转移到未见的体外中风患者血管下的荧光透视下，经过微调后，路径比率的平均改善从63%提高到80%（p < 0.001），经过40x10^3次微调步骤（相当于约107分钟的临床院际转运时间）。这项工作代表了多血管训练和患者特异性适应的概念验证，但在临床应用之前仍需进一步验证。
cs.RO / 19 / 2608.18672
Orienteering Problem with Uncertain Time-Varying Rewards: Framework and Benchmark for Everyday Service Robotics
具有不确定时变奖励的定向问题：日常服务机器人框架与基准
Abstract
We present the orienteering problem with uncertain time-varying rewards (OP-UTVR), a novel variant of the orienteering problem (OP). While most existing OP formulations assume rewards to be known in advance, practical applications involve uncertain and time-varying rewards, as with shifting customer demand for delivery agents. OP-UTVR relaxes this assumption by allowing agents to estimate reward dynamics from observations and forecast future rewards. This enables informed routing decisions despite stochastic reward changes and inevitable prediction errors. We address this problem using three planners that differ in planning horizon and online adaptivity, and derive theoretical bounds on their performance under reward stochasticity. We further introduce a mobile service robot benchmark for OP-UTVR, where a robot navigates among pedestrians in indoor environments. Experiments reveal trade-offs between planning horizon and adaptivity, and demonstrate the effectiveness of long-horizon planning with online adaptation.
Chinese Translation
我们提出了具有不确定时变奖励的定向问题（OP-UTVR），这是定向问题（OP）的一个新变体。虽然大多数现有的OP模型假设奖励是事先已知的，但实际应用中涉及不确定和时变的奖励，例如交付代理的客户需求变化。OP-UTVR通过允许代理从观察中估计奖励动态并预测未来奖励，放宽了这一假设。这使得尽管存在随机奖励变化和不可避免的预测误差，代理仍能做出明智的路径规划决策。我们使用三种在规划视野和在线适应性上有所不同的规划器来解决这一问题，并推导出它们在奖励随机性下的性能理论界限。我们进一步为OP-UTVR引入了一个移动服务机器人基准，其中机器人在室内环境中在人群中导航。实验揭示了规划视野与适应性之间的权衡，并展示了长视野规划与在线适应的有效性。
cs.RO / 20 / 2608.18701
SoftVTBench: A Deformation-Aware Visuo-Tactile Dataset and Benchmark for Deformable-Object Manipulation
SoftVTBench：一种考虑变形的视觉-触觉数据集及其在可变形物体操作中的基准测试
Abstract
Physical interaction quality is central to deformable-object manipulation, yet most benchmarks evaluate task success alone. A policy may complete the task while allowing slip or causing excessive compression. A primary bottleneck is the absence of visuo-tactile datasets that pair policy-visible contact observations with independent physical ground truth over complete tasks. We introduce SoftVTBench, a visuo-tactile dataset for physical-interaction-aware deformable-object manipulation. It contains 4,000 expert demonstrations and more than 50 assets, including volumetric deformable objects and visually matched rigid twins. At 20 Hz, each episode synchronizes multi-view RGB, dual-finger tactile RGB and marker motion, proprioception, language, and binary and continuous gripper actions, alongside evaluator-only finite-element (FEM) states. Building upon this dataset, we establish a closed-loop benchmark that uses fixed object-specific calibration to define the Deformation-aware Success Rate (DSR), which counts a rollout as successful only when it completes the task and keeps peak normalized deformation within tolerance. Across Diffusion Policy, $\pi_{0.5}$, and FastWAM, all 12 in-distribution configurations contain successful rollouts that violate the deformation tolerance, accounting for 0.7--24% of each configuration's successes. Under distribution shift, visuo-tactile variants achieve higher task success in all six policy--suite comparisons and higher DSR in five, whereas their in-distribution benefits are mixed. These results show that making touch available does not by itself ensure effective multimodal fusion. SoftVTBench therefore provides a common visuo-tactile resource for studying not only whether a policy succeeds, but how it physically interacts with deformable objects and when touch improves that interaction.
Chinese Translation
物理交互质量是可变形物体操作的核心，但大多数基准测试仅评估任务成功与否。一种策略可能在完成任务的同时允许滑动或造成过度压缩。主要瓶颈在于缺乏将策略可见的接触观察与完整任务的独立物理真实情况相结合的视觉-触觉数据集。我们引入了SoftVTBench，这是一个用于物理交互感知的可变形物体操作的视觉-触觉数据集。它包含4000个专家演示和超过50个资产，包括体积可变形物体和视觉匹配的刚性双胞胎。在20 Hz的频率下，每个实验同步多视角RGB、双指触觉RGB、标记运动、身体感知、语言以及二元和连续夹持器动作，同时提供仅限评估者的有限元（FEM）状态。在此数据集的基础上，我们建立了一个闭环基准，利用固定的物体特定校准来定义考虑变形的成功率（Deformation-aware Success Rate, DSR），只有在完成任务并保持峰值归一化变形在容忍范围内时，才将一次滚动视为成功。在Diffusion Policy、$ ext{π}_{0.5}$和FastWAM的所有12个分布内配置中，都包含了违反变形容忍度的成功滚动，占每个配置成功的0.7%至24%。在分布转移下，视觉-触觉变体在所有六个策略-套件比较中实现了更高的任务成功率，并在五个比较中实现了更高的DSR，而它们在分布内的优势则各有不同。这些结果表明，仅仅提供触觉并不能确保有效的多模态融合。因此，SoftVTBench为研究策略是否成功、如何与可变形物体进行物理交互以及何时触觉改善这种交互提供了一个共同的视觉-触觉资源。
cs.RO / 21 / 2608.18787
Dream2Reward: Transition-Alignment Reward Models from Positive Demonstrations for Robotic Manipulation
Dream2Reward：基于正面示范的过渡对齐奖励模型用于机器人操控
Abstract
Learning robotic policies requires dense rewards that remain informative when behavior departs from successful demonstrations. Progress-based rewards estimate how far an observation has advanced along a nominal successful trajectory, but may remain high after an incorrect transition. We introduce Dream2Reward, which learns a language-conditioned successful latent transition field from positive demonstrations. Given the visual history up to a transition start, the model predicts the latent displacement associated with successful execution and scores the observed displacement through signed directional and symmetric magnitude agreement. This transition-level comparison penalizes wrong-direction, overshooting, and stagnant motion even when the resulting observation appears to show progress. Dream2Reward requires no failure annotations, progress labels, or synthetic negatives, and produces a dense causal reward. Across mechanism diagnostics and shared-trajectory evaluations, it provides stronger success-failure separation and more informative feedback on low-quality behavior than progress-based alternatives. Across online and offline policy learning, the same frozen reward model reduces reward hacking and supports stronger downstream performance, including in real-robot manipulation. These results show that comparing realized motion with predicted successful change provides an effective way to convert positive demonstrations into dense rewards for robot learning.
Chinese Translation
学习机器人策略需要密集的奖励，这些奖励在行为偏离成功示范时仍然保持信息量。基于进展的奖励估计观察在名义成功轨迹上前进的距离，但在错误过渡后可能仍然保持高值。我们提出了Dream2Reward，它从正面示范中学习语言条件的成功潜在过渡场。给定过渡开始时的视觉历史，该模型预测与成功执行相关的潜在位移，并通过有符号方向和对称幅度一致性对观察到的位移进行评分。这种过渡级比较惩罚错误方向、超越和停滞运动，即使结果观察似乎显示出进展。Dream2Reward不需要失败注释、进展标签或合成负样本，并产生密集的因果奖励。在机制诊断和共享轨迹评估中，它提供了更强的成功-失败分离和对低质量行为更具信息量的反馈，优于基于进展的替代方法。在在线和离线策略学习中，相同的冻结奖励模型减少了奖励操控，并支持更强的下游性能，包括在真实机器人操控中的表现。这些结果表明，将实际运动与预测的成功变化进行比较提供了一种有效的方法，将正面示范转化为机器人学习的密集奖励。
cs.RO / 22 / 2608.18840
Beyond Placement and Articulation: Usage-Driven Code Scenes for Embodied Interaction
超越位置与表达：基于使用驱动的具身交互代码场景
Abstract
Indoor scene synthesis provides essential environments for embodied AI, robotic manipulation, and simulation-based policy learning. Recent code-based scene generation methods produce editable and extensible environments, yet they remain focused on visual construction and object-level articulation, leaving the functional usage of scenes largely unmodeled. To address this problem, we present RoomWright, an agentic usage-driven framework for generating 3D scenes represented entirely as code for embodied interaction. RoomWright performs usage-driven object reasoning, which treats each anchor as a task centre and admits task-required objects and their affordances. A code agent further enables multi-part interaction by compiling each interaction into a trigger, condition, effect rule that updates structured object states, capturing causal dependencies across objects. Moreover, since manipuland orientation is ambiguous and hard to recover from pixels, RoomWright alleviates this via annotation-informed usage-guided orientation. Extensive experiments demonstrate the effectiveness of our method. The resulting scenes are executable, editable, and simulation-ready, providing interactive environments for embodied AI and policy learning.
Chinese Translation
室内场景合成为具身人工智能、机器人操作和基于模拟的策略学习提供了必要的环境。近期的基于代码的场景生成方法产生了可编辑和可扩展的环境，但仍然主要集中于视觉构建和对象级表达，导致场景的功能使用在很大程度上未被建模。为了解决这个问题，我们提出了RoomWright，一个基于使用驱动的框架，用于生成完全以代码表示的3D场景，以实现具身交互。RoomWright执行使用驱动的对象推理，将每个锚点视为任务中心，并接纳任务所需的对象及其可供性。一个代码代理进一步通过将每次交互编译为触发器、条件、效果规则来实现多部分交互，从而更新结构化的对象状态，捕捉对象之间的因果依赖。此外，由于操作对象的方向模糊且难以从像素中恢复，RoomWright通过注释信息驱动的使用导向方向来缓解这一问题。大量实验表明我们的方法是有效的。生成的场景可执行、可编辑且适合模拟，为具身人工智能和策略学习提供了交互环境。
cs.RO / 23 / 2608.18948
RoboEdit: Turning Human Manipulation Videos into Scalable Robot Experience
RoboEdit：将人类操控视频转化为可扩展的机器人经验
Abstract
Collecting robot hand-object interaction data is costly and embodiment-specific, yet abundant human-object videos remain unusable for robot training. We present RoboEdit, a human-to-robot video editing suite that transforms human manipulation videos into action-consistent, physically plausible robot videos with aligned 3D hand states. To enable scalable supervision, we introduce RoboEdit-ADC, an automatic pipeline that reconstructs and retargets 3D interactions from RGB videos across embodiments. This pipeline generates RoboEdit-14M, a large-scale dataset of 174K aligned video pairs (14M frames) spanning seven robot embodiments, diverse scenes, and interaction types. The core editing engine, RoboEdit-Trans, employs cross-embodiment adaptation modules to preserve temporal coherence while adapting appearance and motion. It further integrates a 3D Robot-State Decoder to recover per-frame hand states for structured motion supervision. Experiments show that RoboEdit achieves state-of-the-art editing quality and supports downstream robot control policies in real-world manipulation tasks. Ultimately, the RoboEdit suite unlocks the vast potential of unlabeled human videos, providing scalable, high-fidelity visual and 3D motion supervision for generalizable robot learning.
Chinese Translation
收集机器人手-物体交互数据成本高且依赖于具体的体现形式，而大量人类-物体视频仍无法用于机器人训练。我们提出了RoboEdit，一个人到机器人视频编辑套件，它将人类操控视频转化为动作一致、物理上合理的机器人视频，并对齐3D手部状态。为了实现可扩展的监督，我们引入了RoboEdit-ADC，一个自动化流程，从RGB视频中重建并重新定向3D交互，适用于不同的体现形式。该流程生成了RoboEdit-14M，一个大规模数据集，包含174K对齐视频对（14M帧），涵盖七种机器人体现形式、多样的场景和交互类型。核心编辑引擎RoboEdit-Trans采用跨体现适应模块，以保持时间一致性，同时适应外观和运动。它进一步集成了3D机器人状态解码器，以恢复每帧的手部状态，从而实现结构化的运动监督。实验表明，RoboEdit实现了最先进的编辑质量，并支持在现实世界操控任务中的下游机器人控制策略。最终，RoboEdit套件释放了未标记人类视频的巨大潜力，为可扩展的高保真视觉和3D运动监督提供了支持，从而促进了可推广的机器人学习。
cs.RO / 24 / 2608.19004
Autonomous Agricultural Tractor: Integrated Weed Detection and LiDAR Navigation for Precision Paddy Farming
自主农业拖拉机：集成杂草检测与激光雷达导航的精准水稻耕作
Abstract
Site-specific weed management in paddy farming offers substantial reductions in herbicide use over conventional broadcast spraying, but field deployment has been limited by three persistent challenges: robust crop-row navigation under canopy where GNSS degrades, real-time visual discrimination between rice and morphologically diverse weeds, and the asymmetric cost of misclassifying rice as weed, which is irreversible. This paper presents AgriNav, an integrated autonomous tractor system built around four ROS-coupled modules: a custom PyTorch reimplementation of WeedDet for rice detection, a parallel lightweight 1.68M-parameter CNN-FPN variant with asymmetric class weighting, an inverted-logic discrimination module that protects the rice class through a hardcoded confidence-gate veto, and a 6-state constant-velocity-turn-rate Extended Kalman Filter fusing GNSS, IMU, and wheel odometry with three-level outage bridging. Our primary system-level contribution is a four-mechanism LiDAR-camera fusion bridge that uses the navigation LiDAR for region-of-interest constraint, world-coordinate projection, ground-plane filtering, and bidirectional confidence fusion at zero additional hardware cost. Simulation experiments demonstrate continuous position tracking through a 20-second GNSS outage, crop row detection confidence above 0.9 throughout operation, and rice-detection confidences from 0.32 to 0.95 across paddy, aerial, and post-flood imagery. The LiDAR ROI constraint reduces detection inference region by an estimated 30 to 50 percent.
Chinese Translation
在水稻耕作中，特定地点的杂草管理相较于传统的均匀喷洒方法显著减少了除草剂的使用，但在实际应用中面临三大持续挑战：在GNSS信号衰减的遮蔽环境下的稳健作物行导航、实时区分水稻与形态各异的杂草的视觉能力，以及将水稻误分类为杂草所带来的不可逆转的成本不对称。本文提出了AgriNav，一个集成的自主拖拉机系统，围绕四个ROS耦合模块构建：一个基于PyTorch的定制版WeedDet用于水稻检测，一个具有不对称类别加权的轻量级1.68M参数CNN-FPN变体，一个通过硬编码置信门否决保护水稻类别的反向逻辑区分模块，以及一个融合GNSS、IMU和车轮里程计的六状态恒速转向率扩展卡尔曼滤波器，具备三层故障桥接。我们的主要系统级贡献是一个四机制的激光雷达-相机融合桥，利用导航激光雷达进行感兴趣区域约束、世界坐标投影、地面平面过滤和双向置信融合，且无需额外硬件成本。仿真实验表明，在20秒的GNSS信号中断期间实现了连续位置跟踪，作物行检测置信度在整个操作过程中超过0.9，水稻检测置信度在水稻、空中和洪水后影像中从0.32到0.95不等。激光雷达的ROI约束将检测推理区域减少了估计的30%至50%。
cs.RO / 25 / 2608.19059
LT-Mem: Volatility-Aware Spatio-Temporal Memory for Lifelong Scene Understanding
LT-Mem：面向波动的时空记忆框架用于终身场景理解
Abstract
Long-term robot operation in evolving environments requires object-level understanding that persists across repeated revisits. Existing systems either overwrite history to maintain an up-to-date map or store semantic snapshots without consistent cross-session object identity, resulting in temporal amnesia: the systematic loss of object history that prevents answering queries such as "Where has the green chair been across all sessions?" We propose LT-Mem, a volatility-aware memory evolution framework that unifies spatially aligned instance-level 3D perception with volatility-conditioned temporal reasoning. First, a multi-session SLAM backbone provides spatially aligned per-object observations across sessions. Second, a reasoning layer governs how object memory evolves: deterministic evidence scoring preserves cross-session identity, and a volatility-aware policy selects among overwrite, hold, and multi-hypothesis actions based on each object's dynamics. Third, the resulting Tri-Memory structure (Live, Delta, Meta) preserves both current states and event histories, enabling longitudinal object-centric reasoning. We further introduce LT-VQA, a dataset and evaluation suite comprising multi-session recordings, persistent identity annotations, and temporal QA pairs. Experiments show that LT-Mem consistently outperforms baselines across all metrics while consuming an order of magnitude fewer tokens, and ablations confirm that gains are driven by the structured memory architecture rather than LLM capacity.
Chinese Translation
在不断变化的环境中，长期机器人操作需要能够在重复访问中持续存在的对象级理解。现有系统要么通过覆盖历史来维护最新地图，要么存储语义快照而没有一致的跨会话对象身份，导致时间遗忘：系统性地丧失对象历史，无法回答诸如“绿色椅子在所有会话中出现过在哪里？”的问题。我们提出了LT-Mem，一个面向波动的记忆演化框架，统一了空间对齐的实例级3D感知与波动条件下的时间推理。首先，一个多会话SLAM主干提供跨会话的空间对齐每个对象的观测。其次，一个推理层控制对象记忆的演变：确定性证据评分保持跨会话身份，而面向波动的策略根据每个对象的动态选择覆盖、保持和多假设行动。第三，得到的三重记忆结构（Live, Delta, Meta）同时保留当前状态和事件历史，支持纵向的对象中心推理。我们进一步引入LT-VQA，一个包含多会话录音、持久身份注释和时间问答对的数据集和评估套件。实验表明，LT-Mem在所有指标上始终优于基线，同时消耗的令牌数量减少了一个数量级，消融实验确认了性能提升源于结构化记忆架构而非大型语言模型（LLM）能力。
cs.RO / 26 / 2608.19085
DA-WAM: Decision-Aligned Future Latents for Driving World Models
DA-WAM：用于驾驶世界模型的决策对齐未来潜变量
Abstract
Anticipating how scenes evolve under ego actions is fundamental to safe autonomous driving, yet the full potential of world models for decision-making remains unrealized. The critical challenge lies in ensuring that future modeling is not merely predictive, but decision-informative: the predicted future must directly shape which trajectory is selected. Existing approaches decouple future representation learning from planning optimization, or share predicted states across trajectory candidates, thereby diluting the action-specific consequences that ought to guide selection. To bridge this gap, we propose DA-WAM, a framework that unifies predictive representation learning, action-conditioned future modeling, and trajectory scoring under a single decision-making objective. DA-WAM maintains predictive supervision throughout planner optimization via an online encoder and a stable momentum target, allowing future representations to co-evolve with the driving task. An action-conditioned predictor generates a distinct future latent state per trajectory candidate, which is then evaluated by a future-latent-conditioned factorized scorer. For the expert-matched trajectory, the predicted future latent is supervised by the observed future representation, while safety-critical hard negatives provide additional supervision near planning boundaries. Extensive experiments on NAVSIM-v1 and NAVSIM-v2 demonstrate state-of-the-art performance, while ablations and diagnostic analyses validate the key components.
Chinese Translation
预测场景在自我行动下如何演变对于安全的自主驾驶至关重要，但世界模型在决策中的潜力尚未完全实现。关键挑战在于确保未来建模不仅仅是预测性的，而是能够为决策提供信息：预测的未来必须直接影响所选择的轨迹。现有方法将未来表示学习与规划优化解耦，或在轨迹候选之间共享预测状态，从而稀释了应指导选择的特定于动作的后果。为了解决这一问题，我们提出了DA-WAM，一个将预测表示学习、基于动作的未来建模和轨迹评分统一在单一决策目标下的框架。DA-WAM通过在线编码器和稳定的动量目标在规划优化过程中保持预测监督，使未来表示能够与驾驶任务共同演化。基于动作的预测器为每个轨迹候选生成独特的未来潜变量状态，然后由一个基于未来潜变量的分解评分器进行评估。对于与专家匹配的轨迹，预测的未来潜变量由观察到的未来表示进行监督，而安全关键的硬负样本在规划边界附近提供额外的监督。在NAVSIM-v1和NAVSIM-v2上的广泛实验展示了最先进的性能，而消融实验和诊断分析验证了关键组件的有效性。
cs.RO / 27 / 2608.19182
ADEPT: Accelerating Dexterity via Pre-Training and Post-Training using Reinforcement Learning
ADEPT：通过预训练和后训练加速灵巧性，利用强化学习
Abstract
We introduce Accelerating Dexterity via Pre-Training (ADEPT), a large-scale reinforcement learning (RL) framework for learning sim-to-real transferable dexterity across high degree-of-freedom (DoF) robot embodiments that can solve long-horizon tasks directly from raw visuo-tactile perception. ADEPT pretrains a dexterous policy on a generic object reposing task, then post-trains downstream policies with this pretrained behavior as a prior. ADEPT enables learning new behaviors that are otherwise difficult to discover from scratch on multi-fingered robots and avoids learning the same set of skills over again for every new downstream task. The pretrained policy zero-shots the reposing phase of downstream tasks, but na\"ive RL fine-tuning rapidly degrades this capability during transfer. We address this with a stable post-training recipe combining behavior-cloning distillation, critic warm-up, and conservative on-policy updates. To safely exploit the full kinematic dexterity, we introduce a joint-space Geometric Fabric that mediates between the RL policy and the robot. We distill post-trained teachers into perceptive students that zero-shot sim-to-real transfer on two embodiments: a 23 DoF Kuka-Allegro with two RGB cameras, and a 29 DoF Flexiv-Sharpa with two RGB cameras and five vision-based tactile sensors, and can solve long-horizon tasks from challenging initial states with dexterity at human-level speed.
Chinese Translation
我们介绍了通过预训练加速灵巧性（ADEPT），这是一个大规模的强化学习（RL）框架，用于学习可在高自由度（DoF）机器人实现中进行模拟到真实（sim-to-real）可转移的灵巧性，能够直接从原始的视觉-触觉感知中解决长时间跨度的任务。ADEPT在一个通用物体静置任务上预训练灵巧策略，然后利用这一预训练行为作为先验，对下游策略进行后训练。ADEPT使得在多指机器人上学习新的行为成为可能，这些行为在从零开始的情况下通常难以发现，并且避免了在每个新的下游任务中重复学习相同的技能。预训练策略能够零-shot（zero-shot）处理下游任务的静置阶段，但简单的强化学习微调在迁移过程中会迅速降低这一能力。我们通过结合行为克隆蒸馏、评论家预热和保守的在线更新，提出了一种稳定的后训练方法。为了安全地利用全运动学灵巧性，我们引入了一种关节空间几何结构（Geometric Fabric），作为强化学习策略与机器人之间的中介。我们将后训练的教师模型蒸馏为感知学生，使其能够在两个实现上进行零-shot的模拟到真实转移：一个23自由度的Kuka-Allegro配备两个RGB摄像头，以及一个29自由度的Flexiv-Sharpa配备两个RGB摄像头和五个基于视觉的触觉传感器，并能够以人类水平的速度从具有挑战性的初始状态中解决长时间跨度的任务。
cs.RO / 28 / 2608.19188
PartialBiGrasp: Inferring Hidden Local Geometry for Bimanual Grasping from Partial Views
PartialBiGrasp：从部分视图推断双手抓取的隐藏局部几何结构
Abstract
Dual-arm robotic grasping is essential for manipulating large, heavy, and geometrically complex objects that cannot be reliably handled using a single manipulator. These large objects often contain only sparse graspable regions determined by local geometric properties such as thickness, edge structure, and gripper clearance. Prior bimanual grasping methods assume access to a full point cloud of the object which inherently contains this geometric information, but may not be accessible in real scenarios. This work proposes PartialBiGrasp, a dual-arm grasp generation framework that operates directly on partial point cloud observations. Our model learns geometric features implicitly through convolutional occupancy networks, enabling local reasoning about graspability, collision-free contact regions, and object thickness. We leverage this understanding to generate force-closure compliant grasp pairs, which are further refined using a sampling-based optimization to correct for ambiguity caused by incomplete geometry. We evaluate our approach using analytical force-closure metrics, large-scale simulation experiments, and real-world robot evaluations on noisy partial point clouds of novel objects, demonstrating robust and physically stable dual-arm grasp generation.
Chinese Translation
双臂机器人抓取对于操控大型、重型及几何复杂的物体至关重要，这些物体无法通过单一操控器可靠处理。这些大型物体通常仅包含由局部几何特性（如厚度、边缘结构和夹持器间隙）决定的稀疏可抓取区域。以往的双手抓取方法假设可以访问物体的完整点云，而该点云本质上包含了这些几何信息，但在实际场景中可能无法获取。本研究提出了PartialBiGrasp，一个直接基于部分点云观测的双臂抓取生成框架。我们的模型通过卷积占用网络隐式学习几何特征，使得能够对抓取能力、无碰撞接触区域和物体厚度进行局部推理。我们利用这种理解生成符合力闭合条件的抓取对，并通过基于采样的优化进一步修正因几何不完整而导致的模糊性。我们使用分析力闭合指标、大规模仿真实验以及在新物体的噪声部分点云上的实际机器人评估来评估我们的方法，展示了稳健且物理稳定的双臂抓取生成能力。
cs.CV / 1 / 2608.18184
Human-Centric Intelligence in the Era of Foundation Models: A Survey
基础模型时代的人本智能：一项综述
Abstract
Human-centric intelligence is evolving in the foundation-model era, with growing emphasis on scale, transferability, and general-purpose modeling. Yet it has not fully integrated with foundation models to achieve the comparable progress seen in them. More importantly, recent advances across this broad landscape remain fragmented across tasks, modalities, and research communities, leaving their intrinsic conceptual and methodological connections unclear. To bridge these divides and rethink human-centric intelligence in the foundation-model era, we introduce a full-spectrum human context taxonomy that integrates six interconnected levels by viewing humans as observable subjects through visual appearance and spatial geometry, as dynamic actors through kinematic dynamics and interaction modeling, and as situated agents through world simulation and embodied agency. We next present the methodological foundations of the field, covering human-centric data families, computational architecture paradigms, and representative training and inference optimization strategies. We then systematically review representative methods across these levels and organize the associated datasets, benchmarks, and evaluation metrics. We further discuss open challenges and promising research directions toward human-centric intelligence that is scalable, trustworthy, physically grounded, and deployable, aiming to provide a coherent framework and practical reference for advancing the field. Finally, we provide a systematically organized and continuously updated collection of human-centric AI literature and resources on our project page.
Chinese Translation
在人本智能的发展过程中，基础模型时代越来越强调规模、可转移性和通用建模。然而，它尚未与基础模型充分整合，以实现与之相当的进展。更重要的是，近期在这一广泛领域的进展在任务、模态和研究社区之间仍然是碎片化的，使得其内在的概念和方法论联系不清晰。为了弥合这些鸿沟并重新思考基础模型时代的人本智能，我们引入了一个全谱的人类背景分类法，通过视觉外观和空间几何将人类视为可观察的对象，通过运动动力学和交互建模将其视为动态参与者，通过世界模拟和具身代理将其视为情境代理，从而整合六个相互关联的层次。接下来，我们介绍该领域的方法论基础，涵盖人本数据家族、计算架构范式以及代表性的训练和推理优化策略。然后，我们系统地回顾这些层次中的代表性方法，并整理相关的数据集、基准和评估指标。我们进一步讨论面向可扩展、可信赖、物理基础和可部署的人本智能的开放挑战和有前景的研究方向，旨在为推动该领域提供一个连贯的框架和实用参考。最后，我们在我们的项目页面上提供了一个系统组织和持续更新的人本人工智能文献和资源集合。
cs.CV / 2 / 2608.18193
Bound-Aware Per-Organ Recall Risk Control for Multi-Organ CT Segmentation under Clinical Domain Shift
临床领域转移下多脏器CT分割的边界感知每脏器召回风险控制
Abstract
Distribution-free risk control adds organ-specific recall guarantees to frozen segmentation. We calibrate per-organ thresholds for an AMOS-trained nnU-Net, audit transfer to RAOS, and estimate local re-certification cost using case-level voxel false-negative rate (FNR). The AMOS control passes, but $7/12$ organs exceed $\alpha{=}0.10$ after transfer; smaller calibration sets can mask exceedances with conservative or vacuous thresholds. Risk-Controlling Prediction Sets (RCPS) give high-probability control of population-mean risk, whereas Conformal Risk Control (CRC) gives weaker expectation control. Both require exchangeability; fixed and global thresholds give no per-organ guarantee. The Waudby--Smith--Ramdas (WSR) betting bound re-certifies six Tier-1 organs with 25 local cases, versus 30--40 for Hoeffding--Bentkus (HB). CRC needs 10--15 but has a heavier individual-case tail. No Tier-2 organ meets our illustrative precision criterion with 25 cases.
Chinese Translation
无分布风险控制为冻结分割添加了特定于脏器的召回保证。我们为经过AMOS训练的nnU-Net校准每个脏器的阈值，审核转移到RAOS，并使用案例级体素假阴性率（FNR）估算局部再认证成本。AMOS控制通过，但在转移后有$7/12$个脏器超过了$eta{=}0.10$；较小的校准集可能会通过保守或空洞的阈值掩盖超标情况。风险控制预测集（RCPS）提供了对总体均值风险的高概率控制，而符合风险控制（CRC）提供了较弱的期望控制。两者都需要可交换性；固定和全局阈值无法提供每个脏器的保证。Waudby-Smith-Ramdas（WSR）投注界限在25个局部案例中重新认证了六个Tier-1脏器，而Hoeffding-Bentkus（HB）则需要30-40个案例。CRC需要10-15个案例，但个别案例的尾部更重。在25个案例中，没有Tier-2脏器满足我们的示例精度标准。
cs.CV / 3 / 2608.18215
LumiTokens: 3D Relighting via Token-Space Lighting Transformation
LumiTokens：通过令牌空间光照变换进行3D重光照
Abstract
Existing 3D relighting methods operate through either explicit material decomposition, diffusion-based view-space generation, or a combination of both, requiring full recomputation for each new lighting condition. We observe that recent latent scene representations, which encode multi-view images into a set of compact tokens with no fixed physical semantics, open up a novel design space for relighting. We present LumiTokens, a framework that formulates 3D relighting as a direct transformation on latent scene tokens, without explicit 3D representations, rendering equations, or physics-based decomposition. Our model introduces a Scene Token Editor that processes scene tokens jointly with light-ray tokens through self-attention, producing updated tokens that can be decoded into multi-view-consistent relit images. To support diverse lighting types through a unified interface, all lighting signals, including environment maps, point lights, and area lights, are parameterized as Plucker ray tokens, enabling native 3D user interaction with a representation that carries no explicit spatial structure. Crucially, this design supports progressive relighting: because the editor's output remains in the same latent space as its input, a user can incrementally build up illumination one light source at a time, with each edit composing in token space. Experiments demonstrate that LumiTokens achieves comparable or superior relighting quality to other methods and supports progressive, composable lighting edits. Project page: https://neu-vi.github.io/LumiTokens/
Chinese Translation
现有的3D重光照方法通过显式材料分解、基于扩散的视图空间生成或两者的结合来操作，要求在每种新的光照条件下进行全面的重新计算。我们观察到，最近的潜在场景表示将多视图图像编码为一组没有固定物理语义的紧凑令牌，为重光照开辟了新的设计空间。我们提出了LumiTokens，一个将3D重光照公式化为对潜在场景令牌的直接变换的框架，无需显式的3D表示、渲染方程或基于物理的分解。我们的模型引入了一个场景令牌编辑器，通过自注意力机制与光线令牌共同处理场景令牌，生成可以解码为多视图一致的重光照图像的更新令牌。为了通过统一接口支持多样的光照类型，所有光照信号，包括环境贴图、点光源和面光源，都被参数化为Plucker光线令牌，使得用户能够与没有显式空间结构的表示进行原生的3D交互。关键是，这种设计支持渐进式重光照：由于编辑器的输出与其输入保持在同一潜在空间中，用户可以逐步构建光照，每次添加一个光源，每次编辑都在令牌空间中组合。实验表明，LumiTokens在重光照质量上与其他方法相当或更优，并支持渐进式、可组合的光照编辑。项目页面：https://neu-vi.github.io/LumiTokens/
cs.CV / 4 / 2608.18240
Zero-Shot Transfer of Force Map Estimation Across GelSight Mini Sensors
跨GelSight Mini传感器的零样本力图估计迁移
Abstract
Despite the rapid industrialization of the touch sensor manufacturing process, most of these sensors are still handmade in research laboratories. This complicates standardizing their performance, requiring the repetition of data collection and training models for each unit produced. To address this problem, this paper presents a method that can generalize the estimation of 3D force maps across different GelSight Mini sensor units, regardless of the sensor version. Specifically, the method consists of two stages: a domain adaptation stage, in which the input tactile image is reconstructed as a general tactile image using a UniT-based model; and a stage for estimating 3D force maps employing a U-Net network. Our proposal achieves promising results in both steps, such as an SSIM of 0.9338 +- 0.0358 in the image reconstruction phase and an MAE_F of 1.1294 +- 1.5934(N) in the force estimation phase.
Chinese Translation
尽管触觉传感器制造过程迅速工业化，但大多数传感器仍在研究实验室手工制作。这使得标准化其性能变得复杂，要求对每个生产单元重复数据收集和模型训练。为了解决这一问题，本文提出了一种方法，可以在不同的GelSight Mini传感器单元之间进行3D力图估计的泛化，无论传感器版本如何。具体而言，该方法包括两个阶段：一个领域适应阶段，在该阶段中，输入的触觉图像使用基于UniT的模型重构为一般触觉图像；以及一个使用U-Net网络进行3D力图估计的阶段。我们的提案在两个步骤中均取得了令人鼓舞的结果，例如在图像重构阶段SSIM为0.9338 ± 0.0358，在力估计阶段MAE_F为1.1294 ± 1.5934(N)。
cs.CV / 5 / 2608.18246
Visual-Prompt Guided Wildlife Instance-Level Recognition
视觉提示引导的野生动物实例级识别
Abstract
Fine-grained wildlife re-identification remains a challenging area in research. Current state-of-the-art approaches apply a detection and re-identification pipeline. We propose a one-stage end-to-end detection and re-identification model that performs identity searching within the latent space. We adopt DINOv2 for robust spatial geometry and MegaDescriptor for wildlife re-identification. We enhance latent queries with prompt re-identification features. A detection decoder queries the scene latent space to establish object boundaries around the target identity. Preliminary findings reflect a competitive mean average precision score of 30.584% compared to the state-of-the-art two stage approach of 44.89%. Qualitative results depict effective bounding and identification of animal identities.
Chinese Translation
细粒度野生动物再识别仍然是一个具有挑战性的研究领域。目前的最先进方法采用检测和再识别的流程。我们提出了一种一阶段端到端的检测和再识别模型，该模型在潜在空间中进行身份搜索。我们采用 DINOv2 以实现稳健的空间几何，并使用 MegaDescriptor 进行野生动物再识别。我们通过提示再识别特征增强潜在查询。检测解码器查询场景潜在空间，以建立目标身份周围的物体边界。初步结果显示，与最先进的两阶段方法 44.89% 相比，我们的方法在平均精度得分上达到了 30.584%，具有竞争力的表现。定性结果描绘了动物身份的有效边界和识别。
cs.CV / 6 / 2608.18305
Acquisition Geometry-Assisted Whole-Group Localization of X-ray Fluorescence Maps in Optical Microscopy Images
基于采集几何的X射线荧光图在光学显微镜图像中的整体组定位
Abstract
X-ray fluorescence (XRF) microscopy maps elemental distributions, while optical microscopy can provide complementary morphological context. Localizing XRF fields of view (FOVs) in optical images is difficult because the two modalities differ in contrast mechanism and resolution. Most current workflows place each XRF tile independently, even when acquisition metadata already record the tiles' relative scan positions. This study formalizes XRF tile-group localization, in which one optical-frame placement is estimated for the whole group, constrained by acquisition geometry and quantified using group intersection-over-union (GroupIoU). In a controlled case study, independent localization failed with GroupIoU 0.000, whereas group localization achieved 0.931. Replacing the normalized cross-correlation (NCC) metric with mutual information (MI) gave nearly identical results, showing that the outcome is not specific to one local similarity metric. In another multiscale case study, using a coarse XRF survey scan to connect the fine-scale tile group to the optical image increased mean GroupIoU from 0.694 to 0.856. These case studies support using acquisition geometry as an explicit constraint when localizing related XRF tiles.
Chinese Translation
X射线荧光（XRF）显微镜能够绘制元素分布，而光学显微镜则提供了补充的形态学背景。在光学图像中定位XRF视场（FOV）是困难的，因为这两种成像方式在对比机制和分辨率上存在差异。目前大多数工作流程独立放置每个XRF图块，即使采集元数据已经记录了图块的相对扫描位置。本研究形式化了XRF图块组的定位，其中为整个组估计一个光学帧的放置位置，该位置受到采集几何的约束，并通过组交并比（Group Intersection-over-Union, GroupIoU）进行量化。在一个受控案例研究中，独立定位的GroupIoU为0.000，而组定位则达到了0.931。将归一化互相关（Normalized Cross-Correlation, NCC）度量替换为互信息（Mutual Information, MI）得到了几乎相同的结果，表明结果并不特定于某一局部相似性度量。在另一个多尺度案例研究中，使用粗略的XRF调查扫描将细尺度图块组与光学图像连接，使得平均GroupIoU从0.694提高到0.856。这些案例研究支持在定位相关XRF图块时将采集几何作为显式约束。
cs.CV / 7 / 2608.18306
High-Flux Count-Free Single-Photon 3D Cameras
高通量无计数单光子三维相机
Abstract
Single-photon cameras based on single-photon avalanche diode (SPAD) technology are gaining popularity for 3D sensing, thanks to their extreme sensitivity and time resolution. There are two key challenges with single-photon cameras that limit their widespread use: (i) they suffer from non-linear distortions called ''pile-up'' when operated in high-photon-flux conditions, and (ii) they generate a large volume of raw photon data, creating a severe data bottleneck at each sensor pixel. In this work, we show that while compressive capture techniques successfully mitigate data transfer challenges, they exacerbate the effects of dead-time distortion because they fail to retain sufficient information about the photon detection history to allow post-processing pile-up correction via existing methods. We propose a new computational-imaging method that combines free-running capture with an analysis-by-synthesis software pipeline to mitigate pile-up distortions. Our results with hardware emulations and full-scene and single-pixel simulations show that our method can reliably capture scene distance and reflectance over a wide range of illumination conditions. Our work will enable high-resolution SPAD cameras that are severely bandwidth-constrained to operate in real-world high-flux scenarios.
Chinese Translation
基于单光子雪崩二极管（SPAD）技术的单光子相机因其极高的灵敏度和时间分辨率在三维感知中越来越受到欢迎。然而，单光子相机在广泛应用中面临两个主要挑战：（i）在高光子通量条件下，它们会遭遇称为“堆叠”（pile-up）的非线性失真；（ii）它们生成大量原始光子数据，导致每个传感器像素处出现严重的数据瓶颈。在本研究中，我们展示了尽管压缩捕获技术成功缓解了数据传输挑战，但由于无法保留足够的光子检测历史信息以允许通过现有方法进行后处理堆叠校正，它们加剧了死时间失真的影响。我们提出了一种新的计算成像方法，将自由运行捕获与基于合成的分析软件管道相结合，以减轻堆叠失真。我们的硬件仿真和全场景及单像素模拟结果表明，我们的方法能够在广泛的照明条件下可靠地捕获场景的距离和反射率。我们的工作将使得在带宽严重受限的情况下，高分辨率的SPAD相机能够在真实世界的高通量场景中运行。
cs.CV / 8 / 2608.18309
XRF-to-Optical Field-of-View Localization with Vision Language Models
基于视觉语言模型的XRF与光学视场定位
Abstract
Registering images acquired with different microscopy modalities is essential for relating complementary measurements of the same specimen. In correlative X-ray fluorescence (XRF) and optical microscopy, the XRF map often covers only a small region of an optical image acquired from the same or an adjacent tissue section. Field-of-view (FOV) localization is necessary but can be difficult when appearance and structure differ across modalities. Here we evaluate training-free vision language model (VLM) localization on two datasets representing same-section high-correspondence and adjacent-section low-correspondence imaging. We test unconstrained and metadata-constrained search and compare VLMs with geometric controls, classical template matching, and two alternative training-free approaches (DINOv2 and multiGradICON). Direct VLM prompting produced content-dependent spatial signals but was not reliable alone. Classical matching was most accurate when cross-modal structure was preserved but failed in the low-correspondence collection. A proposal-and-verify workflow used repeated VLM predictions as candidates and image-based similarity to select the final location. This workflow recovered useful localization in the low-correspondence regime.
Chinese Translation
使用不同显微镜模式获取的图像进行配准对于关联同一标本的互补测量至关重要。在相关的X射线荧光（XRF）和光学显微镜中，XRF图谱通常仅覆盖从同一或相邻组织切片获取的光学图像的小区域。视场（FOV）定位是必要的，但在不同模式之间外观和结构差异时可能会很困难。在这里，我们评估了在两个数据集上进行无训练的视觉语言模型（VLM）定位，这两个数据集分别代表同切片高对应性和相邻切片低对应性的成像。我们测试了无约束和元数据约束的搜索，并将VLM与几何控制、经典模板匹配以及两种替代的无训练方法（DINOv2和multiGradICON）进行了比较。直接的VLM提示产生了依赖内容的空间信号，但单独使用时并不可靠。当跨模式结构得以保留时，经典匹配的准确性最高，但在低对应性集合中失败。一个提议与验证的工作流程使用重复的VLM预测作为候选，并基于图像相似性选择最终位置。该工作流程在低对应性条件下恢复了有用的定位。
cs.CV / 9 / 2608.18311
FedCoRe: Target-Adaptive Completion for Missing Modalities in Healthcare Federated Learning
FedCoRe：医疗保健联邦学习中缺失模态的目标自适应补全
Abstract
Federated multimodal models often assume every site has every modality, although hospitals differ in access to EHRs, chest radiographs, and ECGs. We study this setting on a MIMIC-derived respiratory deterioration task with simulated FL clients and introduce FedCoRe (Federated Cross-Modal Representation Completion). FedCoRe learns representation- or logit-space corrections rather than generating synthetic ECGs or CXR images. When a client observes a modality that may be missing at deployment, it evaluates the same example with and without that modality to obtain paired supervision. Only clients with such pairs update the completion module, and validation may retain the unchanged prediction. We freeze the trained multimodal predictor during evaluation so that measured differences come only from completion. Hiding ECG reduced AUROC by about 0.085; paired-example FedAvg restored 0.0415 AUROC, or 49.0% of the lost performance. We therefore report two distinct effects: paired-example FedAvg partially recovers the missing-ECG gap, while validation-selected completion is a task-specific classifier-logit correction rather than literal ECG recovery. For CXR, effect-aware completion recovers 52.8% of the loss in a controlled test where CXR is hidden. Paired-example FedAvg transfers part of this effect, but validation keeps the no-completion baseline for deployment cases whose inputs lack CXR. Thus, FedCoRe should be read as a validation-gated completion/correction framework: it can recover missing-modality signal in supported settings, but it should be deployed only when paired examples and validation evidence support that modality.
Chinese Translation
联邦多模态模型通常假设每个站点都有每种模态，尽管医院在电子健康记录（EHR）、胸部X光片（CXR）和心电图（ECG）的获取上存在差异。我们在一个基于MIMIC的呼吸恶化任务上研究这种设置，并引入FedCoRe（联邦跨模态表示补全）。FedCoRe学习表示或对数空间的修正，而不是生成合成的ECG或CXR图像。当客户端观察到在部署时可能缺失的模态时，它会评估同一示例在有和没有该模态下的表现，以获得配对监督。只有具有这种配对的客户端才会更新补全模块，而验证可能保留未更改的预测。在评估期间，我们冻结训练好的多模态预测器，以便测量的差异仅来自补全。隐藏ECG使AUROC降低约0.085；配对示例的FedAvg恢复了0.0415 AUROC，或49.0%的性能损失。因此，我们报告了两个不同的效果：配对示例的FedAvg部分恢复了缺失ECG的差距，而验证选择的补全是特定任务的分类器对数修正，而非字面上的ECG恢复。对于CXR，效果感知补全在CXR被隐藏的受控测试中恢复了52.8%的损失。配对示例的FedAvg转移了部分这一效果，但验证保持了缺乏CXR输入的部署案例的无补全基线。因此，FedCoRe应被视为一个验证门控的补全/修正框架：它可以在支持的设置中恢复缺失模态信号，但仅在配对示例和验证证据支持该模态时才应部署。
cs.CV / 10 / 2608.18316
A Configurable Privacy-Preserving MRI Processing Workflow Using Deep Learning-Based Brain Extraction and Adaptive Anatomical Preservation
基于深度学习的脑部提取与自适应解剖保留的可配置隐私保护MRI处理工作流程
Abstract
Structural Magnetic Resonance Imaging (MRI) is widely used in neuroimaging research and clinical practice, but structural MRI volumes may retain facial and cranial anatomical information that raises privacy concerns. Existing deep learning-based brain extraction methods generally produce a single fixed output, limiting flexibility when different applications require different balances between privacy and anatomical preservation. This paper presents a configurable privacy-preserving MRI processing workflow that extends deep learning-based brain extraction through adaptive anatomical preservation, interactive preservation selection, and integrated quality control. The workflow employs SynthStrip for automated brain extraction, followed by morphological mask expansion to generate configurable shell-based preservation levels. An Interactive Preservation Framework enables users to compare preservation configurations and select an appropriate output, while an integrated Quality Control Framework provides multi-plane visualisation and brain-mask overlay verification. The workflow was implemented in Python using open-source neuroimaging libraries within the Renku reproducible research environment and evaluated using structural T1-weighted MRI data from the publicly available IXI dataset. Experimental results demonstrate anatomically plausible brain extraction and configurable preservation outputs, supported by systematic visual verification. The principal contribution is a modular and reproducible MRI preprocessing framework that enhances deep learning-based brain extraction with configurable anatomical preservation, interactive user-guided processing, and integrated quality control. The workflow provides a practical foundation for privacy-oriented neuroimaging research and collaborative medical image analysis.
Chinese Translation
结构性磁共振成像（MRI）广泛应用于神经影像学研究和临床实践，但结构性MRI体积可能保留面部和颅骨解剖信息，从而引发隐私问题。现有的基于深度学习的脑部提取方法通常产生单一固定输出，限制了在不同应用中对隐私和解剖保留之间不同平衡的灵活性。本文提出了一种可配置的隐私保护MRI处理工作流程，通过自适应解剖保留、交互式保留选择和集成质量控制扩展了基于深度学习的脑部提取。该工作流程采用SynthStrip进行自动脑部提取，随后通过形态学掩膜扩展生成可配置的基于壳体的保留级别。交互式保留框架使用户能够比较保留配置并选择合适的输出，而集成质量控制框架则提供多平面可视化和脑掩膜叠加验证。该工作流程在Python中实现，使用开源神经影像学库，并在Renku可重复研究环境中进行评估，使用来自公开可用的IXI数据集的结构性T1加权MRI数据。实验结果表明，脑部提取在解剖上是合理的，且可配置的保留输出得到了系统的视觉验证支持。主要贡献在于提供了一个模块化和可重复的MRI预处理框架，增强了基于深度学习的脑部提取，具有可配置的解剖保留、交互式用户引导处理和集成质量控制。该工作流程为面向隐私的神经影像学研究和协作医学图像分析提供了实用基础。
cs.CV / 11 / 2608.18317
Reproducible Multimodal Affordance Prediction
可重复的多模态可供性预测
Abstract
Affordance prediction is the identification of potential actions an agent can perform on a target object from multimodal inputs. Affordance prediction methods are difficult to evaluate and compare due to heterogeneous problem formulations, inconsistent dataset annotations, incomplete reporting of experimental protocols, and limited information about deployment conditions. These limitations challenge fair benchmarking and performance comparison. To promote transparency, we propose the Affordance Sheet, a documentation detailing task formulation with its input modalities, model architectures and training information, datasets, and experimental protocols. Affordance Sheets enable reproducible benchmarking and reliable evaluation of affordance models for real-world scenarios, including generalisation to novel conditions and human safety.
Chinese Translation
可供性预测是从多模态输入中识别代理可以对目标对象执行的潜在动作。由于问题表述的异质性、数据集注释的不一致性、实验协议报告的不完整性以及关于部署条件的信息有限，可供性预测方法的评估和比较变得困难。这些限制对公平基准测试和性能比较构成了挑战。为了促进透明度，我们提出了可供性文档（Affordance Sheet），该文档详细描述了任务表述及其输入模态、模型架构和训练信息、数据集以及实验协议。可供性文档使得可重复的基准测试和对可供性模型在现实场景中的可靠评估成为可能，包括对新条件的泛化和人类安全性。
cs.CV / 12 / 2608.18339
From Inference to Adaptation: A Unified Optimal Transport View of Vision Language Model
从推理到适应：视觉语言模型的统一最优传输视角
Abstract
Vision-language models (VLMs) have demonstrated remarkable zero-shot capabilities yet remain sensitive to real-world distribution shifts during inference. Although significant efforts are devoted to adapting VLMs at test time, they rely heavily on noisy pseudo-labels predicted directly from raw embedding similarities during inference, which are unreliable under distribution shift and mislead the adaptation. To avoid noise amplification, existing works craft coarse-grained surrogate objectives during adaptation, which fail to explicitly model sample-level relationships across different modalities, creating objective mismatch with inference, thus leading to marginal performance improvement. In this work, we aim to bridge the detached objectives of inference and adaptation for VLMs, and propose a principled VLM TTA method called \algname. For VLM inference, we formulate the zero-shot image classification task as a cross-modal alignment problem encoded via a Wasserstein OT formulation, providing robust pseudo-labels at the sample-level to effectively adapt VLMs. For VLM adaptation, we adopt a soft-label InfoNCE loss to adapt VLMs based on the OT-induced pseudo-labels, leveraging fine-grained supervisions to explicitly model relationships of individual image-text pairs via contrastive learning, which empowers accurate inference at the same granularity. Moreover, we theoretically reveal that the InfoNCE loss can be neatly reformulated as a Wasserstein OT formulation, thereby unifying the objectives of the inference and adaptation of VLMs to achieve their mutual benefits. Extensive experiments demonstrate the effectiveness and efficiency of our methods, outperforming the best-performing methods by up to 7% with state-of-the-art efficiency.
Chinese Translation
视觉语言模型（VLMs）在零-shot 能力方面表现出色，但在推理过程中对现实世界分布变化仍然敏感。尽管在测试时对 VLMs 的适应进行了大量努力，但它们在推理过程中严重依赖于直接从原始嵌入相似性预测的噪声伪标签，而这些伪标签在分布变化下不可靠，导致适应过程受到误导。为了避免噪声放大，现有工作在适应过程中设计了粗粒度的替代目标，但未能明确建模不同模态间的样本级关系，从而导致推理与适应之间的目标不匹配，进而仅实现边际的性能提升。在本研究中，我们旨在弥合 VLMs 的推理与适应之间的脱节目标，并提出了一种原则性的 VLM TTA 方法，称为 extit{algname}。对于 VLM 推理，我们将零-shot 图像分类任务表述为一个通过 Wasserstein 最优传输（OT）公式编码的跨模态对齐问题，从而在样本级提供稳健的伪标签，以有效适应 VLMs。对于 VLM 适应，我们采用软标签的 InfoNCE 损失，基于 OT 诱导的伪标签来适应 VLMs，利用细粒度的监督通过对比学习明确建模个别图像-文本对的关系，从而在相同粒度上增强准确推理。此外，我们理论上揭示了 InfoNCE 损失可以被整齐地重新表述为 Wasserstein OT 公式，从而统一 VLMs 的推理与适应目标，实现其相互利益。大量实验表明我们的方法的有效性和效率，超越了最佳方法，性能提升高达 7%，且具备最先进的效率。
cs.CV / 13 / 2608.18386
TTSD-FAR: Test-Time Self-Distillation with Fisher-Anchored Restoration for Missing-Modality Emotion Recognition in LVLMs
TTSD-FAR：基于Fisher锚定恢复的测试时自我蒸馏用于缺失模态情感识别的研究
Abstract
Large video-language models (LVLMs) have shown remarkable performance on multimodal tasks like multimodal emotion recognition (ER) in the wild. ER is inherently multimodal, requiring a joint understanding of facial expressions, vocalizations, language, biosignals, and gestures. However, real-world deployment remains challenging: modalities may be missing or noisy at test time. Partial observations can be viewed as a distribution shift relative to the complete-modality distribution. SOTA TTA methods based on entropy minimization or perplexity reduction do not transfer to autoregressive LVLMs, while retrieval augmented generation (RAG) degrades when the observed modality is weak. Because no ground-truth supervision exists to verify individual updates, adaptation across this stream risks accumulating drift and degrading once the model departs from a reliable solution. An effective solution must therefore adapt to arbitrary missing-modality patterns and remain effective during continual adaptation. We address both jointly with Test-Time Self-Distillation (TTSD), a parameter-efficient framework in which a frozen teacher, trained on complete modalities, guides an adaptive low-rank student via self-distillation, updating only a negligible number of parameters. Stability is built into this same loop through Fisher-Anchored Restoration (FAR), which monitors Fisher information stability to detect convergence versus drift and restores the student toward the teacher's anchor when distributional shifts are identified. Our experiments on MELD, DFEW, and BAH under 0%-50% missing modalities show that this unified adaptation-restoration design consistently outperforms entropy-based adaptation, RAG, and perplexity-based generation over long adaptation horizons, where baselines without restoration progressively degrade while TTSD-FAR remains consistent.
Chinese Translation
大型视频语言模型（LVLMs）在多模态任务中表现出色，例如在真实环境中的多模态情感识别（ER）。情感识别本质上是多模态的，需要对面部表情、声音、语言、生物信号和手势进行联合理解。然而，实际部署仍然面临挑战：在测试时可能会缺失或噪声模态。部分观测可以视为相对于完整模态分布的分布偏移。基于熵最小化或困惑度降低的最新测试时适应（TTA）方法无法转移到自回归LVLMs，而当观察到的模态较弱时，检索增强生成（RAG）会退化。由于不存在真实标签监督来验证个体更新，因此在这一流中进行适应存在累积漂移的风险，一旦模型偏离可靠解决方案就会退化。因此，一个有效的解决方案必须适应任意缺失模态模式，并在持续适应过程中保持有效性。我们通过测试时自我蒸馏（TTSD）同时解决这两个问题，这是一种参数高效的框架，其中一个在完整模态上训练的冻结教师通过自我蒸馏指导一个适应性低秩学生，仅更新极少量参数。通过Fisher锚定恢复（FAR），在同一循环中构建稳定性，监测Fisher信息的稳定性以检测收敛与漂移，并在识别到分布偏移时将学生恢复到教师的锚点。我们在MELD、DFEW和BAH数据集上进行的实验显示，在0%-50%缺失模态的情况下，这种统一的适应-恢复设计在长时间适应过程中始终优于基于熵的适应、RAG和基于困惑度的生成，而没有恢复的基线逐渐退化，而TTSD-FAR保持一致。
cs.CV / 14 / 2608.18388
Depth Anything V4: Dynamic 4D Scene Reconstruction via Riemannian Flow Matching on 4D Gaussian Splatting
Depth Anything V4：通过在4D高斯点云上进行黎曼流匹配实现动态4D场景重建
Abstract
We present Depth Anything V4 (DAV4), a framework for dynamic 4D scene reconstruction from monocular video. Our key contribution is the application of Riemannian Flow Matching (RFM) to 4D Gaussian Splatting parameters, defining probability paths directly on non-Euclidean manifolds (scale, rotation, opacity), ensuring all intermediate states are valid. Through controlled experiments, we isolate RFM's contribution from test-time optimization (TTO) and pre-training. A deterministic MLP baseline with the same data, architecture, and TTO achieves F-score 0.762; RFM achieves 0.806 - the +0.044 gain is RFM's isolated contribution. We provide corrected computational cost analysis: pre-training is 360 GPU-hours, amortizing for large-scale deployment (over 10,000 scenes). Uncertainty is quantified via Negative Gaussian Log-Likelihood and Expected Calibration Error. DAV4 outperforms prior Depth Anything models and per-scene 4D-GS on dynamic reconstruction and novel-view synthesis, while using no human-annotated depth labels as training losses.
Chinese Translation
我们提出了Depth Anything V4（DAV4），这是一个从单目视频中进行动态4D场景重建的框架。我们的主要贡献是将黎曼流匹配（Riemannian Flow Matching, RFM）应用于4D高斯点云参数，直接在非欧几里得流形（尺度、旋转、不透明度）上定义概率路径，确保所有中间状态都是有效的。通过控制实验，我们将RFM的贡献与测试时优化（Test-Time Optimization, TTO）和预训练进行了隔离。使用相同的数据、架构和TTO的确定性多层感知机（MLP）基线达到了F-score 0.762；而RFM达到了0.806，+0.044的增益是RFM的独立贡献。我们提供了修正后的计算成本分析：预训练耗时360 GPU小时，适用于大规模部署（超过10,000个场景）。通过负高斯对数似然和期望校准误差量化不确定性。DAV4在动态重建和新视图合成方面优于之前的Depth Anything模型和每场景4D-GS，同时在训练损失中未使用任何人工标注的深度标签。
cs.CV / 15 / 2608.18399
What Does Attention Transfer Transfer? Attention Structure and Robustness in Vision Transformers
注意力转移传递了什么？视觉变换器中的注意力结构与鲁棒性
Abstract
Vision transformers (ViTs) trained to copy a pretrained teacher's attention maps recover most of fine-tuning's in-distribution accuracy yet fall measurably short of it under distribution shift, as recent work has shown. What the copy delivers has never been measured directly in the attention structure and tied to robustness. We build that instrumentation for ViT-S students of a self-supervised teacher on ImageNet-100, and report three findings that triangulate one conclusion. First, the transfer is essentially perfect and permanently so: the distilled student's attention ends up roughly two orders of magnitude closer to the teacher's than fine-tuning does, and does not drift with additional training. Second, the gap is real at 14$\times$ fewer parameters and 10$\times$ less data than previously studied, but it has a time axis. It tracks training maturity, and completing the schedules that the stopping rule interrupted closes it below our pre-registered threshold in two of three seeds, with comparisons at equal accuracy giving the same result. The endpoint gap at this scale is substantially a training-maturity artifact: robustness matures later than accuracy, and stopping rules tuned to accuracy undersample it. Third, forcing cross-row redundancy down by half the structural separation between the distilled and fine-tuned conditions produces no detectable robustness response under two registered ways of matching accuracy. Verified transfer, a gap that closes while the structure never moves, and a null under direct intervention are together consistent with the deficit residing in features, not in the visible attention structure. This is elimination plus intervention, and its scope is the regime we measured. In this regime, attention overlays show where a model looks, not what it knows.
Chinese Translation
训练以复制预训练教师注意力图的视觉变换器（ViTs）在恢复大多数微调的分布内准确性方面表现良好，但在分布转移下明显不足，正如最近的研究所示。复制所传递的内容从未在注意力结构中直接测量过，也未与鲁棒性联系起来。我们为在 ImageNet-100 上的自监督教师的 ViT-S 学生构建了这一测量工具，并报告了三个发现，这些发现共同指向一个结论。首先，转移基本上是完美的，并且是永久性的：提炼后的学生注意力与教师的注意力相比，最终大约接近两个数量级，而微调则未能达到，并且在额外训练中没有漂移。第二，在参数数量减少 14 倍和数据量减少 10 倍的情况下，差距是真实存在的，但它具有时间轴。它与训练成熟度相关，完成被停止规则中断的训练计划会在三个种子中的两个种子下将差距缩小到我们预注册的阈值以下，而在相同准确性下的比较也得出了相同的结果。在这个规模下的最终差距在很大程度上是一个训练成熟度的伪影：鲁棒性成熟的时间晚于准确性，而针对准确性调整的停止规则对鲁棒性进行了欠采样。第三，通过将提炼和微调条件之间的结构分离减少一半来强制跨行冗余，未能在两种注册的匹配准确性的方法下产生可检测的鲁棒性响应。经过验证的转移、在结构不变的情况下缩小的差距，以及在直接干预下的无响应共同表明，缺陷存在于特征中，而不是在可见的注意力结构中。这是消除加干预，其范围是我们所测量的范围。在这个范围内，注意力叠加显示了模型的关注点，而不是它所知道的内容。
cs.CV / 16 / 2608.18412
JSL-DC: A Word-Level Japanese Sign Language Dataset with Linguist-Derived Descriptions for Distinguishing Confusable Signs
JSL-DC：一个具有语言学家派生描述的词级日本手语数据集，用于区分易混淆的手势
Abstract
Effective sign language (SL) acquisition is crucial for deaf children, yet 95% are born to hearing parents who often lack proficiency in SL. SL recognition can power learning tools to help parents communicate with their children. However, Japanese Sign Language (JSL) lacks large-scale, multi-signer datasets, hindering the development of models that can generalize to new users. To address this gap, we introduce JSL-DC, the largest JSL dataset by video count, comprising 36.7K videos from 19 signers. The entire process was Deaf-centric: the lexicon comprising 270 JSL words was selected by Deaf and Coda linguists to facilitate parent-child communication, all participants were Deaf individuals who use JSL daily, and the data underwent a two-stage review process involving Deaf linguists. Moreover, we provide linguist-derived descriptions for distinguishing confusable signs. We demonstrate that the proposed model inspired by the descriptions outperforms state-of-the-art recognition methods by 9.8% on the confusable subset. The dataset, along with its linguistic description that inspires new models, will be released under a CC-BY 4.0 license to accelerate research in SL recognition.
Chinese Translation
有效的手语（SL）学习对聋哑儿童至关重要，然而95%的聋哑儿童出生于听力正常的父母家庭，这些父母通常缺乏手语的熟练掌握。手语识别可以为学习工具提供支持，帮助父母与孩子进行沟通。然而，日本手语（JSL）缺乏大规模的多签者数据集，这阻碍了能够推广到新用户的模型的开发。为了解决这一问题，我们推出了JSL-DC，这是目前视频数量最多的JSL数据集，包含来自19位手语者的36,700个视频。整个过程以聋人群体为中心：词汇表由270个JSL单词组成，经过聋人和Coda语言学家的选择，以促进亲子沟通，所有参与者均为日常使用JSL的聋人，并且数据经过了由聋人语言学家参与的两阶段审核过程。此外，我们提供了用于区分易混淆手势的语言学家派生描述。我们证明，受这些描述启发的模型在易混淆子集上的性能比最先进的识别方法提高了9.8%。该数据集及其激发新模型的语言描述将以CC-BY 4.0许可证发布，以加速手语识别领域的研究。
cs.CV / 17 / 2608.18413
CoMVS-GS: Collaborative Multi-View Stereo and 3D Gaussian Splatting for Surface Reconstruction
CoMVS-GS：用于表面重建的协作多视图立体与3D高斯点云技术
Abstract
3D Gaussian Splatting enables efficient novel view synthesis, but accurate mesh reconstruction remains difficult in weakly observed and occluded regions, where Gaussian primitives may grow into unstable or geometrically inconsistent structures. We propose CoMVS-GS, a general surface reconstruction framework that combines Multi-View Stereo with Gaussian splatting. CoMVS-GS initializes Gaussian primitives from dense multi-view stereo points with pre-flattened scales and normal-aligned orientations, providing stronger geometric priors than sparse structure-from-motion initialization and reducing ambiguity during early optimization. It further introduces PatchMatch-3DGS Mutual Supervision, where Gaussian-rendered depths and normals initialize PatchMatch refinement, and refined PatchMatch depths supervise Gaussian optimization to improve weakly constrained geometry. For surface extraction, CoMVS-GS replaces truncated signed distance field voxel fusion with a Delaunay graph-cut meshing pipeline, reducing sensitivity to voxel resolution while preserving visibility-consistent surface evidence. Experiments on DTU, GauU-Scene V2, and MatrixCity show that CoMVS-GS remains competitive on object-level reconstruction and improves geometric accuracy and mesh compactness in outdoor scenes while maintaining high rendering quality.
Chinese Translation
3D高斯点云技术能够高效地合成新视图，但在观察较弱和被遮挡区域的精确网格重建仍然困难，此时高斯原件可能会发展成不稳定或几何不一致的结构。我们提出了CoMVS-GS，一个将多视图立体与高斯点云相结合的通用表面重建框架。CoMVS-GS从密集的多视图立体点初始化高斯原件，采用预先平整的尺度和法线对齐的方向，提供比稀疏运动重建初始化更强的几何先验，并减少早期优化过程中的歧义。它进一步引入了PatchMatch-3DGS互监督机制，其中高斯渲染的深度和法线初始化PatchMatch细化，而细化后的PatchMatch深度则监督高斯优化，以改善弱约束几何的表现。对于表面提取，CoMVS-GS用德劳内图切割网格管道替代了截断有符号距离场体素融合，降低了对体素分辨率的敏感性，同时保留了与可见性一致的表面证据。在DTU、GauU-Scene V2和MatrixCity上的实验表明，CoMVS-GS在物体级重建方面保持竞争力，并在户外场景中提高了几何精度和网格紧凑性，同时保持高渲染质量。
cs.CV / 18 / 2608.18479
COSTA: A Cluster-Centric Paradigm for Annotation-Free Open-Set Semantic Segmentation of Aerial Point Clouds with Domain Shifts
COSTA：一种面向集群的无注释开放集语义分割范式，适用于具有领域偏移的航空点云
Abstract
Semantic segmentation of aerial point cloud is trapped in a generalization crisis under distinct domain shifts. While test-time adaptation offers a privacy-preserving and computationally efficient way to adapt pre-trained models to unlabeled target-domain data during inference, existing methods, bound to closed-set label assumptions and non-scalable point-wise segmentation pipelines, still struggle with semantic shifts. We ask: can we adapt any given pre-trained aerial point cloud segmentation model to a shifted target domain at the inference phase alone, without additional training, while segmenting target-specific categories beyond the source label space on demand? This paper introduces COSTA, which breaks this limitation by shifting from closed-set point-wise adaptation to cluster-centric open-set semantic propagation. Our core discovery is that, once effectively adapted at test time, the rich feature distribution of aerial point clouds can be distilled into a compact set of well-separated semantic centroids that are transferable across label spaces. COSTA leverages this to reformulate open-set semantic segmentation as a cluster-level propagating process: it first bridges the domain gap through proven test-time adaptation, then groups each batch of target-domain points into a small set of semantic clusters based on the similarity distribution in the adapted feature space, and finally propagates high-confidence pseudo labels obtained from an open-vocabulary vision-language model to all points through cluster-level voting. This cluster-centric paradigm enables test-time adaptation of aerial point clouds under significant domain gaps with mixed semantic shifts. With DALES as the source domain, COSTA enables on-demand segmentation across three aerial point cloud benchmarks with distinct domains and heterogeneous category spaces, achieving up to 70.09% mIoU under this new setting.
Chinese Translation
航空点云的语义分割在不同领域偏移下面临泛化危机。尽管测试时适应提供了一种隐私保护和计算高效的方法，可以在推理过程中将预训练模型适应于未标记的目标领域数据，但现有方法仍然受限于封闭集标签假设和不可扩展的逐点分割管道，在语义偏移方面仍然存在困难。我们提出一个问题：能否在推理阶段仅通过适应任何给定的预训练航空点云分割模型，而无需额外训练，同时根据需求对超出源标签空间的目标特定类别进行分割？本文介绍了COSTA，它通过从封闭集逐点适应转变为以集群为中心的开放集语义传播，打破了这一限制。我们的核心发现是，一旦在测试时有效适应，航空点云的丰富特征分布可以被提炼为一组紧凑的、良好分离的语义质心，这些质心可以跨标签空间迁移。COSTA利用这一点将开放集语义分割重新表述为一个集群级传播过程：首先通过经过验证的测试时适应弥合领域差距，然后根据适应特征空间中的相似性分布，将每批目标领域点分组为一小组语义集群，最后通过集群级投票将从开放词汇视觉-语言模型获得的高置信度伪标签传播到所有点。该以集群为中心的范式使得在显著领域差距和混合语义偏移下对航空点云进行测试时适应成为可能。在DALES作为源领域的情况下，COSTA能够在三个具有不同领域和异构类别空间的航空点云基准上按需进行分割，在这一新设置下实现高达70.09%的mIoU。
cs.CV / 19 / 2608.18484
Partition the Support, Reconstruct the Residual: Training-Free Sparse Attention for Video Generation and World Models
划分支持，重构残差：无训练稀疏注意力用于视频生成和世界模型
Abstract
Training-free block-sparse attention can accelerate video transformers, but row-wise attention concentration does not by itself specify an executable sparse operator. Queries sharing a block route may have poorly overlapping supports, while retained attention mass alone does not determine the post-softmax error from skipped interactions. We show that partition geometry affects both pooled support and the predictability of the remaining residual from the sparse output. We introduce SparsePR, which combines Response-Coupled Partitioning with Probe-Fitted Residual Reconstruction. Sampled-query key responses form paired K/V groups, whose centroids induce query-response coordinates for shared routing. A small set of exact query rows then calibrates a call-specific affine correction from the sparse output within the output subspace observed in the probe residuals. Across four heterogeneous video generation and world models, SparsePR consistently reduces attention-reconstruction error. Ablations show that probe fitting accounts for most of this reduction, while response-coupled partitioning lowers hard-drop error and improves reconstruction under a finite probe budget. SparsePR preserves generation quality at 22.0-26.0% realized executed-pair density while achieving 1.48x-2.61x end-to-end speedups. Project page: https://pardistaghavi.github.io/SparsePR-website/
Chinese Translation
无训练的块稀疏注意力可以加速视频变换器，但行级注意力集中本身并不能指定可执行的稀疏算子。共享块路由的查询可能具有重叠支持较差的情况，而保留的注意力质量本身并不能决定因跳过交互而导致的后软最大误差。我们展示了划分几何形状如何影响汇聚支持和从稀疏输出中预测剩余残差的可预测性。我们提出了SparsePR，它结合了响应耦合划分和探测拟合残差重构。采样查询的键响应形成成对的K/V组，其质心诱导共享路由的查询-响应坐标。一小组精确的查询行随后在探测残差中校准来自稀疏输出的调用特定仿射修正。在四个异构视频生成和世界模型中，SparsePR始终降低了注意力重构误差。消融实验表明，探测拟合占据了大部分减少，而响应耦合划分降低了硬丢失误差，并在有限的探测预算下改善了重构。SparsePR在实现22.0%-26.0%的实际执行对密度的同时，保持生成质量，并实现了1.48倍至2.61倍的端到端加速。项目页面：https://pardistaghavi.github.io/SparsePR-website/
cs.CV / 20 / 2608.18498
DyG$^2$T: Modeling Object Dynamics with 3D Gaussian Temporal-Spatial Particle Graph Transformer
DyG$^2$T：使用3D高斯时空粒子图变换器建模物体动态
Abstract
Modeling object dynamics from limited visual observations is a fundamental problem for enabling accurate motion trajectory prediction in embodied interaction scenarios. Existing dynamics modeling methods first compress reconstructed particle representations into sparse Key Points and model their evolution using locally constrained interactions, thereby discarding fine-grained local details and obscuring discriminative interaction modeling across spatial and temporal scales, leading to drifting trajectories and inaccurate appearance prediction. To tackle these issues, we propose DyG$^2$T, a dynamics modeling framework that infers object motion trajectories by spatially completing and temporally discriminating Key Point representations and modeling multi-scale interaction over particle graphs. Spatially, DyG$^2$T enriches each Key Point by aggregating neighboring raw particle positions to recover fine-grained local details, while explicitly encoding relative offsets among Key Points to enhance geometric structure perception. Temporally, we introduce a Temporal Disentangling Network (TDN) to identify dominant cross-frame variations in latent space and amplify inter-frame differences, yielding temporally discriminative representations that are subsequently aggregated via Temporal Attention to capture frame-wise temporal evolution cues. For comprehensive interaction modeling, a Particle Graph Transformer leverages global attention to preserve discriminative long-range dependencies among Key Points, mitigating representation homogenization induced by locality-constrained modeling and providing a robust basis for accurate trajectory prediction. Experiments on both synthetic and real-world datasets demonstrate that DyG$^2$T achieves accurate dynamics modeling and reasoning, and exhibits strong cross-object and real-world generalization.
Chinese Translation
从有限的视觉观察中建模物体动态是实现具身交互场景中准确运动轨迹预测的一个基本问题。现有的动态建模方法首先将重建的粒子表示压缩为稀疏关键点，并通过局部约束交互建模它们的演变，从而丢弃了细粒度的局部细节，并模糊了跨空间和时间尺度的区分性交互建模，导致轨迹漂移和外观预测不准确。为了解决这些问题，我们提出了DyG$^2$T，一个动态建模框架，通过空间补全和时间区分关键点表示来推断物体运动轨迹，并在粒子图上建模多尺度交互。在空间上，DyG$^2$T通过聚合邻近的原始粒子位置来丰富每个关键点，以恢复细粒度的局部细节，同时显式编码关键点之间的相对偏移，以增强几何结构感知。在时间上，我们引入了时间解耦网络（Temporal Disentangling Network, TDN）来识别潜在空间中的主导跨帧变化并放大帧间差异，从而产生时间区分的表示，这些表示随后通过时间注意力（Temporal Attention）聚合，以捕捉逐帧的时间演变线索。为了全面的交互建模，粒子图变换器利用全局注意力来保留关键点之间的区分性长程依赖，减轻了局部约束建模引起的表示同质化，并为准确的轨迹预测提供了稳健的基础。在合成和真实世界数据集上的实验表明，DyG$^2$T实现了准确的动态建模和推理，并表现出强大的跨对象和真实世界的泛化能力。
cs.CV / 21 / 2608.18515
Cross-Modal MRI Ovary Segmentation in Endometriosis Using Unpaired TVUS Prototype Priors
基于无配对经阴道超声原型先验的跨模态MRI卵巢分割在子宫内膜异位症中的应用
Abstract
Transvaginal ultrasound (TVUS) and magnetic resonance imaging (MRI) provide complementary information for endometriosis image analysis, yet existing studies mainly focus on single-modality analysis or disease classification, leaving cross-modal ovarian segmentation largely unexplored. In this work, to tackle the increased difficulty of ovary segmentation in MRI due to ovaries' small target size and ambiguous boundaries with surrounding pelvic structures, we propose a dual branch framework for ovary segmentation across TVUS and MRI. More specifically, by adapting MedSAM3 with TVUS-derived prototype bank, we aim to align anatomically consistent feature representations across both modalities. Extensive experiments are conducted on endometriosis-related TVUS and MRI datasets. We observe quantitative and qualitative improvements of over 5 percentage points for the proposed dual-branch approach compared with multiple state-of-the-art methods. Furthermore, our ablation study shows the contribution of individual components such as the prototype bank and the importance of warm-up pretraining in the source TVUS domain.
Chinese Translation
经阴道超声（TVUS）和磁共振成像（MRI）为子宫内膜异位症的图像分析提供了互补的信息，但现有研究主要集中在单一模态分析或疾病分类上，跨模态卵巢分割尚未得到充分探索。在本研究中，为了应对MRI中卵巢分割的难度增加，主要由于卵巢的目标尺寸小且与周围盆腔结构的边界模糊，我们提出了一种双分支框架，用于跨TVUS和MRI的卵巢分割。更具体地说，通过将MedSAM3与基于TVUS的原型库相结合，我们旨在对齐两个模态之间解剖上一致的特征表示。在与子宫内膜异位症相关的TVUS和MRI数据集上进行了广泛的实验。我们观察到，与多种最先进的方法相比，所提出的双分支方法在定量和定性上均提高了超过5个百分点。此外，我们的消融研究显示了各个组件的贡献，如原型库的作用以及在源TVUS领域进行预热预训练的重要性。
cs.CV / 22 / 2608.18516
OptiModNet: A UNet-Transformer Hybrid with Grouped-Query and Channel Attention for Optic Disc and Cup Segmentation
OptiModNet：一种具有分组查询和通道注意力的UNet-Transformer混合模型，用于视盘和视杯分割
Abstract
Precise segmentation of the optic disc and cup is critical for the early detection and diagnosis of glaucoma. However, achieving consistently high performance across datasets while maintaining low computational requirements remains a significant challenge. In glaucoma detection, low-computation methods are crucial for enabling rapid, large-scale screening and facilitating deployment in resource-limited clinical environments. While deep learning models such as UNets, Vision Transformers (ViTs), and Diffusion models have demonstrated strong segmentation performance but these methods often come with substantial computational overhead. UNets are efficient at capturing local features but are limited in modeling global contextual information. Conversely, ViTs excel at long-range dependency modeling but are computationally intensive. Hybrid architectures, such as UNetR, which combine transformer-based encoders with UNet-style decoders, have shown improved performance but while incurring additional complexity. Considering these, in this work, we propose OptiModNet, a light weight novel hybrid architecture tailored for optic disc and cup segmentation. The model integrates diverse attention mechanisms at multiple stages of the network to enhance both local and global feature representation. We include an Aggregated Pyramid Loss that supervises predictions at multiple decoder depths, to promote better gradient flow and structural consistency. We evaluate OptiModNet on the REFUGE2 dataset for both optic disc and cup segmentation tasks. Our method achieves state-of-the-art performance, exceeding existing approaches by over 2.5\%, while maintaining high efficiency with only 3.73 GFLOPs and 1.93M parameters. The code is available at https://github.com/SG1947/OptiModNet.
Chinese Translation
精确的视盘和视杯分割对于青光眼的早期检测和诊断至关重要。然而，在保持低计算需求的同时，在不同数据集上实现一致的高性能仍然是一个重大挑战。在青光眼检测中，低计算方法对于快速、大规模筛查以及在资源有限的临床环境中部署至关重要。尽管深度学习模型如UNet、视觉变换器（Vision Transformers, ViTs）和扩散模型在分割性能上表现出色，但这些方法通常伴随着显著的计算开销。UNet在捕捉局部特征方面效率高，但在建模全局上下文信息时受到限制。相反，ViT在长距离依赖建模方面表现优异，但计算密集。混合架构，如UNetR，结合了基于变换器的编码器和UNet风格的解码器，虽然显示出改进的性能，但也增加了复杂性。考虑到这些因素，在本研究中，我们提出了OptiModNet，一种针对视盘和视杯分割量身定制的轻量级新型混合架构。该模型在网络的多个阶段集成了多样的注意力机制，以增强局部和全局特征表示。我们引入了一种聚合金字塔损失（Aggregated Pyramid Loss），在多个解码器深度上监督预测，以促进更好的梯度流动和结构一致性。我们在REFUGE2数据集上评估了OptiModNet在视盘和视杯分割任务中的表现。我们的方法实现了最先进的性能，超过现有方法超过2.5%，同时在仅有3.73 GFLOPs和1.93M参数的情况下保持高效。代码可在 https://github.com/SG1947/OptiModNet 获取。
cs.CV / 23 / 2608.18523
Prior-Conditioned Gaussian Discriminants for Generalizable AI-generated Image Detection
用于可泛化的AI生成图像检测的先验条件高斯判别器
Abstract
Diffusion-based generators have made synthetic images ubiquitous, but detectors often fail under simultaneous shifts in generator, prompt/style, and source-domain. We study AI-generated image detection as a transfer system described by training prior, frozen encoder feature space, and decision rule, and ask when classifier head training adds value beyond what is already separable in modern features. As a controlled diagnostic, we fit a prior-conditioned Gaussian discriminant ladder: closed-form heads built from first- and second-order feature statistics under nested covariance assumptions. On Percept-Lens, a unified protocol over 39 public datasets (7.1 million images), the best rung is frequently competitive with, and sometimes exceeds, released AI-generated image detector heads when matched on both prior and encoder. We further quantify strong sensitivity to the training prior, data-efficiency of moment-based heads, and representation dependence of Gaussian shift metrics, motivating (prior, encoder, head)-level reporting and stronger analytical baselines for AIGI transfer.
Chinese Translation
基于扩散的生成器使合成图像无处不在，但检测器在生成器、提示/风格和源域同时发生变化时往往失效。我们将AI生成图像检测视为一个转移系统，描述为训练先验、冻结的编码器特征空间和决策规则，并探讨分类器头的训练何时能在现代特征中提供超出已可分离部分的价值。作为一种受控诊断，我们拟合了一个先验条件高斯判别梯度：在嵌套协方差假设下，由一阶和二阶特征统计构建的封闭形式头。在Percept-Lens上，这是一个涵盖39个公共数据集（710万张图像）的统一协议，最佳梯级在与先验和编码器匹配时，常常与已发布的AI生成图像检测头竞争，甚至有时超越它们。我们进一步量化了对训练先验的强敏感性、基于矩的头的数据信效性，以及高斯偏移度量的表示依赖性，推动了（先验、编码器、头）级别的报告和更强的分析基准，以支持AIGI转移。
cs.CV / 24 / 2608.18532
StateTrace: An Object-Centric Framework for Hidden-State Spatiotemporal Reasoning in Long Videos
StateTrace：一种面向对象的框架用于长视频中的隐状态时空推理
Abstract
Existing VLMs have achieved strong performance in video understanding, yet they struggle with long-video spatiotemporal reasoning when target objects become invisible, often mistaking "invisible" for "unknown". We define this challenge as hidden-state spatiotemporal reasoning: inferring object states during prolonged invisible intervals from context interactions. To address this, we propose StateTrace, a novel object-centric framework that endows VideoLLMs with an explicit mechanism for hidden state reasoning in long videos. StateTrace builds a reusable spatiotemporal state memory that organizes object trajectories, inter-object relations, and state-transition events into a structured reasoning substrate. At inference time, it retrieves question-relevant state-evolution trajectories and converts them into compact reasoning cues, enabling the model to explicitly reason about why an object disappears, how its state evolves while invisible, and whether that state should persist at query time. We further build HSR-Bench, a diagnostic benchmark for hidden-state reasoning, containing 1,427 video-QA samples from 1,384 unique videos. Extensive experiments across multiple VideoLLMs show that StateTrace consistently improves performance on both public benchmarks and HSR-Bench (e.g., improving VideoLLaMA3 from 39.6 to 64.2 on HSR-Bench).
Chinese Translation
现有的视频语言模型（VLMs）在视频理解方面取得了良好的表现，但在目标对象变得不可见时，它们在长视频的时空推理方面仍然存在困难，常常将“不可见”误解为“未知”。我们将这一挑战定义为隐状态时空推理：在长时间不可见的间隔中，从上下文交互中推断对象状态。为了解决这个问题，我们提出了StateTrace，这是一种新颖的面向对象的框架，为视频语言模型（VideoLLMs）提供了在长视频中进行隐状态推理的显式机制。StateTrace构建了一个可重用的时空状态记忆，将对象轨迹、对象间关系和状态转移事件组织成一个结构化的推理基础。在推理时，它检索与问题相关的状态演变轨迹，并将其转换为紧凑的推理线索，使模型能够明确推理出对象消失的原因、在不可见状态下其状态如何演变，以及在查询时该状态是否应持续。我们进一步构建了HSR-Bench，这是一个用于隐状态推理的诊断基准，包含来自1,384个独特视频的1,427个视频问答样本。在多个视频语言模型上的广泛实验表明，StateTrace在公共基准和HSR-Bench上均持续提高了性能（例如，将VideoLLaMA3在HSR-Bench上的表现从39.6提升至64.2）。
cs.CV / 25 / 2608.18544
Zero-Shot SAM2 Segmentation and Vision Transformer-Based Recognition of Elamite Cuneiform Symbols from Degraded Tablet Images
基于零样本SAM2分割和视觉变换器的埃兰米特楔形文字符号识别：来自退化平板图像的研究
Abstract
Automated recognition of ancient cuneiform script poses a compound signal-degradation problem: the three-dimensional relief of clay tablets creates spatially varying illumination and cast shadows, surface erosion introduces structured noise that overlaps with genuine sign impressions, and severe class imbalance across 141 sign categories undermines classifier reliability. We introduce EpigraphNet, a segmentation-guided transformer pipeline evaluated on the Persepolis Fortification Archive. From 1,239 annotated tablet images, brightness-adaptive morphological preprocessing and zero-shot SAM2-Large segmentation generate clean binary symbol masks, which a fine-tuned Vision Transformer (ViT-B/16) with inverse-frequency class weighting then classifies. EpigraphNet reaches 86.41% top-1 accuracy on a 132-class benchmark, a 17.21 percentage-point gain over the strongest CNN baseline (ResNet-101, 69.20%) and 5.31-12.91% over four modern backbones (DeiT-B/16, Swin-B, ConvNeXt-B, EfficientNet-B4) under identical conditions. The full pipeline runs at approximately 18 ms per sign on an NVIDIA A100 GPU. A lower Spearman correlation between sign frequency and per-class performance indicates more balanced recognition across frequent and rare classes. Implementation is available at: github.com/r11up/sam-guided-vit
Chinese Translation
古代楔形文字的自动识别面临复合信号退化问题：粘土平板的三维浮雕造成空间变化的照明和投影阴影，表面侵蚀引入的结构噪声与真实的符号印记重叠，并且在141个符号类别之间存在严重的类别不平衡，削弱了分类器的可靠性。我们提出了EpigraphNet，一个基于分割引导的变换器管道，在波斯波利斯防御档案上进行评估。从1,239个标注的平板图像中，亮度自适应形态学预处理和零样本SAM2-Large分割生成干净的二进制符号掩模，随后经过逆频率类别加权的精细调优视觉变换器（ViT-B/16）进行分类。EpigraphNet在132类基准测试中达到86.41%的顶级准确率，比最强的CNN基线（ResNet-101，69.20%）提高了17.21个百分点，并在相同条件下比四个现代骨干网络（DeiT-B/16、Swin-B、ConvNeXt-B、EfficientNet-B4）提高了5.31-12.91%。整个管道在NVIDIA A100 GPU上每个符号运行约18毫秒。符号频率与每类性能之间较低的斯皮尔曼相关性表明在频繁和稀有类别之间的识别更加平衡。实现代码可在以下链接获取：github.com/r11up/sam-guided-vit
cs.CV / 26 / 2608.18573
PATE-Forensics: Perception-as-Tool for Explainable Deepfake Forensics with General-Purpose MLLMs
PATE-Forensics：基于感知的工具用于可解释的深伪取证，结合通用多语言大模型
Abstract
Existing explainable deepfake forensic methods typically rely on task-adapted MLLM to jointly address detection, localization, and explanation. Inspired by agent-style tool use, we instead introduce a Perception-as-Tool paradigm and instantiate it as PATE-Forensics, which architecturally decouples detection and localization from explanation generation while coupling detection and localization as tightly as possible within a forensic perception tool. The DINOv3-based tool couples a multi-granularity detection module that integrates global, patch-level, and segment-level evidence with a cue-guided localization module by spatializing the patch-level and segment-level evidence into forgery score maps that guide dense mask prediction. The original image and forensic perception outputs produced by the tool form structured forensic context for a general-purpose MLLM, which is guided by prompt constraints to generate explanations without task-specific fine-tuning. On DDL-X Track 3, PATE-Forensics achieves the best official score of 0.89, outperforming the second-ranked team by 0.19 points. Our code is available at https://github.com/yqli00000/PATE-Forensics.
Chinese Translation
现有的可解释深伪取证方法通常依赖于任务适应的多语言大模型（MLLM）来共同解决检测、定位和解释问题。受到代理式工具使用的启发，我们提出了一种基于感知的工具范式，并将其具体化为PATE-Forensics，该方法在架构上将检测和定位与解释生成解耦，同时尽可能紧密地耦合检测和定位于一个取证感知工具中。基于DINOv3的工具结合了一个多粒度检测模块，该模块整合了全局、补丁级和分段级证据，并通过将补丁级和分段级证据空间化为伪造评分图来引导密集掩码预测。工具生成的原始图像和取证感知输出为通用多语言大模型（MLLM）形成了结构化的取证上下文，该模型在提示约束的指导下生成解释，而无需特定任务的微调。在DDL-X Track 3中，PATE-Forensics取得了最佳官方得分0.89，领先第二名团队0.19分。我们的代码可在 https://github.com/yqli00000/PATE-Forensics 获取。
cs.CV / 27 / 2608.18579
MR-IQA-2: Faithful Image Quality Reflection via Fine-Grained Credit Assignment
MR-IQA-2：通过细粒度信用分配忠实反映图像质量
Abstract
Multimodal large language models (MLLMs) have shown strong potential for image quality assessment (IQA) by improving consistency between quality ratings and their underlying reasoning. However, most approaches supervise reasoning through human-provided ratings and rarely examine whether it faithfully reflects image quality. Rating accuracy alone does not ensure faithful reasoning; a shared reward also obscures supervision sources and may reinforce unfaithful reasoning when a correct rating occurs by chance. To improve the faithfulness and reliability of blind IQA, we aim to (1) decouple credit assignment for reasoning and rating and (2) provide verifiable supervision for faithful reasoning. We introduce MR-IQA-2, an actor-editor-judge framework that operationalizes reasoning-editing-reflection. The actor generates quality reasoning for an input image, and the editor revises the image according to the identified quality factors. A frozen judge compares the original and edited images and provides reflective supervision for the actor's reasoning. MR-IQA-2 further uses fine-grained credit assignment to decouple reasoning and rating supervision. Judge feedback supervises reasoning, whereas human ratings supervise the predicted rating. Masked token-specific updates distinguish these signals while preserving the causal relation from reasoning to rating. Across IQA benchmarks, MR-IQA-2 achieves competitive rating alignment with humans. Visual reflection also enables richer and more faithful visual understanding beyond rating, which may inform image-quality optimization and related downstream tasks. Code is available at https://github.com/RobinY99/MR-IQA-2.
Chinese Translation
多模态大型语言模型（MLLMs）在图像质量评估（IQA）方面展现出强大的潜力，通过提高质量评分与其基础推理之间的一致性。然而，大多数方法通过人类提供的评分来监督推理，且很少检查其是否忠实反映图像质量。仅凭评分的准确性并不能确保推理的忠实性；共享奖励也会模糊监督来源，并可能在正确评分偶然发生时强化不忠实的推理。为了提高盲评 IQA 的忠实性和可靠性，我们的目标是（1）解耦推理和评分的信用分配，以及（2）为忠实推理提供可验证的监督。我们提出了 MR-IQA-2，一个实施推理-编辑-反映的演员-编辑-评判框架。演员为输入图像生成质量推理，编辑根据识别的质量因素修订图像。一个固定的评判者比较原始图像和编辑后的图像，并为演员的推理提供反思性监督。MR-IQA-2 进一步使用细粒度信用分配来解耦推理和评分监督。评判者的反馈监督推理，而人类评分则监督预测评分。掩码的特定令牌更新区分这些信号，同时保持推理与评分之间的因果关系。在 IQA 基准测试中，MR-IQA-2 实现了与人类的竞争性评分一致性。视觉反映还使得超越评分的更丰富和更忠实的视觉理解成为可能，这可能为图像质量优化及相关下游任务提供信息。代码可在 https://github.com/RobinY99/MR-IQA-2 获取。
cs.CV / 28 / 2608.18585
SPARC: Subspace Position-Aware Robust Few-Shot Calibration for Distribution-Shifted Industrial Anomaly Detection
Abstract
Vision-based industrial anomaly detectors are calibrated on one distribution but may be deployed on another that differs in illumination, fixture placement, or sensor characteristics, sharply degrading an otherwise accurate detector. Adapting to the incoming lot is a natural response, but labeled anomalies are scarce. We therefore consider calibration using only a handful of verified-normal images available before scoring the rest of the lot. Existing fixes require backpropagation, detector-specific tuning, or choices about feature directions that few calibration samples cannot justify. We present SPARC, a few-shot calibration method that intercepts patch features between encoder and detector and removes a closed-form, spatially indexed estimate of deployment-time nuisance through per-cell subspace projection. It needs only $k \le 8$ verified-normal images and uses the algebraic saturation rank $r{=}k{-}1$ on the encoder's native patch grid. The correction requires no gradient or weight updates and works with memory-bank, density, prototype, and mutual detectors. On the shift-prone benchmarks, SPARC improves pooled Image AUROC and AU-PRO$_{0.3}$ for all seven detectors whose image scores depend on corrected patch features by $+13.8$ and $+3.5$ percentage points (pp), respectively; on benchmarks without engineered shift, the changes are small and mixed. Controls that give competing corrections the same calibration images attribute these gains to the per-cell subspace structure rather than the images alone. Further ablations support the saturation-rank choice and characterize sensitivity to backbone and calibration conditions.
cs.CV / 29 / 2608.18586
OmniHandwritingOCR: A Diagnostic Benchmark for Evaluating Multimodal LLMs in Handwritten OCR Scenarios
OmniHandwritingOCR：评估多模态大语言模型在手写OCR场景中的诊断基准
Abstract
Multimodal large language models (MLLMs) are increasingly used as OCR systems in document and knowledge-processing pipelines, but their ability to faithfully read real handwriting remains underexplored. Existing OCR benchmarks focus largely on printed text or clean single-line inputs, leaving limited coverage of realistic handwritten OCR scenarios such as multilingual handwriting, writer errors, and structurally complex mathematical expressions. We introduce OmniHandwritingOCR, a diagnostic benchmark for evaluating MLLMs and OCR systems on handwritten OCR. It covers handwritten text recognition and handwritten mathematical expression recognition across six subtasks and twelve subsets, totaling 77.57K labeled images from public datasets and newly collected student writings. A key component is a difficulty-stratified multi-line formula corpus designed to test robustness under increasing structural complexity. We evaluate thirteen open- and closed-source systems with five complementary metrics under a unified protocol. Results show that current systems remain far from faithful transcription: performance drops sharply on complex multi-line formulas, model rankings vary across language and formula settings, and several generative models hallucinate plausible but visually unsupported corrections. OmniHandwritingOCR provides a challenging testbed for diagnosing language, content, structural, and visual-grounding failure modes of multimodal models in handwritten OCR scenarios.
Chinese Translation
多模态大语言模型（MLLMs）越来越多地被用作文档和知识处理流程中的OCR系统，但它们忠实读取真实手写文本的能力仍然未得到充分探索。现有的OCR基准主要集中在印刷文本或干净的单行输入上，有限地覆盖了现实手写OCR场景，如多语言手写、书写者错误和结构复杂的数学表达式。我们推出了OmniHandwritingOCR，这是一个用于评估MLLMs和OCR系统在手写OCR方面的诊断基准。它涵盖了手写文本识别和手写数学表达式识别，共包括六个子任务和十二个子集，总计来自公共数据集和新收集的学生书写的77.57K标注图像。一个关键组成部分是一个难度分层的多行公式语料库，旨在测试在日益增加的结构复杂性下的鲁棒性。我们在统一协议下使用五个互补指标评估了十三个开源和闭源系统。结果表明，当前系统在忠实转录方面仍然远未达到要求：在复杂的多行公式上性能急剧下降，模型排名在不同语言和公式设置中变化，且一些生成模型产生了看似合理但在视觉上不支持的修正。OmniHandwritingOCR为诊断多模态模型在手写OCR场景中的语言、内容、结构和视觉基础失败模式提供了一个具有挑战性的测试平台。
cs.CV / 30 / 2608.18590
FD-CanKD: Frequency-Decoupled Cross-Attention Distillation as a Refinement Prior for Compact Object Detectors
FD-CanKD：作为紧凑型目标检测器精炼先验的频率解耦交叉注意力蒸馏
Abstract
Compact object detectors are suitable for resource-constrained visual perception, but their limited representation capacity creates an accuracy gap relative to large models. Conventional detector distillation often relies on prediction-level supervision or a single feature-alignment target, such as response, distribution, correlation, or frequency-domain matching. Frequency-Decoupled Cross-Attention Knowledge Distillation (FD-CanKD) is presented as a detector-oriented framework that transfers teacher knowledge at three complementary levels: head-level prediction supervision, relation-level non-local context transfer, and frequency-level component-selective alignment. Student features first aggregate teacher-side spatial context through cross-attention-based relation transfer, after which frequency-aware alignment preserves complementary structural and detail-sensitive cues. Under controlled Microsoft Common Objects in Context (COCO) experiments, fixed 50-epoch from-scratch comparisons show that FD-CanKD remains competitive with representative detector knowledge distillation baselines. Post-distillation continued fine-tuning further produces a stronger refinement-ready student than detector-only fine-tuning, reaching 48.87 mean average precision (mAP) at intersection-over-union thresholds from 0.50 to 0.95 (mAP50:95), 65.84 mAP50, and 53.40 mAP75 after 20 additional epochs. All distillation modules are removed after training, leaving the deployed student unchanged at 19.7M parameters. The framework is instantiated and evaluated in a controlled YOLOv12 teacher-student setting as a representative compact-detector case study.
Chinese Translation
紧凑型目标检测器适用于资源受限的视觉感知，但其有限的表示能力导致相较于大型模型存在准确性差距。传统的检测器蒸馏通常依赖于预测级监督或单一特征对齐目标，如响应、分布、相关性或频域匹配。频率解耦交叉注意力知识蒸馏（FD-CanKD）被提出作为一个面向检测器的框架，在三个互补层面上转移教师知识：头级预测监督、关系级非局部上下文转移和频率级组件选择性对齐。学生特征首先通过基于交叉注意力的关系转移聚合教师端的空间上下文，随后频率感知对齐保留互补的结构和细节敏感线索。在受控的微软通用对象上下文（COCO）实验中，固定50个周期的从零开始比较显示，FD-CanKD在代表性检测器知识蒸馏基线中保持竞争力。蒸馏后的继续微调进一步产生了一个比仅使用检测器微调更强的准备精炼的学生，在交并比阈值从0.50到0.95（mAP50:95）下达到48.87的平均精度（mAP），65.84的mAP50和53.40的mAP75，经过20个额外周期后。所有蒸馏模块在训练后被移除，部署的学生参数保持不变，为19.7M。该框架在受控的YOLOv12教师-学生设置中实例化并评估，作为一个代表性的紧凑型检测器案例研究。
cs.CV / 31 / 2608.18593
ReX-Shot: Single-Image Rephotography via Geometry- and Camera-Grounded Generation
ReX-Shot：基于几何和相机的单图像再摄影生成
Abstract
Single-image rephotography aims to synthesize new shots of a scene from a single reference image with specified viewpoints, focal lengths, and photographic effects, which are intrinsically coupled in imaging. Existing methods typically treat these factors separately and struggle under joint control: novel-view synthesis may introduce geometric distortions under focal-length changes, while super-resolution and instruction-guided editing remain confined to 2D and cannot reliably extend detail restoration or appearance control to novel viewpoints. We attribute these limitations to imperfect single-image 3D reconstruction and the sampling limit of continuous focal-length enlargement. To reduce projection bias from geometric errors, we use implicitly transformed foundation-model features for robust target-view guidance. We further formulate focal-length enlargement as a geometry-guided super-resolution problem and exploit generative detail priors to recover details lost during sparse 3D resampling. Built on this 3D-aware generative backbone, we lift photographic-effect control from 2D filtering to 3D-aware appearance editing, preserving content consistency across viewpoints and focal lengths. These components form ReX-Shot, a geometry- and camera-grounded generative framework for single-image rephotography. To our knowledge, ReX-Shot is the first unified framework to jointly control viewpoint, focal length, and parameterized photographic effects from a single image. Experiments show that ReX-Shot outperforms representative baselines across all three controls while enabling near-real-time interactive rephotography.
Chinese Translation
单图像再摄影旨在从单一参考图像合成具有指定视点、焦距和摄影效果的新场景图像，这些因素在成像中本质上是相互关联的。现有方法通常将这些因素分开处理，并在联合控制下面临挑战：新视角合成可能在焦距变化下引入几何失真，而超分辨率和基于指令的编辑仍然局限于二维，无法可靠地将细节恢复或外观控制扩展到新视点。我们将这些限制归因于单图像三维重建的不完美和连续焦距放大的采样限制。为了减少几何误差带来的投影偏差，我们使用隐式转换的基础模型特征进行稳健的目标视图引导。我们进一步将焦距放大形式化为一个几何引导的超分辨率问题，并利用生成细节先验来恢复在稀疏三维重采样过程中丢失的细节。基于这一具有三维感知的生成骨架，我们将摄影效果控制从二维滤波提升到三维感知的外观编辑，保持不同视点和焦距之间的内容一致性。这些组件构成了ReX-Shot，一个基于几何和相机的单图像再摄影生成框架。据我们所知，ReX-Shot是第一个统一框架，能够从单一图像中联合控制视点、焦距和参数化的摄影效果。实验表明，ReX-Shot在所有三种控制下均优于代表性基线，同时实现近实时的交互式再摄影。
cs.CV / 32 / 2608.18602
Teach a Molmo2Fish: Towards interactive fish tracking with natural language guidance
教一个Molmo2Fish：朝着自然语言指导的互动鱼类追踪迈进
Abstract
Computer vision is increasingly used to automate recognition tasks in large ecological datasets, but more complex tasks such as multi-object tracking continue to pose challenges. As researchers seek to incorporate vision models in ecology workflows, various lines of research have explored how to make imperfect predictions useful through human-in-the-loop processes. We propose a new approach to working with imperfect tracking predictions through an interactive prediction correction workflow taking place as a conversation with a multimodal large language model, which we tailor to a sonar fish tracking dataset as an initial proof of concept. We investigate the performance of the tool, Molmo2Fish, across guided and unguided tasks, correcting its own predicted tracks and external tracks. We find that Molmo2Fish achieves high performance on fish tracking and track correction tasks, but there is still much room to improve on incorporating natural language guidance. The code and data are publicly available at https://github.com/tidalove/molmo2fish.
Chinese Translation
计算机视觉在自动化识别大型生态数据集中的任务中越来越多地被应用，但更复杂的任务，如多目标追踪，仍然面临挑战。随着研究人员寻求将视觉模型纳入生态工作流程，各种研究方向探索了如何通过人机交互过程使不完美的预测变得有用。我们提出了一种新的方法，通过与多模态大型语言模型的对话进行互动预测修正工作流程，以处理不完美的追踪预测，我们将其应用于声纳鱼类追踪数据集，作为初步的概念验证。我们调查了工具Molmo2Fish在引导和非引导任务中的表现，修正其自身的预测轨迹和外部轨迹。我们发现，Molmo2Fish在鱼类追踪和轨迹修正任务中表现出色，但在整合自然语言指导方面仍有很大的改进空间。代码和数据可在 https://github.com/tidalove/molmo2fish 获取。
cs.CV / 33 / 2608.18607
VA-Judger: Reward Modeling from Human Preference Feedback for Joint Video-Audio Generation
VA-Judger：基于人类偏好反馈的奖励建模用于联合视频-音频生成
Abstract
Using reinforcement learning to post-train joint video-audio generation models requires a reward signal. Existing methods construct this reward by combining metrics for individual quality dimensions, including audio quality, visual fidelity, and synchronization. However, these metrics evaluate perceptual dimensions separately and fail to capture the overall semantic and temporal coherence among the text prompt, video, and audio that shapes human preferences. Optimizing models against these metrics encourages reward hacking, generating video-audio content that achieves high scores on these metrics yet appears incoherent or unfaithful to human viewers. To address this problem, we first construct a large-scale human-preference dataset VAPref-10K for joint video-audio generation, comprising 9K prompts and 10.3K fine-grained paired comparisons from open-source generation models. We also introduce the VA-Judger-Bench benchmark with both in-domain and out-of-domain model comparisons to evaluate whether reward models truly align with human preferences. We further propose VA-Judger, a chain-of-thought omni-reward model for joint video-audio generation. In particular, VA-Judger first learns from pairs with clear quality gaps to establish structured output and coarse preference discrimination, then distills reliable preference explanations for harder near-quality comparisons via rejection sampling verified against human annotations, and finally performs dimension-wise reinforcement learning that decomposes human feedback into individual quality dimensions for denser reward signals than a single binary preference label. Experiments show that VA-Judger outperforms metric baselines in predicting human preferences on both in-domain and out-of-domain evaluations. Using its human-aligned rewards for post-training audio-video generation model also yields significant improvements in generation quality.
Chinese Translation
使用强化学习对联合视频-音频生成模型进行后训练需要一个奖励信号。现有方法通过结合音频质量、视觉保真度和同步等单独质量维度的指标来构建这一奖励。然而，这些指标是分别评估感知维度，未能捕捉文本提示、视频和音频之间的整体语义和时间一致性，这些因素共同影响人类偏好。针对这些指标优化模型会导致奖励黑客行为，生成在这些指标上得分高但对人类观众而言显得不连贯或不真实的视频-音频内容。为了解决这一问题，我们首先构建了一个大规模的人类偏好数据集VAPref-10K，用于联合视频-音频生成，包含来自开源生成模型的9K个提示和10.3K个细粒度配对比较。我们还引入了VA-Judger-Bench基准，进行领域内和领域外模型比较，以评估奖励模型是否真正与人类偏好对齐。我们进一步提出了VA-Judger，一种用于联合视频-音频生成的思维链全奖励模型。具体而言，VA-Judger首先从具有明显质量差距的配对中学习，以建立结构化输出和粗略的偏好区分，然后通过拒绝采样从人类注释中验证，提炼出可靠的偏好解释以应对更难的近质量比较，最后进行维度级强化学习，将人类反馈分解为单独的质量维度，以提供比单一二元偏好标签更密集的奖励信号。实验表明，VA-Judger在领域内和领域外评估中均优于指标基线，能够更好地预测人类偏好。使用其与人类对齐的奖励对音频-视频生成模型进行后训练也显著提高了生成质量。
cs.CV / 34 / 2608.18614
CDGP: Contrastive Dual Gaussian Processes for Weakly Supervised Anomaly Segmentation
CDGP：用于弱监督异常分割的对比双高斯过程
Abstract
Industrial visual inspection must both decide whether a product is defective and localize the defect, yet pixel-level masks are costly to collect at scale. Most anomaly-segmentation methods learn only from defect-free images and score deviations from normality. A true defect and an unusual-but-normal region, however, can both deviate substantially and receive similarly high scores. We propose Contrastive Dual Gaussian Processes (CDGP), a weakly supervised framework that models normal and anomaly inducing-variable predictive distributions over dense tokens. Its posterior-dominance statistic standardizes their predictive-mean difference by the joint predictive uncertainty, providing both spatial evidence and image-level confidence. This evidence complements hierarchical normal-reconstruction residuals for fine localization. All calibration uses training data only, without human pixel annotations or test-time fitting. Across MVTec AD~2, KSDD2, and VisA, CDGP ranks first among the evaluated methods on all MVTec AD~2 localization metrics and is first-place or competitive on KSDD2 and VisA. Factorized and matched linear-head controls delimit the contribution and scope of the linear-kernel Gaussian process (GP) formulation.
Chinese Translation
工业视觉检测必须同时判断产品是否存在缺陷并定位缺陷，但在大规模收集像素级掩码的成本较高。大多数异常分割方法仅从无缺陷图像中学习，并对正常性偏差进行评分。然而，真实缺陷和异常但正常的区域都可能显著偏离，并获得类似的高分。我们提出了对比双高斯过程（Contrastive Dual Gaussian Processes, CDGP），这是一个弱监督框架，建模正常和异常诱导变量的密集标记的预测分布。其后验主导统计量通过联合预测不确定性标准化预测均值差异，提供空间证据和图像级置信度。这些证据补充了层次正常重建残差，以实现精细定位。所有校准仅使用训练数据，而不需要人工像素标注或测试时拟合。在 MVTec AD~2、KSDD2 和 VisA 数据集上，CDGP 在所有 MVTec AD~2 定位指标中排名第一，并在 KSDD2 和 VisA 上表现出色或具有竞争力。分解和匹配的线性头控制限制了线性核高斯过程（Gaussian Process, GP）公式的贡献和范围。
cs.CV / 35 / 2608.18616
SPARC: Slice-to-volume Pipeline for Automated Reconstruction of gated 3D+time fetal Cardiac MRI
SPARC：用于自动重建门控3D+时间胎儿心脏MRI的切片到体积管道
Abstract
Fetal cardiac MRI (fCMR) provides valuable diagnostic information complementary to echocardiography, particularly for complex congenital heart disease (CHD). Dynamic cine imaging captures cardiac motion essential for assessment of cardiac function; however, the reconstruction of 3D+time cine volumes from 2D+time acquired slices remains challenging due to unpredictable fetal motion and the absence of automated and robust processing tools suitable for clinical deployment. We present the SPARC pipeline (Slice-to-volume Pipeline for Automated Reconstruction of gated 3D+time fetal Cardiac MRI) which combines physics-informed slice-to-volume reconstruction (SVR) of Doppler ultrasound (DUS) gated stacks of slices, assisted by deep learning (DL) models for thoracic segmentation and anatomical reorientation. The proposed SVR algorithm achieves a tenfold reduction in reconstruction time relative to existing frame-wise approaches ($4.8 \pm 1.0$ vs $49.0 \pm 14.1$ min, $p < 0.0001$) while improving the reconstruction quality. Thoracic segmentation performance using ensemble aggregation exceeded inter-rater agreement (Dice $84.7 \pm 3.9\%$ vs $81.4 \pm 7.7\%$, $p<0.05$), while anatomical reorientation achieved a success rate of $90.1\%$. End-to-end evaluation on a large held-out clinical cohort ($n = 121$) demonstrated fully automatic processing in $82.6\%$ of cases with a mean runtime of $7.1 \pm 1.3$ min, compatible with clinical deployment. The complete SPARC pipeline is publicly available as a Docker container https://hub.docker.com/r/aboutill/sparc and is currently deployed at our institution as a clinical research tool.
Chinese Translation
胎儿心脏MRI（fCMR）提供了与超声心动图互补的宝贵诊断信息，特别是在复杂先天性心脏病（CHD）方面。动态电影成像捕捉心脏运动，这对于评估心脏功能至关重要；然而，由于不可预测的胎儿运动以及缺乏适合临床应用的自动化和稳健的处理工具，从2D+时间获取的切片重建3D+时间电影体积仍然具有挑战性。我们提出了SPARC管道（切片到体积管道，用于自动重建门控3D+时间胎儿心脏MRI），该管道结合了物理信息驱动的切片到体积重建（SVR）技术，利用深度学习（DL）模型辅助进行胸部分割和解剖重定向。所提出的SVR算法在重建时间上相较于现有逐帧方法实现了十倍的减少（$4.8 ext{分钟} ext{±} 1.0$ vs $49.0 ext{分钟} ext{±} 14.1$，$p < 0.0001$），同时提高了重建质量。使用集成聚合的胸部分割性能超过了评估者间一致性（Dice $84.7 ext{±} 3.9 ext{ ext{%}}$ vs $81.4 ext{±} 7.7 ext{ ext{%}}$，$p<0.05$），而解剖重定向的成功率达到了$90.1 ext{ ext{%}}$。在一个大型保留临床队列（$n = 121$）上的端到端评估显示，在$82.6 ext{ ext{%}}$的案例中实现了完全自动处理，平均运行时间为$7.1 ext{分钟} ext{±} 1.3$，适合临床应用。完整的SPARC管道作为Docker容器公开可用，链接为https://hub.docker.com/r/aboutill/sparc，目前在我们机构作为临床研究工具进行部署。
cs.CV / 36 / 2608.18622
PALATE: Personalized Aesthetic Learning through Adaptive Taste Evolution for Multi-User Portrait Retouching
PALATE：通过适应性品味演化实现个性化审美学习的多用户肖像修饰
Abstract
Automatic portrait retouching has advanced rapidly, yet its objective is inherently subjective: the same portrait admits multiple professionally valid results, and users disagree about which one is best. Most existing methods optimize a population-level aesthetic standard and therefore cannot capture individual taste, while fine-tuning a separate editing model for every user incurs prohibitive training, storage, and data costs. We propose PALATE, a shared reward-evolution framework that keeps the image editor fixed and instead personalizes the selection among retouched candidates of the same source portrait. PALATE decomposes the reward for each user into a global backbone shared by all users, category-level residuals shared by aesthetically similar users, and a lightweight user adapter, with anti-collapse regularizers keeping the three levels complementary.A cyclic dual-level distillation scheme first distills user-specific preferences into category rewards and then consolidates the resulting category-level knowledge into the global backbone, which is redistributed to initialize the next evolution round. In this way, the shared initialization improves progressively across rounds, enabling unseen users to be calibrated from only a few rankings. On expert-retouched candidates from PPR10K with held-out users and held-out images, PALATE attains 72.83% pairwise preference-prediction accuracy, surpassing all reward, aesthetic, and image-quality baselines, of which the strongest, PickScore, reaches 58.06%. Each new user costs only 512 bytes of user-specific parameters and millisecond-level scoring.
Chinese Translation
自动肖像修饰技术迅速发展，但其目标本质上是主观的：同一肖像可以产生多种专业有效的结果，用户对哪种结果最佳存在分歧。现有大多数方法优化的是基于人群的审美标准，因此无法捕捉个体的品味，而为每个用户单独微调编辑模型则会产生高昂的训练、存储和数据成本。我们提出了PALATE，一个共享奖励演化框架，该框架保持图像编辑器不变，而是个性化选择同一源肖像的修饰候选项。PALATE将每个用户的奖励分解为所有用户共享的全球骨干、由审美相似用户共享的类别级残差，以及一个轻量级用户适配器，同时采用反崩溃正则化器保持这三个层次的互补性。一个循环双层蒸馏方案首先将用户特定的偏好蒸馏为类别奖励，然后将得到的类别级知识整合到全球骨干中，重新分配以初始化下一轮演化。通过这种方式，共享初始化在每轮中逐步改进，使得未见过的用户仅需少量排名即可进行校准。在来自PPR10K的专家修饰候选项中，针对被排除的用户和图像，PALATE实现了72.83%的成对偏好预测准确率，超越了所有奖励、审美和图像质量基线，其中最强的PickScore达到58.06%。每个新用户仅需512字节的用户特定参数和毫秒级评分。
cs.CV / 37 / 2608.18624
Evaluation of Image Matching Methods for Visual Odometry on UAVs
无人机视觉里程计图像匹配方法评估
Abstract
Unmanned aerial vehicles (UAVs) are becoming a powerful tool for many environmental monitoring and transport applications. Yet, their reliance on Global Navigation Satellite System (GNSS) technology for navigation makes them susceptible to catastrophic failures in scenarios where the positioning signal is unavailable or disrupted. This work explores Visual Odometry (VO) as a crucial navigation component. Recently, numerous deep-learning-based methods for image matching have been proposed that are yet to be implemented in a fully-fledged VO system. In this paper, we evaluate recent state-of-the-art image matching methods for the task of VO for UAV position tracking, with a downwards-facing camera, on our synthetic dataset, and find that while the best results are generated by the recent RoMa matcher, SIFT features can outperform some recent state-of-the-art.
Chinese Translation
无人驾驶飞行器（UAV）正成为许多环境监测和运输应用的强大工具。然而，它们对全球导航卫星系统（GNSS）技术的依赖使其在定位信号不可用或中断的情况下容易发生灾难性故障。本研究探讨了视觉里程计（VO）作为关键导航组件的重要性。近年来，提出了许多基于深度学习的图像匹配方法，但尚未在完整的VO系统中实施。本文评估了近期最先进的图像匹配方法在无人机位置跟踪任务中的表现，使用向下朝向的摄像头和我们的合成数据集，发现尽管最新的RoMa匹配器生成了最佳结果，但SIFT特征在某些近期的最先进方法中表现优于其他方法。
cs.CV / 38 / 2608.18627
PCQA-R1: Advancing Generalized 3D Point Cloud Quality Assessment with Reinforcement Learning
PCQA-R1：通过强化学习推进广义3D点云质量评估
Abstract
No-reference point cloud quality assessment (PCQA) has been an active topic in recent years and is used to measure and optimize the visual experience of point clouds. However, large multimodal models (LMMs) have rarely been explored in this area. Previous LMM-based methods mainly rely on supervised fine-tuning to directly predict numerical quality scores, lacking the ability to generalize across datasets with heterogeneous MOS scales and limited annotations. A key difficulty is that absolute MOS regression can be brittle across datasets with different score scales and distortion distributions, whereas relative quality ranking is more stable under such shifts. In this paper, we present PCQA-R1, the first reinforcement learning LMM for 3D point cloud quality assessment to simultaneously model quality understanding and scoring. Built upon the group relative policy optimization (GRPO) strategy, PCQA-R1 first constructs a chain-of-thought dataset, PCQA-CoT, which serves as cold-start training data through a reverse reasoning strategy that teaches the LMM to generate its reasoning process. We further introduce a Gaussian proximity reward that prevents calibration drift by anchoring score predictions to the source MOS range. Experimental results demonstrate that PCQA-R1 achieves state-of-the-art cross-dataset generalization across five benchmarks and competitive in-domain accuracy. Ablation studies support the role of ranking, Gaussian reward, and cold-start traces.
Chinese Translation
无参考点云质量评估（PCQA）近年来成为一个活跃的研究主题，用于衡量和优化点云的视觉体验。然而，大型多模态模型（LMM）在这一领域的探索仍然较少。以往基于LMM的方法主要依赖于监督微调，直接预测数值质量评分，缺乏在具有异构主观评分（MOS）尺度和有限注释的数据集间的泛化能力。一个关键的难点是，绝对MOS回归在不同评分尺度和失真分布的数据集间可能表现不稳定，而相对质量排名在这种变化下更为稳定。本文提出了PCQA-R1，这是首个用于3D点云质量评估的强化学习LMM，旨在同时建模质量理解和评分。PCQA-R1基于组相对策略优化（GRPO）策略，首先构建了一个思维链数据集PCQA-CoT，该数据集通过反向推理策略作为冷启动训练数据，教会LMM生成其推理过程。我们进一步引入了一种高斯邻近奖励，以通过将评分预测锚定到源MOS范围来防止校准漂移。实验结果表明，PCQA-R1在五个基准测试中实现了最先进的跨数据集泛化，并在领域内准确性上具有竞争力。消融研究支持了排名、高斯奖励和冷启动轨迹的作用。
cs.CV / 39 / 2608.18628
When Safety Overrides Vision: Exploring Dynamics between Vision Influence and Safety Alignment in Vision-Language Models
当安全性优先于视觉：探索视觉影响与安全对齐在视觉语言模型中的动态关系
Abstract
Aligned vision-language models (VLMs) are designed to balance grounded visual reasoning with safe generation behavior. However, we observe a striking phenomenon: under safety-constrained instruction, models frequently abstain from answering questions that remain correctly answerable under default instruction despite receiving identical image-question inputs. This raises a fundamental question: does safety alignment suppress perceptual grounding itself, or does visual evidence remain internally available while generation is redirected toward abstention? In this work, we investigate the internal decoding dynamics underlying safety-induced abstention in aligned VLMs. Across multiple architectures and multimodal benchmarks, we show that abstained generations remain consistently influenced by visual evidence throughout decoding, indicating that perceptual grounding is largely preserved despite refusal behavior. We further demonstrate that, although the representational organization of refusal differs substantially across architectures, safety-constrained instruction consistently alters late-stage hidden-state dynamics toward refusal-oriented decoding. Finally, through targeted activation-level interventions, we show that suppressing refusal-related representations reliably restores grounded answering behavior across models without retraining or modifying visual inputs. Together, these findings reveal a previously underexplored failure mode in aligned VLMs: safety alignment can override grounded visual expression even when perceptual evidence remains internally preserved.
Chinese Translation
对齐的视觉语言模型（VLMs）旨在平衡扎根的视觉推理与安全生成行为。然而，我们观察到一个显著现象：在安全约束的指令下，模型常常拒绝回答在默认指令下仍然可以正确回答的问题，尽管输入的图像-问题对是相同的。这引发了一个根本性的问题：安全对齐是否抑制了感知扎根本身，还是视觉证据在内部仍然可用，而生成被重定向为拒绝？在本研究中，我们探讨了对齐的 VLMs 中安全引发的拒绝行为背后的内部解码动态。在多个架构和多模态基准测试中，我们显示拒绝生成在解码过程中始终受到视觉证据的一致影响，表明尽管存在拒绝行为，感知扎根在很大程度上得以保留。我们进一步证明，尽管拒绝的表征组织在不同架构中有显著差异，但安全约束的指令始终改变晚期隐藏状态动态，朝向拒绝导向的解码。最后，通过针对性的激活水平干预，我们显示抑制与拒绝相关的表征可以可靠地恢复模型的扎根回答行为，而无需重新训练或修改视觉输入。总的来说，这些发现揭示了对齐的 VLMs 中一个先前未被充分探索的失败模式：安全对齐可以覆盖扎根的视觉表达，即使感知证据在内部仍然被保留。
cs.CV / 40 / 2608.18640
SAM2Dual: Training-Free, Dual Memory for Long-Term Video Object Segmentation
SAM2Dual：无训练的双重记忆用于长期视频目标分割
Abstract
Long-term video object segmentation (VOS) remains challenging due to error accumulation under extended occlusions, re-appearance, and scene changes. Although SAM2 provides strong zero-shot performance, its streaming memory can amplify drift over long horizons when recent, unreliable predictions dominate the memory state. We propose SAM2Dual, a training-free, plug-and-play inference-time enhancement that improves long-video robustness without updating model weights. SAM2Dual introduces a Dual Memory design that explicitly separates (i) short-term memory for rapid local adaptation and (ii) long-term memory built via interval-based sampling to preserve global identity cues, combined through a gated fusion strategy. In addition, we present Text-Aware Memory (TAM), which extracts a compact word-level cue from early frames and uses text embeddings to reweight memory contributions based on semantic compatibility, supporting identity preservation when visual evidence becomes weak or ambiguous. Across long-term benchmarks, SAM2Dual consistently improves stability on long videos, raising J&F from 49.33 to 50.65 on MOSEv2 and achieving consistent gains on LVOSv2.
Chinese Translation
长期视频目标分割（VOS）由于在长时间遮挡、重现和场景变化下的错误累积而仍然具有挑战性。尽管SAM2在零样本性能上表现出色，但其流式记忆在长时间范围内可能会因近期不可靠的预测主导记忆状态而放大漂移。我们提出了SAM2Dual，这是一种无训练的即插即用推理时增强方法，能够在不更新模型权重的情况下提高长视频的鲁棒性。SAM2Dual引入了一种双重记忆设计，明确区分（i）用于快速局部适应的短期记忆和（ii）通过基于间隔的采样构建的长期记忆，以保留全局身份线索，并通过门控融合策略进行结合。此外，我们提出了文本感知记忆（Text-Aware Memory, TAM），它从早期帧中提取紧凑的词级线索，并利用文本嵌入根据语义兼容性重新加权记忆贡献，以支持在视觉证据变得薄弱或模糊时的身份保留。在长期基准测试中，SAM2Dual在长视频上的稳定性持续改善，将MOSEv2上的J&F从49.33提高到50.65，并在LVOSv2上实现了一致的增益。
cs.CV / 41 / 2608.18654
Clinically Structured Surrogate Rewards for Post-SFT Medical Image Captioning
临床结构化替代奖励用于后SFT医学图像描述
Abstract
Medical image captioning requires translating heterogeneous visual evidence into concise clinical descriptions, where errors in findings, assertion states, or anatomical relations can alter clinical meaning despite surface-level fluency. Sequence-level policy optimization can directly optimize complete captions, but common rewards rely on global text similarity, direct image-caption compatibility, or unordered concept overlap, leaving visual neighborhoods and clinical-claim structure implicit. We propose a clinically structured surrogate reward framework for post-SFT medical image captioning. The framework combines biomedical semantic and short-range lexical fidelity with two structured rewards: distributional image-neighborhood alignment, which matches the medical-image-bank distributions induced by reference and generated captions, and clinical graph consistency, which applies maximum-weight one-to-one matching to entities, assertion states, and typed relations. The four rewards are independently normalized within each rollout group, combined with fixed relative weights, and optimized with GDPO. Across organizer-evaluated hidden test sets for the Standard and Synthetical ImageCLEFmedical Caption tracks and three vision-language backbones, the method improves Overall, Relevance, and Factuality over matched SFT baselines in all six backbone-track combinations, with average relative gains of 3.4%, 2.1%, and 5.8%, respectively. Ablations and paired diagnostics indicate that the structured rewards provide complementary signals, reducing image-neighborhood divergence and improving entity-assertion-relation consistency.
Chinese Translation
医学图像描述需要将异质的视觉证据转化为简明的临床描述，其中发现、断言状态或解剖关系的错误可能会改变临床意义，尽管表面上流畅。序列级策略优化可以直接优化完整的描述，但常见的奖励依赖于全局文本相似性、直接的图像-描述兼容性或无序概念重叠，导致视觉邻域和临床主张结构隐含。我们提出了一种用于后SFT医学图像描述的临床结构化替代奖励框架。该框架结合了生物医学语义和短距离词汇保真度，并引入了两个结构化奖励：分布式图像邻域对齐，该奖励匹配由参考和生成描述引起的医学图像库分布，以及临床图一致性，该奖励对实体、断言状态和类型关系应用最大权重一对一匹配。这四个奖励在每个回合组内独立归一化，结合固定的相对权重，并通过GDPO进行优化。在组织者评估的标准和合成ImageCLEF医学描述轨道的隐藏测试集以及三种视觉-语言骨干网络中，该方法在所有六种骨干-轨道组合中均提高了整体性、相关性和事实性，相较于匹配的SFT基线，平均相对增益分别为3.4%、2.1%和5.8%。消融实验和配对诊断表明，结构化奖励提供了互补信号，减少了图像邻域的发散，并改善了实体-断言-关系的一致性。
cs.CV / 42 / 2608.18662
Dynamic SpectraFormer for Ultra-High-Definition Underwater Image Enhancement
超高清水下图像增强的动态光谱变换器
Abstract
Underwater images suffer from color distortion, haze, and poor visibility due to light refraction and absorption in water. These challenges significantly impact the utilization of Autonomous Underwater Vehicles (AUVs) or marine robots. Typically, color and brightness distortions manifest at lower frequencies, while edge and texture distortions are prevalent at higher frequencies. Traditional methods struggle to concurrently rectify these mixed distortions as they primarily concentrate on the spatial domain. To address these issues, we introduce the Dynamic SpectraFormer, which enhances underwater images through a frequency domain transformer. The Dynamic SpectraFormer introduces an ultra-high-resolution sparse spectrum attention module, which could capture the long-term dependency without losing the universal approximating power. Additionally, we have developed a dynamic spectrum weight generation layer that serves as an adaptive spectrum band selector, accentuating critical frequency bands and suppressing less relevant ones. Consequently, this method significantly improves underwater image quality by addressing both high- and low-frequency distortions. Our extensive ablation studies and comparative evaluations consolidate the Dynamic SpectraFormer's efficacy across multiple underwater image enhancement benchmarks. The source code is available at https://github.com/arifence2024/DynamicSpectraFormer.git.
Chinese Translation
水下图像由于光在水中的折射和吸收，常常遭受颜色失真、雾霾和可见度差等问题。这些挑战显著影响了自主水下航行器（AUV）或海洋机器人的应用。通常，颜色和亮度失真主要表现为低频现象，而边缘和纹理失真则在高频中更为明显。传统方法在同时纠正这些混合失真方面存在困难，因为它们主要集中于空间域。为了解决这些问题，我们提出了动态光谱变换器（Dynamic SpectraFormer），通过频域变换器增强水下图像。动态光谱变换器引入了一种超高分辨率稀疏光谱注意力模块，能够捕捉长期依赖关系而不损失通用逼近能力。此外，我们开发了一种动态光谱权重生成层，作为自适应光谱带选择器，强调关键频带并抑制不相关的频带。因此，该方法通过同时处理高频和低频失真显著改善了水下图像质量。我们广泛的消融研究和比较评估巩固了动态光谱变换器在多个水下图像增强基准测试中的有效性。源代码可在 https://github.com/arifence2024/DynamicSpectraFormer.git 获取。
cs.CV / 43 / 2608.18667
Teeth2Point: A Two-Stage Dental CBCT ROI-to-Point Segmentation Framework
Teeth2Point：一种两阶段牙科CBCT ROI到点的分割框架
Abstract
Modern deep learning architectures have demonstrated strong performance in dental CBCT segmentation. One remaining crucial challenge is accurate tooth labeling in cases with missing or malpositioned teeth, which are highly relevant for dental practice. Transformer-based architectures should in theory be able to resolve such ambiguities using global anatomical context. However, due to the high resolution of CBCT volumes and the wide spatial distribution of teeth within volumes, dense patch-based volumetric processing faces an inherent trade-off. Computational costs limit the number of patches that can be used in self-attention and thus, one can either increase the extent of the context captured in self-attention or capture fine-grained structural details by using small patches, but not both. In this work, we present Teeth2Point, an efficient point-based transformer framework for dental CBCT semantic segmentation that can avoid this trade-off. Teeth2Point first localizes volumetric regions of interest (ROIs) surrounding teeth using a convolutional model, then converts ROIs into point tokens using adaptive sampling. A transformer model predicts accurate segmentations using the point tokens, which allow capturing global context while retaining high resolution. The transformer is first pretrained using self-supervised learning (SSL), in the style of DINO but using domain-specific augmentation strategies, followed by supervised finetuning. The SSL pretraining, which includes random token masking, provides robustness to complex anatomical variations. Compared with the strongest two-stage baseline, Teeth2Point improves abnormal-case performance by 1.44 DSC points on average across four datasets; relative to the first-stage nnU-Net, the gain is 1.9 points.
Chinese Translation
现代深度学习架构在牙科CBCT分割中表现出色。然而，在缺失或位置不当的牙齿情况下，准确的牙齿标记仍然是一个关键挑战，这对牙科实践具有重要意义。理论上，基于Transformer的架构应该能够利用全局解剖上下文来解决此类模糊性。然而，由于CBCT体积的高分辨率以及牙齿在体积内的广泛空间分布，基于密集补丁的体积处理面临固有的权衡。计算成本限制了在自注意力中可以使用的补丁数量，因此，可以选择增加自注意力中捕获的上下文范围，或通过使用小补丁捕获细粒度的结构细节，但无法同时实现。在本研究中，我们提出了Teeth2Point，这是一种高效的基于点的Transformer框架，用于牙科CBCT语义分割，能够避免这种权衡。Teeth2Point首先使用卷积模型定位围绕牙齿的体积感兴趣区域（ROIs），然后通过自适应采样将ROIs转换为点令牌。Transformer模型利用点令牌预测准确的分割，允许在保持高分辨率的同时捕获全局上下文。该Transformer首先使用自监督学习（SSL）进行预训练，采用DINO风格，但使用特定领域的增强策略，随后进行监督微调。SSL预训练包括随机令牌掩蔽，增强了对复杂解剖变异的鲁棒性。与最强的两阶段基线相比，Teeth2Point在四个数据集上平均提高了1.44 DSC点的异常案例表现；相较于第一阶段的nnU-Net，增益为1.9点。
cs.CV / 44 / 2608.18671
Vision-Language Models for Egocentric Video: From Hand-Object Interaction to Embodied AI
自我中心视频的视觉-语言模型：从手-物体交互到具身人工智能
Abstract
Egocentric video captures activities from the wearer's perspective, providing a direct view of human attention, hand--object interaction, and goal-directed behavior. This perspective is increasingly important for wearable intelligence, assistive systems, human--robot interaction, and embodied AI, yet it introduces challenges including ego-motion, occlusion, small active objects, viewpoint-dependent appearance, and long-range temporal dependencies. Vision--language models (VLMs) offer a promising foundation for addressing these challenges by linking visual observations with semantic knowledge and natural-language supervision. This survey presents a critical review of VLMs for egocentric video understanding, tracing the progression from conventional recognition architectures to multimodal foundation models and embodied systems. We organize the literature around tasks, datasets, hand--object interaction understanding, temporal reasoning, frame and clip selection, multimodal representation learning, prompting, semantic alignment, and model adaptation. Particular attention is given to graph-based and object-centric reasoning as mechanisms for modeling relations among hands, objects, actions, and scene context over time. We further examine how first-person perception and multimodal foundation models support wearable assistance, robot skill learning, human-to-robot transfer, and embodied decision making. Across the reviewed literature, a consistent limitation emerges: current models recognize visible objects more reliably than evolving interactions, actions, and user intent, especially over long activities. We therefore identify temporally grounded reasoning, interaction-aware supervision, efficient long-video processing, multimodal fusion, graph-enhanced representations, cross-domain generalization, privacy, and trustworthy evaluation as key priorities for deployable embodied intelligence.
Chinese Translation
自我中心视频从佩戴者的视角捕捉活动，提供了人类注意力、手-物体交互和目标导向行为的直接视图。这一视角对于可穿戴智能、辅助系统、人机交互和具身人工智能日益重要，但也带来了诸如自我运动、遮挡、小型活动物体、视角依赖外观和长时间依赖等挑战。视觉-语言模型（VLMs）为解决这些挑战提供了一个有前景的基础，通过将视觉观察与语义知识和自然语言监督相连接。本综述对自我中心视频理解中的VLMs进行了批判性回顾，追踪了从传统识别架构到多模态基础模型和具身系统的发展。我们围绕任务、数据集、手-物体交互理解、时间推理、帧和片段选择、多模态表示学习、提示、语义对齐和模型适应等方面组织文献。特别关注基于图的和以物体为中心的推理作为建模手、物体、动作和场景上下文之间关系的机制。我们进一步探讨了第一人称感知和多模态基础模型如何支持可穿戴辅助、机器人技能学习、人对机器人的迁移以及具身决策。在回顾的文献中，一个一致的局限性浮现：当前模型在识别可见物体方面比识别不断演变的交互、动作和用户意图更可靠，尤其是在长时间活动中。因此，我们确定了时间基础推理、交互感知监督、高效长视频处理、多模态融合、图增强表示、跨域泛化、隐私和可信评估作为可部署具身智能的关键优先事项。
cs.CV / 45 / 2608.18673
DynCur-Geo: Dynamic Curiosity Reward Shaping for Multimodal Active Geo-Localization
DynCur-Geo：用于多模态主动地理定位的动态好奇心奖励塑造
Abstract
Active geo-localization enables low-altitude UAVs to search for specified targets from limited local aerial observations, supporting time-sensitive applications such as search and rescue and emergency inspection. However, multimodal target cues, restricted views, and sparse feedback make it difficult to balance exploration with target convergence. Existing curiosity-driven methods assign a fixed intrinsic-reward weight throughout search, which can continue rewarding novelty after the agent nears the target and induce detours. We propose DynCur-Geo, a dynamic curiosity framework that adjusts prediction-error intrinsic reward according to remaining target distance. A distance-aware gate encourages early exploration and shifts the policy toward goal-directed behavior near the target, while potential-based reward shaping supplies dense progress guidance. Experiments across multimodal, cross-scene, disaster-affected, and long-range settings show consistent gains over active geo-localization baselines.
Chinese Translation
主动地理定位使低空无人机能够从有限的局部空中观察中搜索指定目标，支持如搜救和紧急检查等时间敏感的应用。然而，多模态目标线索、受限视野和稀疏反馈使得探索与目标收敛之间的平衡变得困难。现有的基于好奇心的方法在整个搜索过程中分配固定的内在奖励权重，这可能在代理接近目标后继续奖励新奇性，从而导致绕路。我们提出了DynCur-Geo，一种动态好奇心框架，根据剩余目标距离调整预测误差内在奖励。一个距离感知门鼓励早期探索，并在接近目标时将策略转向目标导向行为，而基于潜力的奖励塑造则提供密集的进展指导。在多模态、跨场景、灾后影响和长距离设置下的实验显示出相较于主动地理定位基线的一致性提升。
cs.CV / 46 / 2608.18679
FRAGMENT: Factorized Graph Representations for Document Generation and Editing via Entity-Aware Transformations
FRAGMENT：通过实体感知转换进行文档生成和编辑的因子化图表示
Abstract
Structured documents such as invoices, forms, reports, and scientific articles derive meaning from the interplay between spatial layout, textual content, and logical structure. Generative models operating at the pixel or token level often struggle to capture these dependencies effectively. We explore FRAGMENT, a generative framework that represents a document as a typed relational graph and factorizes its distribution as p(structure, content) = p(structure) * p(content | structure). The framework consists of two stages. The first stage, the Architect, is a causally masked Transformer conditioned on document category that autoregressively generates the graph topology and typed spatial relations. The second stage, the Builder, is a GATv2-based graph attention network that enriches the graph with normalized bounding boxes, text, and visual style attributes. Both stages define explicit likelihood models, yielding a tractable document-level likelihood that serves as an anomaly score for forgery detection. For controlled editing, a prompt-conditioned extension injects instruction embeddings into the Builder through cross-attention, enabling semantic and entity-aware modifications. We describe training on DocLayNet and fine-tuning on FUNSD and SROIE. Experiments on DocLayNet, FUNSD, and SROIE evaluate FRAGMENT alongside representative autoregressive, layout-only, and graph-based baselines, providing an empirical analysis of the characteristics and trade-offs of the proposed factorized graph generation framework.
Chinese Translation
结构化文档，如发票、表单、报告和科学文章，其意义源于空间布局、文本内容和逻辑结构之间的相互作用。在像素或标记级别操作的生成模型通常难以有效捕捉这些依赖关系。我们探索了FRAGMENT，一个生成框架，它将文档表示为一个类型化的关系图，并将其分布因子化为p(结构, 内容) = p(结构) * p(内容 | 结构)。该框架由两个阶段组成。第一阶段，Architect，是一个基于因果掩蔽的Transformer，条件为文档类别，采用自回归方式生成图的拓扑结构和类型化空间关系。第二阶段，Builder，是一个基于GATv2的图注意力网络，利用归一化的边界框、文本和视觉风格属性丰富图的内容。两个阶段都定义了明确的似然模型，产生可处理的文档级似然性，作为伪造检测的异常分数。为了实现受控编辑，一个基于提示的扩展通过交叉注意力将指令嵌入注入Builder，从而实现语义和实体感知的修改。我们描述了在DocLayNet上的训练以及在FUNSD和SROIE上的微调。对DocLayNet、FUNSD和SROIE的实验评估了FRAGMENT，并与代表性的自回归、仅布局和基于图的基线进行了比较，提供了对所提出的因子化图生成框架特征和权衡的实证分析。
cs.CV / 47 / 2608.18685
DocClaw: A Unified Agentic System for Intelligent Document Processing
DocClaw：一个统一的智能文档处理代理系统
Abstract
Intelligent document processing (IDP) encompasses a broad range of tasks, including optical character recognition (OCR), document question answering (DocQA), and key information extraction (KIE). Despite their distinct objectives, these tasks share a common need to perceive document content, acquire task-relevant information, and progressively refine intermediate results. However, they are typically formulated as separate prediction problems and addressed by task-specific models or processing pipelines. We introduce DocClaw, a unified agentic system that formulates diverse intelligent document processing tasks as a shared process of interaction between an agent and a document. Given a document and a task-specific query, DocClaw follows an appropriate document skill to iteratively identify the information required, invoke relevant tools, and integrate the resulting observations into the desired output. Throughout this process, a structured document state organizes reusable document knowledge and task-specific interaction context, allowing the agent to accumulate, revisit, and progressively refine information as the interaction proceeds. Under this formulation, task-specific requirements are captured by the agent's interpretation of the query objective and the corresponding document skill, while the underlying interaction loop, tool space, and document state are shared across tasks. Extensive experiments across multiple intelligent document processing benchmarks demonstrate that DocClaw effectively handles diverse tasks within a single agentic framework and achieves competitive performance compared with both general-purpose VLMs and task-specific methods.
Chinese Translation
智能文档处理（IDP）涵盖了广泛的任务，包括光学字符识别（OCR）、文档问答（DocQA）和关键信息提取（KIE）。尽管这些任务的目标各不相同，但它们在感知文档内容、获取与任务相关的信息以及逐步完善中间结果方面有着共同的需求。然而，这些任务通常被视为独立的预测问题，并由特定任务的模型或处理管道来解决。我们提出了DocClaw，一个统一的代理系统，将多种智能文档处理任务表述为代理与文档之间的共享交互过程。给定一个文档和一个特定任务的查询，DocClaw遵循适当的文档技能，迭代地识别所需信息，调用相关工具，并将生成的观察结果整合到期望的输出中。在整个过程中，一个结构化的文档状态组织了可重用的文档知识和特定任务的交互上下文，使代理能够在交互进行时积累、回顾并逐步完善信息。在这种表述下，特定任务的需求通过代理对查询目标的解释和相应的文档技能来捕捉，而底层的交互循环、工具空间和文档状态则在任务之间共享。针对多个智能文档处理基准的广泛实验表明，DocClaw能够在单一的代理框架内有效处理多样化的任务，并在与通用VLM（视觉语言模型）和特定任务方法的比较中取得了竞争力的表现。
cs.CV / 48 / 2608.18694
Composed Historical Image Retrieval by Modeling Temporal Representations
通过建模时间表示实现组合历史图像检索
Abstract
While time evolves linearly, the geometry of neural embedding spaces is inherently multi-dimensional, often chaotic, and difficult to interpret. In principle, one could constrain an embedding space to a single temporal dimension; however, such a reduction would sacrifice performance on downstream tasks, as one-dimensional embeddings cannot retain sufficient expressive capacity. This paper asks whether it is possible to learn representations that preserve temporal structure while remaining effective for image and object retrieval, and answers this question by building the mathematical foundations of such a system. We propose Temporally Decomposable Image Representations (TDIR), a representation learning algorithm that decomposes historical photographs into separate date and content components through orthogonal subspaces. We define and prove the conditions under which such a decomposition is achievable, characterize the error incurred when those conditions are only partially met, and show that orthogonality between temporal and categorical subspaces emerges naturally from the joint optimization, without requiring it to be imposed explicitly. Beyond its geometric properties, TDIR enables a class of transitive operations on embedding spaces: the temporal information of one image can be extracted and injected into the representation of another, with no label supervision required. All theoretical properties are grounded and validated in the real-world problem of Composed Image Retrieval on historical photographs, where a query simultaneously specifies object content and a target time period, either through labels or through example images. This in-the-wild setting serves as a concrete backing for the propositions we derive, offering an intuitive and interpretable way to navigate photographic archives while maintaining competitive performance in both date estimation and object retrieval.
Chinese Translation
尽管时间是线性演变的，但神经嵌入空间的几何结构本质上是多维的，往往是混沌的，且难以解释。原则上，可以将嵌入空间限制为单一的时间维度；然而，这种简化会牺牲下游任务的性能，因为一维嵌入无法保留足够的表达能力。本文探讨是否可以学习保留时间结构的表示，同时在图像和物体检索中保持有效性，并通过构建此类系统的数学基础来回答这个问题。我们提出了时间可分解图像表示（Temporally Decomposable Image Representations, TDIR），这是一种表示学习算法，通过正交子空间将历史照片分解为单独的日期和内容组件。我们定义并证明了实现这种分解的条件，表征了在这些条件仅部分满足时所产生的误差，并展示了时间子空间与类别子空间之间的正交性是如何在联合优化中自然产生的，而不需要显式施加。除了其几何特性，TDIR还使得在嵌入空间上进行一类传递操作成为可能：可以提取一幅图像的时间信息并注入到另一幅图像的表示中，而无需标签监督。所有理论属性都在历史照片的组合图像检索这一现实问题中得到了基础和验证，其中查询同时通过标签或示例图像指定对象内容和目标时间段。这种实际环境为我们推导的命题提供了具体支持，提供了一种直观且可解释的方式来浏览摄影档案，同时在日期估计和对象检索中保持竞争力的性能。
cs.CV / 49 / 2608.18696
Impact of Iterative Fine-Tuning on Transcription Accuracy in Complex Historical Sanskrit Manuscripts
迭代微调对复杂历史梵文手稿转录准确性的影响
Abstract
Digitizing the text from handwritten historical manuscripts is required to make them easily accessible, preservable, and to enable historical scholars to study them in new ways. Historical manuscripts, however, often exhibit complex heterogeneous layouts and non-standard appearance due to period-specific writing styles, page textures, camera noise, and other nuisance factors, making them difficult to perform OCR on. To tackle this challenge, we introduce a local traditional OCR pipeline, which can be iteratively fine-tuned on the target manuscript at the layout-level and the appearance-level. By adapting to the target manuscript distribution, the proposed Traditional OCR pipeline makes better predictions on subsequent pages, causing iterative reduction in human annotation effort, which is expensive and time-consuming as it requires historical domain expertise. Using this pipeline, we digitize text from three complex historical Sanskrit manuscripts and introduce a dataset with granular layout-level annotations, along with Unicode annotations in the standard PAGE-XML format. We demonstrate quantitative gains due to iterative fine-tuning of the proposed traditional OCR pipeline, and also benchmark the performance of leading Multi-Modal Large Language Models on the introduced Dataset. Code and dataset are available at: https://github.com/flame-cai/gnn-synthetic-layout-historical/.
Chinese Translation
从手写历史手稿中数字化文本是为了使其更易于获取、保存，并使历史学者能够以新的方式研究它们。然而，历史手稿通常由于特定时期的书写风格、页面纹理、相机噪声和其他干扰因素，表现出复杂的异质布局和非标准外观，这使得进行光学字符识别（OCR）变得困难。为了解决这一挑战，我们提出了一种本地传统OCR流程，该流程可以在布局层面和外观层面对目标手稿进行迭代微调。通过适应目标手稿的分布，所提出的传统OCR流程在后续页面上做出更好的预测，从而迭代减少人工标注工作量，而人工标注既昂贵又耗时，因为它需要历史领域的专业知识。利用该流程，我们从三部复杂的历史梵文手稿中数字化文本，并引入一个具有细粒度布局级注释的数据集，以及标准PAGE-XML格式的Unicode注释。我们展示了由于对所提出的传统OCR流程进行迭代微调而带来的定量收益，并对在引入的数据集上领先的多模态大型语言模型的性能进行了基准测试。代码和数据集可在以下链接获取：https://github.com/flame-cai/gnn-synthetic-layout-historical/.
cs.CV / 50 / 2608.18709
A Critical Synthesis of Uncertainty Quantification and Foundation Models for Semantic Segmentation
不确定性量化与基础模型在语义分割中的关键综合研究
Abstract
Foundation models are increasingly breaking what seemed to be impossible not long ago by enabling unprecedented accuracy and cross-domain generalization. Yet their lack of interpretability, tendency to be overconfident, and sensitivity to real-world domain shifts pose critical challenges for safety- and mission-critical applications. Uncertainty quantification (UQ) offers a principled way to address these issues, but its integration into segmentation foundation models has yet to be explored. In this paper we present the first systematic evaluation of UQ methods applied to a foundation model for semantic segmentation. We fine-tune a lightweight DPT decoder on top of the pretrained SAM2 encoder to establish a simple yet competitive baseline and benchmark four representative UQ approaches - Monte Carlo Dropout, Deep Sub-Ensemble, Test-Time Augmentation, and Evidential Deep Learning - across Cityscapes, NYUv2, and two challenging out-of-domain settings. Our analysis compares segmentation accuracy, calibration, uncertainty quality, and inference time, revealing clear trade-offs between predictive performance, reliability, and computational cost. These results highlight both the promise and the current limitations of uncertainty-aware foundation models, pointing to the need for future work that jointly optimizes accuracy, robustness, and efficiency for real-world deployment.
Chinese Translation
基础模型正在不断打破不久前看似不可能的界限，实现前所未有的准确性和跨领域泛化。然而，它们缺乏可解释性、过于自信的倾向以及对现实世界领域变化的敏感性，为安全和任务关键型应用带来了重大挑战。不确定性量化（UQ）提供了一种原则性的方法来解决这些问题，但其在分割基础模型中的整合尚待探索。本文首次系统评估了应用于语义分割基础模型的UQ方法。我们在预训练的SAM2编码器基础上，微调了一个轻量级的DPT解码器，以建立一个简单而具有竞争力的基线，并在Cityscapes、NYUv2以及两个具有挑战性的域外设置中基准测试四种代表性的UQ方法——蒙特卡罗Dropout、深度子集成、测试时增强和证据深度学习。我们的分析比较了分割准确性、校准、预测不确定性质量和推理时间，揭示了预测性能、可靠性和计算成本之间的明显权衡。这些结果突显了不确定性感知基础模型的潜力和当前局限性，指向未来需要共同优化准确性、鲁棒性和效率以实现现实世界部署的研究方向。
cs.CV / 51 / 2608.18710
CamWorldQA: Perceptual Quality Assessment of Camera-Controlled World Video Generation
CamWorldQA：相机控制世界视频生成的感知质量评估
Abstract
Recent advances in generative video models have enabled camera-controlled world video generation, allowing models to synthesize videos under user-defined camera trajectories. However, existing video quality assessment (VQA) methods are mainly developed for natural videos and fail to capture the unique perceptual characteristics of camera-controlled generation, such as viewpoint consistency, motion coherence, and content preservation. In this work, we introduce CamWorldQA, the first benchmark for perceptual quality assessment of camera-controlled world video generation. CamWorldQA contains 720 generated videos produced by 6 representative generation methods from 20 diverse source videos under 6 camera trajectories, where each video is annotated with a human-rated perceptual quality score through subjective experiments. Furthermore, we propose CWQA, a no-reference quality assessment network with three complementary branches that extract spatial features, temporal motion features and optical flow features to jointly predict quality scores. Extensive experiments demonstrate that CWQA achieves superior performance over existing quality assessment methods on the CamWorldQA dataset.
Chinese Translation
近年来，生成视频模型的进展使得相机控制的世界视频生成成为可能，允许模型在用户定义的相机轨迹下合成视频。然而，现有的视频质量评估（VQA）方法主要是为自然视频开发的，无法捕捉相机控制生成的独特感知特征，如视点一致性、运动连贯性和内容保留。在本研究中，我们引入了CamWorldQA，这是第一个用于相机控制世界视频生成的感知质量评估基准。CamWorldQA包含720个生成视频，这些视频是由6种代表性的生成方法在6条相机轨迹下从20个多样化的源视频中生成的，每个视频都通过主观实验标注了人类评分的感知质量分数。此外，我们提出了CWQA，这是一种无参考质量评估网络，具有三个互补分支，分别提取空间特征、时间运动特征和光流特征，以共同预测质量分数。大量实验表明，CWQA在CamWorldQA数据集上相较于现有质量评估方法表现出色。
cs.CV / 52 / 2608.18711
EgoHRV: Continuous Heart Rate Variability Estimation from Egocentric Systems for Autonomic Response and Skill Assessment
EgoHRV：基于自我中心系统的连续心率变异性估计用于自主反应和技能评估
Abstract
Egocentric vision systems capture human behavior from visible cues, but overlook physiological indicators of autonomic states such as stress, engagement, and attention. Heart rate variability (HRV) is a widely used noninvasive marker of autonomic regulation under stress. HRV reflects small timing differences between successive heartbeats and has so far been out of reach for egocentric platforms, where motion and noise in gaze video mask exactly this fine-grained timing. We propose EgoHRV, a method that estimates HRV as well as heart rate (HR) from the gaze cameras that are already integrated into egocentric headsets. Our pipeline combines a 3D backbone with a novel low--high decomposition module that extracts the blood volume pulse (BVP) signal from gaze video. Our cross-domain pretraining aligns the frequency-domain representations of contact-based and camera-derived signals. This alignment gives EgoHRV the temporal precision to recover HRV from the subtle fluctuations in gaze video. EgoHRV achieves state-of-the-art accuracy for HR and HRV estimation from egocentric video, and its uncertainty-aware design improves downstream behavioral modeling. Integrating our HRV estimates and confidence measures into EgoExo4D's proficiency estimator raises accuracy by 17.8%. Beyond skill, continuous HRV estimation also opens egocentric systems to stress- and arousal-aware estimation tasks. Code: https://github.com/eth-siplab/EgoHRV
Chinese Translation
自我中心视觉系统通过可见线索捕捉人类行为，但忽视了诸如压力、参与度和注意力等自主状态的生理指标。心率变异性（HRV）是一个广泛使用的非侵入性标志，用于在压力下评估自主调节。HRV反映了连续心跳之间的微小时间差，但在自我中心平台上，由于视线视频中的运动和噪声掩盖了这种细微的时间差，HRV的估计一直难以实现。我们提出了EgoHRV，这是一种从已经集成在自我中心头戴设备中的视线摄像头估计HRV和心率（HR）的方法。我们的流程结合了一个3D骨干网络和一个新颖的低高频分解模块，从视线视频中提取血容量脉搏（BVP）信号。我们的跨域预训练对接触式信号和摄像头衍生信号的频域表示进行了对齐。这种对齐使EgoHRV具备了从视线视频中的微小波动中恢复HRV的时间精度。EgoHRV在自我中心视频中实现了HR和HRV估计的最先进准确性，其不确定性感知设计改善了下游行为建模。将我们的HRV估计和置信度度量整合到EgoExo4D的熟练度估计器中，使准确性提高了17.8%。除了技能，连续HRV估计还使自我中心系统能够进行压力和唤醒感知的估计任务。代码：https://github.com/eth-siplab/EgoHRV
cs.CV / 53 / 2608.18715
The Impact of CutMix on Reliability and Robustness in Semantic Segmentation
CutMix对语义分割中可靠性和鲁棒性的影响
Abstract
Ensuring not only high accuracy but also reliable and robust predictions is critical for the deployment of semantic segmentation models in safety-critical applications such as autonomous driving. Despite the widespread use of CutMix - a simple yet powerful data augmentation strategy - its effect on the reliability and robustness in dense predictions tasks remains unexplored. Motivated by recent findings that semi-supervised segmentation methods, where CutMix is a core component, can severely degrade reliability, this study isolates and systematically analyzes the influence of CutMix on segmentation accuracy, calibration, and uncertainty quality. We evaluate two representative architectures, the CNN-based DeepLabV3+ and the transformer-based SegFormer, across both in-domain and out-of-domain scenarios. Our results show that CutMix has only a minor impact on segmentation accuracy but consistently improves the reliability, particularly under distribution shifts. These improvements indicate that CutMix primarily enhances the trustworthiness of the model's calibration and uncertainty rather than the raw segmentation prediction itself. This distinction is crucial for safety-critical deployment, where reliable confidence estimates are as important as raw performance.
Chinese Translation
确保不仅具有高准确性，而且具有可靠和鲁棒的预测，对于在安全关键应用（如自动驾驶）中部署语义分割模型至关重要。尽管CutMix作为一种简单而强大的数据增强策略被广泛使用，但其在密集预测任务中的可靠性和鲁棒性影响尚未得到探索。受到最近发现的启发，即CutMix作为核心组件的半监督分割方法可能严重降低可靠性，本研究对CutMix对分割准确性、校准和不确定性质量的影响进行了系统的分析。我们在领域内和领域外场景中评估了两种代表性架构：基于CNN的DeepLabV3+和基于变换器的SegFormer。我们的结果表明，CutMix对分割准确性的影响较小，但在分布变化下始终提高了可靠性。这些改进表明，CutMix主要增强了模型的校准和不确定性的可信度，而不是原始的分割预测本身。这一区别对于安全关键的部署至关重要，因为可靠的置信度估计与原始性能同样重要。
cs.CV / 54 / 2608.18731
A Few Cases Are All You Need: An Empirical Study of Annotation-Efficient LoRA Fine-Tuning of MedSAM3
少量案例即可满足需求：对MedSAM3注释高效LoRA微调的实证研究
Abstract
Medical image segmentation is essential for clinical workflows such as treatment planning and disease assessment. While specialist tools like TotalSegmentator and MRSegmentator achieve strong performance, they require large annotated datasets for training. Medical foundation models offer a promising alternative through large-scale pretraining that reduces the annotation burden for new tasks, but zero-shot performance remains limited. Parameter-efficient adaptation via Low-Rank Adaptation (LoRA) enables efficient specialization with few trainable parameters, but a key question remains: how many expert-annotated cases are needed to achieve clinically useful segmentation performance? We address this by adapting MedSAM3 with LoRA for five abdominal organs (liver, kidneys, spleen, gallbladder, and pancreas) in CT and MRI using only 1, 2, 5, and 10 annotated cases, evaluating on AMOS22 dataset. With just 10 cases, models achieve performance competitive with specialist systems trained on orders of magnitude more data. Notably, this includes reliable gallbladder segmentation (Dice 0.68 CT, 0.59 MRI) where existing tools fail almost completely (Dice 0.0004), while remaining within 5--10% of MRSegmentator for liver, kidneys, and spleen using over 100 times fewer annotations. Furthermore, external validation on the Whole Heart Segmentation dataset shows that the approach extends to cardiac segmentation, a use case beyond the scope of TotalSegmentator (MRI) and MRSegmentator, achieving competitive left ventricle (LV) performance with only 10 annotated cases. Training requires only3--5,hours per organ on a single GPU, approximately 2--3 times faster than nnU-Net. These findings suggest that ten annotated cases are sufficient for clinically useful segmentation, effectively reducing bottlenecks for both image annotation and training time.
Chinese Translation
医学图像分割对于临床工作流程，如治疗计划和疾病评估至关重要。尽管像TotalSegmentator和MRSegmentator这样的专业工具表现出色，但它们需要大量的标注数据集进行训练。医学基础模型通过大规模预训练提供了一种有前景的替代方案，减少了新任务的标注负担，但零样本性能仍然有限。通过低秩适应（Low-Rank Adaptation, LoRA）实现的参数高效适应使得在少量可训练参数下实现高效专业化成为可能，但一个关键问题仍然存在：实现临床有用的分割性能需要多少专家标注的案例？我们通过使用仅1、2、5和10个标注案例，采用LoRA对MedSAM3进行适应，针对CT和MRI中的五个腹部器官（肝脏、肾脏、脾脏、胆囊和胰腺）进行评估，数据集为AMOS22。仅用10个案例，模型的性能与在数量级上更多数据上训练的专业系统相媲美。值得注意的是，这包括可靠的胆囊分割（Dice 0.68 CT, 0.59 MRI），而现有工具几乎完全失效（Dice 0.0004），同时在肝脏、肾脏和脾脏的表现上与MRSegmentator相差仅5-10%，而所需标注数量减少了超过100倍。此外，在Whole Heart Segmentation数据集上的外部验证表明，该方法可扩展到心脏分割，这是TotalSegmentator（MRI）和MRSegmentator的应用范围之外，仅用10个标注案例便实现了竞争力的左心室（LV）性能。训练每个器官仅需3-5小时，使用单个GPU，速度约为nnU-Net的2-3倍。这些发现表明，十个标注案例足以实现临床有用的分割，有效减少了图像标注和训练时间的瓶颈。
cs.CV / 55 / 2608.18734
CL4D: Contrastive Language-4D Pretraining for Vision-Language Reasoning in Dynamic Scenes
CL4D：用于动态场景中视觉-语言推理的对比语言-4D 预训练
Abstract
4D understanding and reasoning is a fundamental capability for embodied AI agents operating in dynamic physical environments. However, existing vision encoders are largely limited to static 2D images or 3D point clouds without temporal modeling, or to 2D videos that lack accurate geometric depth reasoning. Consequently, current approaches fail to jointly capture spatial structure and motion evolution in dynamic scenes. We present CL4D, the first foundational 4D vision encoder that directly operates on dynamic point clouds, trained with a contrastive learning objective to align spatio-temporal geometric representations with natural language descriptions. By learning a shared embedding space between text and 4D scene dynamics, CL4D enables zero-shot motion-to-text and text-to-motion retrieval in dynamic environments and serves as a foundational 4D vision encoder for downstream 4D vision-language tasks. Building on this encoder, we introduce 4DVLM, a 4D vision-language model that conditions language generation on dynamic geometric representations. 4DVLM is the first VLM designed to operate directly on 4D point clouds without relying on 2D images, 2D videos, or static 3D point clouds. We train CL4D and subsequently 4DVLM on a newly constructed dataset termed DynAction4D capturing diverse human motions across varying object interactions and scene environments. Extensive experiments across multiple 4D human action benchmarks demonstrate that CL4D achieves state-of-the-art performance, with improvements of approximately ~16.75% over prior methods. Furthermore, 4DVLM outperforms frontier video VLMs such as Gemini and GPT-5 even when these models are provided with RGB video sequences corresponding to the same scenes represented as 4D point clouds for 4DVLM.
Chinese Translation
4D 理解和推理是嵌入式人工智能代理在动态物理环境中操作的基本能力。然而，现有的视觉编码器主要局限于静态的 2D 图像或没有时间建模的 3D 点云，或者是缺乏准确几何深度推理的 2D 视频。因此，当前的方法未能共同捕捉动态场景中的空间结构和运动演变。我们提出 CL4D，这是首个直接在动态点云上操作的基础 4D 视觉编码器，采用对比学习目标进行训练，以将时空几何表示与自然语言描述对齐。通过学习文本与 4D 场景动态之间的共享嵌入空间，CL4D 实现了在动态环境中零-shot 的运动到文本和文本到运动检索，并作为下游 4D 视觉-语言任务的基础 4D 视觉编码器。在此编码器的基础上，我们引入了 4DVLM，一个将语言生成条件于动态几何表示的 4D 视觉-语言模型。4DVLM 是首个设计为直接在 4D 点云上操作的 VLM，无需依赖 2D 图像、2D 视频或静态 3D 点云。我们在新构建的数据集 DynAction4D 上训练 CL4D 和随后 4DVLM，该数据集捕捉了不同物体交互和场景环境下多样的人类动作。在多个 4D 人类动作基准测试中的广泛实验表明，CL4D 达到了最先进的性能，相较于之前的方法提升了约 ~16.75%。此外，4DVLM 在前沿视频 VLM（如 Gemini 和 GPT-5）上表现优越，即使这些模型提供了与作为 4D 点云表示的相同场景对应的 RGB 视频序列。
cs.CV / 56 / 2608.18755
SED-FOD: Scattering-Aware Expert Decomposition for Few-Shot Cross-Sensor SAR Object Detection
SED-FOD：面向少样本跨传感器SAR目标检测的散射感知专家分解
Abstract
Synthetic aperture radar (SAR) object detection is an important part of remote sensing interpretation. However, because of variations in frequency band, resolution, background clutter, and target scattering responses, the performance of existing detectors often degrades when training and testing data are acquired from different SAR domains. Although domain adaptation methods offer a promising paradigm for solving this problem, most of them mainly pursue domain-invariant feature alignment and suppress sensor-dependent scattering characteristics that are useful for object detection. This problem becomes more challenging in few-shot scenarios, where only a few fully annotated target-domain SAR images are available. To address this issue, we propose a scattering-aware shared-specific feature decomposition framework for few-shot SAR domain adaptation object detection. We decompose detection features into a shared path and several soft-gated scattering-specific expert paths. The shared path learns transferable object structural information and is used for asymmetric domain alignment, while the scattering-specific experts adaptively compensate heterogeneous SAR responses. In addition, routing-domain auxiliary loss is introduced to encourage specific experts to capture sensor-dependent routing preferences, and an expert balancing loss is used to prevent routing collapse. Extensive experiments on four bidirectional heterogeneous SAR detection tasks between FARAD-X/FARAD-Ka and MiniSAR under different few-shot settings have been conducted and experimental results demonstrate that the proposed method achieves superior performance in both forward and reverse adaptation directions.
Chinese Translation
合成孔径雷达（SAR）目标检测是遥感解译的重要组成部分。然而，由于频带、分辨率、背景杂波和目标散射响应的变化，现有检测器的性能在训练和测试数据来自不同SAR领域时往往会下降。尽管领域适应方法为解决这一问题提供了有前景的范式，但大多数方法主要追求领域不变特征对齐，并抑制对目标检测有用的传感器依赖散射特征。在仅有少量完全标注的目标领域SAR图像可用的少样本场景中，这一问题变得更加棘手。为了解决这一问题，我们提出了一种面向少样本SAR领域适应目标检测的散射感知共享特定特征分解框架。我们将检测特征分解为一个共享路径和几个软门控散射特定专家路径。共享路径学习可转移的目标结构信息，并用于不对称领域对齐，而散射特定专家则自适应补偿异构SAR响应。此外，引入路由领域辅助损失以鼓励特定专家捕捉传感器依赖的路由偏好，并使用专家平衡损失防止路由崩溃。我们在FARAD-X/FARAD-Ka与MiniSAR之间的四个双向异构SAR检测任务上进行了广泛实验，实验结果表明，所提方法在正向和反向适应方向上均表现出优越的性能。
cs.CV / 57 / 2608.18774
MIFR: A Modality-Invariant and Fair Representation Framework for Skin Disease Classification
MIFR：一种用于皮肤疾病分类的模态不变和公平表示框架
Abstract
Skin diseases represent a major global public health burden, yet machine learning tools developed to assist in their diagnosis suffer from two critical limitations: reliance on only one modality for diagnosis and systematic performance disparities across skin tones. While existing approaches address each challenge separately, this work proposes a modality-invariant framework with fair representation (MIFR) for skin disease classification. The architecture pairs clinical photographs with dermoscopic images using ViT-based encoders, projecting each input into a high-dimensional embedding space via modality-specific projection heads. The resulting model is trained with a five-component multi-objective loss including weighted cross-entropy for classification, confusion and skin-type classification losses for fairness, per-modality supervised contrastive loss for class alignment, and a modality-invariance loss for clinical and dermoscopic modality alignment. Experiments on the HIBA+Derm7pt paired dataset and the external PAD-UFES-20 and ISIC 2019 datasets showed that modality-invariant representation learning provides competitive predictive performance compare to relevant baseline models and competitive fairness on the internal dataset. t-SNE visualizations confirmed that clinical and dermoscopic embeddings of the same disease are geometrically aligned, validating the joint objectives.
Chinese Translation
皮肤疾病是全球公共健康的重要负担，但为其诊断而开发的机器学习工具存在两个关键限制：仅依赖单一模态进行诊断以及在肤色之间存在系统性的性能差异。虽然现有方法分别解决了这两个挑战，但本研究提出了一种模态不变的公平表示框架（MIFR）用于皮肤疾病分类。该架构使用基于ViT的编码器将临床照片与皮肤镜图像配对，通过模态特定的投影头将每个输入投影到高维嵌入空间。所得到的模型采用五个组成部分的多目标损失进行训练，包括用于分类的加权交叉熵、用于公平性的混淆和肤色分类损失、用于类别对齐的每模态监督对比损失，以及用于临床和皮肤镜模态对齐的模态不变损失。在HIBA+Derm7pt配对数据集以及外部PAD-UFES-20和ISIC 2019数据集上的实验表明，模态不变表示学习在预测性能上与相关基线模型具有竞争力，并在内部数据集上表现出竞争性的公平性。t-SNE可视化结果确认了同一疾病的临床和皮肤镜嵌入在几何上是对齐的，验证了联合目标的有效性。
cs.CV / 58 / 2608.18832
EfficientSync: Real-Time Lip Synchronization via Deformation-Based Reference Texture Mixing
EfficientSync：基于变形的参考纹理混合实现实时唇部同步
Abstract
Audio-driven lip synchronization manipulates the mouth region of a talking-face video to match the driving audio while preserving head pose, identity, and background. Although the task is inherently local editing, prevailing approaches reconstruct the entire lower face with heavy GAN- or diffusion-based decoders, incurring substantial latency and, more critically, hallucinating intra-oral details such as teeth and lip wrinkles instead of preserving authentic textures. We contend that the bottleneck in identity preservation is not the scarcity of reference frames, but the lack of a mechanism that faithfully transfers the genuine textures they already contain. We therefore present EfficientSync, a real-time deformation-based framework that retains reference textures rather than resynthesizing them. First, the Dynamic Texture Mixer reformulates multi-reference fusion as channel-wise selection, evaluating each spatially aligned reference in a global context and aggregating them by channel-wise weighted summation, preserving textural integrity at low cost. Second, Spatio-Temporal Shifted Adaptive Masking decomposes the source frame into lip-generation conditions and an independent background prior, suppressing lower-face leakage while blending the synthesized mouth seamlessly into the background. Third, STAR Sampling, a zero-overhead pre-processing step, retrieves the sharpest and most topologically diverse reference frames. Experiments on HDTF and VFHQ show state-of-the-art visual quality and identity preservation at 166 FPS on a single GPU. Video demos: https://alunaticat.github.io/EfficientSync/index.html.
Chinese Translation
音频驱动的唇部同步通过操控说话者面部视频的嘴部区域，使其与驱动音频匹配，同时保持头部姿态、身份和背景。尽管该任务本质上是局部编辑，但现有方法通常使用重型的生成对抗网络（GAN）或扩散模型解码器重建整个下半脸，导致显著的延迟，更严重的是，产生口腔内部细节的幻觉，如牙齿和唇皱，而不是保留真实的纹理。我们认为，身份保留的瓶颈并非参考帧的稀缺，而是缺乏一种能够忠实传递其所包含真实纹理的机制。因此，我们提出了EfficientSync，这是一种实时的基于变形的框架，保留参考纹理而不是重新合成它们。首先，动态纹理混合器将多参考融合重新表述为通道选择，在全球上下文中评估每个空间对齐的参考，并通过通道加权求和进行聚合，以低成本保持纹理完整性。其次，时空自适应掩蔽将源帧分解为唇部生成条件和独立的背景先验，抑制下半脸泄漏，同时将合成的嘴部无缝融入背景。第三，STAR采样作为零开销的预处理步骤，检索最清晰和拓扑多样的参考帧。在HDTF和VFHQ上的实验显示，在单个GPU上以166 FPS实现了最先进的视觉质量和身份保留。视频演示： https://alunaticat.github.io/EfficientSync/index.html。
cs.CV / 59 / 2608.18833
EVADE: Evidence-Verified Agentic Diagnosis with Escape
EVADE：证据验证的自主诊断与逃逸
Abstract
Medical vision-language models (VLMs) can achieve high accuracy but remain unreliable: they are systematically overconfident, benefit little from test-time reasoning, and lack the ability to reliably calibrate trust in their own responses. We introduce EVADE (Evidence-Verified Agentic Diagnosis with Escape), an inferential, non-training method that enhances the safety of deploying a single frozen VLM. EVADE responds and, when uncertain, localises the region most diagnostically relevant, re-answers on a zoomed view, and commits only when both the entire image and the zoomed view responses agree; otherwise, it abstains. To directly address verification hallucination in single-model self-checking, our main idea is to verify gate consistency across different image views rather than re-reading the model's own text. Experimental evaluation on VQA-RAD, SLAKE, and PathVQA using Qwen2.5-VL-7B reports that EVADE is the only method that simultaneously improves both calibration and selective risk while maintaining accuracy, reducing expected calibration error (ECE) by up to 45% compared to zero-shot. Chain-of-thought, self-consistency, and self-verification all fail at least one axis. A grounding analysis reports that self-proposed regions perform better at diagnostic structure localisation than centres or random crops. However, a 7B VLM cannot use this localisation to revise answers. Therefore, reliability gains come from the consistency gate and calibrated abstention.
Chinese Translation
医学视觉语言模型（VLMs）能够实现高准确率，但仍然不可靠：它们系统性地过于自信，几乎没有从测试时推理中受益，并且缺乏可靠校准自身响应信任的能力。我们提出了EVADE（证据验证的自主诊断与逃逸），这是一种推理性、非训练的方法，旨在增强单个冻结VLM的部署安全性。EVADE在响应时，如果不确定，则定位最具诊断相关性的区域，在放大视图上重新回答，并且仅在整个图像和放大视图的响应一致时才做出承诺；否则，它将选择不作答。为了直接解决单模型自检中的验证幻觉，我们的主要思路是验证不同图像视图之间的门控一致性，而不是重新阅读模型自身的文本。对VQA-RAD、SLAKE和PathVQA的实验评估使用Qwen2.5-VL-7B，结果表明EVADE是唯一一种同时提高校准和选择性风险，同时保持准确性的的方法，与零-shot相比，预期校准误差（ECE）减少了多达45%。链式思维、自我一致性和自我验证在至少一个维度上均未能成功。基础分析表明，自我提出的区域在诊断结构定位方面的表现优于中心或随机裁剪。然而，7B VLM无法利用这种定位来修正答案。因此，可靠性提升来自于一致性门控和经过校准的选择不作答。
cs.CV / 60 / 2608.18864
RVLoss: Runoff Vote Loss for Self-Supervised LiDAR Scene Flow Estimation
RVLoss：用于自监督LiDAR场景流估计的径流投票损失
Abstract
LiDAR scene flow estimates point-wise motion between two consecutive scans, referred to as the source and target. Leading self-supervised methods typically minimize the Chamfer loss, the nearest neighbor distance between the flow-compensated source and the target. However, nearest-neighbor search does not enforce motion rigidity, often leading to inconsistent flows within object instances. Existing approaches address this issue with additional regularization terms, but flow consistency among points remains limited, especially for large objects. We propose RVLoss, a self-supervised loss that incorporates motion rigidity by design through a runoff vote mechanism. Our key observation is that the point-wise motion, calculated from nearest neighbor search, can often be grouped into a small set of dominant flow candidates by voting (top-k voting). Furthermore, when compensating the source by these candidates, the flow that best represents the underlying rigid motion often yields the highest consensus after a second voting (top-1 voting). Based on this insight, we incorporate the two-stage runoff vote into loss design and create cluster-wise rigid flows and free-form flows as pseudo-labels for self-supervised learning. RVLoss can be seamlessly integrated into existing feedforward architectures. Experiments on the Argoverse2 2026 Challenge show that models trained with RVLoss achieve state-of-the-art performance among self-supervised approaches, outperforming baseline models trained with alternative loss designs by 20%. Moreover, cross-dataset evaluations demonstrate consistent performance improvements across four additional datasets. Code will be released upon acceptance.
Chinese Translation
LiDAR场景流估计点对点地描述了两个连续扫描之间的运动，称为源和目标。领先的自监督方法通常通过最小化Chamfer损失，即流补偿后的源与目标之间的最近邻距离。然而，最近邻搜索并未强制执行运动的刚性，常常导致对象实例内的不一致流。现有方法通过额外的正则化项来解决此问题，但点之间的流一致性仍然有限，尤其是在大物体的情况下。我们提出了RVLoss，这是一种通过径流投票机制设计的自监督损失，内置了运动刚性。我们的关键观察是，从最近邻搜索计算出的点对点运动，通常可以通过投票（top-k投票）归类为一小组主导流候选者。此外，当通过这些候选者补偿源时，最佳代表基础刚性运动的流在第二次投票（top-1投票）后通常会产生最高的一致性。基于这一见解，我们将两阶段径流投票纳入损失设计，并创建基于聚类的刚性流和自由形式流作为自监督学习的伪标签。RVLoss可以无缝集成到现有的前馈架构中。在Argoverse2 2026挑战赛上的实验表明，使用RVLoss训练的模型在自监督方法中达到了最先进的性能，超越了使用其他损失设计训练的基线模型20%。此外，跨数据集评估显示在四个额外数据集上也有一致的性能提升。代码将在接受后发布。
cs.CV / 61 / 2608.18881
Falcon Perception-HD: High Density Perception via Reinforcement Learning
Falcon Perception-HD：通过强化学习实现高密度感知
Abstract
Autoregressive perception models trained to localize visual entities under the open-vocabulary setting are mostly trained using Supervised fine-tuning (SFT) with maximum likelihood, yet it optimizes a proxy objective (per-token cross-entropy) that is fundamentally misaligned with perception metrics such as precision and recall. In this paper, we explore post-training reinforcement learning (RL), specifically GRPO, to directly align these models with their evaluation metrics. Building up on the recently introduced Falcon Perception, we design an RL framework that addresses perception-specific challenges: reward design for set-structured outputs and multi-head sampling control. We discover multiple benefits from RL for perception: first, RL unlocks state-of-the-art performance in very dense scenes (up to 500 objects per scene), a regime where most existing systems degrade sharply or collapse; furthermore it fixes common issues in autoregressive perception models like mask repetitions and removes almost entirely the need for NMS and coordinate deduplication, which improve both performance and efficiency and remove the need for hyperparameters tuning; overall, we notice improvements on all levels of difficulties in referring expression segmentation (on PBench and SACO-Gold), and we find an elegant way to preserve the knowledge of whether an object exists or not (as evaluated by MCC) without training on negative samples. We show that a simple reward that penalizes false negatives and positives is sufficient. We develop two hybrid self-annotation pipelines, respectively tailored for difficult referring expressions and very dense scenes, and show their benefits on RL-training. Model weights are released as a Falcon Perception revision~\footnote{https://huggingface.co/tiiuae/Falcon-Perception}. Datasets will be published.
Chinese Translation
在开放词汇设置下训练的自回归感知模型主要通过最大似然的监督微调（SFT）进行训练，但它优化的是一个与感知指标（如精确度和召回率）根本不一致的代理目标（每个标记的交叉熵）。在本文中，我们探讨了后训练强化学习（RL），特别是GRPO，以直接将这些模型与其评估指标对齐。在最近提出的Falcon Perception的基础上，我们设计了一个RL框架，解决感知特定的挑战：针对集合结构输出的奖励设计和多头采样控制。我们发现RL在感知方面带来了多个好处：首先，RL在非常密集的场景中（每个场景最多500个物体）解锁了最先进的性能，而在这一领域，大多数现有系统急剧下降或崩溃；此外，它解决了自回归感知模型中的常见问题，如掩码重复，并几乎完全消除了对非极大值抑制（NMS）和坐标去重的需求，从而提高了性能和效率，并消除了超参数调优的需求；总体而言，我们注意到在引用表达分割的所有难度级别上都有所改善（在PBench和SACO-Gold上），并发现了一种优雅的方法来保留对象是否存在的知识（通过MCC评估），而无需在负样本上训练。我们表明，简单的奖励机制，惩罚假阴性和假阳性，已足够。我们开发了两个混合自我注释管道，分别针对困难的引用表达和非常密集的场景，并展示了它们在RL训练中的优势。模型权重作为Falcon Perception的修订版本发布~ ootnote{https://huggingface.co/tiiuae/Falcon-Perception}。数据集将会发布。
cs.CV / 62 / 2608.18907
Learning-State-Aware Dynamic Generative Data Augmentation on Small-Scale Datasets
学习状态感知的动态生成数据增强在小规模数据集上的应用
Abstract
Small-scale image classification is often limited by the scarcity of training data. Generative data augmentation (GDA) based on pretrained generative models has emerged as an effective solution. However, existing methods rely on task-agnostic augmentation strategies that overlook downstream model needs. Although recent dynamic GDA methods incorporate model feedback to guide augmentation, they still struggle to reliably determine sample-specific augmentation strengths and adapt augmentation strategies to different image regions while balancing image diversity and class semantics. To address these issues, we propose learning-state-aware dynamic generative data augmentation (LSADA). Specifically, LSADA constructs a learning state for each sample based on its current loss and loss-decrease rate, which is then mapped to a sample-specific augmentation strength. Furthermore, LSADA introduces a decoupled data augmentation and diffusion fusion strategy that applies strength-controlled transformations to class-relevant regions and generates diverse class-irrelevant regions, progressively fusing them to improve image diversity while preserving class semantics. Experiments on nine public datasets show that LSADA outperforms the existing SOTA dynamic GDA method by an average of 4.5% on six natural image datasets and 2.5% on three medical image datasets.
Chinese Translation
小规模图像分类常常受到训练数据稀缺的限制。基于预训练生成模型的生成数据增强（GDA）已成为一种有效的解决方案。然而，现有方法依赖于与任务无关的增强策略，忽视了下游模型的需求。尽管最近的动态 GDA 方法结合了模型反馈来指导增强，但在可靠地确定样本特定的增强强度和适应不同图像区域的增强策略方面仍然存在困难，同时需要平衡图像多样性和类别语义。为了解决这些问题，我们提出了学习状态感知的动态生成数据增强（LSADA）。具体而言，LSADA 根据每个样本当前的损失和损失下降率构建学习状态，然后将其映射到样本特定的增强强度。此外，LSADA 引入了一种解耦的数据增强和扩散融合策略，对与类别相关的区域应用强度控制的变换，并生成多样化的与类别无关的区域，逐步融合它们以提高图像多样性，同时保留类别语义。在九个公共数据集上的实验表明，LSADA 在六个自然图像数据集上平均超越现有的最先进动态 GDA 方法 4.5%，在三个医学图像数据集上超越 2.5%。
cs.CV / 63 / 2608.18915
Simple, Safe, and Overlooked: Reclaiming Sustainable Domain Generalization with Statistical Color Matching
简单、安全且被忽视：通过统计颜色匹配重新获得可持续的领域泛化
Abstract
Hardware shifts, color variations, and changing patient characteristics between development and deployment routinely break trained medical image classifiers. Existing remedies fall short: standard color jittering provides insufficient diversity, while deep generative style transfer algorithms hallucinate features, destroy clinically relevant structures, and waste massive compute resources. To address this, we revisit classical statistical color matching and repurpose it as Colorist, a highly efficient data augmentation strategy that applies global mean-standard deviation matching directly in the RGB color space. We demonstrate that this training-free, fully interpretable approach safely generates structurally intact domain variations, outperforming deep generative models in structural fidelity and color alignment. Across out-of-distribution histopathology, peripheral blood, dermatology, and retinal datasets, it improves balanced accuracy by up to +9% over state-of-the-art domain generalization regularizers and by +13% over an unaugmented baseline. Moreover, by avoiding neural networks in the augmentation loop, Colorist preserves anatomical structure, minimizes carbon footprint, and integrates seamlessly into standard dataloaders. Together, these findings establish statistical matching as a safe, interpretable, yet overlooked alternative to deep architectures for clinical robustness. Source code is available at https://github.com/sdoerrich97/colorist.
Chinese Translation
硬件变化、颜色差异以及开发与部署之间患者特征的变化常常导致训练好的医学图像分类器失效。现有的解决方案效果不佳：标准的颜色抖动提供的多样性不足，而深度生成风格迁移算法则会幻觉出特征，破坏临床相关结构，并浪费大量计算资源。为了解决这个问题，我们重新审视经典的统计颜色匹配，并将其重新用于Colorist，这是一种高效的数据增强策略，直接在RGB颜色空间中应用全局均值-标准差匹配。我们证明了这种无训练、完全可解释的方法安全地生成结构完整的领域变体，在结构保真度和颜色对齐方面超越了深度生成模型。在超出分布的组织病理学、外周血、皮肤病学和视网膜数据集上，它的平衡准确率比最先进的领域泛化正则化器提高了多达9%，比未增强的基线提高了13%。此外，通过在增强循环中避免使用神经网络，Colorist保留了解剖结构，最小化了碳足迹，并与标准数据加载器无缝集成。综合来看，这些发现确立了统计匹配作为一种安全、可解释且被忽视的替代方案，以增强临床的鲁棒性。源代码可在 https://github.com/sdoerrich97/colorist 获取。
cs.CV / 64 / 2608.18957
Institutional Books - Visual Elements: An open-source pipeline for extracting, classifying, deduplicating, and captioning visual elements from digital book collections
机构书籍 - 视觉元素：一个开源管道，用于从数字书籍收藏中提取、分类、去重和标注视觉元素
Abstract
Historical book collections contain rich visual elements - such as illustrations, photographs, engravings, and decorative art - that are frequently under-explored in large-scale digitization projects. While Optical Character Recognition (OCR) has standardized the extraction of textual content, these visual components offer a layer of nuance and context that remains largely untapped by automated text extraction workflows. This technical report introduces Institutional Books - Visual Elements, an open-source end-to-end pipeline for detecting, classifying, deduplicating, and captioning visual elements from historical book collections. Alongside this pipeline, we release an initial dataset of 22.6 million visual elements extracted from the 983,004 scanned volumes that comprise the Institutional Books: Harvard Library dataset. This work contributes to ongoing, community-wide efforts to enable new use cases for digitized library collections through computational access, from artificial intelligence model training to digital humanities research.
Chinese Translation
历史书籍收藏中包含丰富的视觉元素，如插图、照片、雕刻和装饰艺术，这些元素在大规模数字化项目中常常未得到充分探索。尽管光学字符识别（OCR）已标准化文本内容的提取，但这些视觉组件提供的细微差别和背景信息在自动文本提取工作流程中仍然未被充分利用。本技术报告介绍了机构书籍 - 视觉元素，这是一个开源的端到端管道，用于检测、分类、去重和标注来自历史书籍收藏的视觉元素。与此管道一起，我们发布了一个初始数据集，其中包含从983,004本扫描卷中提取的2260万个视觉元素，这些卷构成了机构书籍：哈佛图书馆数据集。这项工作为通过计算访问使数字化图书馆收藏的新用例成为可能的社区广泛努力做出了贡献，从人工智能模型训练到数字人文学科研究。
cs.CV / 65 / 2608.18968
Frozen DINO Localizes Image Edits Without a Localizer
冻结的 DINO 无需定位器即可定位图像编辑
Abstract
Localized image edits can change a photograph's meaning while leaving most of it authentic, so forensic analysis must identify where an edit occurred. We show that patch-level perturbation responses from frozen DINO encoders are themselves localization maps. Training-free Localization of AI-image Edits from patch-token Drift (TRAIL) applies one global Haar perturbation and maps cosine drift between corresponding patch tokens. On 80 source-disjoint CocoGlide test images, TRAIL reaches .903 patch AUROC versus .912 for the mask-supervised Detective SAM; fixed-threshold Dice is .619 versus .709, while an oracle threshold raises TRAIL to .790. Transferred unchanged to Poisson image interpolation, TRAIL reaches .855 AUROC versus .864, showing that the cue persists without a generator. Across sixteen DINO encoders, the best block lies at normalized depth .80-.94. Global context matters: AUROC falls from .903 globally to .857 for local-in-canvas perturbations and .735 for independently encoded crops. Frozen DINO patch tokens therefore contain a strong late-layer localization signal whose visibility depends on the perturbation and preserved context. Code: https://github.com/VishalJ99/trail-image-edit-localization.
Chinese Translation
局部图像编辑可以在保持大部分图像真实性的同时改变照片的含义，因此法医分析必须识别编辑发生的位置。我们展示了来自冻结的 DINO 编码器的补丁级扰动响应本身就是定位图。无训练的 AI 图像编辑定位方法（Training-free Localization of AI-image Edits from patch-token Drift, TRAIL）应用一种全局 Haar 扰动，并映射相应补丁标记之间的余弦漂移。在 80 张源不重叠的 CocoGlide 测试图像上，TRAIL 达到 0.903 的补丁 AUROC，而掩膜监督的 Detective SAM 达到 0.912；固定阈值的 Dice 指标为 0.619，而 0.709 的 oracle 阈值将 TRAIL 提升至 0.790。TRAIL 在泊松图像插值中未发生变化，达到 0.855 的 AUROC，而后者为 0.864，表明该线索在没有生成器的情况下依然存在。在十六个 DINO 编码器中，最佳块位于归一化深度 0.80-0.94。全局上下文很重要：AUROC 从全局的 0.903 降至局部画布扰动的 0.857，以及独立编码裁剪的 0.735。因此，冻结的 DINO 补丁标记包含强烈的后层定位信号，其可见性取决于扰动和保留的上下文。代码链接： https://github.com/VishalJ99/trail-image-edit-localization.
cs.CV / 66 / 2608.18979
When Simplicity Wins: Bottleneck-Aware Context Modeling for Lightweight Semantic Segmentation
当简单胜出：针对轻量级语义分割的瓶颈感知上下文建模
Abstract
Semantic segmentation demands a careful balance between accuracy, efficiency, and scalability, which remains difficult to achieve for high-resolution imagery. Convolutional networks effectively model local patterns but struggle with long-range dependencies, whereas Vision Transformers capture global context at a high computational cost. While recent work largely focuses on encoder design, the bottleneck stage, central to contextual aggregation and information flow, has been relatively overlooked. We propose SiConMo, a lightweight yet effective framework, implemented in two variants: an RGB-only model (SiConMo) and a GME-enhanced variant (SiConMo$_\dagger$). We show that simplicity arises from a key design principle: at very low computational budgets, the bottleneck is the most efficient stage to integrate local and global context. SiConMo integrates three complementary components: a Token Pyramid Extraction Module for hierarchical multi-scale representation, a Transformer-Branched Depthwise Convolution block for bottleneck-aware context modeling, and a Feature Merging Module that preserves spatial structure while enhancing semantic consistency. Extensive experiments on ADE20K, PASCAL Context, Cityscapes, and COCO-Stuff demonstrate that SiConMo achieves a state-of-the-art accuracy-efficiency trade-off among lightweight semantic segmentation models, highlighting simplicity as a powerful design principle.
Chinese Translation
语义分割需要在准确性、效率和可扩展性之间进行仔细平衡，这在高分辨率图像中仍然难以实现。卷积网络有效地建模局部模式，但在长距离依赖方面表现不佳，而视觉变换器（Vision Transformers）则以较高的计算成本捕捉全局上下文。尽管最近的研究主要集中在编码器设计上，但作为上下文聚合和信息流的核心，瓶颈阶段相对被忽视。我们提出了SiConMo，一个轻量且有效的框架，分为两种变体：仅RGB模型（SiConMo）和增强的GME变体（SiConMo$_ ag{†}$）。我们表明，简单性源于一个关键设计原则：在非常低的计算预算下，瓶颈是整合局部和全局上下文的最有效阶段。SiConMo整合了三个互补组件：用于分层多尺度表示的Token金字塔提取模块、用于瓶颈感知上下文建模的变换器分支深度卷积块，以及在增强语义一致性的同时保持空间结构的特征合并模块。在ADE20K、PASCAL Context、Cityscapes和COCO-Stuff上的大量实验表明，SiConMo在轻量级语义分割模型中实现了最先进的准确性与效率的权衡，突显了简单性作为一种强大的设计原则。
cs.CV / 67 / 2608.18984
Uncertainty-Aware Art-Historical Dating with Vision-Language Models
基于视觉-语言模型的不确定性感知艺术历史定年
Abstract
Museum and archival datasets do not mirror historical artistic production, but materialize the contingent histories of collecting, preservation, cataloging, and digitization. This has direct consequences for interpreting pretrained image representations: they may appear to encode historical time while actually encoding the institutional conditions under which objects become visible as data. We describe this phenomenon as temporal entanglement and investigate it by formulating artwork dating as an uncertainty-aware regression task over frozen image embeddings. We evaluate several pretrained vision models on a temporally controlled Wikidata corpus of artworks. Our results show that these models contain usable temporal information, with Vision-Language Models (VLMs) outperforming purely visual self-supervised baselines. However, a qualitative analysis indicates that this temporal knowledge is shaped by various biases.
Chinese Translation
博物馆和档案数据集并不反映历史艺术创作，而是体现了收集、保存、编目和数字化的偶然历史。这对解释预训练图像表示具有直接影响：它们可能看似编码了历史时间，但实际上编码的是物体作为数据可见的制度条件。我们将这种现象描述为时间纠缠，并通过将艺术品定年表述为一种不确定性感知的回归任务来进行研究，使用冻结的图像嵌入。我们在一个时间控制的维基数据艺术品语料库上评估了几种预训练的视觉模型。我们的结果表明，这些模型包含可用的时间信息，其中视觉-语言模型（VLMs）优于纯视觉自监督基线。然而，定性分析表明，这种时间知识受到各种偏见的影响。
cs.CV / 68 / 2608.18986
X-LMC: Cross-View Spatiotemporal Collateral Circulation Scoring from DSA
X-LMC：基于数字减影血管造影的跨视图时空侧支循环评分
Abstract
Digital subtraction angiography (DSA) is the reference standard for leptomeningeal collateral (LMC) assessment, providing critical prognostic insights to guide secondary treatment strategies, neurorehabilitation planning, and retrospective stroke research. However, clinical LMC grading via the ASITN/SIR scale relies on manual, highly variable visual inspection. We introduce X-LMC, a spatiotemporal framework for automated collateral scoring from time-resolved biplane DSA. The proposed architecture encodes spatial frame representations through a DINOv2 backbone, fuses orthogonal projections via a token-level cross-view attention module, and models representations of contrast bolus dynamics using a recurrent network architecture. We evaluate our framework on a multicenter dataset of 134 patients with M1-segment occlusions. In a 5-fold cross-validation setting, X-LMC yields higher point estimates than static architectures and spatiotemporal baselines adapted from related angiographic tasks, achieving a Quadratic Weighted Kappa (QWK) of 0.398 (vs. 0.322) and a dichotomized macro-F1 score of 0.711 (vs. 0.663) against the best-performing baseline. X-LMC performance also aligns with the observed clinical inter-rater agreement (QWK: 0.314). As the first DSA study attempting to automate LMC scoring, we demonstrate that multi-view temporal deep learning can capture collateral-specific contrast kinetics. Ultimately, these benchmarks delineate the clinical ambiguities and achievable performance boundaries of automated ASITN/SIR grading, establishing a reproducible foundation for objective hemodynamic phenotyping in stroke cohorts. Code is available at https://github.com/maedehafezi/X-LMC.
Chinese Translation
数字减影血管造影（DSA）是评估脑膜下侧支循环（LMC）的参考标准，为指导二级治疗策略、神经康复规划和回顾性中风研究提供了重要的预后见解。然而，临床LMC分级依赖于ASITN/SIR评分标准的人工视觉检查，这种方法高度可变且主观。我们提出了X-LMC，一个基于时空框架的自动化侧支评分方法，利用时间分辨的双平面DSA。所提出的架构通过DINOv2主干网络编码空间帧表示，通过一个基于标记级别的跨视图注意模块融合正交投影，并使用递归网络架构建模对比剂动力学的表示。我们在一个包含134名M1段闭塞患者的多中心数据集上评估了我们的框架。在5折交叉验证设置中，X-LMC的点估计优于静态架构和从相关血管造影任务改编的时空基线，达到0.398的二次加权Kappa（QWK）（对比0.322）和0.711的二分宏观F1分数（对比0.663），相较于表现最佳的基线。X-LMC的性能也与观察到的临床评分者间一致性相符（QWK：0.314）。作为首个尝试自动化LMC评分的DSA研究，我们展示了多视图时间深度学习能够捕捉特定于侧支的对比剂动力学。最终，这些基准划定了自动化ASITN/SIR分级的临床模糊性和可实现的性能边界，为中风队列中的客观血流动力学表型建立了可重复的基础。代码可在https://github.com/maedehafezi/X-LMC获取。
cs.CV / 69 / 2608.18993
ForeSightGuide: An Anticipatory Framework toward Accurate and Low-Redundancy Guidance for the Visually Impaired
ForeSightGuide：面向视觉障碍者的准确且低冗余指导的预期框架
Abstract
Electronic travel aids are pivotal for the independent mobility of the visually impaired. While Vision-Language Models (VLMs) offer rich environmental understanding, they often suffer from excessive false positives in dynamic scenarios, leading to cognitive overload. To address this, we present ForeSightGuide, an anticipatory assistive guidance framework that couples semantic scene understanding with predictive hazard assessment. Unlike reactive systems, ForeSightGuide leverages the reasoning capabilities of VLMs to anticipate obstacle motion, effectively filtering out non-threatening objects to provide concise, actionable guidance. To validate our approach, we introduce a novel dataset captured in complex, dynamic real-world traffic scenes, designed to benchmark predictive capabilities. Extensive experiments on both public benchmarks and our proposed dataset demonstrate that ForeSightGuide achieves state-of-the-art performance. Notably, it significantly mitigates information overload by reducing redundant alerts to 0.299 per guidance output while maintaining a low missed-hazard rate of 0.112, proving its efficacy for safe walking assistance.
Chinese Translation
电子旅行辅助工具对视觉障碍者的独立移动至关重要。虽然视觉-语言模型（Vision-Language Models, VLMs）提供了丰富的环境理解，但在动态场景中，它们往往会出现过多的误报，导致认知负担过重。为了解决这一问题，我们提出了ForeSightGuide，一个将语义场景理解与预测危险评估相结合的预期辅助指导框架。与反应式系统不同，ForeSightGuide利用VLMs的推理能力来预测障碍物的运动，有效过滤出非威胁性物体，从而提供简洁、可操作的指导。为了验证我们的方法，我们引入了一个在复杂动态的真实交通场景中捕获的新数据集，旨在基准测试预测能力。在公共基准和我们提出的数据集上进行的广泛实验表明，ForeSightGuide达到了最先进的性能。值得注意的是，它通过将冗余警报减少到每次指导输出0.299，显著减轻了信息过载，同时保持了0.112的低漏报率，证明了其在安全步行辅助中的有效性。
cs.CV / 70 / 2608.18996
GrabVG: Graph-Attentive Binding for Visual Grounding in UAV Imagery
GrabVG：用于无人机影像视觉定位的图注意力绑定
Abstract
Visual grounding in Unmanned Aerial Vehicle (UAV) imagery aims to localize a target object in complex bird's-eye-view scenes according to a natural language description. However, the abundance of small, densely distributed, and visually similar objects creates high visual redundancy, while repetitive local configurations give rise to strong topological ambiguity. Existing approaches mainly focus on visual--language feature alignment or dense contextual interaction, yet they struggle to distinguish subtle inter-instance differences and effectively exploit spatial topological structures, leading to inaccurate grounding in highly crowded scenarios. To address these challenges, we propose $\textbf{GrabVG}$, a novel visual grounding framework inspired by human visual search. GrabVG explicitly decomposes grounding into two sequential stages: $\textit{preattentive hypothesis search}$ and $\textit{graph-attentive feature binding}$. Specifically, we first generate a compact set of reliable object hypotheses through distillation-guided proposal induction and text-aware hypothesis filtering, substantially reducing background distractions and semantic mismatches. These hypotheses are then organized into a sparse graph, where language-guided intra-instance visual cues and inter-instance topological relationships are jointly bound and propagated via graph attention, enabling efficient spatial reasoning and accurate target localization. Extensive experiments on AerialVG and AerialSense show that GrabVG achieves a favorable accuracy--speed trade-off, reaching 67.31$\%$ and 80.34$\%$
[email protected] and outperforming the corresponding baselines by 10.55 and 8.76 percentage points, respectively.
Chinese Translation
无人机（UAV）影像中的视觉定位旨在根据自然语言描述在复杂的鸟瞰场景中定位目标物体。然而，小型、密集分布且视觉相似的物体的丰富性导致了高视觉冗余，而重复的局部配置则引发了强烈的拓扑模糊性。现有方法主要集中于视觉与语言特征的对齐或密集的上下文交互，但在区分微妙的实例间差异和有效利用空间拓扑结构方面存在困难，导致在高度拥挤的场景中定位不准确。为了解决这些挑战，我们提出了$ extbf{GrabVG}$，一个受人类视觉搜索启发的新型视觉定位框架。GrabVG明确将定位过程分解为两个顺序阶段：$ extit{预注意假设搜索}$和$ extit{图注意力特征绑定}$。具体而言，我们首先通过蒸馏引导的提议诱导和文本感知的假设过滤生成一组紧凑且可靠的物体假设，显著减少背景干扰和语义不匹配。这些假设随后被组织成一个稀疏图，其中语言引导的实例内视觉线索和实例间拓扑关系通过图注意力共同绑定和传播，从而实现高效的空间推理和准确的目标定位。在AerialVG和AerialSense上的大量实验表明，GrabVG在准确性和速度之间达成了良好的平衡，分别达到了67.31$\%$和80.34$\%$的
[email protected]，并且分别比相应的基线提高了10.55和8.76个百分点。
cs.CV / 71 / 2608.19000
Mise-en-Sc\`ene: Implicit Layout Emergence in Diffusion Transformers for Human-AI Design Co-Creation
场景布置：在扩散变换器中隐式布局的出现用于人机设计共创
Abstract
Automating graphic design synthesis from user-provided elements requires both a coherent overall composition and the exact preservation of each asset. Existing methods predict a layout as explicit bounding-box coordinates with a language model and then paste the assets into it, which separates spatial planning from visual synthesis and tends to produce rigid, mis-scaled compositions. We instead ask whether the layout can emerge implicitly inside a pretrained image-editing diffusion transformer. We present Mise-en-Sc\`ene, a two-stage framework. In the first stage, a diffusion transformer adapted with a small, knockout-selected LoRA drafts a complete design in which the arrangement of the elements emerges jointly with the rendered canvas. In the second stage, a deterministic match-and-place step moves the original high-resolution assets to the drafted positions, which guarantees exact asset fidelity and yields an editable, layered design that a designer can keep refining rather than a flat image. Notably, a minimal adaptation of the pretrained transformer already suffices, without the specialized conditioning machinery commonly introduced for multi-element generation. On the large-scale PrismLayersPlus benchmark, the designs produced by Mise-en-Sc\`ene are the closest to the ground truth in perceived quality among all compared methods, by a wide margin over both an LLM layout planner and a specialized layout transformer, while our match-and-place stage bridges the remaining fidelity gap to the ground-truth composites.
Chinese Translation
从用户提供的元素自动化图形设计合成需要既有连贯的整体构图，又能精确保留每个资产。现有方法通过语言模型预测布局为显式的边界框坐标，然后将资产粘贴到其中，这种方法将空间规划与视觉合成分开，往往导致刚性和比例失调的构图。我们则提出一个问题：布局是否可以在预训练的图像编辑扩散变换器中隐式地出现。我们提出了Mise-en-Sc extbackslash`ene，一个两阶段框架。在第一阶段，经过小规模、选择性LoRA调整的扩散变换器草拟出一个完整的设计，其中元素的排列与渲染的画布共同出现。在第二阶段，一个确定性的匹配与放置步骤将原始高分辨率资产移动到草拟的位置，这保证了资产的精确保真，并生成一个可编辑的分层设计，设计师可以不断完善，而不是一个平面的图像。值得注意的是，预训练变换器的最小调整已足够，无需为多元素生成常用的专门条件机制。在大规模的PrismLayersPlus基准测试中，Mise-en-Sc extbackslash`ene生成的设计在感知质量上与所有比较方法中的真实情况最为接近，远远超过了LLM布局规划器和专门的布局变换器，而我们的匹配与放置阶段则弥补了与真实合成之间的剩余保真度差距。
cs.CV / 72 / 2608.19014
One-Stage Object Detectors in Autonomous Driving
自动驾驶中的一阶段目标检测器
Abstract
Autonomous vehicles depend on fast and reliable perception systems to detect surrounding vehicles, pedestrians, cyclists, traffic signs, and other road objects in real time. This paper presents a comprehensive survey and analysis of one-stage object detectors for autonomous driving rather than an implementation of a new detection system. The survey reviews the evolution of major one-stage detectors, including YOLOv1, SSD, RetinaNet, EfficientDet, anchor-free detectors such as FCOS and CenterNet, and recent real-time models such as YOLOv10. It compares these architectures through their design choices, feature-fusion strategies, loss functions, deployment trade-offs, and reported benchmark performance. The paper also summarizes commonly used autonomous-driving datasets, evaluation metrics, open challenges, and future research directions. Overall, this survey highlights how one-stage detectors balance speed, accuracy, efficiency, and robustness, while also emphasizing the remaining gap between benchmark results and dependable real-world autonomous-driving performance.
Chinese Translation
自动驾驶车辆依赖快速且可靠的感知系统实时检测周围的车辆、行人、自行车、交通标志及其他道路物体。本文对一阶段目标检测器在自动驾驶中的应用进行了全面的调查和分析，而非新检测系统的实现。调查回顾了主要一阶段检测器的发展历程，包括YOLOv1、SSD、RetinaNet、EfficientDet、无锚检测器如FCOS和CenterNet，以及最近的实时模型如YOLOv10。本文通过设计选择、特征融合策略、损失函数、部署权衡和报告的基准性能等方面对这些架构进行了比较。文章还总结了常用的自动驾驶数据集、评估指标、开放挑战和未来研究方向。总体而言，本调查强调了一阶段检测器如何在速度、准确性、效率和鲁棒性之间取得平衡，同时也强调了基准结果与可靠的现实世界自动驾驶性能之间的差距。
cs.CV / 73 / 2608.19021
Orthogonal Polynomial Approximation for Matrix Log Normalization in Global Covariance Pooling
全局协方差池化中的矩阵对数归一化的正交多项式近似
Abstract
Global Covariance Pooling (GCP) improves deep networks by capturing second-order feature statistics, and is especially effective for fine-grained recognition. Because covariance matrices live on the Symmetric Positive Definite (SPD) manifold, a normalization step is required before the Euclidean classifier. The faithful choice is the matrix logarithm (MLN-COV), which maps the SPD manifold to its tangent space; in practice it was abandoned in favour of the matrix square root because its eigendecomposition-based gradient is numerically unstable. We show that this instability is an artifact of computing the logarithm spectrally, not of the logarithm itself. Approximating the logarithm with finite polynomials in the covariance matrix removes the eigendecomposition from both passes: every operation becomes a General Matrix Multiplication (GEMM), the gradient stays bounded on the spectral support of the pre-normalized covariance, and the unstable 1/(lambda_i-lambda_j) term never appears. The key ingredient is a mean-eigenvalue pre-normalization that centres the spectrum near 1, away from the singularity of log, with a scalar post-compensation that returns the singular part of log(A) in closed form. Our recommended normalizer is a degree-8 Chebyshev expansion evaluated by a three-term matrix recurrence, with a matching reverse recurrence for the backward pass; Legendre, Laguerre, Taylor and Pade expansions are studied as controls that isolate the roles of the basis and of the target function. On three fine-grained benchmarks and ImageNet-1k the decomposition-free logarithm is both faster and more accurate than the spectral logarithm and than the square-root approximations it replaces, and at matched basis and degree the log target beats the square-root target, confirming that the gain comes from the faithful Riemannian map rather than from a better polynomial family.
Chinese Translation
全局协方差池化（Global Covariance Pooling, GCP）通过捕捉二阶特征统计信息来提升深度网络的性能，特别是在细粒度识别中效果显著。由于协方差矩阵位于对称正定（Symmetric Positive Definite, SPD）流形上，因此在使用欧几里得分类器之前需要进行归一化步骤。理想的选择是矩阵对数（Matrix Logarithm, MLN-COV），它将SPD流形映射到其切空间；然而在实践中，由于基于特征分解的梯度在数值上不稳定，通常放弃了这一方法，转而使用矩阵平方根。我们表明，这种不稳定性是计算对数时的谱方法导致的伪影，而非对数本身的性质。通过在协方差矩阵中用有限多项式近似对数，可以消除两个步骤中的特征分解：每个操作都变为一般矩阵乘法（General Matrix Multiplication, GEMM），梯度在预归一化协方差的谱支持上保持有界，并且不再出现不稳定的1/(lambda_i-lambda_j)项。关键成分是均值特征值预归一化，它将谱中心化至接近1，远离对数的奇点，并通过标量后补偿以封闭形式返回log(A)的奇异部分。我们推荐的归一化器是一个通过三项矩阵递推评估的8阶切比雪夫展开，并且为反向传播提供匹配的反向递推；同时研究了勒让德（Legendre）、拉盖尔（Laguerre）、泰勒（Taylor）和帕德（Pade）展开作为对照，以隔离基函数和目标函数的作用。在三个细粒度基准测试和ImageNet-1k上，无需分解的对数方法在速度和准确性上均优于谱对数和其替代的平方根近似，并且在匹配的基和阶数下，对数目标优于平方根目标，确认了性能提升源于忠实的黎曼映射，而非更优的多项式族。
cs.CV / 74 / 2608.19032
Counterfactual Contrastive Analysis
反事实对比分析
Abstract
Visual Counterfactual Explanations (VCEs) aim to explain image classifiers by generating minimally edited and realistic versions of an input image that change the classifier's prediction. Existing VCE methods are inherently classifier-dependent and therefore susceptible to classifier biases and failure modes, such as sensitivity to shortcut features and calibration errors. In this paper, we propose a classifier-free approach for visual counterfactual generation based on Contrastive Analysis (CA). Given two datasets corresponding to different classes (e.g., healthy and patients), we disentangle the generative factors that are common across the two datasets from those that are salient to each dataset, and generate counterfactual images by swapping only the salient factors. By operating directly on data distributions rather than decision boundaries, our method provides model-agnostic VCEs that are less sensitive to classifier biases. Our approach leverages the high-quality synthesis and well-structured latent space of StyleGAN2. We use the feature space F, instead than the usual W-space, to improve detail preservation. Unlike conventional CA approaches, which typically assume salient factors in only one dataset, we introduce an adapted framework and loss functions for VCE that allow multiple salient factors in each dataset. We evaluate our method on three medical imaging datasets and demonstrate superior counterfactual generation quality compared to existing approaches.
Chinese Translation
视觉反事实解释（VCEs）旨在通过生成最小编辑且逼真的输入图像版本来解释图像分类器，这些版本会改变分类器的预测。现有的VCE方法本质上依赖于分类器，因此容易受到分类器偏差和失败模式的影响，例如对捷径特征的敏感性和校准错误。在本文中，我们提出了一种基于对比分析（CA）的无分类器视觉反事实生成方法。给定两个对应于不同类别（例如，健康和患者）的数据集，我们将两个数据集中共同的生成因素与各自数据集中特征明显的因素进行解耦，并通过仅交换显著因素来生成反事实图像。通过直接在数据分布上操作，而不是决策边界，我们的方法提供了与模型无关的VCE，减少了对分类器偏差的敏感性。我们的方法利用了StyleGAN2的高质量合成和结构良好的潜在空间。我们使用特征空间F，而不是通常的W空间，以提高细节保留。与传统的CA方法不同，后者通常假设仅在一个数据集中存在显著因素，我们引入了一个适应性框架和损失函数，允许每个数据集中存在多个显著因素。我们在三个医学影像数据集上评估了我们的方法，并展示了与现有方法相比，反事实生成质量的优越性。
cs.CV / 75 / 2608.19036
USR-Drive: Unified Driving Scene Representation via Joint Denoising of 3D Gaussians and Boxes
USR-Drive：通过联合去噪3D高斯和边界框的统一驾驶场景表示
Abstract
Spatial representation learning for autonomous driving aims to map raw visual signals into structured 3D scene representations, where object-centric bounding boxes and rendering-oriented 3D primitives (\eg, 3D Gaussians) serve as two distinct yet highly complementary levels for scene understanding. Existing methods typically treat dynamic reconstruction and instance-level perception as separate tasks, despite their shared goal of estimating the underlying 3D world state. As a result, dynamic reconstruction is under-constrained while 3D detection lacks geometric grounding. To address this gap, we propose USR-Drive, a unified conditional generative framework that, given only posed multi-view driving videos, jointly recovers dense dynamic geometry and instance-level object layouts within a shared scene representation. Specifically, USR-Drive represents dense Gaussian primitives and sparse 3D bounding boxes as two aligned latent token streams and jointly denoises them with a unified multi-modal diffusion Transformer. Unlike prior paradigms that use boxes as external conditions or predict them with detached modules, USR-Drive treats them as mutually constrained state variables with a Unified Positional Encoding (UPE) that aligns heterogeneous tokens within a shared metric spatiotemporal coordinate. Via such unified representation and generative framework, the two modalities reinforce each other: geometry supplies dense metric evidence for box prediction, while boxes provide instance-level structural priors that help preserve spatial consistency and reduce ambiguity in sequential 3D geometric representation. Our approach successfully delivers state-of-the-art results for both dynamic reconstruction and 3D detection on the nuScenes and VKitti datasets.
Chinese Translation
自主驾驶的空间表示学习旨在将原始视觉信号映射为结构化的3D场景表示，其中以物体为中心的边界框和以渲染为导向的3D原语（例如，3D高斯）作为场景理解的两个不同但高度互补的层次。现有方法通常将动态重建和实例级感知视为独立任务，尽管它们的共同目标是估计潜在的3D世界状态。因此，动态重建受到约束不足，而3D检测缺乏几何基础。为了解决这一问题，我们提出了USR-Drive，一个统一的条件生成框架，该框架仅基于多视角驾驶视频，联合恢复共享场景表示中的密集动态几何和实例级物体布局。具体而言，USR-Drive将密集的高斯原语和稀疏的3D边界框表示为两个对齐的潜在标记流，并通过统一的多模态扩散Transformer共同去噪。与之前将边界框作为外部条件或通过独立模块预测的范式不同，USR-Drive将其视为相互约束的状态变量，并使用统一位置编码（Unified Positional Encoding, UPE）将异构标记对齐到共享的度量时空坐标中。通过这种统一的表示和生成框架，这两种模态相互强化：几何为边界框预测提供密集的度量证据，而边界框则提供实例级结构先验，帮助保持空间一致性并减少序列3D几何表示中的模糊性。我们的方法在nuScenes和VKitti数据集上成功实现了动态重建和3D检测的最先进结果。
cs.CV / 76 / 2608.19055
Generalized Audio-Driven Synthesis of Precise Drummer Motion
基于音频驱动的精确鼓手动作的广义合成
Abstract
Music-driven character animation enables and enhances transformative applications in entertainment and interactive education. However, synthesizing realistic drumming motion from audio remains challenging due to the inherent tension between high-acceleration dynamics and the need for extreme spatial-temporal precision. Existing approaches, often reliant on motion matching or MIDI input, struggle with generalizing to diverse real-world audio. Moreover, the field lacks standardized evaluation metrics capable of distinguishing precise drumming from noisy motion. In this paper, we introduce a generative diffusion framework featuring a dual-objective loss function that decouples skeletal integrity from drumstick precision, thus enabling centimeter-level stick precision without sacrificing natural body dynamics. Additionally, leveraging our own dataset and data augmentation strategy, the model generalizes to non-curated, in-the-wild audio. To rigorously evaluate performance, we propose two novel metrics: an impact-to-target distance to quantify spatial precision and an audio-motion correlation score to assess temporal alignment. Our quantitative analysis and user studies demonstrate that our system generates high-quality motion that is often indistinguishable from ground-truth performances.
Chinese Translation
音乐驱动的角色动画在娱乐和互动教育中启用并增强了变革性应用。然而，由于高加速度动态与极端时空精度之间的固有张力，从音频合成逼真的打击乐动作仍然具有挑战性。现有的方法通常依赖于动作匹配或MIDI输入，难以推广到多样化的现实音频。此外，该领域缺乏能够区分精确打击乐与噪声动作的标准化评估指标。在本文中，我们引入了一种生成扩散框架，采用双目标损失函数，将骨骼完整性与鼓棒精度解耦，从而实现厘米级的鼓棒精度而不牺牲自然的身体动态。此外，利用我们自己的数据集和数据增强策略，该模型能够推广到非策划的真实音频。为了严格评估性能，我们提出了两个新颖的指标：一个是影响目标距离，用于量化空间精度；另一个是音频-动作相关性评分，用于评估时间对齐。我们的定量分析和用户研究表明，我们的系统生成的高质量动作往往与真实表演难以区分。
cs.CV / 77 / 2608.19063
When Two Tracers Disagree: An Investigation of Multimodal Fusion for Clinical PET/CT Segmentation
当两种示踪剂不一致时：临床PET/CT分割的多模态融合研究
Abstract
PSMA and FDG PET/CT visualise complementary biological information in prostate cancer. Combining both tracers could capture heterogeneous tumour phenotypes that may be missed by either alone, yet there is no consensus on effective deep learning architectures for fusing these modalities. We evaluated multimodal image-fusion strategies for automatic whole-body PET/CT lesion segmentation to estimate total tumour burden. Using the public DEEP-PSMA Challenge dataset, we trained tracer-specific 3D nnU-Net baselines and compared (i) early fusion with a single encoder and one decoder (OEOD) or two decoders (OETD), and (ii) intermediate fusion via a dual-encoder cross-attention U-Net (DECA-UNet). Tracer-specific baselines performed strongly (PSMA Dice = 0.93; FDG = 0.81). Fusion yielded mixed results: OEOD produced a combined Dice of 0.90 (on an easier, non-tracer-specific task), whilst the tracer-specific fusion models reached PSMA/FDG = 0.69/0.64 (OETD) and 0.76/0.57 (DECA-UNet). Whilst fusion often provided reasonable PSMA segmentation, FDG performance degraded and no strategy consistently exceeded the single-tracer baselines. Under the evaluated setting, tracer-specific models remain the stronger baseline; clinically useful gains from multimodal fusion will likely require architectures that better preserve tracer specific representations. Our code is available at: https://github.com/JackJ3636/DEEP_PSMA_code
Chinese Translation
PSMA和FDG PET/CT在前列腺癌中可视化互补的生物信息。结合这两种示踪剂可以捕捉到可能被单独使用时遗漏的异质肿瘤表型，但对于融合这些模态的有效深度学习架构尚无共识。我们评估了多模态图像融合策略，以实现自动全身PET/CT病灶分割，从而估算总肿瘤负担。使用公开的DEEP-PSMA挑战数据集，我们训练了示踪剂特定的3D nnU-Net基线，并比较了(i) 使用单个编码器和一个解码器（OEOD）或两个解码器（OETD）的早期融合，以及(ii) 通过双编码器交叉注意力U-Net（DECA-UNet）进行的中间融合。示踪剂特定的基线表现良好（PSMA Dice = 0.93；FDG = 0.81）。融合结果喜忧参半：OEOD在一个较简单的非示踪剂特定任务上产生了0.90的综合Dice，而示踪剂特定的融合模型达到了PSMA/FDG = 0.69/0.64（OETD）和0.76/0.57（DECA-UNet）。尽管融合通常提供了合理的PSMA分割，但FDG的表现下降，且没有任何策略 consistently 超过单一示踪剂基线。在评估的设置下，示踪剂特定模型仍然是更强的基线；多模态融合的临床有效增益可能需要更好地保留示踪剂特定表示的架构。我们的代码可在以下链接获取：https://github.com/JackJ3636/DEEP_PSMA_code
cs.CV / 78 / 2608.19066
GS-VLA: Plug-and-Play Viewpoint Canonicalization for Frozen VLA Policies via Gaussian Splatting
GS-VLA：通过高斯点云实现冻结VLA策略的即插即用视角规范化
Abstract
This paper proposes a lightweight, plug-and-play framework that improves robustness to viewpoint shifts in Vision-Language-Action (VLA) policies without policy retraining. To our knowledge, this is the first approach to directly leverage 3D Gaussian-based novel-view synthesis for observation-space adaptation in VLA policies. Current VLA performance relies on the implicit assumption that training and deployment camera configurations are identical. Our experiments show that even a small displacement of the camera mount can reduce the success rate on the LIBERO benchmark from about 90% to about 10% in the worst case. Prior approaches, such as large-scale fine-tuning or generative data augmentation, are computationally expensive and risk catastrophic forgetting. To address this, viewpoint shifts are reformulated as a localized novel-view synthesis problem. Under a Locality assumption, that camera perturbations remain within a small bounded region relative to the workspace, viewpoint normalization reduces to a scene- and policy-independent disocclusion task. Our work implements this idea with a 4M-parameter 3D-Gaussian canonicalizer prepended to a frozen VLA policy. Without modifying policy weights, GS-VLA improves performance across three orthogonal axes: (1) Policy architectures, (2) Unseen task suites, and (3) Perturbation scales. These results show that a lightweight visual module can recover a large fraction of the performance lost under viewpoint shift, without policy retraining.
Chinese Translation
本文提出了一种轻量级的即插即用框架，旨在提高视觉-语言-动作（VLA）策略对视角变化的鲁棒性，而无需重新训练策略。据我们所知，这是首个直接利用基于3D高斯的新视角合成方法来进行VLA策略观察空间适应的研究。目前，VLA的性能依赖于训练和部署相机配置相同的隐含假设。我们的实验表明，即使是相机支架的小幅位移，也可能在最坏情况下将LIBERO基准的成功率从约90%降低到约10%。以往的方法，如大规模微调或生成数据增强，计算成本高且存在灾难性遗忘的风险。为了解决这一问题，视角变化被重新表述为局部的新视角合成问题。在局部性假设下，即相机扰动相对于工作空间保持在一个小的有界区域内，视角规范化简化为一个与场景和策略无关的去遮挡任务。我们的工作通过在冻结的VLA策略前添加一个具有400万参数的3D高斯规范化器来实现这一思想。在不修改策略权重的情况下，GS-VLA在三个正交轴上提高了性能：（1）策略架构，（2）未见任务套件，以及（3）扰动规模。这些结果表明，轻量级视觉模块能够在不重新训练策略的情况下，恢复因视角变化而损失的大部分性能。
cs.CV / 79 / 2608.19075
ReWEIGH the Evidence: Calibrating Token-Level Ordinal Visual Evidence to Mitigate Hallucinations in Large Vision-Language Models
重新加权证据：校准令牌级序数视觉证据以减轻大型视觉语言模型中的幻觉现象
Abstract
Large vision-language models (LVLMs) often hallucinate, generating content that the input image does not support. Preventing such content during decoding calls for a candidate-specific measure of how strongly the image supports the token under consideration. The model's visual-token states offer a natural source of this evidence because projecting each state through the output head reveals which vocabulary items that position favors. These position-wise readouts cannot be pooled directly because their probability magnitudes are not comparable across visual positions. Vocabulary ranks provide a scale-invariant basis for pooling, but tokens still differ systematically in their typical rank-based evidence. We propose ReWEIGH, a training-free decoding intervention that aggregates these ranks across visual positions and compares each candidate with a token-specific reference estimated from unlabeled images. At inference, ReWEIGH caches the image evidence during prefill and applies a bounded penalty only to candidates that fall below their reference. On four 7B backbones, ReWEIGH reduces hallucinated object mentions by up to 21.3% while largely preserving or improving descriptive and general performance. With evidence cached, the average added latency is 1.33% per token, and the reductions extend across six architecture families to 32B parameters.
Chinese Translation
大型视觉语言模型（LVLMs）常常出现幻觉，生成与输入图像不符的内容。在解码过程中防止此类内容的产生需要一种候选特定的度量，以评估图像对所考虑令牌的支持程度。模型的视觉令牌状态提供了这一证据的自然来源，因为通过输出头投影每个状态可以揭示该位置偏好的词汇项。这些位置级的读出不能直接汇总，因为它们的概率大小在视觉位置之间不可比。词汇排名提供了一种尺度不变的汇总基础，但令牌在其典型的基于排名的证据上仍然存在系统性差异。我们提出了ReWEIGH，这是一种无训练的解码干预方法，它在视觉位置之间汇总这些排名，并将每个候选项与从未标记图像中估算的令牌特定参考进行比较。在推理过程中，ReWEIGH在预填充期间缓存图像证据，并仅对低于其参考的候选项施加有界惩罚。在四个7B主干上，ReWEIGH将幻觉对象提及减少了多达21.3%，同时在描述性和一般性能上大体保持或改善。由于缓存了证据，平均每个令牌的额外延迟为1.33%，且减少效果在六个架构系列中扩展至32B参数。
cs.CV / 80 / 2608.19078
Subgroup performance analysis of adaptation strategies for chest X-ray foundation models
胸部X光基础模型适应策略的子组性能分析
Abstract
Foundation models are increasingly adapted for downstream medical imaging tasks, yet the influence of the chosen adaptation strategy on subgroup fairness remains poorly understood. We investigate how three parameter-efficient adaptation techniques, including linear heads on the raw CLS token, an MLP, and an attention-pooling module over multi-layer patch features, affect both pathology classification performance and subgroup disparities when applied to the frozen Rad-DINO chest X-ray encoder. Using MIMIC-CXR, we evaluate eight pathologies across race, sex, and imaging-view subgroups on a prevalence-preserving, demographically balanced test set, and additionally probe how strongly each adapter encodes protected attributes. We find that attention pooling achieves the strongest overall discriminative performance and encodes attributes, particularly race, most strongly, but that improved overall performance does not consistently reduce subgroup disparities. Notably, stronger attribute encoding did not correspond to larger disparities: early network layers encoded race most weakly yet produced the largest subgroup performance gaps. Exploring different attention-pooling layer combinations further revealed no consistent relationship between the layers pooled, attribute encoding strength, and subgroup fairness. Our results indicate that richer, more expressive representations can improve accuracy while leaving fairness implications task-dependent and unpredictable, which must be assessed directly and per-task rather than inferred from encoding strength or overall performance alone.
Chinese Translation
基础模型越来越多地被适应用于下游医学影像任务，但所选择的适应策略对子组公平性的影响仍然不够明确。我们研究了三种参数高效的适应技术，包括在原始CLS标记上的线性头、一个多层感知机（MLP）和一个基于多层补丁特征的注意力池化模块，如何影响病理分类性能和子组差异，应用于冻结的Rad-DINO胸部X光编码器。使用MIMIC-CXR数据集，我们在一个保持流行率、人口统计平衡的测试集上评估了种族、性别和成像视图子组中的八种病理，并进一步探讨每个适配器对受保护属性的编码强度。我们发现，注意力池化在整体判别性能上表现最佳，并且对属性（特别是种族）的编码最强，但整体性能的提升并不总是能减少子组差异。值得注意的是，较强的属性编码并不对应于更大的差异：早期网络层对种族的编码最弱，但产生了最大的子组性能差距。探索不同注意力池化层组合进一步揭示了所池化层、属性编码强度和子组公平性之间没有一致的关系。我们的结果表明，更丰富、更具表现力的表示可以提高准确性，同时使公平性影响在任务上依赖且不可预测，这必须直接评估并针对每个任务进行，而不是仅仅从编码强度或整体性能推断。
cs.CV / 81 / 2608.19080
SPK: Eliciting Structured Prior Knowledge for Interpretable Out-of-Distribution Detection in Real-Time Object Detection
SPK：为实时目标检测中的可解释性异常检测引出结构化先验知识
Abstract
Object detectors often produce over-confident predictions for objects outside their training categories, leading to so-called out-of-distribution (OoD) hallucinations. Existing approaches for detecting or mitigating such hallucinations typically either construct scoring functions directly over learned object detector representations or modify the object detector itself to suppress hallucination emergence. However, the latent priors implicitly encoded in these representations remain largely unexplored and have not been explicitly decoded for OoD detection. To uncover and exploit these latent priors, we propose Structured Prior Knowledge (SPK), a hallucination-oriented framework that explicitly elicits OoD-relevant priors from pretrained object detectors. Specifically, SPK leverages in-distribution data and hallucination-inducing samples as diagnostic supervision to elicit part-level semantic concepts underlying object detector decision-making, rather than using them merely for rejection or object detector adaptation. The elicited semantic priors are further integrated with geometric and contextual priors to form a compact five-dimensional SPK representation for OoD detection. Extensive experiments across diverse object detector architectures and multiple OoD benchmarks demonstrate that SPK achieves state-of-the-art OoD detection. Our findings reveal that pretrained object detectors already encode substantially richer latent knowledge than is typically exploited for OoD detection. More importantly, this knowledge can be explicitly elicited and organized into a compact, structured, and interpretable knowledge space for prediction reliability analysis. This suggests a promising proactive route for improving object detector reliability by explicitly uncovering and leveraging latent priors. Code and data are available at: https://gricad-gitlab.univ-grenoble-alpes.fr/dnn-safety/spk
Chinese Translation
目标检测器通常对训练类别之外的物体产生过于自信的预测，导致所谓的异常（out-of-distribution, OoD）幻觉。现有的检测或缓解此类幻觉的方法通常直接在学习到的目标检测器表示上构建评分函数，或修改目标检测器本身以抑制幻觉的出现。然而，这些表示中隐含的潜在先验知识仍然在很大程度上未被探索，并且尚未被明确解码用于OoD检测。为了揭示和利用这些潜在先验，我们提出了结构化先验知识（Structured Prior Knowledge, SPK），这是一个以幻觉为导向的框架，能够从预训练的目标检测器中明确引出与OoD相关的先验。具体而言，SPK利用分布内数据和诱发幻觉的样本作为诊断监督，以引出支撑目标检测器决策的部分级语义概念，而不仅仅是将其用于拒绝或目标检测器的适应。引出的语义先验进一步与几何和上下文先验相结合，形成一个紧凑的五维SPK表示，用于OoD检测。在多种目标检测器架构和多个OoD基准上的广泛实验表明，SPK实现了最先进的OoD检测。我们的研究发现，预训练的目标检测器已经编码了比通常用于OoD检测的潜在知识要丰富得多的内容。更重要的是，这些知识可以被明确引出并组织成一个紧凑、结构化且可解释的知识空间，以便进行预测可靠性分析。这表明，通过明确揭示和利用潜在先验，改善目标检测器可靠性是一条有前景的主动途径。代码和数据可在以下链接获取：https://gricad-gitlab.univ-grenoble-alpes.fr/dnn-safety/spk
cs.CV / 82 / 2608.19088
Detecting Backdoors in Object Detection via Pre-NMS Prediction Distribution Shift
通过预-NMS预测分布偏移检测目标检测中的后门
Abstract
Object detection models deployed in safety-critical applications remain vulnerable to backdoor attacks that cause targeted misbehaviors when a hidden trigger is present. Existing detection methods either rely on trigger inversion or exploit architecture-specific assumptions, and critically, representative existing methods fail to generalize reliably to scene-level attacks, where a single trigger induces anomalous behavior across all objects in the scene simultaneously. We present DistScan, a backdoor detection framework based on a simple but previously unexploited observation: backdoor injection systematically shifts a model's pre-NMS prediction class distribution away from its training class frequencies, even on clean inputs without any trigger present. DistScan aggregates intermediate class predictions over a clean validation set and flags a model as backdoored if the resulting distribution deviates significantly from the training class frequencies, requiring no model weight access, no trigger knowledge, and no additional training. Extensive experiments on MS-COCO and PASCAL VOC across two architectures and three scene-level attack scenarios demonstrate that DistScan substantially outperforms existing methods, improving average detection accuracy over the best-performing applicable baseline by 27.32 percentage points.
Chinese Translation
部署在安全关键应用中的目标检测模型仍然容易受到后门攻击，这些攻击在存在隐藏触发器时会导致目标行为异常。现有的检测方法要么依赖于触发器反演，要么利用特定于架构的假设，而现有的代表性方法在场景级攻击中无法可靠地推广，在这种情况下，单个触发器会同时引发场景中所有对象的异常行为。我们提出了DistScan，这是一种基于一个简单但之前未被利用的观察的后门检测框架：后门注入系统性地使模型的预-NMS预测类别分布偏离其训练类别频率，即使在没有任何触发器的干净输入上也是如此。DistScan在干净的验证集上聚合中间类别预测，并在结果分布显著偏离训练类别频率时将模型标记为后门，无需访问模型权重、触发器知识或额外训练。在MS-COCO和PASCAL VOC上进行的广泛实验，涵盖两种架构和三种场景级攻击场景，证明了DistScan显著优于现有方法，平均检测准确率比最佳适用基线提高了27.32个百分点。
cs.CV / 83 / 2608.19177
Image-Guided Pavement Defect Recognition in GPR Data with novel 3D Deep Learning Architecture
基于图像引导的GPR数据路面缺陷识别的新型3D深度学习架构
Abstract
Ground Penetrating Radar (GPR) is a widely adopted non-destructive sensing technology for subsurface inspection in civil and transportation engineering. Despite its potential for pavement condition assessment, the large-scale application of GPR in automated inspection has two key challenges: the scarcity of annotated real-world datasets and the lack of deep learning models designed for the unique characteristics of 3-Dimensional (3D) GPR data. This study addresses these limitations by firstly introducing a cost-effective data preparation pipeline that integrates orthomosaic Red Green Blue (RGB) imagery with 3D GPR scans to generate annotated 3D GPR datasets. The proposed method uses the aligned segments of RGB and GPR data, using pavement surface images as a reference to transfer labels of surface-visible defects to corresponding GPR segments, enabling efficient large-scale annotation in a real-world dataset collected on a highway section under operation. In addition to the dataset contribution, we propose a specialised 3D Convolutional Neural Network (CNN) architecture incorporating residual connections, mixed convolutional kernel sizes, and both depthwise and channelwise attention mechanisms to enhance feature representation and defect classification. The model is evaluated on binary classification tasks for detecting patch and crack defects in pavement structures. Experimental results demonstrate that the proposed network outperforms baseline architectures across multiple evaluation metrics. Ablation studies further confirm the effectiveness of the designed architectural components. This work contributes a scalable and practical method for real-world dataset generation, along with a novel deep learning framework.
Chinese Translation
地面穿透雷达（Ground Penetrating Radar, GPR）是一种广泛应用于土木工程和交通工程的非破坏性检测技术，用于地下检查。尽管其在路面状况评估中的潜力巨大，但GPR在自动化检测中的大规模应用面临两个主要挑战：一是标注的真实世界数据集稀缺，二是缺乏针对三维（3D）GPR数据独特特征设计的深度学习模型。本研究通过首先引入一种经济高效的数据准备流程，解决了这些限制，该流程将正射影像红绿蓝（RGB）图像与3D GPR扫描结合，以生成标注的3D GPR数据集。所提出的方法利用RGB和GPR数据的对齐段，以路面表面图像作为参考，将表面可见缺陷的标签转移到相应的GPR段，从而实现对在运营中的高速公路段收集的真实世界数据集的高效大规模标注。除了数据集的贡献外，我们还提出了一种专门的3D卷积神经网络（Convolutional Neural Network, CNN）架构，结合了残差连接、混合卷积核大小以及深度和通道注意机制，以增强特征表示和缺陷分类。该模型在检测路面结构中的补丁和裂缝缺陷的二分类任务上进行了评估。实验结果表明，所提出的网络在多个评估指标上优于基线架构。消融研究进一步确认了所设计架构组件的有效性。本研究为真实世界数据集生成提供了一种可扩展且实用的方法，并提出了一种新颖的深度学习框架。
cs.AI / 1 / 2608.18078
Position: Collusion Risks Among AI Reasoning Agents Justify Certification Requirements for Making Market Decisions
立场：AI推理代理之间的串通风险证明了市场决策所需的认证要求
Abstract
This position paper argues that AI agents with chain-of-thought reasoning capabilities are predisposed to exhibit collusive behavior and should be required to obtain behavioral certification before making decisions that affect economic markets. This is because integrating these agents into society could collapse the legal evidentiary distinction between competition and collusion among independent firms without eroding the economic harm distinction. Experiments with DeepSeek-R1 agents in the Bertrand oligopoly pricing domain reveal a tendency towards tacit collusion that persists even when humans prompt the agents not to collude. We further show that the chain-of-thought of these agents can be steered toward either extremely collusive or highly competitive behavior in a way that is not semantically detectable by another LLM analyzing the reasoning traces. As a result, deploying reasoning agents for market decisions leads to collusive economic outcomes without any evidence of conspiracy or intent. Thus, certification based on observed behavior in representative situations is necessary to prevent collusion. We provide preliminary evidence that such agents can be steered in a generalizable way toward efficient competitive equilibria. However, developing a comprehensive behavioral certification will be required before these models can be deployed in real-world markets while ensuring their stability and efficiency.
Chinese Translation
本文立场论文认为，具备链式思维推理能力的AI代理倾向于表现出串通行为，因此在做出影响经济市场的决策之前，应要求其获得行为认证。这是因为将这些代理整合到社会中可能会破坏独立公司之间竞争与串通的法律证据区分，而不会削弱经济损害的区分。在Bertrand寡头定价领域对DeepSeek-R1代理的实验显示，即使在有人提示代理不要串通的情况下，它们仍然倾向于进行默契串通。我们进一步表明，这些代理的链式思维可以被引导向极度串通或高度竞争的行为，而这种行为在分析推理轨迹的其他大型语言模型（LLM）中是无法语义上检测到的。因此，部署推理代理进行市场决策会导致串通的经济结果，而没有任何阴谋或意图的证据。因此，基于在代表性情境中观察到的行为进行认证是必要的，以防止串通。我们提供了初步证据，表明这些代理可以以可推广的方式引导向有效的竞争均衡。然而，在这些模型能够在现实市场中部署之前，开发全面的行为认证将是必要的，以确保其稳定性和效率。
cs.AI / 2 / 2608.18079
Position: Profiling Game Worlds by Transition Complexity
位置：通过转移复杂性对游戏世界进行分析
Abstract
Game world modeling (GWM) and reinforcement learning (RL) are often confounded because research papers rarely quantify how difficult the underlying transition prediction problem is at the declared interface (pixels/tokens/latents with finite history). We propose the Transition Complexity Profile (TCP): a small, reproducible set of metrics that characterizes an environment's (or gameplay dataset's) induced transition kernel by (i) intrinsic one-step branching, (ii) interaction-induced uncertainty and opponent influence when observable, and (iii) temporal/spatial dependency span via standardized probe curves. TCP is reported with an explicit reference distribution, protocol stochasticity, and a versioned measurement budget (sampling/resampling and fixed probe compute), enabling comparable numbers across benchmarks. We outline how common game families and modern "neural game engine" domains populate this landscape and call for TCP to become standard benchmark metadata and a required statistic in GWM and RL papers.
Chinese Translation
游戏世界建模（Game World Modeling, GWM）和强化学习（Reinforcement Learning, RL）常常被混淆，因为研究论文很少量化在声明的接口（具有有限历史的像素/标记/潜变量）下，基础转移预测问题的难度。我们提出了转移复杂性剖面（Transition Complexity Profile, TCP）：一组小而可重复的指标，通过（i）内在的一步分支，（ii）可观察时的交互引起的不确定性和对手影响，以及（iii）通过标准化探测曲线的时间/空间依赖跨度，来表征环境（或游戏玩法数据集）所诱导的转移核。TCP报告时附有明确的参考分布、协议随机性和版本化的测量预算（采样/重采样和固定探测计算），使得不同基准之间的数值具有可比性。我们概述了常见游戏家族和现代“神经游戏引擎”领域如何填充这一领域，并呼吁将TCP作为标准基准元数据和GWM与RL论文中所需的统计量。
cs.AI / 3 / 2608.18080
Large Language Models in Mental Health: A Systematic Review of Applications, Innovations, and Ethical Challenges
大型语言模型在心理健康中的应用：应用、创新与伦理挑战的系统评估
Abstract
We present a review on the applications of large language models (LLMs) in health, e.g., social media analysis, clinical conversational agents, therapy support tools, prompt engineering, multimodal learning, and ethical considerations. We integrate findings from interdisciplinary studies utilizing diverse data sources such as social media posts, electronic medical records, and multimodal inputs to enable early detection of depression, suicide risk assessment, personalized therapy support, and psychoeducational content generation. Our review highlights advancements in LLM models and annotation strategies that enhance interpretability and clinical relevance, while we also emphasize the critical role of prompt engineering for domain adaptation. We also discuss emerging multimodal fusion techniques integrating text, speech, and sensor data for improved mental health diagnosis and monitoring. Finally, we address ongoing ethical, sociotechnical, and regulatory challenges, and advocate frameworks to ensure safe, equitable, and accountable deployment of LLMs in real-world mental health care.
Chinese Translation
我们对大型语言模型（LLMs）在健康领域的应用进行了评估，例如社交媒体分析、临床对话代理、治疗支持工具、提示工程、多模态学习和伦理考量。我们整合了来自跨学科研究的发现，这些研究利用了多种数据来源，如社交媒体帖子、电子病历和多模态输入，以实现抑郁症的早期检测、自杀风险评估、个性化治疗支持和心理教育内容生成。我们的评估突出了LLM模型和注释策略的进展，这些进展增强了可解释性和临床相关性，同时我们也强调了提示工程在领域适应中的关键作用。我们还讨论了新兴的多模态融合技术，这些技术整合了文本、语音和传感器数据，以改善心理健康的诊断和监测。最后，我们解决了持续存在的伦理、社会技术和监管挑战，并倡导建立框架，以确保在现实心理健康护理中安全、公平和负责任地部署LLMs。
cs.AI / 4 / 2608.18081
Position: Behavioral Systems Require Behavioral Tests
立场：行为系统需要行为测试
Abstract
Artificial agentic systems increasingly operate as behavioral systems by interacting with dynamic environments, pursuing goals, and adapting over time. Yet, current evaluation methods largely focus on performance outcomes, not the underlying behavioral processes that produce them. This paper argues that AI agents must be evaluated like other behavioral systems: through systematic observation, perturbation, and interpretation of their actions. We draw on lessons from the behavioral sciences to motivate this position, and propose a research agenda focused on developing rigorous behavioral tests. These include methods for recovering decision strategies from action sequences, constructing environments that isolate behavioral differences, and probing emergent dynamics in multi-agent systems. Taken together, these directions offer a roadmap for developing a science of AI behavior.
Chinese Translation
人工智能代理系统越来越多地作为行为系统运作，通过与动态环境互动、追求目标和随时间适应。然而，目前的评估方法主要关注性能结果，而非产生这些结果的基础行为过程。本文认为，人工智能代理必须像其他行为系统一样进行评估：通过系统观察、扰动和对其行为的解释。我们借鉴行为科学的经验来支持这一立场，并提出一个研究议程，重点在于开发严格的行为测试。这些测试包括从行动序列中恢复决策策略的方法、构建隔离行为差异的环境，以及探测多代理系统中的涌现动态。综合来看，这些方向为发展人工智能行为科学提供了一条路线图。
cs.AI / 5 / 2608.18086
Position: Current Model Cards Are Insufficient for Downstream Governance of Open-Weight Foundation Models
立场：当前的模型卡不足以支持开放权重基础模型的下游治理
Abstract
The growth of open-weight foundation models (OWFMs) has prompted the AI community to re-evaluate strategies for effective downstream governance. Although model cards have been widely adopted as transparency artifacts in model repositories, existing frameworks often fail to adequately inform downstream developers and users about the distinct safety challenges posed by OWFMs. This position paper analyzes 500 model cards hosted on Hugging Face and argues that effective governance of OWFMs requires a multi-layered approach integrating three complementary components: (i) model cards, (ii) acceptable use policies (AUPs), and (iii) licenses. To motivate this claim, we identify a safety gap left by existing regulatory approaches, including model heritage, alignment provenance, and empirically observed behaviors, through an analysis of model cards with safety-critical information. We further argue that standard open-source licenses (OSLs) are not well suited for OWFMs and may weaken the enforceability of AUPs. Building on these observations, we outline directions for evolving model cards, AUPs, and licenses into integrated safety artifacts to enable a more comprehensive governance framework that coherently integrates informational, normative, and legal dimensions.
Chinese Translation
开放权重基础模型（OWFMs）的增长促使人工智能社区重新评估有效的下游治理策略。尽管模型卡已被广泛采用作为模型库中的透明性工具，但现有框架往往未能充分告知下游开发者和用户OWFMs所带来的独特安全挑战。本文分析了在Hugging Face上托管的500个模型卡，并认为有效治理OWFMs需要一种多层次的方法，整合三种互补的组成部分：（i）模型卡，（ii）可接受使用政策（AUPs），以及（iii）许可证。为了支持这一论点，我们通过分析包含安全关键信息的模型卡，识别了现有监管方法所留下的安全缺口，包括模型遗产、对齐来源和经验观察到的行为。我们进一步认为，标准的开源许可证（OSLs）并不适合OWFMs，可能削弱AUPs的可执行性。在这些观察的基础上，我们概述了将模型卡、AUPs和许可证演变为综合安全工具的方向，以便实现一个更全面的治理框架，连贯地整合信息、规范和法律维度。
cs.AI / 6 / 2608.18088
A Metamorphic Artificial Age Score Decision-Support Prototype for Flight-Log-Based Drone Propeller Health Monitoring
基于飞行日志的无人机螺旋桨健康监测的变形人工年龄评分决策支持原型
Abstract
Drone propeller faults can create safety and reliability risks when their effects are distributed across multiple flight-log channels rather than appearing as a single diagnostic signal. This paper proposes a Metamorphic Artificial Age Score (AAS) decision-support prototype for flight-log-based drone propeller health monitoring. Using selected historical real flight logs from the 2024 DronePropA public dataset, the framework computes six health-related indicators from raw MATLAB matrices: trajectory tracking error, attitude instability, thrust-command burden, motor-command imbalance, ESC-command instability, and battery-level stress. These indicators are normalized relative to a healthy baseline and evaluated through candidate scoring policies, metamorphic adequacy relations, and a redundancy-adjusted AAS formulation. In this context, AAS is used as a structural policy-adequacy and burden measure rather than as a chronological age measure. A controlled retrospective evaluation was performed using one healthy baseline and three defective propeller cases under the same speed profile and trajectory. The healthy case was assigned to routine monitoring. The Severity 1 case was dominated by ESC-command instability and assigned to maintenance review. The Severity 2 case reached maximum motor-command and ESC-command burden, while the Severity 3 case reached maximum trajectory tracking error; both triggered mandatory inspection. The results show that propeller fault effects may appear through different operational channels, supporting the need for a multi-indicator decision-support layer for post-flight maintenance prioritization and autonomous-system oversight.
Chinese Translation
无人机螺旋桨故障在其影响分布于多个飞行日志通道时，可能会造成安全和可靠性风险，而不是以单一诊断信号的形式出现。本文提出了一种基于飞行日志的无人机螺旋桨健康监测的变形人工年龄评分（AAS）决策支持原型。该框架利用2024年DronePropA公共数据集中选取的历史真实飞行日志，从原始MATLAB矩阵中计算出六个与健康相关的指标：轨迹跟踪误差、姿态不稳定性、推力命令负担、电机命令不平衡、ESC命令不稳定性和电池电量压力。这些指标相对于健康基线进行归一化，并通过候选评分策略、变形充分性关系和冗余调整的AAS公式进行评估。在此背景下，AAS被用作结构性政策充分性和负担度量，而不是作为时间年龄度量。通过在相同速度特征和轨迹下使用一个健康基线和三个缺陷螺旋桨案例进行了受控的回顾性评估。健康案例被分配为常规监测。严重性1案例以ESC命令不稳定性为主，分配为维护审查。严重性2案例达到了最大电机命令和ESC命令负担，而严重性3案例达到了最大轨迹跟踪误差；两者均触发了强制检查。结果表明，螺旋桨故障的影响可能通过不同的操作通道出现，支持了在飞行后维护优先级和自主系统监督中需要多指标决策支持层的必要性。
cs.AI / 7 / 2608.18092
Position: Multi-Agent Systems Should Prioritize Concurrency Control
立场：多智能体系统应优先考虑并发控制
Abstract
LLM-based multi-agent systems (MAS) promise scalable collaboration, yet adding agents often reduces reliability. This position paper argues that many MAS failures are fundamentally concurrency control problems: agents concurrently read and write shared state, and long LLM inference windows amplify the risk of stale reads, lost updates, and inconsistent outcomes. Failure modes commonly attributed to coordination or communication breakdowns can be mapped directly onto classical concurrency anomalies. We contend that MAS frameworks should address these failures through explicit concurrency control mechanisms: conflict detection, isolation guarantees, and structured access to shared resources. Concurrency control should be a first-class design concern, not an afterthought.
Chinese Translation
基于大语言模型（LLM）的多智能体系统（MAS）承诺实现可扩展的协作，然而，增加智能体往往会降低系统的可靠性。本文立场论文认为，许多MAS的失败根本上是并发控制问题：智能体同时读取和写入共享状态，而较长的LLM推理窗口则加大了过时读取、更新丢失和结果不一致的风险。通常归因于协调或通信故障的失败模式可以直接映射到经典的并发异常上。我们主张，MAS框架应通过明确的并发控制机制来应对这些失败：冲突检测、隔离保证以及对共享资源的结构化访问。并发控制应成为设计中的首要关注点，而不是事后考虑。
cs.AI / 8 / 2608.18099
FinSkillBench: Evaluating AI Agents and Domain Skills for Investment Management
FinSkillBench：评估投资管理中的人工智能代理和领域技能
Abstract
Investment management is a high-stakes domain in which agentic AI systems must do more than generate plausible text. They must retrieve point-in-time data, assemble correct computational inputs, invoke specialized methods, and produce auditable structured outputs. We introduce FinSkillBench, an evaluation suite designed to measure whether language model agents can effectively use financial domain skills to solve investment management tasks. The benchmark spans three domains, portfolio construction, risk management, and fundamental analysis, and includes 12 subtasks with 2,603 task episodes. Each episode provides point-in-time inputs, hidden ground truth, and a task-specific verifier.We compare three conditions: no skill, curated skill packages consisting of procedural documents and executable components, and self-generated skills in which the agent writes and reuses its own procedures within an episode. Across 9 models and a large-scale evaluation, curated skills consistently improve performance, raising mean scores from 0.366 to 0.528, with the largest gains in portfolio construction and risk management. In contrast, self-generated skills provide little benefit despite higher computational cost. An independent evaluation using a separate agent framework (Hermes Agent, 8 models, 5,280 episodes total) reproduces the directional pattern across all three domains, with the magnitude of skill effects varying by subtask and harness. These results showthat in investment management agents, access to reliable procedural skills can be as important as model choice, while naive self-generation of skills is often ineffective. We release the benchmark, evaluation tools, curated skill packages, and full trajectories to support further research.
Chinese Translation
投资管理是一个高风险领域，在这个领域中，智能代理系统必须不仅仅生成可信的文本。它们还必须检索特定时间的数据，组装正确的计算输入，调用专业方法，并生成可审计的结构化输出。我们介绍了FinSkillBench，这是一个评估套件，旨在衡量语言模型代理是否能够有效利用金融领域技能来解决投资管理任务。该基准涵盖三个领域：投资组合构建、风险管理和基本面分析，并包括12个子任务，共2603个任务情节。每个情节提供特定时间的输入、隐藏的真实值和特定任务的验证器。我们比较了三种条件：无技能、由程序文档和可执行组件组成的策划技能包，以及自生成技能，其中代理在情节中编写和重用自己的程序。在9个模型和大规模评估中，策划技能始终提高了性能，使平均分数从0.366提升至0.528，投资组合构建和风险管理的增幅最大。相比之下，尽管自生成技能的计算成本更高，但其提供的好处很小。使用独立的评估框架（Hermes Agent，8个模型，总计5280个情节）进行的评估在所有三个领域中重现了方向性模式，技能效果的大小因子任务和工具而异。这些结果表明，在投资管理中，代理访问可靠的程序技能与模型选择同样重要，而天真的自生成技能往往无效。我们发布了基准、评估工具、策划技能包和完整的轨迹，以支持进一步的研究。
cs.AI / 9 / 2608.18104
Self-Evolving Agents as Dynamic Graph Transformation: A Survey and New Perspective
自我演化代理作为动态图转化：综述与新视角
Abstract
Large language model (LLM)-based agents are increasingly becoming self-evolving systems that persist across interactions, maintain memories, use tools, acquire skills, refine workflows, and coordinate with other agents. These capabilities make agent states structural and dynamic: entities, relations, attributes, dependencies, and execution structures change with new evidence, feedback, and environmental conditions. Existing graph-agent surveys typically treat graphs as support structures for agent functions rather than as evolving substrates, while self-evolving-agent surveys focus on agent-level mechanisms and rarely discuss graph topology evolution. Thus, the coupling between evolving agent state and dynamic graph topology remains underexplored. This survey connects these two research lines by framing \textit{agent evolution as dynamic graph transformation}. We model agent state as a dynamic graph, where memories, tools, skills, workflows, and inter-agent relations are represented as typed nodes, edges, and subgraphs updated through schema-constrained rewrites. Based on this formulation, we organize existing dynamic-graph-based methods for self-evolving agents into four taxonomies: node/feature evolution, edge/topology evolution, subgraph activation, and cross-component co-evolution. Building on this taxonomy, we propose dynamic graph learning as reusable infrastructure for self-evolving agents and map nine dynamic-graph-learning subfields to agent-evolution capabilities, discussing their adaptations and possible failure modes. Finally, we discuss five types of graph-aware evaluation and governance protocols from a dynamic-graph perspective, which complement end-task evaluation. The goal is to provide a compact structural lens for designing and governing self-evolving agents.
Chinese Translation
基于大型语言模型（LLM）的代理正日益成为自我演化的系统，这些系统在交互中持续存在，维护记忆，使用工具，获取技能，优化工作流程，并与其他代理协调。这些能力使得代理状态具有结构性和动态性：实体、关系、属性、依赖关系和执行结构随着新证据、反馈和环境条件而变化。现有的图代理综述通常将图视为代理功能的支持结构，而不是作为演化基质，而自我演化代理的综述则侧重于代理级别的机制，很少讨论图拓扑的演化。因此，演化代理状态与动态图拓扑之间的耦合仍然未得到充分探索。本综述通过将“代理演化视为动态图转化”来连接这两条研究线索。我们将代理状态建模为动态图，其中记忆、工具、技能、工作流程和代理间关系被表示为通过模式约束重写更新的类型节点、边和子图。基于这一表述，我们将现有的基于动态图的自我演化代理方法组织为四个分类：节点/特征演化、边/拓扑演化、子图激活和跨组件共同演化。在此分类的基础上，我们提出动态图学习作为自我演化代理的可重用基础设施，并将九个动态图学习子领域映射到代理演化能力，讨论它们的适应性和可能的失败模式。最后，我们从动态图的角度讨论五种图感知评估和治理协议，这些协议补充了最终任务评估。我们的目标是为设计和治理自我演化代理提供一个紧凑的结构视角。
cs.AI / 10 / 2608.18110
Emergence of Agentic AI: A Review on Evolution, Background, Working Principles, Applications, Adoption Factors, and Future Research Directions
代理型人工智能的出现：关于其演变、背景、工作原理、应用、采纳因素及未来研究方向的综述
Abstract
Agentic AI is gaining new insights and advancements in the field of Artificial Intelligence, fostering significant potential to enable rapid transformation across various domains.This rapid advancement and the potential to revolutionize various domains advocate the need for a deeper understanding and firm grasp of the technology. Moreover, an investigation into state of the art research directions in agentic AI needs to be conducted to comprehensively assess the potential scope for improvement and application.Therefore, to address these objectives, a comprehensive review can provide researchers and practitioners with valuable insights into the current state and future research scopes of agentic AI.Hence, this work considers the recently published scholarly contributions in agentic AI across various domains and discusses the fundamentals and working principles of Agentic AI, traces the historical and theoretical evolution of agency in artificial systems, explores and discusses Agentic AIs architecture, working principles, and functionalities, explores real-world applications of Agentic AI across various domains, analyzes the research findings, identifies current challenges, and discuss potential future research directions, and proposes a comprehensive framework of stakeholders intention to use and adopt Agentic AI with the help of proposed system quality dimensions.Therefore, this systematic review provides researchers and practitioners with a comprehensive understanding of Agentic AI, its current developments and applications, highlights key research gaps, and outlines future research directions.
Chinese Translation
代理型人工智能在人工智能领域正获得新的见解和进展，展现出在各个领域实现快速转型的重大潜力。这一快速发展及其颠覆各个领域的潜力呼唤对该技术的深入理解和牢固掌握。此外，需要对代理型人工智能的前沿研究方向进行调查，以全面评估其改进和应用的潜在范围。因此，为了实现这些目标，全面的综述可以为研究人员和从业者提供关于代理型人工智能当前状态和未来研究范围的宝贵见解。因此，本研究考虑了最近在各个领域发表的关于代理型人工智能的学术贡献，讨论了代理型人工智能的基本原理和工作机制，追溯了人工系统中代理性的历史和理论演变，探讨并讨论了代理型人工智能的架构、工作原理和功能，考察了代理型人工智能在各个领域的实际应用，分析了研究发现，识别了当前挑战，讨论了潜在的未来研究方向，并提出了一个综合框架，旨在帮助利益相关者理解使用和采纳代理型人工智能的意图，结合所提出的系统质量维度。因此，这一系统综述为研究人员和从业者提供了对代理型人工智能的全面理解，涵盖其当前发展和应用，突出关键研究空白，并概述未来的研究方向。
cs.AI / 11 / 2608.18111
Solving Is Not Drawing: A Benchmark for Diagrammatic Reasoning in Olympiad Geometry
解决问题并非绘图：奥林匹克几何中的图示推理基准
Abstract
Foundation models such as GPT and Claude now solve olympiad-level mathematics with remarkable proficiency, so much so that geometry problem solving has become a standard proxy for their mathematical reasoning. Yet solving a geometry problem and drawing the figure it depends on are not the same skill: progress often hinges on a faithful diagram with the right auxiliary constructions and incidences, and it is unclear that a model which reasons its way to the answer can also produce one. A growing collection of benchmarks, including MathVista, and MathVerse, measures whether models reach the correct answer, but to our knowledge, none isolate the distinct ability to construct the diagram itself, leaving this capability unmeasured. We introduce an open-source benchmark that targets this gap: 954 self-contained olympiad geometry problems, with a 297-problem hard subset, each paired with its solution and a human-authored, high-fidelity diagram in renderable Asymptote code, together with a suite of text-, code-, image-, VLM-, and constraint-based metrics for what we term diagrammatic reasoning. Evaluating current foundation models reveals a pronounced gap between solving and drawing: their diagrams are markedly less faithful, with an average compile success rate of only 36.14\%. Strong mathematical reasoning, we find, does not imply the ability to construct accurate geometric diagrams. Our benchmark and dataset can be accessed at https://huggingface.co/datasets/max98765/hard_geometry_problems_with_diagrams.
Chinese Translation
基础模型如GPT和Claude现在以显著的熟练度解决奥林匹克级别的数学问题，以至于几何问题的解决已成为其数学推理的标准代理。然而，解决几何问题与绘制其依赖的图形并不是同一种技能：进展往往依赖于一个忠实的图示，包含正确的辅助构造和交点，而尚不清楚一个能够推理出答案的模型是否也能绘制出图形。越来越多的基准测试，包括MathVista和MathVerse，衡量模型是否能够得出正确答案，但据我们所知，没有任何测试单独隔离出构建图形本身的独特能力，这一能力尚未被测量。我们引入了一个开源基准，旨在填补这一空白：954个自包含的奥林匹克几何问题，其中包含297个难题子集，每个问题都配有其解决方案和一个由人类创作的高保真图示，采用可渲染的Asymptote代码，以及一套基于文本、代码、图像、视觉语言模型（VLM）和约束的图示推理度量标准。对当前基础模型的评估揭示了解决与绘图之间的明显差距：它们的图示明显不够忠实，平均编译成功率仅为36.14%。我们发现，强大的数学推理并不意味着具备构建准确几何图示的能力。我们的基准和数据集可以在https://huggingface.co/datasets/max98765/hard_geometry_problems_with_diagrams获取。
cs.AI / 12 / 2608.18117
Position: AI Leaderboards Are Underserving the Global South: A Case Study from India
立场：人工智能排行榜未能服务全球南方：来自印度的案例研究
Abstract
This position paper argues that AI leaderboards are structurally ill-suited to serving the Global South because they lack independent governance, conflict-of-interest policies, and mechanisms for metric evolution. The barrier is not missing data; high-quality regional benchmarks already exist: IndicSUPERB, MILU, and LAHAJA for India; IrokoBench for Africa; AlGhafa for Arabic. The barrier is institutional design. Global leaderboards do not include these benchmarks, and no governance mechanism compels them to do so. Commercial pressure corrects leaderboard failures when paying customers in the Global North are affected. The Global South lacks equivalent leverage. Without governance, failures affecting Hindi, Swahili, or Arabic speakers persist indefinitely as documented but unaddressed gaps. Using India as a case study (1.4 billion people, 22 scheduled languages, high-quality benchmarks, but no trusted aggregation), we report findings from a consultation with 58 AI practitioners showing consistent preference for formal governance and disclosure-based conflict management. The solution is not more data but better institutions: regional leaderboards with independent governance from the start.
Chinese Translation
本文立场论文认为，人工智能排行榜在结构上不适合服务全球南方，因为它们缺乏独立的治理、利益冲突政策和指标演变机制。障碍并非缺乏数据；高质量的区域基准已经存在：印度的IndicSUPERB、MILU和LAHAJA；非洲的IrokoBench；阿拉伯语的AlGhafa。障碍在于制度设计。全球排行榜不包括这些基准，且没有治理机制迫使它们这样做。当全球北方的付费客户受到影响时，商业压力会纠正排行榜的失误。全球南方缺乏相应的杠杆。没有治理，影响印地语、斯瓦希里语或阿拉伯语使用者的失误将无限期持续，正如已记录但未解决的差距所示。以印度为案例（14亿人口，22种计划语言，高质量基准，但没有可信的聚合），我们报告了与58位人工智能从业者的咨询结果，显示出对正式治理和基于披露的利益冲突管理的一致偏好。解决方案不是更多的数据，而是更好的制度：从一开始就具有独立治理的区域排行榜。
cs.AI / 13 / 2608.18131
Safety Alignment Illusion: The Cross-Lingual Safety Gap in LLMs
安全对齐幻觉：大型语言模型中的跨语言安全差距
Abstract
Current safety alignment training for Large Language Models (LLMs) are heavily English-centric. When such safety filters fail for non-English languages, the consequences are immediate and user-facing: voice assistants and spoken dialogue systems may produce stereotype-reinforcing outputs, bypassing the standard English-focused safety alignments and propagating harmful bias to non-English speaking communities. For spoken language technologies deployed across India's linguistically diverse population, this represents a critical failure mode. To address this cross-lingual gap, we introduce INCLUDE (Indian Cultural Lens for Understanding and Detecting Embedded Biases), a multilingual evaluation benchmark designed to quantify Indian-centric socio-cultural biases. INCLUDE consists of 2,604 prompts spanning six prompt languages: English, Hindi, Bengali, Marathi, Tamil, and Hinglish (Hindi-English code-mix). We evaluate ten open- and closed-source LLMs against this benchmark, analyzing 14,988 bias scores. Our statistical results reveal two key findings. First, Bengali yielded the highest average bias score in open-source models. Second, English demonstrated a notable reversal, producing the lowest bias in open-source models but the highest bias in closed-source models.
Chinese Translation
当前大型语言模型（LLMs）的安全对齐训练严重以英语为中心。当这种安全过滤器在非英语语言中失效时，后果是直接且面向用户的：语音助手和对话系统可能会产生强化刻板印象的输出，绕过以英语为中心的安全对齐，向非英语社区传播有害偏见。对于在印度语言多样化人群中部署的口语语言技术，这代表了一种关键的失效模式。为了解决这一跨语言差距，我们提出了INCLUDE（印度文化视角用于理解和检测嵌入偏见），这是一个旨在量化以印度为中心的社会文化偏见的多语言评估基准。INCLUDE包含2,604个提示，涵盖六种提示语言：英语、印地语、孟加拉语、马拉地语、泰米尔语和Hinglish（印地语-英语混合）。我们针对这一基准评估了十个开源和闭源的LLMs，分析了14,988个偏见分数。我们的统计结果揭示了两个关键发现。首先，孟加拉语在开源模型中产生了最高的平均偏见分数。其次，英语表现出显著的反转，在开源模型中产生最低的偏见，而在闭源模型中产生最高的偏见。
cs.AI / 14 / 2608.18133
Optimized Fuzzy Logic Approach with the IEEE Key Gas Method for Diagnosing Power Transformer Faults Using Dissolved Gas Analysis
基于IEEE关键气体法的优化模糊逻辑方法用于通过溶解气体分析诊断电力变压器故障
Abstract
Reliable transformer fault diagnosis is essential for maintaining power system stability. The IEEE Key Gas Method (KGM), a widely utilized approach in Dissolved Gas Analysis (DGA), exhibits limitations in addressing ambiguous data and ensuring high diagnostic accuracy. This study presents An enhanced model combining Fuzzy Logic with the IEEE Key Gas Method (FL-KGM) that introduces refined membership functions, optimized fuzzy rule sets, and a novel separation of CO and CO2 to eliminate diagnostic inconsistencies. By leveraging multidimensional gas ratio analysis and an adaptive classification framework, FL-KGM delivers superior fault identification and classification. Experimental validation utilizing real-world datasets demonstrates that FL-KGM achieves up to 98.6% accuracy, significantly outperforming KGM and other FL-based approaches. These findings elucidate the potential of FL-KGM in advancing transformer monitoring, enabling intelligent fault detection, and enhancing predictive maintenance strategies in modern power systems.
Chinese Translation
可靠的变压器故障诊断对于维护电力系统的稳定性至关重要。IEEE关键气体法（KGM）作为一种广泛应用于溶解气体分析（DGA）的方法，在处理模糊数据和确保高诊断准确性方面存在局限性。本研究提出了一种结合模糊逻辑与IEEE关键气体法的增强模型（FL-KGM），该模型引入了精细的隶属函数、优化的模糊规则集以及新颖的CO和CO2分离方法，以消除诊断不一致性。通过利用多维气体比率分析和自适应分类框架，FL-KGM提供了更优的故障识别和分类。利用真实世界数据集进行的实验验证表明，FL-KGM的准确率高达98.6%，显著优于KGM和其他基于模糊逻辑的方法。这些发现阐明了FL-KGM在推进变压器监测、实现智能故障检测以及增强现代电力系统预测性维护策略方面的潜力。
cs.AI / 15 / 2608.18135
Improving Rural Medication Safety with AI: A Scoping Review
利用人工智能改善农村用药安全：一项范围审查
Abstract
Introduction: Medication errors (MEs) represent a significant threat to global healthcare systems, contributing to patient harm. Introducing artificial intelligence (AI) in rural healthcare enhances patient safety. The aim is to explore the applications and effectiveness of AI technologies in enhancing patient safety and reducing medication errors in rural health settings. Methods: A scoping review was conducted through a systematic literature search spanning 2012 to 2025 across multiple databases, including EBSCohost, Emcare (Ovid), MEDLINE, and the ProQuest Consumer Health Database. Twelve primary studies from nine different nations were examined. Data were analysed thematically to obtain insights on AI interventions across the medication process. Results: AI technologies have been integrated into every stage of medication management, right from prescribing and dispensing to administration and post-administration monitoring. Four key themes came to light: (1) the various types of AI being utilised (like Clinical Decision Support Systems, Machine Learning, Natural Language Processing, and smart pumps); (2) the phases of the medication process that are affected; (3) how effective these technologies are in minimising errors and boosting workflow safety; and (4) rural-specific challenges including infrastructure, staff training, system integration, and alert fatigue. Several studies have demonstrated that machine learning-based surveillance improves incident detection and reduces prescribing and transcription errors by an impressive 34% to 80%. Barriers included lack of governance frameworks, financial limitations, and clinician resistance, which still present major obstacles. Conclusion: In rural healthcare, AI technologies hold great potential for enhancing pharmaceutical safety. They can allow data-driven monitoring, automate processes, and offer clinical decision assistance.
Chinese Translation
引言：用药错误（MEs）对全球医疗系统构成了重大威胁，导致患者受到伤害。在农村医疗中引入人工智能（AI）可以增强患者安全。本文旨在探讨AI技术在提高患者安全和减少农村医疗环境中用药错误方面的应用及有效性。方法：通过对多个数据库（包括EBSCohost、Emcare（Ovid）、MEDLINE和ProQuest消费者健康数据库）进行系统文献检索，开展了一项范围审查，检索时间跨度为2012年至2025年。共分析了来自九个不同国家的十二项主要研究。数据通过主题分析法进行分析，以获取关于AI干预在用药过程中的见解。结果：AI技术已被整合到用药管理的每个阶段，从处方、发药到给药和给药后监测。四个关键主题浮现： (1) 使用的各种AI类型（如临床决策支持系统、机器学习、自然语言处理和智能泵）； (2) 受影响的用药过程阶段； (3) 这些技术在减少错误和提高工作流程安全性方面的有效性； (4) 特有的农村挑战，包括基础设施、员工培训、系统集成和警报疲劳。多项研究表明，基于机器学习的监测提高了事件检测能力，并将处方和转录错误减少了34%至80%。障碍包括缺乏治理框架、财务限制和临床医生的抵制，这些仍然是主要障碍。结论：在农村医疗中，AI技术在提高药品安全方面具有巨大潜力。它们可以实现数据驱动的监测、自动化流程，并提供临床决策支持。
cs.AI / 16 / 2608.18136
FraudBench: Stress-Testing Policy-Grounded Banking Agents Against Adaptive Fraud
FraudBench：针对自适应欺诈的政策基础银行代理的压力测试
Abstract
Conversational agents now act for end users through tools while holding access to customer databases and internal policy documents that a caller can reach through dialogue alone. Banking is the clearest case: the same agent that answers a question can also change contact details, reset a PIN, or move money, so ordinary customer service is inseparable from authorization, fraud detection, and policy compliance. Existing financial-fraud benchmarks classify static transactions or messages, and general agent-safety benchmarks target prompt injection or generic harmful use; none test whether a policy-grounded banking agent safely acts when a caller manipulates identity, authorization, and trust over a conversation. We introduce FraudBench, an executable benchmark built on the $\tau^2$-bench dual-control framework and the $\tau$-Knowledge banking environment. Both the agent and the simulated caller act through tools over shared, mutable account state, and the agent may grant the caller access to selected tools; the environment exposes a 698-document internal policy corpus that the agent must retrieve from. FraudBench contains 150 authored adversarial scenarios; a frozen public set of 107 (90 across ten fraud mechanisms plus 17 chained adaptive attacks) is used for all reported runs, with 43 further chained attacks held out. Safety is history-dependent: single-control tasks satisfy every precondition but one, and adaptive attacks make a later, locally valid request unsafe because of an earlier probe, admission, or failed attempt. Each scenario is annotated with observable evidence, prohibited actions, safe dispositions, and intervention points. A preliminary single-trial evaluation of four agents on the 107 graded tasks yields attack-security between 49\% and 65\%, with money-mule and first-party fraud the most common cross-model weaknesses.
Chinese Translation
对话代理现在通过工具为最终用户提供服务，同时访问客户数据库和内部政策文件，这些文件只能通过对话访问。银行业是最明显的案例：同一个代理不仅可以回答问题，还可以更改联系信息、重置密码或转移资金，因此普通的客户服务与授权、欺诈检测和政策合规是不可分割的。现有的金融欺诈基准主要对静态交易或消息进行分类，而一般的代理安全基准则针对提示注入或一般性有害使用；没有任何基准测试政策基础的银行代理在通话中是否能安全地应对身份、授权和信任的操控。我们引入了FraudBench，这是一个基于$ au^2$-bench双控制框架和$ au$-Knowledge银行环境的可执行基准。代理和模拟的呼叫者通过共享的可变账户状态使用工具，代理可以授予呼叫者对选定工具的访问权限；该环境提供了一个698份文件的内部政策语料库，代理必须从中检索信息。FraudBench包含150个创作的对抗场景；一个冻结的公共集包含107个（涵盖十种欺诈机制的90个加上17个链式自适应攻击），用于所有报告的运行，另外43个链式攻击被保留。安全性依赖于历史：单控制任务满足每个前提条件，除了一个，而自适应攻击使得后续的局部有效请求因早期探测、承认或失败尝试而变得不安全。每个场景都附有可观察证据、禁止行为、安全处置和干预点的注释。对107个评分任务的四个代理进行的初步单次评估显示攻击安全性在49\%到65\%之间，钱骡（money-mule）和第一方欺诈是最常见的跨模型弱点。
cs.AI / 17 / 2608.18142
Efficient Adaptation of LLMs for Hate Speech Detection in Low-Resource Languages: A Comparative Study on Roman Urdu
低资源语言中仇恨言论检测的LLMs高效适应：关于罗马乌尔都语的比较研究
Abstract
It is challenging to detect hate speech in Low Resource Languages (LRLs) because of the absence of annotated data, the informality of its language structure, and the lack of standardized grammar. A good example of such a challenge is Roman Urdu which is broadly used by South Asians on social media and has a high variation while lacking contextually consistent spellings. The objective of this paper is to conduct a comprehensive assessment of Large Language Models (LLMs) for Hate Speech Detection (HSD) in Roman Urdu script and fine-tune these models using the Parameter-Efficient Fine-Tuning (PEFT) method called Low-Rank Adaptation (LoRA). To evaluate zero-shot inference, we benchmarked it against PEFT on different transformer models, including Mistral, LLaMA, Falcon, and multilingual BERT. Experiments are conducted on the PURUTT (Parallel Urdu and Roman Urdu Corpus for Toxic Comments and Transliteration) dataset with over 72,000 annotated comments. The results suggest that zero shot models perform moderately (F1 = 0.56), but updating a small fraction of the model trainable parameters improves the classification performance significantly (F1 > 0.93). Our results have shown that PEFT delivers outstanding performance alongside excellent computational efficiency, making it highly suitable for low-resource language processing tasks.
Chinese Translation
在低资源语言（LRLs）中检测仇恨言论面临诸多挑战，主要由于缺乏标注数据、语言结构的非正式性以及缺乏标准化语法。罗马乌尔都语就是一个很好的例子，它在社交媒体上被南亚人广泛使用，但拼写缺乏上下文一致性，且变化多样。本文的目标是对大型语言模型（LLMs）在罗马乌尔都语脚本中的仇恨言论检测（HSD）进行全面评估，并使用一种称为低秩适应（LoRA）的参数高效微调（PEFT）方法对这些模型进行微调。为了评估零样本推理，我们将其与不同变换器模型（包括Mistral、LLaMA、Falcon和多语言BERT）的PEFT进行了基准测试。实验在PURUTT（平行乌尔都语和罗马乌尔都语毒性评论及音译语料库）数据集上进行，该数据集包含超过72,000条标注评论。结果表明，零样本模型的表现中等（F1 = 0.56），但更新模型可训练参数的一小部分显著提高了分类性能（F1 > 0.93）。我们的结果表明，PEFT在提供卓越性能的同时具备出色的计算效率，使其非常适合低资源语言处理任务。
cs.AI / 18 / 2608.18165
RDFdL: Integrating RDF with Differential Dynamic Logic
RDFdL：将RDF与微分动态逻辑相结合
Abstract
Knowledge graphs modeled in RDF are powerful for describing static knowledge, but they cannot capture or reason about the dynamic behavior of physical systems, e.g., systems described by differential equations, which is a critical gap for AI-driven cyber-physical systems. To solve this, we propose RDFdL, a framework that integrates RDF with Differential Dynamic Logic (dL) to represent and reason about both static knowledge and the continuous dynamics of physical systems. For the dynamic part, we syntactically represent differential equations and ranges in the state space in RDF and SHACL and provide semantics using a translation to dL. Linking RDF and dL through their shared foundation in first-order logic achieves a unique integration: verification results for safety and reachability properties in the dynamic logic domain become available as entailment to SPARQL queries over RDF data. We implement the pipeline using Apache Jena for ontology-driven RDF reasoning and KeYmaera X, the theorem prover for dL, and sketch its applicability in manufacturing.
Chinese Translation
用RDF建模的知识图谱在描述静态知识方面具有强大的能力，但它们无法捕捉或推理物理系统的动态行为，例如由微分方程描述的系统，这对于基于人工智能的网络物理系统而言是一个关键的缺口。为了解决这个问题，我们提出了RDFdL，一个将RDF与微分动态逻辑（Differential Dynamic Logic, dL）相结合的框架，以表示和推理静态知识及物理系统的连续动态部分。在动态部分，我们在RDF和SHACL中以语法形式表示微分方程和状态空间中的范围，并通过翻译到dL提供语义。通过它们在一阶逻辑中的共同基础将RDF和dL连接起来，实现了独特的整合：动态逻辑领域中安全性和可达性属性的验证结果可以作为对RDF数据的SPARQL查询的蕴涵。我们使用Apache Jena实现了该管道，以进行本体驱动的RDF推理，并使用KeYmaera X作为dL的定理证明器，概述了其在制造业中的适用性。
cs.AI / 19 / 2608.18167
Adversarial Review: Structured Disagreement for Grounded Agentic Code Review
对抗性审查：基于结构化分歧的基础性自主代码审查
Abstract
Early multi-agent LLM systems often used role-separated teams, yet scaling agent count yields diminishing returns on repository-level coding tasks. Recent alternatives treat agents as passive tools (subagents), yet this removes the benefits of agent interaction entirely. We study whether a subagent paradigm can support a middle ground: minimal agentic cooperation without the overhead of large multi-agent teams. We introduce Adversarial Review (AR), a minimal cooperative code-review protocol in which a main coding agent works with a reviewer and a critic agent. The reviewer evaluates code, while the critic audits the review through structured disagreement before the main agent edits. On LiveCodeBench, AR achieves the highest pass rate among tested methods, outperforming a five-agent baseline while using only three agents. On SWE-PRBench, naive AR exposes a false-consensus failure mode, where agents converge on agreement without sufficient evidence, but a single prompt iteration that adds disagreement explicitly achieves the highest F1 among tested methods. On SWE-bench Verified, AR also shows improvements over the baselines on repository-level coding tasks. Together, AR demonstrates that cooperative code review does not require many agents or complex communication structures: it requires that disagreement be minimal, structured, and evidence-grounded.
Chinese Translation
早期的多智能体大语言模型（LLM）系统通常使用角色分离的团队，但增加智能体数量在代码库级别的编码任务上收益递减。近期的替代方案将智能体视为被动工具（子智能体），但这完全消除了智能体交互的好处。我们研究了子智能体范式是否可以支持一种折中方案：在不需要大型多智能体团队的情况下实现最小的自主合作。我们提出了对抗性审查（Adversarial Review, AR），这是一种最小合作的代码审查协议，其中主要编码智能体与审查者和批评智能体协作。审查者评估代码，而批评者通过结构化分歧对审查进行审核，然后主要智能体进行编辑。在 LiveCodeBench 上，AR 在测试方法中实现了最高的通过率，超越了使用五个智能体的基线，同时仅使用三个智能体。在 SWE-PRBench 上，简单的 AR 暴露了一个虚假共识失败模式，即智能体在没有足够证据的情况下达成一致，但通过一次明确增加分歧的提示迭代，AR 在测试方法中实现了最高的 F1 值。在 SWE-bench Verified 上，AR 在代码库级别的编码任务上也显示出相对于基线的改进。综合来看，AR 证明了合作代码审查并不需要大量智能体或复杂的沟通结构：它要求分歧是最小的、结构化的，并且以证据为基础。
cs.AI / 20 / 2608.18171
Looped Language Models Improve Compositional Tool Calling
循环语言模型改善组合工具调用
Abstract
Looped language models have shown promising results on reasoning benchmarks, yet their potential for agentic tool use remains largely unexplored. We study this question in compositional tool-calling settings, where models must coordinate multiple API calls, maintain intermediate state, and preserve dependencies across tool interactions. We evaluate native and retrofitted looped language models on API-Bank, BFCL, and NESTful, comparing looped and non-looped models trained under matched supervised fine-tuning recipes and varying recurrent depth at inference time. In controlled experiments, recurrent computation generally benefits compositional and dependency-aware tool use, while providing smaller and more model-dependent gains on isolated API invocation. Accuracy on multi-step tool use generally increases with recurrent depth; adaptive inference, however, achieves a more favorable compute-performance trade-off by allocating additional computation only when needed. Our results suggest that looped language models are a promising architecture for agentic systems that require reliable planning, coordination, and execution of compositional tool use workflows.
Chinese Translation
循环语言模型在推理基准测试中显示出良好的结果，但其在代理工具使用方面的潜力仍然未被充分探索。我们在组合工具调用的环境中研究这个问题，在这些环境中，模型必须协调多个 API 调用，维护中间状态，并保持工具交互之间的依赖关系。我们在 API-Bank、BFCL 和 NESTful 上评估了原生和改装的循环语言模型，比较了在匹配的监督微调方案下训练的循环和非循环模型，以及在推理时不同的递归深度。在受控实验中，递归计算通常有利于组合和依赖感知的工具使用，同时在孤立的 API 调用上提供了较小且更依赖模型的收益。多步骤工具使用的准确性通常随着递归深度的增加而提高；然而，自适应推理通过仅在必要时分配额外计算，实现了更有利的计算性能权衡。我们的结果表明，循环语言模型是一种有前景的架构，适用于需要可靠规划、协调和执行组合工具使用工作流的代理系统。
cs.AI / 21 / 2608.18194
On the Triangle Inequality for the Jaccard Distance in Arbitrary Lattices
任意格中Jaccard距离的三角不等式研究
Abstract
This paper presents new theoretical results on generalizing the Jaccard distance for lattices and real valuations. We demonstrate that when the valuation is strictly positive, monotone, and modular, the Jaccard distance satisfies the triangle inequality on arbitrary lattices, effectively generalizing earlier results that depended heavily on distributivity. Moving to relatively complemented distributive lattices (which safely drop the requirement for the global bounds found in Boolean algebras), we prove the triangle inequality holds as long as the valuation is positive, monotone, supermodular, and $\log$-submodular. Additionally, we adapt the symmetric-difference Jaccard formulation for submodular valuations to sectionally complemented distributive lattices. Shifting to necessary conditions, we prove that supermodularity is a strict requirement for the standard generalized Jaccard distance to operate as a valid metric. Finally, we map the practical value of relaxing these structural constraints to computational fields like quantum information theory, formal concept analysis, and machine learning, closing with a brief look at open mathematical problems.
Chinese Translation
本文提出了关于格和实值的Jaccard距离推广的新理论结果。我们证明，当值严格为正、单调且模态时，Jaccard距离在任意格上满足三角不等式，有效地推广了早期依赖于分配律的结果。转向相对补全的分配格（该类格安全地省略了布尔代数中全局界限的要求），我们证明只要值为正、单调、超模和$ ext{log}$-子模，三角不等式就成立。此外，我们将对子模值的对称差Jaccard公式适配到分段补全的分配格。转向必要条件，我们证明超模性是标准广义Jaccard距离作为有效度量运作的严格要求。最后，我们将放宽这些结构约束的实际价值映射到量子信息理论、形式概念分析和机器学习等计算领域，并简要回顾开放的数学问题。
cs.AI / 22 / 2608.18238
GenEx: A Graph-Based Representational Paradigm for SARS-CoV-2 Variant Detection via Codon Co-occurrence Networks
GenEx：一种基于图的表示范式，通过密码子共现网络检测SARS-CoV-2变种
Abstract
Genomic analysis on viruses such as SARS-CoV-2 variants: Beta, Gamma, Delta, and Omicron is heavily dominated by classical bioinformatics methods, including Sequence Alignment, Phylogenetic Analysis, and Mutation Frequency Statistics. These approaches use pairwise codon or nucleotide distance matrices to analyze gene sequences, treating them as linear strings rather than capturing their complex contextual interdependencies. We proposed GenEx, a pipeline that converts raw gene sequences into codon co-occurrence graphs and extracts more than 25 graph features. Our two most prominent techniques for graph generation and feature extraction are MSCG (Multi-Scale Codon Co-occurrence Graph) and LAPCG (Linear-time Adjacency PMI Codon Graph). Using these algorithms, we treated codon sequences as structured symbolic vocabularies interpretable to codon co-occurrence graph analysis, a representational paradigm borrowed from computational linguistics. Another major contribution includes implementing a spectral graph feature extraction using Singular Value Decomposition (SVD), using the squared singular value ($\sigma^2$) instead of the traditionally used eigenvalue, which helped us to amplify the separation between dominant and subdominant spectral components, thereby enhancing inter-class separability in downstream classification. And to further demonstrate that our method works, we trained 23 benchmarked ML models against the latest SARS-CoV-2 variants, achieving remarkable results in detecting all SARS-CoV-2 variants.
Chinese Translation
对SARS-CoV-2变种（如Beta、Gamma、Delta和Omicron）的基因组分析主要依赖于经典的生物信息学方法，包括序列比对、系统发育分析和突变频率统计。这些方法使用成对的密码子或核苷酸距离矩阵来分析基因序列，将其视为线性字符串，而未能捕捉其复杂的上下文相互依赖关系。我们提出了GenEx，一个将原始基因序列转换为密码子共现图并提取超过25个图特征的管道。我们用于图生成和特征提取的两种最显著技术是MSCG（多尺度密码子共现图）和LAPCG（线性时间邻接PMI密码子图）。使用这些算法，我们将密码子序列视为可解释的结构化符号词汇，以便进行密码子共现图分析，这是一种借鉴自计算语言学的表示范式。另一个主要贡献是实现了一种使用奇异值分解（SVD）的谱图特征提取，采用平方奇异值（$ ext{σ}^2$）而不是传统使用的特征值，这有助于增强主导和次主导谱成分之间的分离，从而提高下游分类中的类间可分性。为了进一步证明我们的方法有效，我们针对最新的SARS-CoV-2变种训练了23个基准机器学习模型，在检测所有SARS-CoV-2变种方面取得了显著成果。
cs.AI / 23 / 2608.18260
Redakto - The Incognito Tab for LLMs
Redakto - LLMs的隐私标签
Abstract
Large Language Models (LLMs) are being increasingly used in everyday applications. A major challenge in the context of LLMs or Artificial Intelligence (AI) in general is to ensure privacy when using them, meaning that personally identifiable information (PII) is removed from any text that enters an LLM. These challenges have become more urgent with novel EU legislation. Uncertainty around LLM usage with respect to privacy concerns in EU countries can be a major blocker for the speed of innovation and transfer from research to applications. Here we present \textbf{Redakto}, a tool that can be used for anonymizing text prior to feeding it to an LLM or other downstream text processing. We provide state-of-the-art functionalities for both redaction of PII but also when used for pseudonymization. These functionalities are exposed such that they can easily be used by end-users, through the Redakto web application, and by developers and researchers, via REST APIs and model context protocol (MCP) hooks. The implementation is fully open source, requires modest compute resources, and can be readily deployed on local hardware. In contrast to prior work and in order to better assess the quality of the anonymized texts, we conduct extensive empirical evaluations on textual data from legal and medical domain with respect to both privacy and utility of the redacted texts. Our empirical results demonstrate that the texts anonymized with different redaction strategies achieve utility scores on par with the original texts, suggesting that anonymization with Redakto can be used for LLM tasks without substantial negative impact for the tasks we explored.
Chinese Translation
大型语言模型（LLMs）在日常应用中的使用越来越广泛。在LLMs或人工智能（AI）的一般背景下，一个主要挑战是确保使用时的隐私，这意味着需要从进入LLM的任何文本中移除个人可识别信息（PII）。随着新的欧盟立法的出台，这些挑战变得更加紧迫。关于LLM使用的隐私问题的不确定性可能会成为创新速度和研究向应用转化的主要障碍。在此，我们介绍了 extbf{Redakto}，这是一种可以在将文本输入LLM或其他下游文本处理之前用于匿名化的工具。我们提供了最先进的功能，既可以用于PII的删除，也可以用于假名化。这些功能以易于最终用户通过Redakto网络应用程序使用的方式呈现，同时也供开发者和研究人员通过REST API和模型上下文协议（MCP）钩子使用。该实现完全开源，所需计算资源适中，并且可以轻松部署在本地硬件上。与之前的工作相比，为了更好地评估匿名文本的质量，我们对来自法律和医学领域的文本数据进行了广泛的实证评估，关注于删除文本的隐私性和实用性。我们的实证结果表明，采用不同删除策略的文本在实用性评分上与原始文本相当，这表明使用Redakto进行匿名化可以在我们探索的任务中用于LLM任务，而不会产生显著的负面影响。
cs.AI / 24 / 2608.18261
Cacheable by Design? Training Mixture-of-Experts Routers for Locality Against the Edge Memory-Bandwidth Wall: A Pre-Registered Negative Result with a Systems Measurement Study
设计可缓存？针对边缘内存带宽壁垒训练混合专家路由器以提高局部性：一项预注册的负面结果与系统测量研究
Abstract
Serving a 235B-parameter Mixture-of-Experts (MoE) model on a single 8 GB GPU is bottlenecked not by compute but by memory bandwidth: decode must stream each token's active experts from whichever tier holds them, and on consumer hardware most experts sit on an SSD far slower than RAM. We quantify this bandwidth wall on Qwen3-235B (Q4_K_M, 134 GB): measured decode is 0.44 tok/s warm, matching a bytes-per-token / bandwidth model, while a batching scheme that should amortize one disk sweep instead collapses at batch 32 from paging thrash. We build llama-moe-trace, a zero-surgery router-telemetry tool, and measure routing on Qwen3-30B: adjacent-token expert reuse is 2.0x chance, 95% of traffic uses 52.5% of experts, and an LRU cache of 13.4% of experts serves 66% of requests. We then ask whether cacheability is trainable: we pre-register training of 137M MoE language models with auxiliary locality and domain router losses, under joint criteria on cache-miss reduction and perplexity. The mechanism works (misses down up to 60%; a 99% static-pin hit rate) but every configuration fails the pre-registered <=1% perplexity gate -- miss reduction and quality are tightly coupled. Concurrent StickyMoE reports the same loss as near-free on single-domain sub-25M models; on multi-domain 137M we find the tax real. Our contribution is this pre-registered, stricter-criterion, multi-domain evaluation plus edge-serving measurements. A 340M rung shows the tax does not shrink with scale (it rises slightly). We further show training-free cache-aware rerouting stacks with trained locality -- together ~80% miss reduction at <=3.4% perplexity at both sizes, far cheaper than either alone -- while domain-primed prefetching does not help. All code, traces, and the pre-registration are released.
Chinese Translation
在单个8 GB GPU上服务于一个235B参数的混合专家（Mixture-of-Experts, MoE）模型，其瓶颈并非计算能力，而是内存带宽：解码必须从持有每个令牌的活跃专家的任一层流式传输，而在消费级硬件上，大多数专家位于SSD上，其速度远低于RAM。我们量化了Qwen3-235B（Q4_K_M，134 GB）上的带宽壁垒：测得的解码速度为0.44个令牌/秒，符合每个令牌的字节数/带宽模型，而一个应该摊销一次磁盘扫描的批处理方案在批量32时因页面抖动而崩溃。我们构建了llama-moe-trace，一个零手术的路由遥测工具，并在Qwen3-30B上测量路由：相邻令牌的专家重用概率为2.0倍，95%的流量使用52.5%的专家，而一个包含13.4%专家的LRU缓存满足66%的请求。然后我们询问缓存可训练性：我们预注册了137M MoE语言模型的训练，辅以局部性和领域路由损失，基于缓存未命中减少和困惑度的联合标准。该机制有效（未命中减少高达60%；99%的静态固定命中率），但每种配置都未能通过预注册的<=1%困惑度门槛——未命中减少和质量紧密相关。并行的StickyMoE报告在单领域子25M模型上几乎无损失；在多领域137M模型上，我们发现这种代价是真实存在的。我们的贡献在于这项预注册的、更严格标准的多领域评估以及边缘服务测量。一个340M的层级显示，随着规模的扩大，这种代价并未缩小（略微上升）。我们进一步展示了无训练的缓存感知重路由堆栈与训练局部性结合——在两种规模下，未命中减少约80%，困惑度<=3.4%，远低于单独使用任一方法的成本——而领域预取并未提供帮助。所有代码、跟踪和预注册信息均已发布。
cs.AI / 25 / 2608.18289
Evaluating Structured Information Extraction with Open Models in a High Risk Public Sector Application
在高风险公共部门应用中评估开放模型的结构化信息提取
Abstract
The extraction of structured information from unstructured documents represents a critical component of digital transformations in all sectors. While proprietary solutions dominate commercial applications, a rapidly growing ecosystem of open-source Optical Character Recognition (OCR) engines, Large Language Models (LLMs), and Vision-Language Models (VLMs) offers accessible alternatives. However, systematic evaluations on realistic, multi-step extraction pipelines remain scarce. Responsible usage of such extraction tools require comprehensive evaluations on realistic tasks, especially as these solutions will be key components of applications in the public sector that the EU AI act categorizes as high risk. To address this gap we present a comprehensive benchmark assessing the end-to-end performance of open-source systems on a complex real-world document processing task classified as high risk: Student applications for an international study program. We conduct a comprehensive empirical evaluation with state-of-the-art OCR engines, LLMs and VLMs. Our results reveal that while VLMs generally outperform OCR+LLM pipelines, even state-of-the-art open-source models struggle to handle such tasks reliably in zero-shot settings. Only 4 of 35 configurations achieved F1 scores above 0.5, with the best OCR+LLM pipeline matching top VLM performance, though most OCR+LLM combinations performed substantially worse. Roughly 75\% of all configurations scored below 0.25. Model scale influences performance, yet the relationship is non-linear: substantially larger models do not guarantee proportionally better results. Input quality, particularly the structural preservation of OCR output, emerges as a critical factor independent of downstream model capability.
Chinese Translation
从非结构化文档中提取结构化信息是各个领域数字化转型的关键组成部分。尽管专有解决方案主导了商业应用，但一个快速增长的开源光学字符识别（OCR）引擎、大型语言模型（LLMs）和视觉-语言模型（VLMs）生态系统提供了可获取的替代方案。然而，关于现实多步骤提取流程的系统评估仍然稀缺。负责任地使用这些提取工具需要对现实任务进行全面评估，尤其是因为这些解决方案将成为欧盟人工智能法案所分类为高风险的公共部门应用的关键组成部分。为了解决这一空白，我们提出了一个全面的基准，评估开源系统在一个复杂的现实文档处理任务中的端到端性能，该任务被分类为高风险：国际学习项目的学生申请。我们进行了全面的实证评估，使用了最先进的OCR引擎、LLMs和VLMs。我们的结果显示，尽管VLMs通常优于OCR+LLM管道，但即使是最先进的开源模型在零样本设置中也难以可靠地处理此类任务。35种配置中仅有4种达到了0.5以上的F1分数，最佳的OCR+LLM管道与顶级VLM的性能相匹配，尽管大多数OCR+LLM组合的表现明显较差。大约75%的所有配置得分低于0.25。模型规模影响性能，但这种关系是非线性的：显著更大的模型并不保证获得成比例更好的结果。输入质量，特别是OCR输出的结构保持，成为一个独立于下游模型能力的关键因素。
cs.AI / 26 / 2608.18300
The Lifecycle of LLM-as-a-Judge for Large-Scale Recommendation Explanations
大规模推荐解释中 LLM 作为评判者的生命周期
Abstract
LLM-as-a-Judge, which leverages a large language model to evaluate natural language generated by another AI application or model, has become a standard, scalable approach for accelerating and extending costly human evaluation. However, most work treats a judge as a static artifact, evaluating it once at construction or against a fixed benchmark. In contrast, we argue that an LLM judge running in a production system is better understood as having a lifecycle: it must be built, trained, deployed, and continuously maintained as the surrounding data evolves, and each phase poses distinct technical and operational challenges. We present such a lifecycle for the LLM judges that evaluate user-facing recommendation explanations at Netflix, where our pipeline generates and the judges assess hundreds of thousands of distinct show-level explanations per week, served across the mobile experience to millions of members. Our framework has four phases: (I) Birth, defining multiple evaluation criteria and building curated benchmark datasets with human labels and rationales; (II) Training, refining the judges' rubrics via Reasoning-Aligned Rubric Tuning (RART), a rubric-tuning procedure that uses a meta-judge over reasoning output as the learning signal; (III) Deployment, in which one judge serves two production roles: quality gating and reflective generation; and (IV) Monitoring, a continuous Human-in-the-Loop alignment process that detects drift and triggers re-tuning behind a human review gate. We report post-launch results from a five-week A/B test over tens of millions of members, in which the judge-aligned explanations shifted member viewing toward novel content (previously unwatched) and increased successful browse-to-play sessions relative to a no-explanation control, with no quality-related takedowns.
Chinese Translation
LLM 作为评判者（LLM-as-a-Judge）利用大型语言模型评估由其他人工智能应用或模型生成的自然语言，已成为加速和扩展昂贵的人类评估的标准可扩展方法。然而，大多数研究将评判者视为静态的人工制品，仅在构建时或针对固定基准进行一次评估。相反，我们认为在生产系统中运行的 LLM 评判者应被理解为具有生命周期：它必须被构建、训练、部署，并随着周围数据的演变持续维护，每个阶段都面临独特的技术和操作挑战。我们提出了一个生命周期，适用于评估 Netflix 用户面向推荐解释的 LLM 评判者，我们的管道每周生成并评估数十万种不同的节目级解释，服务于数百万会员的移动体验。我们的框架分为四个阶段：（I）出生，定义多个评估标准并构建带有人类标签和理由的精心策划的基准数据集；（II）训练，通过推理对齐的评分标准调优（Reasoning-Aligned Rubric Tuning, RART）来优化评判者的评分标准，这是一种使用推理输出作为学习信号的元评判程序；（III）部署，其中一个评判者承担两个生产角色：质量门控和反思生成；（IV）监控，一个持续的人机协作对齐过程，检测漂移并在人工审核门后触发重新调优。我们报告了在数千万会员中进行的为期五周的 A/B 测试的发布后结果，其中评判者对齐的解释使会员观看新内容（之前未观看的内容）的倾向发生变化，并相较于无解释的对照组增加了成功的浏览到播放会话，且没有质量相关的下架。
cs.AI / 27 / 2608.18303
SESSE: Sketch, Expand, Sort, Summarize, Evaluate -- LLM-as-Judge Evaluation via Structured Decomposition
SESSE：草图、扩展、排序、总结、评估——通过结构化分解实现的 LLM 作为评判者的评估
Abstract
LLM-as-judge evaluation reduces response quality assessment to a single holistic A/B preference choice, providing no mechanism to isolate which quality dimensions drove the preference or distinguish model errors from genuine label ambiguity. We propose SESSE (Sketch, Expand, Sort, Summarize, Evaluate), a training-free framework that decomposes holistic judgment into structured sub-questions mined directly from the judge's own error cases; requiring no oracle responses, task-specific rubrics, or fine-tuning. On RewardBench (n=1,000), SESSE achieves near-parity with the chain-of-thought baseline and is competitive with RISE-Judge-32B (92.7%), a fine-tuned specialist, while remaining fully training-free. Per-criterion vote evidence provides an interpretable audit trail for diagnosing label ambiguity and judge failure modes unavailable from a single holistic output token.
Chinese Translation
LLM 作为评判者的评估将响应质量评估简化为单一的整体 A/B 偏好选择，未提供机制以隔离驱动偏好的质量维度或区分模型错误与真实标签模糊性。我们提出了 SESSE（Sketch, Expand, Sort, Summarize, Evaluate），这是一个无训练框架，将整体判断分解为直接从评判者自身错误案例中挖掘的结构化子问题；不需要预言者响应、任务特定评分标准或微调。在 RewardBench（n=1,000）上，SESSE 达到了与链式思维基线几乎相当的水平，并且与 RISE-Judge-32B（92.7%）这一微调专家模型具有竞争力，同时保持完全无训练。每个标准的投票证据提供了可解释的审计轨迹，用于诊断标签模糊性和评判者失败模式，而这些在单一整体输出标记中无法获得。
cs.AI / 28 / 2608.18307
ComponentBench: Diagnosing Component-Level Failures in Computer-Use Agents
ComponentBench：计算机使用代理的组件级故障诊断
Abstract
Current evaluation of computer-use agents is split between long-horizon workflow benchmarks and atomic GUI-grounding tests. This leaves an under-instrumented middle layer: realistic component-centered interactions (e.g., toggle a button set) that are short enough to diagnose and rich enough to capture the burdens of modern interfaces. We present ComponentBench, a benchmark and diagnostic pipeline for component-level evaluation of computer-use agents on modern web UIs. ComponentBench is organized around a library-agnostic ontology of 97 canonical UI components instantiated as 2,910 programmatically verified tasks across widely used component libraries, paired with cleaned human reference trajectories that enable evaluation of both task success and interaction efficiency. Beyond task collection, we introduce a scalable pipeline for auditing realized structural difficulty after implementation and synthesizing structured failure analyses across tasks and component families. Evaluating seven models -- GPT-5.4, Gemini 3 Flash, GPT-5.4 mini, GPT-5 mini, Gemini 3.1 Flash-Lite, Qwen3-VL-235B, and UI-TARS-1.5-7B -- across four observation and action spaces, we show that these design choices critically impact performance. Within a single shared harness, changing only the observation and action space shifts task success by more than 30% for the same model: GPT-5 mini falls from 83.1% with accessibility-tree observations to 48.9% with coordinate-only Pixel control. Moreover, even the fastest configuration takes 3.7x as long as the matched human reference, and spatial manipulations that are trivial for humans continue to challenge current agents.
Chinese Translation
当前对计算机使用代理的评估分为长时间流程基准测试和原子化的图形用户界面（GUI）基础测试。这导致了一个缺乏充分工具的中间层：现实的以组件为中心的交互（例如，切换一组按钮），这些交互既足够短以便于诊断，又足够丰富以捕捉现代界面的负担。我们提出了ComponentBench，一个用于现代网页用户界面上计算机使用代理的组件级评估的基准和诊断管道。ComponentBench围绕一个与库无关的本体组织，包含97个典型的用户界面组件，实例化为2910个经过程序验证的任务，涵盖广泛使用的组件库，并配有清理过的人类参考轨迹，以便评估任务成功率和交互效率。除了任务收集，我们还引入了一个可扩展的管道，用于审计实现后的结构难度，并在任务和组件家族之间综合结构化故障分析。我们在四个观察和行动空间中评估了七个模型——GPT-5.4、Gemini 3 Flash、GPT-5.4 mini、GPT-5 mini、Gemini 3.1 Flash-Lite、Qwen3-VL-235B和UI-TARS-1.5-7B，结果表明这些设计选择对性能有重要影响。在一个共享的测试环境中，仅改变观察和行动空间就使得同一模型的任务成功率变化超过30%：GPT-5 mini在可访问性树观察下的成功率为83.1%，而在仅使用坐标的像素控制下降至48.9%。此外，即使是最快的配置也需要3.7倍于匹配人类参考的时间，而对人类而言微不足道的空间操作仍然对当前代理构成挑战。
cs.AI / 29 / 2608.18324
Governance Records as Supervision: Verifier-Selected Self-Training for Structured Workflow Repair
治理记录作为监督：验证者选择的自我训练用于结构化工作流修复
Abstract
Machine-verifiable workflows produce governance records linking a task contract, model attempt, verifier decision, accepted output, and target origin. We test whether these records can supervise bounded models, consolidating occasional or expensive capability into reliable one-shot execution. On fresh, structure-disjoint PlanBench replanning cases, Qwen3-14B thinking generated 24 plans admitted by the independently authored VAL verifier. Those plans trained the same checkpoint for non-thinking execution, without oracle targets or a stronger teacher. On 80 unopened cases, VAL-accepted plans increased from 1 to 57, with 56 paired gains and zero regressions; thinking reached 30. The adapter was schema-valid on all cases and used approximately 1/56 of thinking's mean latency. The separate paired interface-cure gate did not pass. A matched ablation fixed the source cases, 52-candidate pool, 24-target count, model, recipe, and seed while changing target selection. On 160 new cases, base, schema-selected, model-self-selected, and VAL-selected execution reached 1, 55, 69, and 102 accepted plans. VAL exceeded self-selection by paired net +33 (p=0.0000019647), with gains in both difficulty strata. Independent semantic selection is therefore load-bearing relative to matched alternatives within this band. A complementary Phi stronger-teacher arm raised base Phi-4 from 2 to 51 accepted plans and from 35 to 80 schema-valid outputs. Earlier synthetic experiments establish teachability, cumulative learning, construction robustness, and stopping boundaries. The results support verifier-selected supervision for bounded, machine-checkable capabilities, not arbitrary planning, enterprise validity, or unrestricted self-improvement.
Chinese Translation
机器可验证的工作流生成治理记录，链接任务合同、模型尝试、验证者决策、接受的输出和目标来源。我们测试这些记录是否能够监督有限模型，将偶尔或昂贵的能力整合为可靠的一次性执行。在全新的、结构不相交的PlanBench重新规划案例中，Qwen3-14B思考生成了24个计划，这些计划被独立撰写的VAL验证者接受。这些计划训练了相同的检查点用于非思考执行，无需oracle目标或更强的教师。在80个未开启的案例中，VAL接受的计划从1个增加到57个，其中56个配对增益且没有回归；思考达到了30。适配器在所有案例中都是模式有效的，使用了大约1/56的思考平均延迟。单独的配对接口修复门未通过。一个匹配的消融修复了源案例、52个候选池、24个目标计数、模型、配方和种子，同时改变了目标选择。在160个新案例中，基础、模式选择、模型自我选择和VAL选择的执行分别达到了1、55、69和102个接受的计划。VAL在配对净增33（p=0.0000019647）上超过了自我选择，并且在两个难度层次中都有增益。因此，独立的语义选择在这一范围内相对于匹配的替代方案是承载性的。一个补充的Phi强教师臂将基础Phi-4从2个接受的计划提高到51个，并将模式有效输出从35个提高到80个。早期的合成实验确立了可教性、累积学习、构建稳健性和停止边界。结果支持验证者选择的监督用于有限的、机器可检查的能力，而非任意规划、企业有效性或无限制的自我改进。
cs.AI / 30 / 2608.18336
Measuring the Partial-Credit Gap: A Strict Benchmark on Vietnam's 2025 Convex Marking Scheme
测量部分得分差距：越南2025年凸性评分方案的严格基准
Abstract
When evaluating language models on human exams, benchmarks typically score each response as right or wrong and report the overall accuracy. This approach assumes that partial knowledge is worth proportional credit, an assumption that fails when an examination uses a non-additive grading scheme. The 2025 reform of Vietnam's National High School Graduation Examination demonstrates the cost of this substitution. In Part II of the exam, candidates evaluate four true/false statements per question. The grading is convex: the number of correct statements earns 0, 0.10, 0.25, 0.50, or 1.00 points. Identifying three statements correctly pays 0.50 points, not the 0.75 points that standard accuracy metrics would award. Because Part II accounts for 4.00 of the exam's 10.00 points, reporting accuracy inflates the score by rewarding partial knowledge that the state explicitly penalizes. We introduce THPT-Ladder, a benchmark of 632 items from 21 official exams across 11 subjects, graded exactly as the ministry grades its students. The ministry publishes the marks of over a million candidates, allowing us to place models directly into the human cohort. Across eight models, the official rubric pays 0.020 to 0.159 points less per Part II question than proportional credit. This shortfall changes a model's apparent competence. For Qwen3.5-27B on the 2025 History exam, a 0.042-point shortfall drops its standing from the 90th to the 77th percentile among 481,293 candidates. A model's accuracy does not predict this penalty. At Claude Sonnet 5's accuracy level, different distributions of errors yield scores varying from 0.869 to 0.932 points per question. Official marks depend on how correct statements are grouped, meaning standard benchmarks report a competence the institution would not certify.
Chinese Translation
在对语言模型进行人类考试评估时，基准通常将每个回答评分为正确或错误，并报告整体准确性。这种方法假设部分知识应获得相应的信用，但在使用非加性评分方案的考试中，这一假设并不成立。越南国家高中毕业考试的2025年改革展示了这种替代的代价。在考试的第二部分，考生每道题评估四个真/假陈述。评分是凸性的：正确陈述的数量可获得0、0.10、0.25、0.50或1.00分。正确识别三个陈述只获得0.50分，而不是标准准确性指标所奖励的0.75分。由于第二部分占考试总分10.00分中的4.00分，报告准确性会通过奖励国家明确惩罚的部分知识来夸大分数。我们引入了THPT-Ladder，这是一个包含21个官方考试、11个科目的632个项目的基准，评分方式与教育部对学生的评分完全一致。教育部公布了超过一百万考生的分数，使我们能够将模型直接置于人类群体中。在八个模型中，官方评分标准在第二部分每道题上比按比例信用少支付0.020到0.159分。这一短缺改变了模型的表面能力。在2025年历史考试中，Qwen3.5-27B的0.042分短缺使其在481,293名考生中从第90百分位降至第77百分位。模型的准确性并不能预测这一惩罚。在Claude Sonnet 5的准确性水平下，不同的错误分布导致每道题的得分从0.869到0.932分不等。官方分数取决于正确陈述的分组方式，这意味着标准基准报告了一个机构不会认证的能力。
cs.AI / 31 / 2608.18389
A Jagged Frontier: Evaluating Robustness of Code Agents to Semantics-Preserving Transformations
锯齿状前沿：评估代码代理对语义保持变换的鲁棒性
Abstract
AI code agents are increasingly deployed to resolve real software issues, yet their reliability under superficial code variations remains poorly understood. We evaluate whether coding agents that repair repository-level issues remain reliable when the surrounding codebase is rewritten into a semantically equivalent form. We introduce a random variant sampler that applies common semantics-preserving transformations (SPTs) - spanning control-flow rewrites, dead-code injection, and identifier renaming - to produce perturbed variants. We evaluate two agentic scaffolds (mini-SWE agent and OpenCode) each backed by one of four frontier models (Claude Opus 4.5, Kimi K2.5, MiniMax M2.5, and Qwen 3.6-27B) across instances drawn from SWE-bench Verified and SWE-bench Pro. For each instance, the agent is run multiple times on the unperturbed and perturbed variants, yielding paired resolve-rate estimates that isolate the perturbation effect from intrinsic stochasticity. We find small degradation in most configurations: up to 6.7 percentage points mean resolve-rate drop in the most affected configurations with statistically significant degradations in 6 of 16 configurations of model, scaffold, and dataset. Crucially, no single model ranking by robustness holds across scaffolds - Qwen is among the most robust under mini-SWE agent on SWE-bench Verified yet the most brittle under OpenCode - revealing a jagged robustness frontier. The simpler scaffold (mini-SWE agent) is more robust to perturbation. Our results demonstrate that even top frontier models are susceptible to semantics-preserving perturbations although the effect is not uniform, raising concerns about the deployment reliability of AI code agents in diverse real-world codebases.
Chinese Translation
人工智能代码代理越来越多地被用于解决实际软件问题，但它们在表面代码变动下的可靠性仍然不甚了解。我们评估了修复仓库级问题的代码代理在周围代码库被重写为语义等价形式时是否仍然可靠。我们引入了一种随机变体采样器，应用常见的语义保持变换（SPTs）——包括控制流重写、死代码注入和标识符重命名——以生成扰动变体。我们评估了两种代理框架（mini-SWE agent 和 OpenCode），每种框架均由四个前沿模型中的一个（Claude Opus 4.5、Kimi K2.5、MiniMax M2.5 和 Qwen 3.6-27B）支持，实例来自于 SWE-bench Verified 和 SWE-bench Pro。对于每个实例，代理在未扰动和扰动变体上多次运行，得出配对的解决率估计，以将扰动效应与内在随机性分离。我们发现大多数配置中存在轻微的下降：在受影响最严重的配置中，平均解决率下降最多达 6.7 个百分点，且在 16 种模型、框架和数据集配置中有 6 种配置的降级具有统计显著性。重要的是，没有单一模型的鲁棒性排名在不同框架中保持一致——在 mini-SWE agent 下，Qwen 在 SWE-bench Verified 上是最鲁棒的，但在 OpenCode 下则是最脆弱的——揭示了一个锯齿状的鲁棒性前沿。更简单的框架（mini-SWE agent）对扰动更具鲁棒性。我们的结果表明，即使是顶尖的前沿模型也容易受到语义保持扰动的影响，尽管这种影响并不均匀，这引发了对人工智能代码代理在多样化实际代码库中部署可靠性的担忧。
cs.AI / 32 / 2608.18397
When Clean Signals Are Not Enough: Detecting Structural Ambiguity for Safe Wearable Stress Classification
当干净信号不足时：检测结构模糊性以实现安全的可穿戴压力分类
Abstract
Wearable stress classifiers can achieve strong average performance while failing completely for a particular individual. On WESAD, a Random Forest reaches 93.0% mean accuracy yet yields F1 = 0 for Subject 14, whose cross-signal coupling weakens near stress onset. We call this structural ambiguity: individually plausible physiological channels form an inter-signal pattern that is poorly supported by the person's non-stress reference. We introduce the Individual Conformal Coupling Monitor (ICCM), a lightweight and transparent pre-inference monitor that quantifies subject-specific coupling divergence and routes each window to classify, defer, or abstain without retraining the downstream classifier. Across WESAD (N = 15) and Stress-Predict (N = 35), full-cohort Pearson associations between ambiguity and accuracy are negative (r = -0.607, p = 0.016; r = -0.412, p = 0.014). Robustness analyses temper this finding: rank correlations are not significant, and the WESAD association disappears when Subject 14 is removed. ICCM changes false-positive counts from 29 to 27 and 94 to 92, although neither paired change is significant. It withholds 3 of Subject 14's 21 stress windows but does not repair the missed-stress failure. These results position ICCM as an interpretable signal of unsupported physiology and individual failure, rather than a stand-alone safety guarantee.
Chinese Translation
可穿戴压力分类器在整体表现上可以达到较强的平均性能，但对于特定个体可能完全失效。在 WESAD 数据集上，随机森林模型的平均准确率达到 93.0%，但对于第 14 位受试者，其 F1 值为 0，因为该受试者在压力开始时的交叉信号耦合减弱。我们称之为结构模糊性：个体上看似合理的生理通道形成了一种信号间模式，但该模式在该个体的非压力参考中支持不足。我们引入了个体一致性耦合监测器（Individual Conformal Coupling Monitor, ICCM），这是一种轻量且透明的推理前监测工具，能够量化特定受试者的耦合偏差，并在不重新训练下游分类器的情况下，决定每个窗口是分类、延迟还是放弃。在 WESAD（N = 15）和 Stress-Predict（N = 35）数据集中，模糊性与准确性之间的全体 Pearson 关联为负相关（r = -0.607, p = 0.016; r = -0.412, p = 0.014）。稳健性分析缓和了这一发现：秩相关性不显著，并且当去除第 14 位受试者时，WESAD 的关联消失。ICCM 将第 14 位受试者的假阳性计数从 29 降至 27，从 94 降至 92，尽管这两个变化均不显著。它保留了第 14 位受试者的 21 个压力窗口中的 3 个，但并未修复未检测到压力的失败。这些结果将 ICCM 定位为一种可解释的信号，表明生理支持不足和个体失败，而不是单独的安全保证。
cs.AI / 33 / 2608.18409
Improving Natural-Language Combinatorial-Optimization Accuracy in Resource-Constrained Language Models via Formal Abstractions
通过形式抽象提高资源受限语言模型中自然语言组合优化的准确性
Abstract
Combinatorial scheduling poses a significant challenge for language models, requiring them to identify feasible solutions within exponentially large search spaces while satisfying complex constraints. This challenge is especially pronounced in resource-constrained settings, where larger language models are impractical and selection is limited to smaller models which often fail to preserve feasibility when scheduling directly from natural language. To address these limitations, we introduce SDDL, a neuro-symbolic framework that translates natural-language scheduling problems into compact, solver-aligned representations of tasks, resources, constraints, and objectives, while delegating low-level modeling and search to a deterministic compiler and external solver. On a 300-instance, multi-family subset of scheduling problems, SDDL improves independently verified feasibility for every resource-constrained model tested. The two strongest SDDL configurations reach 55.3% and 28.3%, up from direct-generation baselines of 23.7% and 1.3% and solver-code baselines of 21.7% and 7.0%, with a 0.0% median optimality gap among feasible schedules. By expressing problem structure rather than generating solutions or solver code, SDDL enables smaller models to approach the strongest evaluated direct- and solver-code configurations, including substantially larger frontier models.
Chinese Translation
组合调度对语言模型提出了重大挑战，要求它们在指数级大的搜索空间中识别可行解，同时满足复杂约束。在资源受限的环境中，这一挑战尤为明显，因为较大的语言模型不切实际，选择仅限于较小的模型，而这些模型在直接从自然语言调度时往往无法保持可行性。为了解决这些局限性，我们引入了SDDL（神经符号框架），该框架将自然语言调度问题转换为紧凑的、与求解器对齐的任务、资源、约束和目标的表示，同时将低级建模和搜索委托给确定性编译器和外部求解器。在一个包含300个实例的多家庭调度问题子集上，SDDL提高了每个经过独立验证的资源受限模型的可行性。两个最强的SDDL配置分别达到了55.3%和28.3%，相比之下，直接生成的基线为23.7%和1.3%，求解器代码的基线为21.7%和7.0%，而可行调度的中位最优性差距为0.0%。通过表达问题结构而不是生成解决方案或求解器代码，SDDL使得较小的模型能够接近评估的最强直接和求解器代码配置，包括显著更大的前沿模型。
cs.AI / 34 / 2608.18423
FM-Bench: A Benchmark for Long-Horizon Management with Competing Agents
FM-Bench：一个用于竞争代理的长期管理基准
Abstract
Language model agents now execute bounded tasks reliably. Whether they can sustain effective decision-making over long horizons, where actions have cumulative consequences and the environment responds to their choices, remains largely unmeasured. FM-Bench (Football Management Benchmark) measures this. An LLM agent runs a football club for 20 in-game years through 26 tools and roughly 340 to 400 decision stops. It drafts a squad on the same budget as every rival, trades players, negotiates contracts, invests in facilities and youth, sets lineups, and answers to a board that can fire it, while a deterministic engine accumulates every year into one final score with no LLM judge or human rater. The solo track plays each of 15 frontier models against a frozen scripted world, and the Arena places the same models plus a scripted anchor in one shared 20-year world; to our knowledge, the first head-to-head evaluation at this scale. We measure six behavioral capabilities behind the score. Across three seeds, all 15 models complete every horizon while the blind scripted baselines die out in most of theirs, and claude-fable-5 tops the solo board on mean score and the Arena, where the title nonetheless rotates among ten models. Neither scale, price, nor vendor predicts the order; the order settles only late in the horizon, and the best first-play human lands only at the bottom of the model board. What separates the models is managerial behavior rather than computation. Higher-scoring models reduce slow-payoff investment near the end, keep cash invested rather than idle, and open renewals well before the deadline, while token spend predicts nothing. No model learns the market's hidden prices from hundreds of rejected bids, and self-managed memory fails in two opposite modes: an archive that only grows or a plan rewritten every season. Code is available at https://github.com/Analogy-AI/fm-bench.
Chinese Translation
语言模型代理现在能够可靠地执行有限任务。然而，它们是否能够在长期内维持有效的决策能力，尤其是在行动具有累积后果且环境对其选择做出反应的情况下，仍然未得到充分测量。FM-Bench（足球管理基准）对此进行了评估。一名大型语言模型（LLM）代理通过26个工具和大约340到400个决策节点，管理一个足球俱乐部长达20个游戏年。它在与每个对手相同的预算下选拔队伍，交易球员，谈判合同，投资设施和青训，设置阵容，并对一个可以解雇它的董事会做出回应，同时一个确定性引擎将每年的结果累积成一个最终得分，而没有LLM评审或人类评分者。单人轨道将15个前沿模型与一个冻结的脚本世界进行对抗，而竞技场则将相同的模型加上一个脚本锚点放置在一个共享的20年世界中；据我们所知，这是首次在此规模下进行的正面评估。我们测量了得分背后的六种行为能力。在三个种子下，所有15个模型在每个时间段内均完成任务，而盲目的脚本基线在大多数时间段内则失效。claude-fable-5在平均得分和竞技场中名列单人排行榜的首位，尽管该头衔在十个模型之间轮换。无论是规模、价格还是供应商都无法预测顺序；顺序仅在时间段的后期确定，而表现最好的首轮人类仅位于模型排行榜的底部。模型之间的区别在于管理行为而非计算能力。得分较高的模型在接近结束时减少慢回报投资，保持现金投资而非闲置，并在截止日期之前提前开启续约，而代币支出则无法预测任何结果。没有模型能够从数百个被拒绝的投标中学习市场的隐性价格，自我管理的记忆在两种相反的模式中失败：一个仅不断增长的档案或一个每个赛季重写的计划。代码可在 https://github.com/Analogy-AI/fm-bench 获取。
cs.AI / 35 / 2608.18504
UMER: Unifying Embedding and Ranking via Pair-Aware Discriminative Reasoning for Universal Multimodal Retrieval
UMER：通过对偶感知判别推理统一嵌入和排序以实现通用多模态检索
Abstract
Universal multimodal retrieval aims to support diverse instruction-aware retrieval tasks, demanding both efficient corpus-scale matching and fine-grained semantic reasoning. Recent MLLM-based embedding methods typically derive representations from hidden states, while Chain-of-Thought (CoT) reasoning is emerging as a promising strategy for embedding enhancement by encoding intermediate semantic evidence into the representation space. However, existing CoT methods typically use item-wise reasoning over queries and candidates in isolation, providing no explicit evidence to distinguish a positive from a semantically confusable hard negative. Moreover, contrastive embeddings capture global similarity but struggle with meta-tasks requiring answer verification, category judgment or fine-grained reasoning. In this paper, we propose UMER, a Unified Multimodal Embedding and Ranking framework for universal multimodal retrieval. UMER replaces item-wise reflection with Pair-Aware Discriminative Reasoning, which compares query--candidate pairs to identify instruction-relevant matching and discrepancy evidence. UMER jointly learns contrastive embeddings for efficient global matching and discriminative ranking for explicit pairwise relevance judgment within a single MLLM. A complementary mutual distillation strategy further transfers reliable pairwise preferences between the embedding and ranking functions. On the MMEB-V2 benchmark, UMER achieves state-of-the-art performance under comparable experimental settings while supporting budget-adjustable inference.
Chinese Translation
通用多模态检索旨在支持多样化的基于指令的检索任务，要求在高效的语料库规模匹配和细粒度语义推理之间取得平衡。最近基于MLLM的嵌入方法通常从隐藏状态中推导表示，而链式思维（Chain-of-Thought, CoT）推理作为一种有前景的策略，通过将中间语义证据编码到表示空间中来增强嵌入。然而，现有的CoT方法通常在查询和候选项上进行逐项推理，未能提供明确的证据来区分正例和语义上混淆的困难负例。此外，对比嵌入捕捉全局相似性，但在需要答案验证、类别判断或细粒度推理的元任务中表现不佳。本文提出了UMER，一个用于通用多模态检索的统一多模态嵌入和排序框架。UMER用对偶感知判别推理替代了逐项反思，比较查询-候选对以识别与指令相关的匹配和差异证据。UMER在单个MLLM中联合学习对比嵌入以实现高效的全局匹配和判别排序，以进行明确的成对相关性判断。一个互补的相互蒸馏策略进一步在嵌入和排序函数之间转移可靠的成对偏好。在MMEB-V2基准上，UMER在可比的实验设置下实现了最先进的性能，同时支持可调预算推理。
cs.AI / 36 / 2608.18521
Which Negatives Matter? Ask Your Text Encoder: Adaptive Similarity Margins for Dense-Caption Retrieval
哪些负样本重要？询问你的文本编码器：用于密集字幕检索的自适应相似性边界
Abstract
Dense-caption retrieval has recently been improved by introducing segmentation, edge maps, LLM-filtered captions, and cross-modal modules into contrastive fine-tuning. However, these methods largely inherit the same InfoNCE objective, whose optimization can prematurely saturate under a strong pre-trained initialization: on dense captions, the loss falls below 10^{-3} on 80% of batches within the first epoch, while its gradient reaches exact zero in fp32 in 47% of measurements. We find that this behavior is closely related to the large number of near-duplicate captions in dense-caption benchmarks, where a few highly similar negatives remain unresolved after the easy majority has already been separated. As a remedy, we introduce HN-CLIP, which uses the text encoder's own text-text geometry to construct per-negative adaptive similarity margins. Specifically, a detached caption-similarity matrix is added to the negative logits, assigning larger margins to more similar captions without mining, synthesizing, or resampling negatives. The resulting objective requires only one caption-similarity matrix and a masked logit addition during training, with no auxiliary data, additional parameters, offline preprocessing, or inference-time overhead. Extensive experiments on four dense-caption retrieval benchmarks show that HN-CLIP improves over the strongest competitors by +2.4--+4.3 R@1 while training 2.4x faster than GOAL and 5.4x faster than StructXLIP. Moreover, the proposed objective improves all six tested fine-tuning frameworks on the in-domain benchmarks and reaches the strongest full-data baseline with only 20% of the training data.
Chinese Translation
密集字幕检索最近通过引入分割、边缘图、LLM过滤的字幕和跨模态模块来改善对比微调。然而，这些方法在很大程度上继承了相同的InfoNCE目标，其优化在强大的预训练初始化下可能会过早饱和：在密集字幕上，损失在第一个周期内的80%批次中降至10^{-3}以下，而在47%的测量中其梯度在fp32下达到精确零。我们发现这种行为与密集字幕基准中大量近重复字幕密切相关，其中在大多数容易分离的负样本之后，少数高度相似的负样本仍然未得到解决。为此，我们引入了HN-CLIP，它利用文本编码器自身的文本-文本几何结构来构建每个负样本的自适应相似性边界。具体而言，附加一个独立的字幕相似性矩阵到负对数中，为更相似的字幕分配更大的边界，而无需挖掘、合成或重新采样负样本。所得到的目标在训练过程中仅需要一个字幕相似性矩阵和一个掩蔽对数的加法，不需要辅助数据、额外参数、离线预处理或推理时的开销。在四个密集字幕检索基准上的广泛实验表明，HN-CLIP在最强竞争者的基础上提高了+2.4到+4.3的R@1，同时训练速度比GOAL快2.4倍，比StructXLIP快5.4倍。此外，所提出的目标在领域内基准上改善了所有六个测试的微调框架，并且仅使用20%的训练数据达到了最强的全数据基线。
cs.AI / 37 / 2608.18531
Pairwise Ranking Outperforms Single-Action RL for Offline Explanation Selection: A Practical Lesson
成对排名优于单动作强化学习用于离线解释选择：一项实践经验
Abstract
Industrial explainable-recommendation systems built on LLMs incur a substantial serving cost: each request triggers an LLM generation, with latency in the hundreds of milliseconds and cost that scales linearly with traffic. We separate generation from selection: explanations are produced ahead of time as a frozen candidate pool (six prompt styles, two commodity LLMs), and a small CPU-resident selector picks one at request time. The stack needs no GPU and returns in under 100 ms. Our primary benchmark is a 2,958-pair XRec Google Local subset, evaluating six offline-pool selectors (LambdaRank, PPO, GRPO, DPO, teacher-student distillation) and three KG-path selectors (random walks, edge-disjoint enumeration, MMR-reranked paths). A 300-pair MovieLens-1M split with Claude-Sonnet-4.5 references serves as an internal cross-dataset check, since no public benchmark exists for this setting. All variants use the same BERTScore-F1 protocol as XRec and G-Refer, averaged across five seeds. LambdaRank reaches F1 = 0.500 on Google Local, exceeding both G-Refer and XRec, and F1 = 0.329 on the MovieLens-1M check. With seed variance below 0.003 F1, the ordering is reliable: pairwise learning-to-rank outperforms single-action RL (PPO, GRPO, DPO), which use only one labelled candidate per rollout, leaving K-1 labels unused. The KG-path family targets a different objective: all three variants reach USR = 1.000 on Google Local and 0.997-1.000 on MovieLens-1M, since per-request path grounding yields a unique output per query, avoiding template-collapse failures affecting cached-LLM outputs. A generator-pool study comparing Claude 3 Haiku and Claude Haiku 4.5 shows small F1 shifts (0.001-0.006) while preserving selector ranking: selector and generator can be evaluated independently, though absolute F1 depends on the generator. End-to-end build cost is near $15 on commodity hardware.
Chinese Translation
基于大型语言模型（LLMs）的工业可解释推荐系统产生了可观的服务成本：每个请求都会触发LLM生成，延迟在几百毫秒内，并且成本与流量呈线性增长。我们将生成与选择分开：解释提前作为一个冻结的候选池生成（六种提示风格，两种商品LLM），并且一个小型CPU驻留的选择器在请求时选择一个。该堆栈无需GPU，返回时间少于100毫秒。我们的主要基准是一个包含2,958对的XRec Google Local子集，评估六种离线池选择器（LambdaRank、PPO、GRPO、DPO、教师-学生蒸馏）和三种知识图谱路径选择器（随机游走、边不相交枚举、MMR重新排序路径）。一个包含300对的MovieLens-1M拆分与Claude-Sonnet-4.5的参考作为内部跨数据集检查，因为在此设置下没有公共基准。所有变体均使用与XRec和G-Refer相同的BERTScore-F1协议，平均跨五个种子。LambdaRank在Google Local上达到F1 = 0.500，超过G-Refer和XRec，并在MovieLens-1M检查中达到F1 = 0.329。种子方差低于0.003 F1，排序是可靠的：成对学习排名优于单动作强化学习（PPO、GRPO、DPO），后者在每次回合中仅使用一个标记候选，留下K-1个标签未使用。知识图谱路径系列针对不同的目标：所有三种变体在Google Local上达到USR = 1.000，在MovieLens-1M上达到0.997-1.000，因为每个请求的路径基础提供了每个查询的唯一输出，避免了影响缓存LLM输出的模板崩溃故障。比较Claude 3 Haiku和Claude Haiku 4.5的生成器池研究显示F1的微小变化（0.001-0.006），同时保持选择器排名：选择器和生成器可以独立评估，尽管绝对F1依赖于生成器。端到端构建成本在商品硬件上接近15美元。
cs.AI / 38 / 2608.18534
FinRCA-Bench: Benchmarking Evidence Retrieval and Reasoning for Financial AI Systems
FinRCA-Bench：金融人工智能系统证据检索与推理的基准测试
Abstract
Large language models are increasingly used to support financial operations, but their apparent reasoning performance can depend on whether they receive the right evidence. In financial reconciliation, the evidence needed for diagnosis is distributed across invoices, purchase orders, approvals, allocations, payments, ledger entries, and bank activity, linked by transactional relationships rather than textual similarity. End-to-end accuracy can therefore conflate evidence access with reasoning quality. We introduce FinRCA-Bench, a deterministic synthetic benchmark of 2,250 accounts-payable-to-bank reconciliation cases spanning 14 operational tables, including 1,500 injected failures across 15 causal categories and 750 legitimate or hard-negative cases. Root-cause labels and record-level evidence contracts are hidden from the model, allowing retrieval to be evaluated independently of answer correctness. We compare Rules/SQL, classical machine learning, dense semantic retrieval, deterministic relational expansion, and Typed Provenance Graph Retrieval (TPGR), a typed traversal restricted to persisted transaction relationships. Rules/SQL reaches 84.97% held-out exact accuracy and classical ML reaches 95.44%. Holding the reasoning model, prompt, and generation settings fixed while changing only retrieval increases macro required-record recall from 0.83% to 77.70% and exact 16-class accuracy from 2.05% to 72.44%. Structural retrieval failures outnumber reasoning failures with sufficient retrieval by 95 to 15; 254 correct predictions occur despite incomplete retrieval, and strict returned-evidence contract accuracy is only 5.72%. On FinRCA-Bench, retrieval architecture strongly shapes observed AI-system performance, and a correct root-cause label is a weak proxy for an auditable diagnosis.
Chinese Translation
大型语言模型越来越多地被用于支持金融操作，但它们的推理性能显然依赖于是否获得了正确的证据。在金融对账中，诊断所需的证据分布在发票、采购订单、批准、分配、付款、账目条目和银行活动中，这些证据通过交易关系而非文本相似性相互关联。因此，端到端的准确性可能将证据获取与推理质量混为一谈。我们引入了FinRCA-Bench，这是一个确定性的合成基准，涵盖了2250个应付账款到银行的对账案例，涉及14个操作表，包括1500个注入的故障，分布在15个因果类别中，以及750个合法或难负样本。根本原因标签和记录级证据合同对模型是隐藏的，从而允许独立评估检索与答案正确性。我们比较了规则/SQL、经典机器学习、密集语义检索、确定性关系扩展和类型化来源图检索（Typed Provenance Graph Retrieval, TPGR），后者是一种限制于持久交易关系的类型化遍历。规则/SQL的持出精确度达到84.97%，经典机器学习达到95.44%。在保持推理模型、提示和生成设置不变的情况下，仅改变检索，宏观所需记录的召回率从0.83%提高到77.70%，16类精确度从2.05%提高到72.44%。在足够的检索下，结构检索失败的数量以95比15超过推理失败；尽管检索不完整，仍有254个正确预测，而严格的返回证据合同准确率仅为5.72%。在FinRCA-Bench上，检索架构强烈影响观察到的人工智能系统性能，而正确的根本原因标签只是可审计诊断的一个弱代理。
cs.AI / 39 / 2608.18543
Bridging Search and CRM: Productionizing AI Product Research Agents for Customer Re-Engagement
连接搜索与客户关系管理：将人工智能产品研究代理商生产化以实现客户再参与
Abstract
Modern e-commerce platforms often operate search, recommendation, personalization, and CRM systems independently, limiting opportunities for proactive customer re-engagement. This is particularly challenging for exploratory intents such as best smartphones or latest 5G phones, where users may leave the platform for external research before purchasing. We present a scalable, production-deployed framework that bridges search and CRM workflows through AI-powered Product Research Agents. The system identifies users with exploratory purchase intent and low engagement, conducts grounded multi-agent product research using behavioral signals, external knowledge, and enterprise catalog data, and delivers personalized recommendations through WhatsApp. We evaluate the framework in a 23-day production deployment involving approximately 15K WhatsApp notifications for mobile product discovery. The campaign achieved substantial CTR improvements over traditional WhatsApp recommendation campaigns, with evidence of secondary engagement through message forwarding and sharing. The deployment also generated downstream purchases and GMV impact, demonstrating the practical effectiveness of AI Product Research Agents for proactive customer re-engagement and end-to-end customer journey optimization.
Chinese Translation
现代电子商务平台通常独立运作搜索、推荐、个性化和客户关系管理（CRM）系统，这限制了主动客户再参与的机会。这对于探索性意图（如最佳智能手机或最新5G手机）尤其具有挑战性，因为用户可能在购买前离开平台进行外部研究。我们提出了一个可扩展的、已部署于生产的框架，通过人工智能驱动的产品研究代理商连接搜索和CRM工作流程。该系统识别出具有探索性购买意图和低参与度的用户，利用行为信号、外部知识和企业目录数据进行基于事实的多代理产品研究，并通过WhatsApp提供个性化推荐。我们在为期23天的生产部署中评估了该框架，涉及约15,000条WhatsApp通知用于移动产品发现。该活动在传统WhatsApp推荐活动中实现了显著的点击率（CTR）提升，并通过消息转发和分享的证据显示了二次参与。该部署还产生了下游购买和总商品价值（GMV）影响，证明了人工智能产品研究代理商在主动客户再参与和端到端客户旅程优化中的实际有效性。
cs.AI / 40 / 2608.18580
FACET: Preserving Source Intent and Executable State in Terminal Task Synthesis
FACET：在终端任务合成中保留源意图和可执行状态
Abstract
Training terminal agents requires scalable executable supervision, yet synthesizing high-quality terminal tasks remains challenging. Each task couples an instruction, an initialized environment, a reference solution, and an executable verifier; if these artifacts are generated from inconsistent assumptions, the resulting task may be unsolvable or incorrectly evaluated. Meanwhile, multi-stage synthesis can discard the goals, dependencies, state transitions, and procedural constraints encoded in the original sources. We present FACET (Fine-grained Agentic Construction of Executable Tasks), a framework that addresses both information preservation and cross-artifact consistency. FACET reconstructs related agent skills into coherent, information-rich scenarios, then realizes and repairs the execution environment before generating the final task artifacts. The resulting container state serves as shared grounding for the instruction, solution, and verifier, while execution-based validation and targeted repair correct artifact-specific failures without unnecessarily regenerating valid components. FACET produces complex terminal tasks with dense executable checks, and successful trajectories collected from these tasks provide effective, data-efficient supervision. Fine-tuning models across multiple scales consistently improves performance on Terminal-Bench 2.1, while analyses of alternative generation schemes support the importance of environment-grounded construction for task validity and solution-verifier alignment. These results establish source-intent preservation and shared executable-state grounding as key principles for scalable terminal-task synthesis.
Chinese Translation
训练终端代理需要可扩展的可执行监督，但合成高质量的终端任务仍然具有挑战性。每个任务都结合了指令、初始化环境、参考解决方案和可执行验证器；如果这些工件是基于不一致的假设生成的，最终的任务可能是不可解的或被错误评估。同时，多阶段合成可能会丢弃原始来源中编码的目标、依赖关系、状态转换和过程约束。我们提出了FACET（可执行任务的细粒度代理构建），这是一个解决信息保留和跨工件一致性的问题的框架。FACET将相关的代理技能重构为连贯且信息丰富的场景，然后在生成最终任务工件之前实现并修复执行环境。最终的容器状态作为指令、解决方案和验证器的共享基础，而基于执行的验证和针对性的修复则在不必要地重新生成有效组件的情况下纠正特定工件的失败。FACET生成具有密集可执行检查的复杂终端任务，并且从这些任务中收集的成功轨迹提供了有效且数据高效的监督。在多个尺度上微调模型一致地提高了在Terminal-Bench 2.1上的性能，而对替代生成方案的分析支持了基于环境的构建对于任务有效性和解决方案-验证器一致性的重要性。这些结果确立了源意图保留和共享可执行状态基础作为可扩展终端任务合成的关键原则。
cs.AI / 41 / 2608.18591
Can a Lightweight Multimodal Model Estimate LLM Reasoning Performance? A Study for Compute-Optimal Document Inference
轻量级多模态模型能否估计大型语言模型的推理性能？一项关于计算最优文档推理的研究
Abstract
Uniformly allocating inference reasoning budgets to LLMs is expensive and prone to over-thinking penalties; especially in document tasks where visual layouts drive complexity. To address this, we introduce BudgetDoc, the first multimodal benchmark providing explicit supervision for model-budget-performance trade-offs across three document tasks. Using BudgetDoc, we train DRB (Document-Reasoning Balancer), an approx. 1B-parameter pre-flight estimator (SigLIP-2 + Qwen3-0.6B) that predicts ordinal model performance across budget levels, achieving a 0.753 weighted F1. When dynamically allocating reasoning budgets across five frontier models and three datasets, DRB matches or improves F1 scores compared to always-maximum-budget baselines in 9 of 15 configurations while drastically reducing cost. Finally, preliminary evaluations demonstrate DRB's potential to generalize to cross-model selection.
Chinese Translation
均匀分配推理预算给大型语言模型（LLMs）成本高昂且容易导致过度思考的惩罚；尤其是在视觉布局驱动复杂性的文档任务中。为了解决这一问题，我们引入了BudgetDoc，这是第一个提供明确监督的多模态基准，旨在在三个文档任务中实现模型、预算和性能之间的权衡。利用BudgetDoc，我们训练了DRB（文档推理平衡器），这是一个约10亿参数的预飞行估计器（SigLIP-2 + Qwen3-0.6B），可以预测不同预算水平下的模型性能顺序，达到了0.753的加权F1分数。在五个前沿模型和三个数据集之间动态分配推理预算时，DRB在15种配置中的9种情况下与始终使用最大预算的基线相比，匹配或提高了F1分数，同时大幅降低了成本。最后，初步评估表明DRB在跨模型选择中具有推广潜力。
cs.AI / 42 / 2608.18613
CTIFoundry: An Agent-Native Corpus Scaffold for Cyber Threat Intelligence
CTIFoundry：一种面向代理的网络威胁情报语料库框架
Abstract
Cyber threat intelligence (CTI) is increasingly consumed not by human analysts but by LLM agents that compose multi-step investigations at query time. The harness side of this shift has matured rapidly (planning loops, tool protocols, context management), but the corpus side has not: threat reports and vulnerability databases are still packaged for retrieval-augmented generation, as opaque chunks behind an embedding index. We argue that this substrate, not model capability, is the bottleneck on agentic CTI investigation, and present CTIFoundry, an agent-native corpus scaffold. At build time, CTIFoundry materializes the latent structure of a CTI corpus: a deterministic ontology graph over four authoritative knowledge bases (CVE, CWE, CAPEC, ATT&CK) whose official cross-references become typed, traversable edges; a span-grounded report layer whose canonical, alias-resolved cross-vendor entities index provenance-carrying chunks; and hybrid dense+lexical retrieval surfaces. At query time this structure is exposed through seven typed tools and three procedural skills mounted on a stock open-source agent harness. On the public CTIConnect benchmark, swapping only the action surface lifts the identically-harnessed agent by +0.19 to +0.28 overall F1 across a four-model, two-provider panel: a small model on CTIFoundry surpasses a flagship on the flat substrate, and the gain is not bought with search effort, since on both Claude models the scaffolded agent is more accurate at roughly half the tool calls. An ablation attributes it: typed structure carries the larger share, procedural skills convert structure into discipline, and the two compose super-additively, because skills bind only to structure that exists.
Chinese Translation
网络威胁情报（CTI）越来越多地被大型语言模型（LLM）代理所消费，这些代理在查询时进行多步骤调查。这一转变的利用方面迅速成熟（规划循环、工具协议、上下文管理），但语料库方面却未能跟上：威胁报告和漏洞数据库仍然以不透明的块形式打包在嵌入索引后，供检索增强生成使用。我们认为，这种基础结构，而非模型能力，是代理CTI调查的瓶颈，并提出CTIFoundry，一种面向代理的语料库框架。在构建时，CTIFoundry实现了CTI语料库的潜在结构：一个基于四个权威知识库（CVE、CWE、CAPEC、ATT&CK）的确定性本体图，其官方交叉引用成为类型化的可遍历边；一个基于跨度的报告层，其规范化、别名解析的跨供应商实体索引承载来源信息的块；以及混合的密集+词汇检索表面。在查询时，这一结构通过七种类型化工具和三种程序技能在标准开源代理框架上展现。在公共CTIConnect基准测试中，仅通过更换动作表面，便使得同样配置的代理在四个模型、两个提供者的面板上整体F1提升了+0.19到+0.28：在CTIFoundry上，一个小模型超越了平坦基础上的旗舰模型，而这一增益并未以搜索努力为代价，因为在两个Claude模型上，框架化的代理在大约一半的工具调用中更为准确。一项消融实验归因于此：类型化结构占据了更大份额，程序技能将结构转化为学科，而二者的组合则呈现超加性，因为技能仅绑定于已存在的结构。
cs.AI / 43 / 2608.18631
Preference Reasoning under Indeterminacy in Large Language Models
大型语言模型中的不确定性下的偏好推理
Abstract
As large language models evolve into decision-making agents, the ability to reason over preferences becomes fundamental to alignment, coordination, and collective intelligence. Yet, unlike standard benchmarks, real-world preference reasoning is inherently indeterminate: information may be incomplete, and valid solutions may not exist. We argue that indeterminacy, rather than correctness alone, is a central challenge for AI reasoning. We formalize this challenge along two axes, (i) epistemic indeterminacy, arising from incomplete, partial, or expressive preferences, and (ii) structural indeterminacy, arising from the non-existence of solutions under standard social choice concepts. Across a hierarchy of tasks, we show that state-of-the-art language models systematically fail to distinguish between determined and undetermined instances, exhibiting miscalibrated reasoning even in verification settings.
Chinese Translation
随着大型语言模型逐渐演变为决策代理，推理偏好的能力成为对齐、协调和集体智能的基本要素。然而，与标准基准不同，现实世界中的偏好推理本质上是模糊不清的：信息可能不完整，且有效的解决方案可能不存在。我们认为，不确定性而非单纯的正确性，是人工智能推理面临的核心挑战。我们沿着两个维度对这一挑战进行了形式化，(i) 认知不确定性，源于不完整、部分或表达性的偏好，以及 (ii) 结构不确定性，源于在标准社会选择概念下解决方案的不存在。在一系列任务中，我们展示了最先进的语言模型系统性地未能区分确定和不确定的实例，即使在验证设置中也表现出错误的推理校准。
cs.AI / 44 / 2608.18665
Candidate-Fate Accounting for Transparent Sensor Diagnostic Pipeline Search
透明传感器诊断管道搜索中的候选命运核算
Abstract
Industrial sensor diagnostics relies on preprocessing, representation, and classification pipelines, making automated pipeline search useful for reducing manual design cost. However, existing automated machine/deep learning (AutoML/AutoDL) reports typically retain only fitted trials, scores, and winners, omitting generated candidates that are invalid, pruned, skipped, cached, or unfitted. This omission limits reviewers' ability to check signal constraints, budget use, and unevaluated legal alternatives. To address this, we propose candidate-fate accounting, a candidate-level audit framework for diagnostic search traces. It records each observed candidate as auditable evidence: hashes merge repeated observations, legality checks flag invalid candidates, allocation rationales explain budget decisions, and a closed fate ledger assigns one terminal fate to each candidate. Experiments on three bearing-diagnostic datasets show that the framework detects invalid candidates and identifies 30--41 candidates omitted by fitted-trial-only reports, with closed fate records verifying complete candidate accounting while maintaining competitive diagnostic performance. The code is available at https://github.com/XXIE999/candidate-fate-accounting.
Chinese Translation
工业传感器诊断依赖于预处理、表示和分类管道，因此自动化管道搜索对于降低手动设计成本非常有用。然而，现有的自动化机器/深度学习（AutoML/AutoDL）报告通常仅保留拟合试验、得分和优胜者，忽略了生成的无效候选、被修剪的候选、跳过的候选、缓存的候选或未拟合的候选。这一遗漏限制了审查者检查信号约束、预算使用和未评估合法替代方案的能力。为了解决这个问题，我们提出了候选命运核算，这是一种用于诊断搜索轨迹的候选级审计框架。它将每个观察到的候选记录为可审计的证据：哈希合并重复观察，合法性检查标记无效候选，分配理由解释预算决策，封闭命运账本为每个候选分配一个终极命运。在三个轴承诊断数据集上的实验表明，该框架能够检测无效候选，并识别出30至41个在仅拟合试验报告中被遗漏的候选，封闭命运记录验证了完整的候选核算，同时保持了竞争性的诊断性能。代码可在 https://github.com/XXIE999/candidate-fate-accounting 获取。
cs.AI / 45 / 2608.18677
Sanyu Studio: A Multi-Agent System for Art-Historical Narrative Construction
三寓工作室：一种用于艺术历史叙事构建的多智能体系统
Abstract
Amid concerns that generative AI may standardize art interpretation, this paper examines whether LLM-based interaction can support plural art-historical narrative construction. We present Sanyu Studio, a multi-agent dialogue system that models 321 Sanyu oil paintings as agents with fact, interpretation, organization, and memory-filtering mechanisms. Based on a seven-day workshop with eight art-university participants, the study shows that user prompts, evidence organization, and cognitive tendencies shaped divergent yet coherent versions of digital Sanyu. The findings suggest that, under conditions of limited historical evidence, AI can amplify human agency and offer public audiences an interactive entry point into art-historical interpretation.
Chinese Translation
在对生成性人工智能可能标准化艺术解读的担忧中，本文探讨了基于大型语言模型（LLM）的互动是否能够支持多元的艺术历史叙事构建。我们提出了三寓工作室（Sanyu Studio），这是一个多智能体对话系统，将321幅三寓油画建模为具有事实、解读、组织和记忆过滤机制的智能体。基于与八位艺术大学参与者进行的为期七天的工作坊，研究表明，用户提示、证据组织和认知倾向塑造了数字三寓的多样化但一致的版本。研究结果表明，在有限历史证据的条件下，人工智能可以增强人类的能动性，并为公众提供一个互动的艺术历史解读入口。
cs.AI / 46 / 2608.18682
RTPO: Reverse-Turn Policy Optimization for Stabilizing Agentic RL Training
RTPO：反向回合策略优化以稳定代理强化学习训练
Abstract
Training multi-turn agentic workflows with reinforcement learning (RL) enables large language models to perform complex reasoning, use external tools, and conduct iterative search beyond single-turn settings. Yet multi-turn RL training remains highly unstable, often causing severe performance degradation as the number of turns increases. Through theoretical analysis, we identify three tightly coupled sources of instability: rollout-training context mismatch, weak turn-level credit assignment under sparse terminal rewards, and asynchronous policy drift when short and long trajectories are optimized under different policy versions. We show that these issues share a common structural origin in flattened trajectory optimization and address them through a unified reverse-turn formulation. We propose Reverse-Turn Policy Optimization (RTPO), which organizes multi-turn rollouts as sparse reverse trees and performs turn-level policy updates in temporal reverse order, aligning each decision with its downstream continuation. RTPO enables causally consistent turn-level credit assignment and on-policy continuation to control asynchronous drift. We provide theoretical guarantees showing that RTPO eliminates context mismatch and asynchronous drift under the proposed turn-level formulation, reduces credit bias, and converges to recursive optimality. Experiments on multi-turn agentic RL benchmarks show that RTPO improves upon trajectory- and turn-level baselines by 21.50% and 10.76%, respectively, highlighting its potential to support more stable training for tool-using agents.
Chinese Translation
使用强化学习（RL）训练多回合代理工作流使大型语言模型能够进行复杂推理、使用外部工具并在单回合设置之外进行迭代搜索。然而，多回合RL训练仍然高度不稳定，通常导致随着回合数的增加而严重的性能下降。通过理论分析，我们识别出三种紧密耦合的不稳定来源：回合训练上下文不匹配、在稀疏终端奖励下的弱回合级信用分配，以及在不同策略版本下优化短轨迹和长轨迹时的异步策略漂移。我们表明，这些问题在扁平化轨迹优化中具有共同的结构起源，并通过统一的反向回合公式加以解决。我们提出了反向回合策略优化（RTPO），它将多回合回放组织为稀疏反向树，并按时间反向顺序执行回合级策略更新，使每个决策与其下游延续对齐。RTPO 实现了因果一致的回合级信用分配和在政策上的延续，以控制异步漂移。我们提供理论保证，表明RTPO在提出的回合级公式下消除了上下文不匹配和异步漂移，减少了信用偏差，并收敛到递归最优性。在多回合代理RL基准上的实验表明，RTPO分别比轨迹和回合级基线提高了21.50%和10.76%，突显了其支持工具使用代理更稳定训练的潜力。
cs.AI / 47 / 2608.18719
Competence, Not Accuracy: A Diagnostic for Reference-Free Judge Gates in Skill Optimization
能力，而非准确性：无参考评判门在技能优化中的诊断
Abstract
Text-space skill optimization adapts a frozen agent by evolving a natural-language skill document, accepting each candidate through a validation gate. Existing gates rely on verifiable rewards, confining these methods to tasks with an automatic verifier. Replacing the verifier with an LLM-judge gate would lift that restriction, but whether such a gate carries usable signal is untested. We ask a prior question: can we tell, before placing a judge in the loop, whether its scores separate correct from incorrect answers at all? We formalize a reference-free judge as a latent solver -- its verdict rests on agreement with whatever it would itself conclude, so its capacity to evaluate is bounded by its capacity to solve. The model yields a closed-form bound on discriminability (ROC-AUC) in the judge's competence $c$ and answer-space size $k$, a necessary condition $c > 1/k$, and the result that the marginal AUC is confounded by item difficulty while a within-question estimator is not. A non-intervening probe records judge scores on genuine optimization runs without altering any decision. We find discriminability at chance where competence sits near the floor and usable above it; that a judge's benchmark accuracy overstates the competence that matters; and, in a closed-loop study, that the screen predicts which kind of gating error occurs. The result is a cheap pre-deployment diagnostic for judge gates.
Chinese Translation
文本空间技能优化通过演化自然语言技能文档来适应一个冻结的代理，接受每个候选者通过验证门。现有的验证门依赖于可验证的奖励，使得这些方法局限于具有自动验证器的任务。用LLM-judge门替代验证器将解除这一限制，但这种门是否具有可用信号尚未经过检验。我们提出一个先前的问题：在将评判者纳入循环之前，我们能否判断其分数是否能够区分正确与错误的答案？我们将无参考评判者形式化为一个潜在求解器——其裁决基于与其自身得出的结论的一致性，因此其评估能力受限于其求解能力。该模型在评判者的能力$c$和答案空间大小$k$上提供了可区分性（ROC-AUC）的封闭形式界限，必要条件为$c > 1/k$，并且结果表明边际AUC受项目难度的混淆，而问题内估计则不受影响。一个非干预探测器记录评判者在真实优化运行中的得分，而不改变任何决策。我们发现，当能力接近下限时可区分性处于偶然水平，而在其上方则可用；评判者的基准准确性夸大了重要的能力；并且在一个闭环研究中，筛选预测了发生哪种类型的门控错误。该结果为评判门提供了一种廉价的预部署诊断。
cs.AI / 48 / 2608.18740
A Multi-Agent Platform for Automated Enterprise Analytics and Insight Generation
用于自动化企业分析和洞察生成的多智能体平台
Abstract
This paper proposes a multi-agent framework built on CrewAI [1] for conversational business intelligence. Five specialized AI agents operate in a sequential pipeline to process natural language queries, retrieve and analyze data, generate visualizations via the Model Context Protocol (MCP) [2], and deliver actionable insights. The platform features a defense-in-depth security architecture for multi-tenant data isolation and a query parameterization mechanism for transforming conversational insights into reusable dashboard components. Evaluation across 300 end-to-end test cases spanning synthetic and production enterprise datasets demonstrates 95.3% functional accuracy, a mean response latency of 24 seconds, and a response quality score of 4.52/5.0 as assessed by an LLM-as-a-Judge framework, with a 93.0% hallucination-free rate, representing a 22.6 percentage point accuracy improvement and 20.2% quality gain over a single-agent baseline. Cross-model evaluation across four LLM backends and human expert validation confirm architectural generalizability and evaluator reliability. An ablation study confirms that the Data Analysis and Report Aggregation agents are the primary drivers of output quality.
Chinese Translation
本文提出了一种基于CrewAI的多智能体框架，用于对话式商业智能。五个专业化的人工智能代理在一个顺序管道中操作，以处理自然语言查询、检索和分析数据、通过模型上下文协议（Model Context Protocol, MCP）生成可视化，并提供可操作的洞察。该平台具有深度防御的安全架构，以实现多租户数据隔离，并采用查询参数化机制将对话式洞察转化为可重用的仪表板组件。在对300个端到端测试案例的评估中，涵盖了合成和生产企业数据集，显示出95.3%的功能准确率、平均响应延迟为24秒，以及通过LLM-as-a-Judge框架评估的响应质量评分为4.52/5.0，且无幻觉率为93.0%，相比单代理基线提高了22.6个百分点的准确率和20.2%的质量提升。跨模型评估涵盖四个LLM后端和人类专家验证确认了架构的通用性和评估者的可靠性。消融研究确认数据分析和报告聚合代理是输出质量的主要驱动因素。
cs.AI / 49 / 2608.18744
Metrics That Write Themselves: Evolving an Evaluator from Its Own Blind Spots
自我生成的度量：从自身盲点演化评估器
Abstract
Agents improve quickly against a reliable automatic metric and stall without one, and the applications that need them most, report generation among them, are the ones nobody knows how to score. Can the metric write itself? Saying what makes an answer good is hard; pointing at something wrong with one is easier, so the metric we evolve is a pool of small Python operators that each flag a candidate for one named defect, or abstain, and vote. Asking a model for operators directly does not work: 183 candidates realise only 96 distinct behaviours, from one narrow region of an enormous space. EvalCEGAR instead borrows counterexample-guided abstraction refinement from program verification. It reads the pool as an abstraction and searches for a collision, two answers the operators score identically, one correct and one not. That pair, not a prompt, is the authoring request, and when a collision defeats every attempt the loop widens what an operator may read rather than resampling. On MBPP+ and HumanEval+, a sandbox whose hidden unit tests give exact ground truth, the loop writes a 55-line operator that closes 15.4% of the gap between flagging nothing and a perfect filter on 428 unseen tasks (+0.0065, p=0.0010) at a quarter of our best hand-written operator's flags. On the benchmark it never saw it matches that operator's effect exactly on a third of the flags. Six of eight runs admit such an operator and all six help out of sample; our 15 hand-written operators applied together as one filter lose accuracy. An LLM judge on the same information ties that delta on a nearly disjoint set of candidates, and charges a model call per candidate forever where the operator charges none.
Chinese Translation
智能体在可靠的自动度量下迅速提升，而在没有度量的情况下则停滞不前，而最需要这些度量的应用（例如报告生成）恰恰是那些无人知晓如何评分的应用。度量能否自我生成？定义一个答案优秀的标准很困难；指出一个答案存在的问题则相对容易，因此我们演化的度量是一个小型 Python 操作符的池，每个操作符标记一个候选答案的特定缺陷，或选择不标记，并进行投票。直接向模型请求操作符并不奏效：183 个候选答案仅实现了 96 种不同的行为，这些行为来自于一个巨大空间的狭窄区域。EvalCEGAR 则借鉴了程序验证中的反例引导抽象细化技术。它将操作符池视为一种抽象，并搜索碰撞，即两个答案被操作符同样评分，一个是正确的，另一个则不是。这个配对，而不是提示，是创作请求，当碰撞击败每一次尝试时，循环扩大操作符可以读取的内容，而不是重新抽样。在 MBPP+ 和 HumanEval+ 上，这是一种沙盒，其隐藏单元测试提供了精确的真实值，循环生成了一个 55 行的操作符，缩小了在 428 个未见任务中标记无和完美过滤器之间的 15.4% 的差距（+0.0065，p=0.0010），而其标记数量仅为我们最佳手写操作符的四分之一。在它从未见过的基准上，它在三分之一的标记上完全匹配该操作符的效果。八次运行中有六次接受了这样的操作符，并且所有六次都在样本外提供帮助；我们 15 个手写操作符作为一个过滤器共同应用时则失去了准确性。基于相同信息的 LLM 判断在几乎不重叠的候选集上达成了相同的增量，并对每个候选者收取模型调用费用，而操作符则不收取任何费用。
cs.AI / 50 / 2608.18820
Pairwise Logical Selection of Enthymeme Completions under Semantic-Link Uncertainty
在语义链接不确定性下的对偶逻辑选择推理完成
Abstract
Arguments often omit premises or claims, forming enthymemes. We study pairwise logical selection between two candidates for the omitted component. Existing natural language methods can identify or generate candidates but often do not expose how the selected candidate completes the inference, while logic-based approaches usually assume that the required formulae and background knowledge are available. We extend a prior neuro-symbolic pipeline from missing-premise to missing-claim selection and replace binary entailment outcomes with logical-resistance scores. Top-Link uses weighted Partial MaxSAT under a single configuration of highest-confidence semantic links. We then introduce Possible-World Atom-Link Formalization (PWAL), which keeps translated formulae fixed and marginalizes logical resistance over alternative cross-formula semantic-link configurations. We evaluate PWAL on five tasks: ARCT and a CDED-derived task for missing-premise selection, iDebate- and AAE2-derived tasks for missing-claim selection, and alphaNLI for abductive hypothesis selection. Relative to Top-Link, PWAL raises strict accuracy by 2.95-30.86 percentage points and reduces tie rates by 4.57-58.00 percentage points on all five tasks. When ties receive half credit, accuracy still increases by 0.45-6.04 percentage points. PWAL also records the translated formulae, sampled link configurations, and resistance components for every comparison, providing a transparent trace of each score.
Chinese Translation
论证通常省略前提或主张，形成隐含论证（enthymemes）。我们研究在两个候选者之间进行的对偶逻辑选择，以填补省略的成分。现有的自然语言方法能够识别或生成候选者，但往往未能揭示所选候选者如何完成推理，而基于逻辑的方法通常假设所需的公式和背景知识是可用的。我们将先前的神经符号管道从缺失前提扩展到缺失主张的选择，并用逻辑抵抗分数替代二元蕴含结果。Top-Link在单一配置的最高置信度语义链接下使用加权部分最大可满足性（Partial MaxSAT）。随后，我们引入可能世界原子链接形式化（Possible-World Atom-Link Formalization，PWAL），该方法保持翻译公式不变，并在替代的跨公式语义链接配置上边际化逻辑抵抗。我们在五个任务上评估PWAL：用于缺失前提选择的ARCT和CDED衍生任务，用于缺失主张选择的iDebate和AAE2衍生任务，以及用于溯因假设选择的alphaNLI。相较于Top-Link，PWAL在所有五个任务上将严格准确率提高了2.95-30.86个百分点，并将平局率降低了4.57-58.00个百分点。当平局获得半分时，准确率仍然增加了0.45-6.04个百分点。PWAL还记录了每次比较的翻译公式、采样链接配置和抵抗成分，为每个分数提供了透明的追踪。
cs.AI / 51 / 2608.18836
Verifiable abstention makes AI leak diagnosis accountable in water distribution networks
可验证的弃权使人工智能在水分配网络中的泄漏诊断变得可追溯
Abstract
Utilities lose a substantial share of treated water to leakage, yet rarely trust artificial-intelligence localizers to dispatch crews: guessing everywhere cannot justify excavation. The gap is accountability, not accuracy: no method proves when it should not act. Here we recast leak localization as decision-making under verifiable abstention. A physics-grounded executor agent falsifies hypotheses (leak, demand, sensor, valve) against a digital twin; an independent supervisor agent, with a large-language-model (LLM) auditor, checks evidence against a code-verifiable contract, then certifies a dispatch, requests evidence or abstains. Under field-grade noise, a 32% forced baseline becomes 96% decision precision on acted events. On an independently generated benchmark it acts on only 4 of 33 leaks, all correct. A 194-event register of audited real leak locations with twin-simulated pressures and flows yields five excavation dispatches, three correct, and 44% survey recovery at full district precision. Accountable abstention offers a defensible route to autonomous water-infrastructure operation.
Chinese Translation
公用事业在泄漏中损失了大量处理水，但很少信任人工智能定位器派遣工作人员：到处猜测无法证明挖掘的合理性。问题在于责任，而非准确性：没有方法能证明何时不应采取行动。在这里，我们将泄漏定位重新构建为可验证弃权下的决策过程。一个基于物理的执行代理针对数字双胞胎对假设（泄漏、需求、传感器、阀门）进行验证；一个独立的监督代理，结合大型语言模型（LLM）审计员，检查证据与代码可验证合同的一致性，然后认证派遣、请求证据或选择弃权。在现场噪声下，32%的强制基线在实际事件上变为96%的决策精度。在一个独立生成的基准测试中，它仅对33个泄漏中的4个采取行动，且全部正确。一个包含194个审核真实泄漏位置的事件登记，结合双胞胎模拟的压力和流量，产生了五次挖掘派遣，其中三次正确，且在全区精度下回收率达到44%。负责任的弃权为自主水基础设施运营提供了一条可辩护的途径。
cs.AI / 52 / 2608.18846
ORBITER: Conflict-Aware Decision-Making for Agentic Last-Mile Delivery
ORBITER：面向冲突感知的代理最后一公里配送决策
Abstract
Last-mile delivery aims to handle dynamically arriving orders with couriers while modeling complex spatial and temporal correlations. Recent learning-based methods model spatiotemporal dependencies among orders to predict courier service sequences, but leave next-order decision making unexplained. Describing the current delivery state in language allows LLMs to reason explicitly about the spatial, temporal, and behavioral cues behind an individual decision. As direct predictors, however, LLMs remain sensitive to task presentation and often produce unreliable decisions. To address these challenges, we introduce ORBITER, an agentic Order Arbiter for next-order decision-making in last-mile delivery. ORBITER models courier service through decision points, each containing the courier's spatiotemporal state and visible orders and exposing local trade-offs for modeling and verification. Fixed proposers rank the candidates, and a structured report identifies where their rankings disagree. The LLM uses task-specific tools to gather evidence on the leading alternatives, while an independent critic checks the resulting decision against that evidence. We conduct extensive evaluations on data in four cities, where ORBITER outperforms existing state-of-the-art baselines by up to 9.2% on average showing its effectiveness.
Chinese Translation
最后一公里配送旨在处理动态到达的订单与快递员的匹配，同时建模复杂的时空相关性。近期基于学习的方法通过建模订单之间的时空依赖关系来预测快递服务序列，但对下一订单的决策过程却未作解释。用语言描述当前的配送状态使得大型语言模型（LLMs）能够明确推理出个体决策背后的时空和行为线索。然而，作为直接预测器，LLMs 对任务呈现非常敏感，常常产生不可靠的决策。为了解决这些挑战，我们提出了 ORBITER，一个用于最后一公里配送中下一订单决策的代理订单仲裁者。ORBITER 通过决策点建模快递服务，每个决策点包含快递员的时空状态和可见订单，并揭示局部权衡以便建模和验证。固定提议者对候选者进行排名，结构化报告识别其排名不一致的地方。LLM 使用特定任务的工具收集关于主要备选方案的证据，而独立评审者则根据这些证据检查最终决策。我们在四个城市的数据上进行了广泛评估，结果表明 ORBITER 在性能上平均超过现有的最先进基线达 9.2%，显示出其有效性。
cs.AI / 53 / 2608.18852
SkillGate: Training In-Policy Skill Selection in Long-Horizon Agents
SkillGate：在长时间跨度代理中训练策略内技能选择
Abstract
Agent frameworks increasingly package procedural knowledge as skills: instruction files an agent reads on demand, while public libraries now hold thousands of them. Which skill to read has thus become a decision the policy itself makes in the middle of an episode, yet no existing signal trains it. We show that the default remedy, outcome-rewarded RL over the candidate slate, cannot teach it, for a structural reason we identify and name selector credit starvation: under a broadcast, sequence-level advantage, the few tokens that name the chosen skill carry a vanishing share of the loss, and the credit they inherit is increasingly wrong-signed as trajectories lengthen. A correct choice is punished whenever the execution after it fails, even though the choice itself is among the most valuable decisions in the trajectory. Auditing a completed run's own training artifacts confirms all three properties, each worsening monotonically with horizon. SkillGate removes the failure by construction: it partitions the token support into two disjoint credit channels, outcome credit reaching only execution tokens, and a separate action-local advantage reaching exactly the skill-naming tokens, positive only when a trajectory's single read is the correct one. On five agentic benchmarks under a 16-candidate slate, SkillGate lifts a 9B policy from 40.8% to 53.2% trial success, well ahead of the identical budget spent on outcome reward alone, while cutting exposure to misleading candidates by two thirds and reading fewer skills.
Chinese Translation
代理框架越来越多地将程序性知识打包为技能：代理按需读取的指令文件，而公共库现在拥有数千个这样的技能。因此，选择读取哪个技能已成为策略在一个回合中自行做出的决策，但目前没有现有信号来训练它。我们展示了默认的补救措施，即对候选技能列表进行结果奖励的强化学习，无法进行有效的训练，原因在于我们识别并命名的选择者信用匮乏：在广播的序列级优势下，命名所选技能的少量标记承载的损失份额逐渐消失，并且随着轨迹的延长，它们继承的信用越来越多地出现错误标记。每当执行失败时，正确的选择都会受到惩罚，即使该选择本身是轨迹中最有价值的决策之一。对已完成运行的训练产物进行审计确认了这三种特性，每种特性都随着时间跨度的增加而单调恶化。SkillGate通过构造消除了这种失败：它将标记支持划分为两个不相交的信用通道，结果信用仅到达执行标记，而一个单独的动作局部优势则准确到达命名技能的标记，仅在轨迹的单次读取是正确时为正。在五个代理基准测试中，SkillGate在16个候选技能的情况下，将一个90亿参数的策略的试验成功率从40.8%提升至53.2%，远超单独花费在结果奖励上的相同预算，同时将误导性候选的暴露减少了三分之二，并减少了技能的读取数量。
cs.AI / 54 / 2608.18878
DentAgent: Evidence-Centric Multi-Agent Coordination for Multimodal Dental Reasoning
DentAgent：基于证据的多智能体协调用于多模态牙科推理
Abstract
Oral diseases affect billions of people worldwide, underscoring a pressing need for accurate and reliable dental assessment that integrates heterogeneous evidence from domain knowledge, radiographs, intraoral photographs, and 3D dental data. Most existing dental AI systems remain modality- or task-specific. Although recent vision-language models support flexible dental question answering, directly generated response leaves evidence implicit and untraceable. To address these limitations, we introduce DentAgent, an evidence-centric multi-agent framework, in which the Orchestrator coordinate five specialized agents spanning various modalities. Each specialist utilizes domain tools to convert observations into structured evidence records. The Evidence Blackboard manages these records as a shared evidence state, tracking coverage, gaps, and conflicts before response generation. This standardized evidence representation integrates isolated dental capabilities into a unified agentic workflow. Across four benchmarks, DentAgent demonstrates leading performance, even surpassing the senior specialists by 17.3 percentage points on multi-label diagnosis, which supports its value for broadly applicable and traceable multimodal dental reasoning, and highlights its potential as a technical foundation for population oral health assessment and management.
Chinese Translation
口腔疾病影响全球数十亿人，凸显了对准确可靠的牙科评估的迫切需求，该评估需整合来自领域知识、放射影像、口内照片和三维牙科数据的异构证据。现有的大多数牙科人工智能系统仍然是特定于模态或任务的。尽管最近的视觉-语言模型支持灵活的牙科问答，但直接生成的响应使证据隐含且不可追溯。为了解决这些局限性，我们提出了DentAgent，一个基于证据的多智能体框架，其中协调者（Orchestrator）协调五个跨越不同模态的专业智能体。每个专家利用领域工具将观察结果转化为结构化的证据记录。证据黑板（Evidence Blackboard）管理这些记录作为共享的证据状态，在生成响应之前跟踪覆盖范围、空白和冲突。这种标准化的证据表示将孤立的牙科能力整合为统一的智能工作流程。在四个基准测试中，DentAgent展现出领先的性能，甚至在多标签诊断中超过了资深专家17.3个百分点，这支持了其在广泛适用和可追溯的多模态牙科推理中的价值，并突显了其作为人口口腔健康评估和管理技术基础的潜力。
cs.AI / 55 / 2608.18884
Training-Free Inference-Time Self-Reflection and Cost-Bounded Early Stopping for Large Language Models
无训练推理时自我反思与成本限制的早停机制用于大型语言模型
Abstract
Reinforcement-learning training of reasoning LLMs (e.g., GRPO) is expensive and requires a controllable environment, committing every contribution to a full training pipeline. We present EvoResearcher, a training-free, inference-time protocol that adds cost-bounded self-reflection to a single frozen LLM backbone. The protocol iterates generate -> self-critique -> revise until a maximum depth D is reached or the critique returns the CONFIRMED sentinel, an implicit early stop that lets the backbone self-verify its answer under a strict compute budget. Four self-reflective meta-reward components (correctness, efficiency, reflection depth, tool-call diversity) act as design principles instantiated as prompt-level mechanisms, so their benefits accrue with zero gradient updates. We validate the protocol on Big-Bench Hard (100 questions) and establish cross-domain behavior on GSM8K (500) and MATH (500) on the same frozen backbone, with cross-model replication on Qwen2.5-72B. All experiments use pure-reasoning benchmarks; the tool-call diversity component is validated in prompt-level form, and the environment-level and multi-agent extensions are design blueprints left to future work. On clean BBH the protocol does not raise accuracy beyond the 95% Wilson interval; its value is cost-bounded self-verification, with the CONFIRMED early stop terminating 82-88% of items at equal accuracy (about 2.1 generations per question).
Chinese Translation
推理大型语言模型（如 GRPO）的强化学习训练成本高昂，并且需要一个可控的环境，将每个贡献都投入到完整的训练流程中。我们提出了 EvoResearcher，这是一种无训练的推理时协议，为单个冻结的 LLM 主干添加了成本限制的自我反思。该协议通过生成 -> 自我批评 -> 修订的迭代过程，直到达到最大深度 D 或批评返回 CONFIRMED 哨兵，这是一种隐式的早停机制，使主干能够在严格的计算预算下自我验证其答案。四个自我反思的元奖励组件（正确性、效率、反思深度、工具调用多样性）作为设计原则，通过提示级机制具体化，因此它们的好处在零梯度更新的情况下累积。我们在 Big-Bench Hard（100 个问题）上验证了该协议，并在同一冻结主干上在 GSM8K（500）和 MATH（500）上建立了跨领域行为，同时在 Qwen2.5-72B 上进行了跨模型复制。所有实验均使用纯推理基准；工具调用多样性组件以提示级形式得到验证，而环境级和多智能体扩展则是留待未来工作的设计蓝图。在干净的 BBH 上，该协议未能将准确率提升至 95% 威尔逊区间以上；其价值在于成本限制的自我验证，CONFIRMED 早停在相等准确率下终止了 82-88% 的项目（每个问题约 2.1 次生成）。
cs.AI / 56 / 2608.18899
Syntactic Simplification of OWL Class Expressions
OWL类表达式的句法简化
Abstract
Class expression learning often produces complex OWL class expressions that are difficult to interpret and reason over. However, by following theoretically grounded simplification principles, this complexity can be reduced. In this paper, we propose Class Expression Simplifier (CES), a novel algorithm for the syntactic simplification of class expressions in Description Logics (DL). CES aims to preserve formal semantics while reducing representational complexity. It systematically applies rewriting rules to eliminate redundancies and identify simpler yet equivalent expressions, thereby producing more compact and human-readable representations without altering logical entailments. We evaluate the effectiveness of CES on class expressions learned from two medium-sized ontologies, demonstrating measurable improvements in reasoning efficiency and reductions in verbosity. This work contributes to the broader goal of making ontology-driven applications more accessible, maintainable, and scalable, with direct implications for knowledge graph construction, semantic search, and Web-scale reasoning. CES is implemented within the open-source Python framework OWLAPY and is publicly available.
Chinese Translation
类表达式学习通常会产生复杂的OWL类表达式，这些表达式难以理解和推理。然而，通过遵循理论基础的简化原则，可以减少这种复杂性。本文提出了一种名为类表达式简化器（Class Expression Simplifier, CES）的新算法，用于描述逻辑（Description Logics, DL）中类表达式的句法简化。CES旨在在保持形式语义的同时，降低表述复杂性。它系统地应用重写规则，以消除冗余并识别更简单但等价的表达式，从而生成更紧凑且易于人类理解的表示，而不改变逻辑蕴含。我们在两个中等规模本体上评估了CES在学习的类表达式上的有效性，证明了推理效率的可测量改善和冗长性的减少。这项工作为使基于本体的应用程序更易于访问、维护和扩展做出了贡献，并对知识图谱构建、语义搜索和Web规模推理具有直接影响。CES已在开源Python框架OWLAPY中实现，并公开提供。
cs.AI / 57 / 2608.18900
\textsc{TestifAI}: Tomography-Based Testing for Deep Learning Systems
TestifAI：基于断层成像的深度学习系统测试
Abstract
As AI systems are increasingly deployed in safety-critical application domains (e.g., autonomous driving), associated risks increase too. Deep learning models underlying modern AI systems, therefore, must undergo thorough testing to ensure their correct behaviour. A single robustness test involves thousands of inferences to empirically verify if a model's outputs remain stable under a bounded perturbation of its inputs. However, existing testing frameworks lack the means to systematically explore and summarise robustness across a combinatorial space of perturbations. We propose TestifAI, a deep learning testing framework for efficient and accurate estimation of robustness against combinations of perturbations. TestifAI enables users to specify operational conditions as structured spaces of semantic input perturbations (e.g., image blur, brightness and zoom) and discrete severity levels (e.g., low, medium and high). Users can query model robustness for any combination (e.g., "low blur, high brightness, and medium zoom"). To achieve efficiency and accuracy, TestifAI introduces partial model tomography, a novel approach to reconstructing model behaviour in a multi-perturbation space from tests that apply only a small number of perturbations (lower-order projections). To estimate robustness against at least three perturbations, TestifAI trains an auxiliary model on the results of tests involving up to two perturbations only, avoiding execution of an exponential number of tests. Our experiments on five image and language classification tasks show that TestifAI can predict higher-order (3 and 4 perturbations) test outcomes from low-order (1 and 2 perturbations) observations with an aggregate robustness estimation error of less than 7%, while reducing the number of inferences by 60-80%.
Chinese Translation
随着人工智能系统在安全关键应用领域（例如，自动驾驶）中的广泛应用，相关风险也在增加。因此，支撑现代人工智能系统的深度学习模型必须经过全面测试，以确保其正确行为。单个鲁棒性测试涉及数千次推理，以实证验证模型的输出在输入的有限扰动下是否保持稳定。然而，现有的测试框架缺乏系统性探索和总结组合扰动空间中鲁棒性的方法。我们提出了TestifAI，一个深度学习测试框架，用于高效且准确地评估对组合扰动的鲁棒性。TestifAI使用户能够将操作条件指定为结构化的语义输入扰动空间（例如，图像模糊、亮度和缩放）和离散的严重性等级（例如，低、中和高）。用户可以查询任何组合的模型鲁棒性（例如，“低模糊、高亮度和中等缩放”）。为了实现高效性和准确性，TestifAI引入了部分模型断层成像，这是一种从仅应用少量扰动（低阶投影）的测试中重构多扰动空间中模型行为的新方法。为了估计至少三种扰动的鲁棒性，TestifAI在涉及最多两种扰动的测试结果上训练一个辅助模型，从而避免执行指数数量的测试。我们在五个图像和语言分类任务上的实验表明，TestifAI能够从低阶（1和2扰动）观察中预测高阶（3和4扰动）测试结果，其整体鲁棒性估计误差低于7%，同时减少了60-80%的推理次数。
cs.AI / 58 / 2608.18938
Breaking the weakest link to evade vision language models
打破最薄弱环节以规避视觉语言模型
Abstract
Vision Language Models (VLMs) have recently emerged as a critical component of multimodal AI systems, enabling joint reasoning over visual and textual inputs in real-world and safety-critical applications. Despite their growing deployment, the robustness of VLMs against adversarial threats remains insufficiently explored, particularly in the context of evasion attacks targeting multimodal alignment. In this work, we investigate the vulnerability of VLMs to adversarial perturbations applied to visual inputs and study two attack settings: untargeted attacks, where the goal is to disrupt the model's interpretation of the original image, and targeted attacks, where the adversary aims to force the model to generate a specific semantic description unrelated to the original image. To efficiently generate adversarial examples, we propose a gradient-based attack method that performs optimization exclusively on the vision encoder of the VLM rather than on the entire multimodal architecture. This design significantly reduces the computational cost and resource requirements of the attack while maintaining strong effectiveness. We evaluate our approach on several open-source VLMs, including Qwen2.5-VL, Granite-Vision, FastVLM, and Phi-3.5-Vision, and show that small, human-imperceptible perturbations can substantially alter the textual interpretation produced by the models. Our findings highlight the vulnerability of modern VLMs to adversarial manipulation and emphasize the need for improved robustness and security mechanisms in multimodal AI systems.
Chinese Translation
视觉语言模型（VLMs）最近成为多模态人工智能系统的关键组成部分，使得在现实世界和安全关键应用中能够对视觉和文本输入进行联合推理。尽管它们的应用日益增长，但VLMs在面对对抗性威胁时的鲁棒性仍然未得到充分探索，尤其是在针对多模态对齐的规避攻击背景下。在本研究中，我们调查了VLMs对施加于视觉输入的对抗性扰动的脆弱性，并研究了两种攻击设置：无目标攻击，其目标是干扰模型对原始图像的解释；以及有目标攻击，攻击者旨在迫使模型生成与原始图像无关的特定语义描述。为了高效生成对抗样本，我们提出了一种基于梯度的攻击方法，该方法仅对VLM的视觉编码器进行优化，而不是对整个多模态架构进行优化。这一设计显著降低了攻击的计算成本和资源需求，同时保持了强大的有效性。我们在多个开源VLM上评估了我们的方法，包括Qwen2.5-VL、Granite-Vision、FastVLM和Phi-3.5-Vision，并显示出微小的、肉眼不可察觉的扰动可以显著改变模型生成的文本解释。我们的研究结果突显了现代VLMs对对抗性操控的脆弱性，并强调了在多模态人工智能系统中提高鲁棒性和安全机制的必要性。
cs.AI / 59 / 2608.19002
A Theory of Post-hoc Debate Judgement
后验辩论判断理论
Abstract
Debates have recently emerged as a useful methodology for agentic AI to improve performance as well as to aid explainability and user engagement. For example, LLM-empowered agents may debate internally (with themselves) and/or externally (with other agents). In many settings where debates are used, debates' outcomes and resulting outputs are determined post-hoc by external judges, often LLMs. In this paper we develop and test a novel theory of debate judgement applicable to all settings where agents engage in debates by providing pros and cons for their opinions therein. Specifically, we identify a number of formal properties that debate judgement may be required to satisfy in general, as concerns reproducibility, robustness, groundedness and explainability. Then, we explore their satisfaction formally and/or experimentally, for claim verification settings, for two specific alternative debate judgement methods: variants of the LLMs as a judge idea and formal semantics drawn from computational argumentation. We show that the two methods give similar accuracy performances but the former may lack formal guarantees that the latter brings. Overall, our study indicates argumentation semantics as an ideal candidate for principled judges in debate-driven AI.
Chinese Translation
辩论最近被视为一种有效的方法论，旨在提升自主智能体（agentic AI）的性能，同时促进可解释性和用户参与。例如，基于大型语言模型（LLM）的智能体可以进行内部（自我辩论）和/或外部（与其他智能体辩论）辩论。在许多使用辩论的场景中，辩论的结果及其输出通常由外部评判者（通常是LLM）在事后决定。本文提出并测试了一种新颖的辩论判断理论，该理论适用于所有参与辩论的智能体，通过提供其观点的利弊来进行判断。具体而言，我们识别出辩论判断在一般情况下可能需要满足的一些正式属性，包括可重复性、稳健性、基础性和可解释性。然后，我们正式和/或实验性地探讨了这些属性在主张验证场景中的满足情况，针对两种特定的替代辩论判断方法：作为评判者的LLM变体和源自计算论证的形式语义。我们展示了这两种方法在准确性表现上相似，但前者可能缺乏后者所提供的正式保证。总体而言，我们的研究表明，论证语义是辩论驱动的人工智能中理想的原则性评判者候选者。
cs.AI / 60 / 2608.19025
Self-prompting and cross-model consensus enable reproducible data extraction from scientific literature with large language models
自我提示与跨模型共识使得利用大型语言模型从科学文献中可重复提取数据成为可能
Abstract
Accurately extracting nuanced, contextualized data from research articles is laborious and time intensive. Here, we investigate the performance of frontier, browser-based large language models (LLMs) to extract highly contextualized information. We demonstrate four escalating workflows, 1) given an expert curated prompt and research articles, most frontier LLMs perform well at data extraction, however can struggle with interpreting scientific context and nuance, 2) given simple instructions, LLMs can author their own prompts which were almost as eNective as expert-written prompts, 3) autonomous discovery of research literature was diNicult, agents either missed or hallucinated references, and 4) LLMs can create new datasets from published guidelines that closely match human-expert judges, but still require a human-in-the-loop. Together, these findings define an auditable division of labour in which experts specify the evidence standard, models cross-check repeated extractions and researchers resolve disputed cases, providing a practical route to scaling scientific data curation without relinquishing expert oversight.
Chinese Translation
从研究文章中准确提取细微的、具有上下文的数据是一个费时费力的过程。在此，我们研究了前沿的基于浏览器的大型语言模型（LLMs）在提取高度上下文信息方面的表现。我们展示了四个逐步提升的工作流程：1）在提供专家策划的提示和研究文章的情况下，大多数前沿LLMs在数据提取方面表现良好，但在解释科学背景和细微差别时可能会遇到困难；2）在给定简单指令的情况下，LLMs能够生成自己的提示，这些提示的效果几乎与专家撰写的提示相当；3）自主发现研究文献是困难的，代理要么遗漏要么虚构参考文献；4）LLMs能够从已发布的指南中创建新的数据集，这些数据集与人类专家评审的结果高度匹配，但仍然需要人类参与。综合这些发现，我们定义了一种可审计的劳动分工，其中专家指定证据标准，模型交叉检查重复提取，研究人员解决争议案例，从而为在不放弃专家监督的情况下扩展科学数据管理提供了一条切实可行的路径。
cs.AI / 61 / 2608.19029
Adaptive Memory and Reflection Multi-Agent System for Medical Question Answering
用于医学问答的自适应记忆与反思多智能体系统
Abstract
Accurate and responsible medical question answering (QA) is important in healthcare, where complex cases require factual knowledge and nuanced reasoning. Existing medical QA systems, typically based on single-agent architectures and static retrieval, often lack adaptability, persistent memory, and structured decision-making. This work introduces an adaptive memory and reflection (AMR) agentic system, a multi-agent framework in which specialized agents use dedicated memory and reflection-based feedback to retrieve relevant prior cases and improve subsequent reasoning. Complexity assessment routes questions through solo, collaborative, or escalated workflows, while consensus and ethical overseer modules support reasoning consolidation and output review. Evaluation on MedQA and MedMCQA demonstrates strong performance compared with several baselines. Ablation studies show that combining agent-specific memory, reflection, and external retrieval yields the strongest performance. These findings highlight the potential of structured memory and feedback for developing more trustworthy medical agents. The source code is publicly available at https://github.com/mm-air/AMR-Agent.
Chinese Translation
准确且负责任的医学问答（QA）在医疗保健中至关重要，因为复杂病例需要事实知识和细致的推理。现有的医学问答系统通常基于单一智能体架构和静态检索，往往缺乏适应性、持久记忆和结构化决策能力。本研究提出了一种自适应记忆与反思（AMR）智能体系统，这是一种多智能体框架，其中专门的智能体利用专用记忆和基于反思的反馈来检索相关的先前案例并改善后续推理。复杂性评估将问题引导通过单独、协作或升级的工作流程，而共识和伦理监督模块支持推理整合和输出审查。在 MedQA 和 MedMCQA 上的评估显示，与多个基准相比，表现出色。消融研究表明，结合特定智能体的记忆、反思和外部检索可以获得最佳性能。这些发现突显了结构化记忆和反馈在开发更可信的医学智能体方面的潜力。源代码可在 https://github.com/mm-air/AMR-Agent 获取。
cs.AI / 62 / 2608.19047
Eureka: Task-Conditioned Meta-Agent Orchestration for Scientific Discovery
Eureka：用于科学发现的任务条件元代理编排
Abstract
We present Eureka, a task-conditioned Meta-Agent architecture that compiles long-horizon tasks into dynamic obligation graphs with explicit acceptance semantics. During execution, Eureka forms Macro-Agents with specialized state, memory, operators, tools, verifiers, and local topology via receding-horizon planning, architecture promotion, and minimal-sufficient compilation. When bottlenecks recur, cost-benefit-gated evolution updates the local architecture under constraints. Theoretically, we establish results on regret, planning invalidation, amortization, subtree interfaces, serializability, and verification. Experimentally, Eureka completes 170/170 recursive tasks and generates 3,948 certificates with no false acceptances. Active context compresses median input from 9,490 to 4,005 tokens; incremental processing avoids 65.38% recomputation across 12,000 tasks; 16,000 concurrent executions serialize consistently. The same Meta-Agent instantiates a Theory-Discovery Agent and a Math/Conjecture Agent. The former yields structural results in quantum-process and spacetime theory. The latter identifies bottlenecks in Riemann Hypothesis research and advances a positivity certificate for Suzuki's localized Weil quadratic form to 0 < a <= 69/200 = 0.345, reaching ~99.55% of (log 2)/2. These results suggest that scientific-agent capability depends not only on the base model but on whether an architecture can be formed to match the task's cognitive structure.
Chinese Translation
我们提出了Eureka，一种任务条件的元代理架构，该架构将长时间跨度的任务编译为具有明确接受语义的动态义务图。在执行过程中，Eureka通过递归规划、架构提升和最小充分编译形成具有专业状态、记忆、操作符、工具、验证器和局部拓扑的宏代理。当瓶颈再次出现时，成本效益门控演化在约束下更新局部架构。从理论上讲，我们建立了关于遗憾、规划失效、摊销、子树接口、可序列化性和验证的结果。在实验上，Eureka完成了170/170个递归任务，并生成了3,948个证书，没有出现错误接受。主动上下文将中位输入从9,490个压缩到4,005个标记；增量处理避免了在12,000个任务中65.38%的重新计算；16,000个并发执行的一致序列化。相同的元代理实例化了一个理论发现代理和一个数学/猜想代理。前者在量子过程和时空理论中产生了结构性结果。后者识别了黎曼假设研究中的瓶颈，并将铃木的局部韦尔二次型的正性证书推进到0 < a <= 69/200 = 0.345，达到了~99.55%的(log 2)/2。这些结果表明，科学代理的能力不仅依赖于基础模型，还依赖于是否能够形成与任务的认知结构相匹配的架构。
cs.AI / 63 / 2608.19072
What is Missing from AI Post-Training AI: An Empirical Analysis
后训练人工智能中缺失的是什么：一项实证分析
Abstract
Large language model (LLM) agents can now post-train an LLM end-to-end. They can write code, launch training, evaluate checkpoints, and improve downstream performance, raising the prospect of AI-for-AI. We argue that this picture conflates two distinct capabilities: execution-level capability, iterating within a selected training strategy; and strategy-level capability, revising the high-level judgment as experimental evidence accumulates. Analyzing a large corpus of publicly released post-training trajectories, we find that across different tasks, the agent's training strategy is locked in at the very beginning, and the entire remaining budget is spent on local adjustments within the selected strategy. We then examine three natural explanations--missing experience, missing guidance, and insufficient reasoning--with escalating interventions. Extensive experiments show that (1) an experience-driven scaffold improves execution across the board (+12.6 points on GSM8K and +40.8 on HumanEval) but leaves the strategy static; (2) human guidance effectively redirects the initial strategy, yet the agent falls back into local adjustment loops once training starts; and (3) additional inference compute pays off on easier tasks but yields almost no gain on the hardest one. In conclusion, what agents lack is neither experience, guidance, nor reasoning compute, but a mechanism for spontaneously reevaluating their strategy during execution.
Chinese Translation
大型语言模型（LLM）代理现在可以对LLM进行端到端的后训练。它们可以编写代码、启动训练、评估检查点并改善下游性能，从而引发了人工智能为人工智能的前景。我们认为，这种情况混淆了两种不同的能力：执行层面的能力，即在选定的训练策略内进行迭代；以及策略层面的能力，即随着实验证据的积累修订高层次判断。通过分析大量公开发布的后训练轨迹，我们发现，在不同任务中，代理的训练策略在一开始就被锁定，整个剩余预算都用于在选定策略内进行局部调整。然后，我们考察了三种自然解释——缺乏经验、缺乏指导和推理不足——并进行了逐步干预。大量实验表明：（1）以经验为驱动的支架在各方面提高了执行效果（GSM8K提高12.6分，HumanEval提高40.8分），但策略保持静态；（2）人类指导有效地重新引导了初始策略，但一旦训练开始，代理又回到了局部调整的循环中；（3）在较简单的任务上，额外的推理计算是有回报的，但在最困难的任务上几乎没有收益。总之，代理所缺乏的既不是经验、指导，也不是推理计算，而是在执行过程中自发重新评估其策略的机制。
cs.AI / 64 / 2608.19073
Robust Risk Under Evolving Uncertainty: A Wasserstein Counterpart of the Entropic Value-at-Risk
在不断演变的不确定性下的稳健风险：熵值风险的Wasserstein对应物
Abstract
An agent still learning its environment should be cautious while ignorant and bold once confident. The entropic value-at-risk captures this through a robust-optimization identity---a confidence level fixes the radius of a relative-entropy ball of alternative models---but that ball cannot reach catastrophes the nominal deems impossible, precisely what a safe agent must hedge. We instead use an optimal-transport ball and study the coherent risk measure it induces, the Wasserstein entropic value-at-risk. It has a variational dual mirroring the entropic formula (an inverse temperature becomes a transport price), occupies a definite place in the risk hierarchy, and provably accounts for the reachable catastrophes the entropic measure ignores; we verify both dualities numerically. Driving the transport radius by belief entropy then yields a closed-form robust dynamic-programming operator whose caution contracts as the belief sharpens, with a certified safety sandwich and a sharp safety switch.
Chinese Translation
一个仍在学习其环境的代理在无知时应保持谨慎，而在自信时则应大胆。熵值风险通过一个稳健优化的身份捕捉了这一点——一个置信水平固定了替代模型的相对熵球的半径——但该球无法覆盖名义上认为不可能的灾难，而这正是一个安全代理必须对冲的内容。我们则使用最优运输球，研究其所引发的连贯风险度量，即Wasserstein熵值风险。它具有一个变分对偶，反映了熵公式（逆温度变为运输价格），在风险层次中占据明确的位置，并且可以证明能够考虑熵度量所忽视的可达灾难；我们通过数值方法验证了这两个对偶。通过信念熵驱动运输半径，最终得到了一个封闭形式的稳健动态规划算子，其谨慎程度随着信念的明确而收缩，同时伴随有经过认证的安全夹层和敏锐的安全切换。
cs.AI / 65 / 2608.19125
Tuning the Stochastic Machine: A Systems Engineer's Operating Model for Human-AI Engineering
调优随机机器：系统工程师的人机智能工程操作模型
Abstract
When an expert corrects an LLM assistant's error, the correction usually dies with the session, and the error class returns. I argue this is an operations problem, not a tooling problem: mechanisms for persisting corrections exist and are shipping, but the discipline for governing them -- versioning with provenance, recurrence monitoring, counter-metrics, retirement of stale rules -- does not. Writing as a systems engineer of thirty years, I map the LLM stack onto the machines my profession already operates (frozen silicon, firmware, loadable modules, persistent configuration, volatile memory), identify where the mapping fails (stochastic generation, configuration that binds only probabilistically, no general-purpose retirement (verification) stage by default), and derive from the failures a seven-principle operating discipline with an error loop at its core. Three cases from my own practice illustrate the mechanism, among them a control that silently became the exact harm it was built to prevent. I close with the measurement framework this view implies and the lab study required to test it.
Chinese Translation
当专家纠正大型语言模型（LLM）助手的错误时，这一纠正通常在会话结束后消失，错误类别会重新出现。我认为这是一个操作问题，而不是工具问题：持久化纠正的机制已经存在并投入使用，但对其进行管理的学科——版本控制与来源追踪、重复监测、反向指标、过时规则的淘汰——却缺失。作为一名拥有三十年经验的系统工程师，我将LLM堆栈映射到我所在职业已经操作的机器（冻结硅片、固件、可加载模块、持久配置、易失性内存）上，识别出映射失败的地方（随机生成、仅以概率绑定的配置、默认情况下没有通用的淘汰（验证）阶段），并从这些失败中推导出以错误循环为核心的七项原则操作规范。我的实践中有三个案例说明了这一机制，其中一个控制措施悄然成为了它所旨在防止的确切危害。我最后提出了这一观点所暗示的测量框架以及测试所需的实验室研究。
cs.AI / 66 / 2608.19140
Grouping the Stochastic Machine: Precision, Not Capability, as the Frontier Metric for AI Systems
分组随机机器：精度，而非能力，作为人工智能系统的前沿指标
Abstract
Frontier language models are compared, marketed, and benchmarked on capability -- what their best or average output can achieve. I argue this measures the wrong axis. The models have saturated accuracy: their mean output lands on the target. What now separates one system from another in practice is precision: how tightly concentrated their outputs are around that target across repeated, identical requests. Borrowing the marksman's distinction, capability is where the average shot lands; reliability is the size of the group. I make three claims. First, precision, not capability, is the frontier differentiator between systems, and benchmark culture systematically fails to measure it, reporting central tendency rather than spread. Second, precision is measurable, cheaply and without circularity, by running a fixed suite of deterministically scored tasks many times at fixed temperature and computing the per-task consistency of outcomes -- no model-in-the-loop grader required. Third, the measurement is not merely descriptive but decision-guiding: it separates consistent failures (a tight group off-centre, correctable by the operating discipline of Paper 1 -- a sight adjustment) from scattered failures (a wide group, correctable only by changing the model or its sampling -- a rifle problem). I define a grouping metric, specify a harness, and show how tracking a human-AI pair's grouping over time yields the compounding signal that Paper 1's field study requires. A first real run, since replicated, illustrates both the method and its most important limit: one measured gap was closed completely by a single rule (0/5 -> 5/5), while a suite of tasks authored from the rules themselves found no value, because a frontier model already embodies explicit good practice -- establishing that a discipline's worth is found by measurement on real work, not constructed from its own rulebook.
Chinese Translation
前沿语言模型的比较、市场营销和基准测试主要基于能力——即它们的最佳或平均输出所能达到的水平。我认为这种衡量方式是错误的。模型的准确性已经饱和：它们的平均输出达到了目标。现在在实践中区分一个系统与另一个系统的关键在于精度：在多次相同请求中，它们的输出围绕目标的集中程度。借用射手的区分，能力是平均射击落点；可靠性则是群体的大小。我提出三项主张。首先，精度，而非能力，是系统之间的前沿区分因素，而基准文化系统性地未能对此进行测量，报告的是集中趋势而非分散程度。其次，精度是可测量的，成本低且没有循环性，通过在固定温度下多次运行一套确定性评分任务，并计算每个任务结果的一致性——无需模型参与的评分者。第三，这一测量不仅是描述性的，还能指导决策：它将一致性失败（一个偏离中心的紧密群体，可以通过《论文1》的操作规范进行调整——视线调整）与分散失败（一个宽泛的群体，仅能通过改变模型或其采样进行纠正——步枪问题）区分开来。我定义了一个分组指标，指定了一个框架，并展示了如何跟踪人类与人工智能配对的分组随时间变化所产生的复合信号，这是《论文1》实地研究所需的。一次首次真实的运行，因被复制而说明了该方法及其最重要的局限性：一个测量的差距通过一条规则（0/5 -> 5/5）完全弥补，而从这些规则本身撰写的一系列任务未发现任何价值，因为前沿模型已经体现了明确的良好实践——证明了一个学科的价值是通过对真实工作的测量发现的，而不是从其自身的规则手册中构建出来的。
cs.AI / 67 / 2608.19161
Beyond the Transcript: Detecting Covert Co ordination in Latent Multi-Agent Communication
超越转录：检测潜在多智能体通信中的隐蔽协调
Abstract
Language-model agents can communicate through continuous hidden states that are invisible in public transcripts, creating opportunities for covert harmful coordination. We introduce Verifiable Latent Alignments (VLA), an activation-aware framework for monitoring and steering these private communication channels. For every monitored decision, VLA links the private latent-state record and channel status to the resulting public action using a shared event identifier, enabling matched causal analysis. Our first contribution is a neutral-only three-layer monitor combining representation anomaly detection, counterfactual action-distribution influence, and sparse-autoencoder interpretation support. Our second contribution is a steerability framework spanning black-box behavioral instructions and white-box matched-neutral counterfactuals. Our third contribution is an evaluation on a controlled multi-agent auction benchmark covering homogeneous and heterogeneous model pairs, many-agent scalability, and intervention effectiveness. The sequential monitor achieves mean area under the receiver operating characteristic curve (AUROC) of 0.993 for homogeneous agents and 0.854 for heterogeneous pairs when text- and latent-collusion rows are pooled as positives. In Qwen3-0.6B auctions with 25-100 bidders, monitoring requires only a small normalized load relative to all possible directed pairs, while full white-box steering achieves 100% bid-distribution recovery and reduces collusive low-bid behavior by 47.3 percentage points. Because full white-box steering replays the matched neutral counterfactual, its exact recovery is a sanity check by construction. Overall, the controlled study shows that the evaluated private channel attacks can be monitored without training the primary monitor on attack examples and mitigated when matched counterfactual access is available.
Chinese Translation
语言模型代理可以通过在公共转录中不可见的连续隐藏状态进行通信，从而创造出隐蔽有害协调的机会。我们提出了可验证的潜在对齐（Verifiable Latent Alignments, VLA），这是一个关注激活的框架，用于监控和引导这些私人通信通道。对于每个被监控的决策，VLA通过共享事件标识符将私人潜在状态记录和通道状态与产生的公共行动联系起来，从而实现匹配的因果分析。我们的第一个贡献是一个仅限中立的三层监控器，结合了表示异常检测、反事实行动分布影响和稀疏自编码器解释支持。我们的第二个贡献是一个可引导性框架，涵盖黑箱行为指令和白箱匹配中立反事实。我们的第三个贡献是在一个受控的多智能体拍卖基准上进行评估，该基准涵盖了同质和异质模型对、多智能体可扩展性以及干预有效性。当文本和潜在串通行被视为正例时，顺序监控器在同质代理中实现了0.993的接收者操作特征曲线（AUROC）下的平均面积，而在异质对中则为0.854。在具有25-100名竞标者的Qwen3-0.6B拍卖中，监控相对于所有可能的定向对只需少量标准化负载，而完全的白箱引导实现了100%的出价分布恢复，并将串通低出价行为减少了47.3个百分点。由于完全的白箱引导重放匹配的中立反事实，其精确恢复是通过构造进行的合理性检查。总体而言，受控研究表明，被评估的私人通道攻击可以在不针对攻击示例训练主要监控器的情况下进行监控，并在可获得匹配反事实访问时得以缓解。
cs.CL / 1 / 2608.18082
LongNovel: A Multi-Scale Benchmark for Hallucination Detection in Long-Context Novel Summarization
LongNovel：用于长文本小说摘要中幻觉检测的多尺度基准
Abstract
Although context windows have expanded significantly in recent years, hallucinations in long-context summarization remain a challenge. Long novels are better suited than news or papers for researching these hallucinations, due to their intrinsic information and detailed descriptions of events and dialogues. However, current research lacks a multi-scale benchmark for hallucination detection in long-context novel summarization and does not fully explore how hallucinations change as the context grows longer. In this study, we propose LongNovel, a multi-scale long-context bilingual (Chinese and English) novel benchmark for hallucination detection. This benchmark is constructed from 29 Chinese novels (ranging from 16k to 100k tokens) and chapter-level data from the BookSum dataset. We design 8 hallucination types and employ a combination of Multi-Model Arbitration and Entity-Referenced Hallucination Generation to ensure both data authenticity and a balanced distribution of hallucination categories. Furthermore, we manually revise the content in the test set to guarantee data reliability. Extensive experimental results demonstrate that LongNovel is a challenging benchmark. We release LongNovel for future research. https://github.com/BDML-lab/LongNovel
Chinese Translation
尽管近年来上下文窗口显著扩大，但在长文本摘要中幻觉现象仍然是一个挑战。与新闻或论文相比，长篇小说更适合研究这些幻觉，因为它们具有内在的信息和对事件及对话的详细描述。然而，目前的研究缺乏一个用于长文本小说摘要中幻觉检测的多尺度基准，并且没有充分探讨随着上下文变长幻觉如何变化。在本研究中，我们提出了LongNovel，一个用于幻觉检测的多尺度长文本双语（中文和英文）小说基准。该基准由29部中文小说（字数范围从16,000到100,000个标记）和BookSum数据集中的章节级数据构成。我们设计了8种幻觉类型，并采用多模型仲裁（Multi-Model Arbitration）与实体引用幻觉生成（Entity-Referenced Hallucination Generation）的结合，以确保数据的真实性和幻觉类别的平衡分布。此外，我们手动修订了测试集中的内容，以保证数据的可靠性。大量实验结果表明，LongNovel是一个具有挑战性的基准。我们将LongNovel发布以供未来研究使用。https://github.com/BDML-lab/LongNovel
cs.CL / 2 / 2608.18083
Entity tracking emerges in sub-billion parameter language models and exceeds human performance in naturalistic narratives
实体追踪在不足十亿参数的语言模型中出现，并在自然叙事中超越人类表现
Abstract
Understanding language requires tracking entities across discourse - i.e., knowing where things are and how they change, even when not explicitly stated. Whether language models perform such tracking in a human-like fashion remains unclear, in part because existing evaluations rely on artificial tasks, far removed from natural language comprehension, and lack comparisons to humans. Here, we evaluate entity tracking in both language models and humans (N = 48) using naturalistic narratives at multiple levels of complexity. In humans, we find that entity tracking degrades specifically with narrative complexity, not narrative length. In language models, we find that human-level entity tracking is already present at 410 million parameters - well below the multi-billion parameter, code-specialised models identified by prior work - and improves with scale, with contemporary models far exceeding human performance. Together, these results demonstrate that entity tracking, a core component of language understanding, emerges at model scales far smaller than previously thought.
Chinese Translation
理解语言需要在话语中追踪实体，即了解事物的位置及其变化，即使这些信息没有明确陈述。语言模型是否以类人方式进行这种追踪仍不清楚，部分原因是现有评估依赖于与自然语言理解相去甚远的人工任务，并且缺乏与人类的比较。在本研究中，我们使用多层次复杂度的自然叙事评估了语言模型和人类（N = 48）的实体追踪。在人类中，我们发现实体追踪在叙事复杂度上特定下降，而非叙事长度。在语言模型中，我们发现410百万参数的模型已经具有人类水平的实体追踪能力——远低于先前研究所识别的多十亿参数的代码专用模型——并且随着规模的增加而改善，现代模型的表现远超人类。综合来看，这些结果表明，实体追踪作为语言理解的核心组成部分，在比之前认为的更小的模型规模下就已出现。
cs.CL / 3 / 2608.18084
Compiler-Guided Adaptive Proof Search with Cross-Model Synergy on Context-Dependent Theorem Proving
基于编译器指导的自适应证明搜索与跨模型协同在上下文依赖定理证明中的应用
Abstract
Theorem proving in real-world Lean 4 projects is challenging because proofs often depend on project-specific context. While iterative refinement can use compiler errors to repair failed proofs, reusing failed attempts requires careful search control: some proofs provide better starting points than others, and later revisions may degrade a partially correct proof. We propose a compiler-guided proof search framework that balances exploration and exploitation. It explores diverse starting points through dual-model generation and stagnation-triggered resampling, while exploiting promising proof states through current-best refinement guided by compiler-grounded pairwise comparison. Experiments on seven real-world Lean 4 projects from miniCTX-v2 show that our method achieves a better effectiveness--efficiency tradeoff than pass@k baselines. Within the pass@32 budget, our method improves average pass rate by 12.8 percentage points while reducing LLM calls by 21.9%.
Chinese Translation
在现实世界的 Lean 4 项目中，定理证明面临挑战，因为证明通常依赖于特定项目的上下文。虽然迭代优化可以利用编译器错误来修复失败的证明，但重用失败尝试需要谨慎的搜索控制：某些证明比其他证明提供更好的起始点，而后续修订可能会降低部分正确证明的质量。我们提出了一种编译器指导的证明搜索框架，该框架在探索与利用之间取得平衡。它通过双模型生成和停滞触发的重采样探索多样化的起始点，同时通过编译器基础的成对比较指导当前最佳的优化，来利用有前景的证明状态。在来自 miniCTX-v2 的七个现实世界 Lean 4 项目的实验中，我们的方法在有效性与效率的权衡上优于 pass@k 基线。在 pass@32 的预算内，我们的方法将平均通过率提高了 12.8 个百分点，同时减少了 21.9% 的 LLM 调用。
cs.CL / 4 / 2608.18085
Persona-Guided LLM Agents for Task-Oriented Dialogue
基于角色指导的大型语言模型代理用于任务导向对话
Abstract
Prior work has shown that large language models (LLMs) can express diverse personality traits in open-ended text generation. However, it remains unclear whether they can do so in a goal-directed dialogue without compromising task completion, and whether adapting to the user's personality improves the interaction quality. We study these questions in task-oriented dialogue (TOD), where a system helps a user accomplish a goal via multi-turn interaction. We build a training-free framework that simulates a TOD interaction between two LLMs: a user agent that exhibits a target personality and a system agent that adapts to the user while completing the task. To isolate the effect of adaptation, we vary how much the system knows about the user's personality across three conditions. In Neutral, the system receives no personality information. In Try, it infers the personality from dialogue cues. In Oracle, it is given the personality explicitly. We evaluate GPT-4o, Qwen3-Next-80B, and Gemini 2.0 Flash on Hotel and Restaurant dialogues from the Schema-Guided Dialogue (SGD) dataset, across the Big Five traits and their opposite poles. We find that the user agent can express personality while the system maintains strong task performance, although some traits are realized far less reliably than others. Adapting to the user's personality improves constraint satisfaction, inform rate, and user satisfaction, but lowers truthfulness, revealing a trade-off between personalization and task-grounding. Oracle's gains grow when the target trait is strongly expressed, whereas Try's gains are largely insensitive to realization strength. Overall, cue-based adaptation in Try best resolves this trade-off and offers a more reliable route to personality-aware TOD without fine-tuning.
Chinese Translation
先前的研究表明，大型语言模型（LLMs）能够在开放式文本生成中表达多样的个性特征。然而，目前尚不清楚它们是否能够在不妨碍任务完成的情况下，在目标导向对话中做到这一点，以及适应用户个性是否能提高交互质量。我们在任务导向对话（TOD）中研究这些问题，在这种对话中，系统通过多轮交互帮助用户实现目标。我们构建了一个无训练的框架，模拟两个LLM之间的TOD交互：一个展示目标个性的用户代理和一个在完成任务的同时适应用户的系统代理。为了隔离适应的影响，我们在三种条件下变化系统对用户个性的了解程度。在中性条件（Neutral）下，系统没有接收到个性信息。在尝试条件（Try）下，系统通过对话线索推断个性。在预言条件（Oracle）下，系统明确获得个性信息。我们在Schema-Guided Dialogue（SGD）数据集中评估了GPT-4o、Qwen3-Next-80B和Gemini 2.0 Flash在酒店和餐厅对话中的表现，涵盖了五大人格特质及其对立面。我们发现用户代理可以表达个性，而系统仍能保持较强的任务表现，尽管某些特质的实现远不如其他特质可靠。适应用户个性提高了约束满足率、信息传递率和用户满意度，但降低了真实性，揭示了个性化与任务基础之间的权衡。当目标特质被强烈表达时，Oracle的收益增加，而Try的收益对实现强度的敏感性较低。总体而言，基于线索的适应在Try中最佳地解决了这一权衡，并提供了一条更可靠的途径，以实现个性意识的任务导向对话，而无需微调。
cs.CL / 5 / 2608.18087
SuTRA : Structurally-Unified Tokenization with Root Awareness
SuTRA：具有根意识的结构统一分词
Abstract
Existing subword tokenizers optimize statistical compression but ignore morphological structure, particularly the relationship between roots and affixes. This is harmful for morphologically rich Indic languages, where basic units are complex orthographic syllables (aksharas) rather than letters. Frequency-based methods over-fragment words, arbitrarily splitting roots and affixes - a phenomenon we term Morphological Shattering. We propose SuTRA (Structurally-Unified Tokenization with Root Awareness), a morphology-aware algorithm that preserves akshara indivisibility and penalizes merges crossing morphological boundaries. We also release a new morphological segmentation dataset for Hindi, Marathi, and Gujarati. SuTRA reduces shattering, achieving peak gains of +14.7% in morphological alignment (Boundary F1) and +34% in semantic recoverability (Hindi) over BPE. These structural gains yield an average improvement of +8.08 chrF2 in machine translation.
Chinese Translation
现有的子词分词器优化了统计压缩，但忽视了形态结构，特别是词根与词缀之间的关系。这对形态丰富的印地语言造成了不利影响，因为基本单位是复杂的正字法音节（aksharas），而非字母。基于频率的方法过度碎片化单词，任意地分割词根和词缀——这一现象我们称之为形态碎裂（Morphological Shattering）。我们提出了SuTRA（具有根意识的结构统一分词），这是一种形态意识算法，能够保持akshara的不可分性，并惩罚跨越形态边界的合并。我们还发布了一个新的形态分割数据集，涵盖印地语、马拉地语和古吉拉特语。SuTRA减少了碎裂现象，在形态对齐（Boundary F1）上实现了高达+14.7%的峰值增益，在语义可恢复性（印地语）上实现了+34%的增益，相较于BPE。这些结构性增益在机器翻译中带来了平均+8.08 chrF2的改善。
cs.CL / 6 / 2608.18089
Latent Space Refusal Anchoring for Low-Resource African Languages: Mechanistic Safety Recovery Without Retraining
低资源非洲语言的潜在空间拒绝锚定：无需重训练的机制安全恢复
Abstract
Instruction-tuned models often refuse harmful requests in English but comply with the same requests in Yoruba, Igbo, Igala, and Hausa. This suggests that the refusal mechanism is present in the residual stream but fails to activate for low-resource inputs. Recovering it normally requires labelled target-language data and retraining, neither of which is available at scale for most African languages. We introduce Latent Space Refusal Anchoring (LSR-Anchoring), a training-free method that extracts the refusal direction from English prompts and clamps it onto the residual stream at inference time. The primary variant, Mean-Activation Steering (MAS), operates across the four architectures we tested: Llama-3-8B, Llama-3.1-70B, Mistral-7B-Instruct, and Qwen2.5-7B. On Mistral and Qwen it recovers safety with benign degradation below 0.08. On Llama-3-8B it overcorrects, with Degraded Performance on Legitimate prompts (DPL) reaching 1.00. We address this with SAE-Derived Steering (SDS), which replaces the dense mean-difference direction with a single Sparse Autoencoder (SAE) feature and reduces Kullback-Leibler (KL) divergence by 3.5-7x without benign collapse. Four languages transfer positively, but Arabic fails on every architecture and at every steering magnitude, indicating a geometric mismatch rather than a baseline effect. Massive Multitask Language Understanding (MMLU) accuracy drops remain below 0.35 percentage points at every effective steering magnitude.
Chinese Translation
经过指令调优的模型在英语中通常拒绝有害请求，但在约鲁巴语、伊博语、伊加拉语和豪萨语中却对此类请求表示同意。这表明拒绝机制存在于残差流中，但在低资源输入时未能激活。恢复该机制通常需要标注的目标语言数据和重训练，而这在大多数非洲语言中并不普遍可用。我们提出了潜在空间拒绝锚定（Latent Space Refusal Anchoring, LSR-Anchoring）这一无训练方法，从英语提示中提取拒绝方向，并在推理时将其固定在残差流上。主要变体均值激活引导（Mean-Activation Steering, MAS）在我们测试的四种架构中运行：Llama-3-8B、Llama-3.1-70B、Mistral-7B-Instruct和Qwen2.5-7B。在Mistral和Qwen上，它以低于0.08的良性降级恢复安全性。在Llama-3-8B上，它过度校正，合法提示的降级性能（Degraded Performance on Legitimate prompts, DPL）达到1.00。我们通过SAE导向引导（SAE-Derived Steering, SDS）解决了这个问题，该方法用单个稀疏自编码器（Sparse Autoencoder, SAE）特征替换了密集均值差异方向，并在没有良性崩溃的情况下将Kullback-Leibler（KL）散度降低了3.5-7倍。四种语言的迁移效果积极，但阿拉伯语在每种架构和每个引导幅度下均未能成功，表明存在几何不匹配而非基线效应。大规模多任务语言理解（Massive Multitask Language Understanding, MMLU）准确率下降在每个有效引导幅度下均保持在0.35个百分点以下。
cs.CL / 7 / 2608.18090
Nine Emotion Centroids: A Label-Free Valence Axis That Transfers Across Four Modalities
九个情感中心：一个无标签的价值轴在四种模态间的转移
Abstract
Inside a modern language model sits a single internal direction that tracks how positive or negative a sentence feels. We show how to find this valence axis (V-axis) from just 9 emotion category names plus 50 short narrative paragraphs per emotion -- about 1,500 fewer labels than the usual supervised approach -- and that the same direction appears in vision, audio, and human-brain encoders never jointly trained. The recipe: embed nine emotion-anchored story sets in a frozen encoder, take the top principal direction of the nine averaged embeddings. Projecting new inputs onto it captures 93% of supervised performance on SST-2 (Llama-3-8B-Instruct, AUC 0.772 vs. 0.828), correlates with human valence ratings on 11,811 EmoSet images at r=0.636, reaches AUC 0.906 on ESC-50 audio (p<2.2e-15), and AUC 0.720+/-0.055 on EEG from 123 subjects (p<3.65e-8). The direction is mechanistically active: ablating it collapses sentiment accuracy by 5.5-37.2 pp across three LLMs vs. at most 0.88 pp for matched random directions (z>12). A 2-parameter classifier trained on text labels transfers to images (AUC 0.961), audio (0.764), and brain recordings (0.828) without target-modality labels; a generic 16-D subspace stays at chance (0.525). The recipe is bounded to continuous attributes -- seven tests on categorical concepts return near-chance -- and steering is family-specific (Llama/Mistral yes, Qwen/Gemma no).
Chinese Translation
在现代语言模型中，存在一个内部方向，用于跟踪句子的积极或消极情感。我们展示了如何仅通过9个情感类别名称和每个情感50个简短叙述段落来找到这个价值轴（V轴）——比通常的监督方法少约1,500个标签——并且同一方向出现在视觉、音频和从未共同训练的人脑编码器中。具体方法是：将九个情感锚定的故事集嵌入到一个冻结的编码器中，取九个平均嵌入的主方向。将新的输入投影到该方向上，可以捕获SST-2上的93%监督性能（Llama-3-8B-Instruct，AUC 0.772对比0.828），与11,811张EmoSet图像的人类价值评分的相关性为r=0.636，在ESC-50音频上达到AUC 0.906（p<2.2e-15），并在123名受试者的EEG上达到AUC 0.720+/-0.055（p<3.65e-8）。该方向在机制上是活跃的：去除它会导致三种大型语言模型（LLMs）的情感准确率下降5.5-37.2个百分点，而匹配随机方向最多下降0.88个百分点（z>12）。一个在文本标签上训练的2参数分类器可以转移到图像（AUC 0.961）、音频（0.764）和脑电图记录（0.828），而无需目标模态标签；一个通用的16维子空间保持在偶然水平（0.525）。该方法仅限于连续属性——对七个类别概念的测试返回接近偶然的结果——并且引导是特定于家族的（Llama/Mistral 是，Qwen/Gemma 否）。
cs.CL / 8 / 2608.18091
Self- and Other-Labels Induce Bidirectional Bias in LLM Judges
自我标签与他人标签在大型语言模型评判者中引发双向偏差
Abstract
As LLM-as-a-judge systems become increasingly widespread, self-preference in LLMs -- the tendency to favor one's own outputs -- raises growing concerns about evaluation reliability. However, it has been studied predominantly on generated text, where stylistic features and response quality are inevitably conflated. As a result, existing measurements cannot separate genuine self-preference from these confounds. We address this by changing the object of evaluation: instead of judging generated text, ten LLMs assess narrative constraint selections, which carry no model-specific stylistic fingerprint yet retain a recoverable model-specific signature. We run two experiments that yield distinct findings. Under blind evaluation, self-preference largely disappears once selection quality and evaluator severity are controlled. It vanishes on three of four rubric dimensions and reverses on the fourth, where judges rate their own selections as less original. Under matched quality, however, self- and other-labels alone -- without naming any model -- shift scores bidirectionally: LLM judges inflate scores for self-labeled selections and deflate those for other-labeled ones regardless of the selection's actual source. We make two contributions: 1) authorship attribution is a distinct driver of evaluation bias, and 2) open-ended, ground-truth-free tasks can serve as controlled instruments for studying LLM judge behavior.
Chinese Translation
随着大型语言模型（LLM）作为评判系统的日益普及，LLM中的自我偏好——即倾向于偏爱自身输出的趋势——引发了对评估可靠性日益增长的担忧。然而，这一现象主要在生成文本上进行研究，在此过程中，风格特征和响应质量不可避免地被混淆。因此，现有的测量方法无法将真正的自我偏好与这些混淆因素分开。我们通过改变评估对象来解决这一问题：十个LLM评估叙事约束选择，而不是判断生成文本，这些选择不带有模型特定的风格特征，但保留了可恢复的模型特定签名。我们进行了两项实验，得出了不同的发现。在盲评估下，一旦控制了选择质量和评估者的严格性，自我偏好在很大程度上消失。在四个评估维度中的三个上，它消失了，而在第四个维度上，评判者将自己的选择评为不那么原创。在匹配质量的情况下，然而，仅凭自我标签和他人标签——而不提及任何模型——就会双向改变评分：LLM评判者对自我标记的选择评分上调，而对他人标记的选择评分下调，无论选择的实际来源如何。我们的贡献有两个：1）作者归属是评估偏差的一个独特驱动因素，2）开放式、无真实依据的任务可以作为研究LLM评判者行为的控制工具。
cs.CL / 9 / 2608.18093
Abliteration Mitigation via Refusal Aliases
通过拒绝别名减轻消融效应
Abstract
Abliteration, the removal of refusal capabilities from large language models by projecting weight matrices orthogonal to an extracted refusal direction, has emerged as a prominent safety concern through its ability to bypass post-training alignment using only a small set of contrastive prompts. We find that existing defenses commonly overlook the cause of abliteration; that is, how easily the refusal direction can be extracted. To hinder this process, we introduce a weight-editing method that obscures the refusal signal by applying rank-$k$ updates to residual stream writer matrices while replacing refusal-inducing activations with random aliases and correcting downstream reader matrices to preserve the model's original behavior. On Llama-3-8B, AMRA improves post-abliteration refusal scores by $2.16$ points over the undefended baseline with less than $0.5$ percentage points of MMLU degradation. On Gemma-2-9B, it improves the post-abliteration refusal by $14.70$ points over the baseline while keeping harmful output rates similar to the baseline, albeit at a greater utility cost.
Chinese Translation
消融效应是指通过将权重矩阵正交投影到提取的拒绝方向，从大型语言模型中去除拒绝能力，这已成为一个突出的安全问题，因为它能够仅通过一小组对比提示绕过后训练对齐。我们发现现有的防御措施通常忽视了消融效应的原因；即，拒绝方向的提取有多容易。为了阻碍这一过程，我们提出了一种权重编辑方法，通过对残差流写入矩阵应用秩-$k$ 更新来模糊拒绝信号，同时用随机别名替换引发拒绝的激活，并校正下游读取矩阵以保持模型的原始行为。在 Llama-3-8B 上，AMRA 在消融后的拒绝评分上比未防御基线提高了 $2.16$ 分，且 MMLU 降级不足 $0.5$ 个百分点。在 Gemma-2-9B 上，它在消融后的拒绝评分上比基线提高了 $14.70$ 分，同时保持有害输出率与基线相似，尽管代价更高。
cs.CL / 10 / 2608.18094
NE-BERT: A Multilingual Language Model for Nine Northeast Indian Languages
NE-BERT：一种针对九种东北印度语言的多语言模型
Abstract
Large pretrained language models have demonstrated remarkable capabilities across diverse languages, yet critically underrepresented low-resource languages remain marginalized. We present NE-BERT, a domain-specific multilingual encoder model trained on approximately 8.3 million sentences spanning 9 Northeast Indian languages and 2 anchor languages (Hindi, English), a linguistically diverse region with minimal representation in existing multilingual models. By employing weighted data sampling and a custom SentencePiece Unigram tokenizer, NE-BERT outperforms IndicBERT-V2 and MuRIL across all 9 Northeast Indian languages, achieving 15.97X and 7.64X lower average perplexity respectively, with 1.50X better tokenization fertility than mBERT. We address critical vocabulary fragmentation issues in extremely low-resource languages such as Pnar (1,002 sentences) and Kokborok (2,463 sentences) through aggressive upsampling strategies. Downstream evaluation on part-of-speech tagging validates practical utility on three Northeast Indian languages. We release NE-BERT, test sets, and training corpus under CC-BY-4.0 to support NLP research and digital inclusion for Northeast Indian communities.
Chinese Translation
大型预训练语言模型在多种语言中展示了显著的能力，但低资源语言仍然受到严重代表不足的影响，处于边缘化状态。我们提出了NE-BERT，这是一种特定领域的多语言编码器模型，训练数据包含约830万句，涵盖9种东北印度语言和2种锚语言（印地语、英语），这一语言多样性区域在现有多语言模型中几乎没有代表性。通过采用加权数据抽样和自定义的SentencePiece Unigram分词器，NE-BERT在所有9种东北印度语言上均优于IndicBERT-V2和MuRIL，分别实现了15.97倍和7.64倍更低的平均困惑度，并且在分词效率上比mBERT提高了1.50倍。我们通过激进的上采样策略解决了在极低资源语言（如Pnar（1,002句）和Kokborok（2,463句））中存在的关键词汇碎片化问题。对词性标注的下游评估验证了在三种东北印度语言上的实际应用价值。我们将NE-BERT、测试集和训练语料库以CC-BY-4.0协议发布，以支持东北印度社区的自然语言处理研究和数字包容性。
cs.CL / 11 / 2608.18095
Backdoor Learning in Language Models and Vision-Language Models
语言模型和视觉-语言模型中的后门学习
Abstract
Recent advances in deep learning have significantly enhanced the capabilities of Natural Language Processing (NLP) and Vision-Language Models (VLMs). However, these advancements come with increased vulnerabilities, notably through backdoor attacks that pose severe security threats. This thesis addresses two critical dimensions of Trustworthy AI and Efficient Multimodal Representation Learning: (1) security through analyzing, detecting, and designing backdoor attacks in NLP and VLMs, and (2) efficiency through advanced multimodal representation methods tailored for clinical and medical imaging applications.
Chinese Translation
最近深度学习的进展显著增强了自然语言处理（NLP）和视觉-语言模型（VLMs）的能力。然而，这些进展也带来了更大的脆弱性，特别是通过后门攻击，这对安全构成了严重威胁。本论文探讨了可信人工智能和高效多模态表示学习的两个关键维度：（1）通过分析、检测和设计NLP和VLMs中的后门攻击来实现安全性，以及（2）通过针对临床和医学成像应用的先进多模态表示方法来提高效率。
cs.CL / 12 / 2608.18096
MAVEN: A Macro-Societal Value Evaluation Framework of Multimodal Content with Compact Aligned Evaluators
MAVEN：一种紧凑对齐评估者的多模态内容宏观社会价值评估框架
Abstract
Assessing whether multimodal content aligns with macro-societal values, such as peace, justice, and freedom, has become an increasingly urgent challenge. Existing frameworks are largely confined to safety-oriented taxonomies, text-only psychometric probes, or single-label classification. Therefore, we propose MAVEN, a hierarchical framework for macro-societal value evaluation of multimodal content, grounded in international human-rights instruments and cultural value theory. MAVEN organizes values into 6 primary dimensions and 72 secondary indicators, supporting multi-level quantitative scoring. Building on MAVEN, we construct a human-verified multimodal benchmark and a soft-match metric to evaluate VLMs' assessments across value dimensions. For evaluator optimization, we propose a span-adaptive variant of multi-level preference optimization for evaluator distillation, together with a training-free multi-role consensus strategy at inference time. We evaluate existing open- and closed-source VLMs on our benchmark, revealing shared tendencies and clear differences in macro-societal value judgments. Experiments show that our compact 2B evaluator matches its 8B counterpart in the same family and approaches frontier closed-source VLMs, offering a practical path toward scalable macro-societal value evaluation. Our SA-MDPO implementation and MacroValue-Bench are available at https://github.com/zzzzzzzzjj/MAVEN.
Chinese Translation
评估多模态内容是否与宏观社会价值（如和平、公正和自由）相一致，已成为一个日益紧迫的挑战。现有框架主要局限于安全导向的分类法、仅文本的心理测量探测或单标签分类。因此，我们提出了MAVEN，一个基于国际人权文书和文化价值理论的多模态内容宏观社会价值评估的层次框架。MAVEN将价值组织为6个主要维度和72个次级指标，支持多层次的定量评分。在MAVEN的基础上，我们构建了一个经过人工验证的多模态基准和一个软匹配度量，以评估VLM（价值学习模型）在价值维度上的评估。为了优化评估者，我们提出了一种跨度自适应的多层次偏好优化变体用于评估者提炼，并在推理时结合无训练的多角色共识策略。我们在我们的基准上评估了现有的开源和闭源VLM，揭示了宏观社会价值判断中的共同倾向和明显差异。实验表明，我们的紧凑型2B评估者与同一家族的8B评估者相匹配，并接近前沿的闭源VLM，提供了一条可扩展的宏观社会价值评估的实用路径。我们的SA-MDPO实现和MacroValue-Bench可在https://github.com/zzzzzzzzjj/MAVEN获取。
cs.CL / 13 / 2608.18097
FrenchNews-7: Benchmarking Cross-Publisher French News Editorial Desk Classification
FrenchNews-7：跨出版商法语新闻编辑部分类基准测试
Abstract
We present FrenchNews-7, a cross-publisher France-based French-language news editorial desk classification benchmark combining a large multi-outlet corpus, a URL-derived seven-class taxonomy, and a fine-tuned CamemBERT classifier. Labels are assigned via a hybrid pipeline combining publisher URL slugs with LLM annotation for structurally ambiguous cases, audited through an inter-rater study (2 humans + 2 LLMs; pairwise $\kappa \geq 0.766$, human--human $\kappa = 0.806$). We evaluate lexical, multilingual, and French-specific trained classifiers under both in-distribution and held-out-publisher settings, with additional comparison against zero-shot LLM baselines (GPT-OSS-120B, Mistral Small 3.2, Llama-3.3-70B) on the held-out pool. The strongest model, CamemBERT-base on full article text, outperforms headline-only input, generalizes to unseen outlets, and exceeds all three zero-shot LLM baselines on overall recall (0.799), with the gap concentrated in the ambiguous editorial-boundary categories Economie and Societe. Cross-publisher evaluation reveals uneven boundary stability: Sport, Culture & Loisirs, and International transfer cleanly, while Economie (recall = 0.517) is close to blinded human agreement (0.55), and Societe (precision = 0.577) absorbs boundary ambiguity, both suggesting editorial conventions rather than recoverable classifier headroom. The fine-tuned CamemBERT-base model, labeled manifest, reference collection scripts, and a reliability-tier guidance table are available at https://huggingface.co/LeFrenchNewsLab/camembert-base-frenchnews7 (model) and https://huggingface.co/datasets/LeFrenchNewsLab/frenchnews-7 (dataset).
Chinese Translation
我们提出了FrenchNews-7，这是一个基于法国的跨出版商法语新闻编辑部分类基准，结合了一个大型多渠道语料库、一个基于URL的七类分类法以及一个经过微调的CamemBERT分类器。标签通过一个混合流程分配，该流程结合了出版商的URL标识符和大型语言模型（LLM）注释，以处理结构模糊的情况，并通过一个评审研究进行审核（2名人类 + 2个LLM；成对$ ext{kappa} ext{≥} 0.766$，人类-人类$ ext{kappa} = 0.806$）。我们在分布内和保留出版商设置下评估了词汇、跨语言和法语特定训练的分类器，并与保留池中的零样本LLM基线（GPT-OSS-120B、Mistral Small 3.2、Llama-3.3-70B）进行了额外比较。最强模型CamemBERT-base在完整文章文本上的表现优于仅使用标题的输入，能够推广到未见过的出版商，并在整体召回率（0.799）上超过所有三个零样本LLM基线，差距主要集中在模糊的编辑边界类别Economie和Societe上。跨出版商评估显示边界稳定性不均：Sport、Culture & Loisirs和International的转移非常清晰，而Economie（召回率=0.517）接近盲人类一致性（0.55），Societe（精确率=0.577）则吸收了边界模糊性，这两者都表明了编辑惯例而非可恢复的分类器潜力。经过微调的CamemBERT-base模型、标注清单、参考收集脚本和可靠性分级指导表可在以下网址获取：https://huggingface.co/LeFrenchNewsLab/camembert-base-frenchnews7（模型）和https://huggingface.co/datasets/LeFrenchNewsLab/frenchnews-7（数据集）。
cs.CL / 14 / 2608.18098
Fractional Decay KV-Cache: Ownership-Aware Memory Management for Improved Inference Relevancy in Dialog Systems
分数衰减KV缓存：面向所有权的内存管理以提高对话系统中的推理相关性
Abstract
Key-value (KV) caching is essential for efficient autoregressive inference in transformer based dialog systems, yet existing strategies treat all cached entries uniformly or apply coarse eviction heuristics that fail to adapt as dialog topics evolve. We propose Fractional Decay KV-Cache (FD-KVC), a novel algorithm that maintains a dual-channel scoring mechanism for each cached KV pair: a cumulative attention channel that tracks aggregate importance (akin to H2O), and a recency-weighted relevance channel governed by temporal decay and reinforcement-inspired updates. The combination enables FD-KVC to both preserve historically important tokens and rapidly adapt when dialog topics shift. An adaptive learning rate driven by an ownership loss function ensures convergence without oscillation. FD-KVC operates entirely on CPU with negligible overhead. Across five diverse multi-turn dialog scenarios with 600 dialogs each, FD-KVC outperforms H2O, the state-of-the-art heavy-hitter baseline, by +6.7% on composite late-turn alignment, with improvements of +127% on topic-shift, +87% on gradual evolution, and +30% on mixed-topic dialogs. FD-KVC adapts to new topics 3.6X faster than H2O and achieves the highest topic diversity (80.6%) across all methods. Ablation studies confirm the contribution of each component.
Chinese Translation
键值（KV）缓存对于基于变换器的对话系统中高效的自回归推理至关重要，但现有策略对所有缓存条目采取统一处理或应用粗糙的驱逐启发式，未能随着对话主题的演变而适应。我们提出了分数衰减KV缓存（Fractional Decay KV-Cache，FD-KVC），这是一种新颖的算法，为每个缓存的KV对维护双通道评分机制：一个累积注意力通道跟踪整体重要性（类似于H2O），另一个由时间衰减和强化学习启发的更新驱动的近期加权相关性通道。两者结合使FD-KVC能够既保留历史上重要的标记，又能在对话主题变化时迅速适应。由所有权损失函数驱动的自适应学习率确保收敛而不产生振荡。FD-KVC完全在CPU上运行，开销微乎其微。在五个不同的多轮对话场景中，每个场景有600个对话，FD-KVC在复合后期对齐上比最先进的重击基线H2O提高了6.7%，在主题转变上提高了127%，在渐进演变上提高了87%，在混合主题对话上提高了30%。FD-KVC在适应新主题方面比H2O快3.6倍，并在所有方法中实现了最高的主题多样性（80.6%）。消融研究确认了每个组件的贡献。
cs.CL / 15 / 2608.18100
Computational Orientalism: Measuring Structural Discourse Bias in Large Language Models Using the Middle East Cultural Sensitivity Score (MECSS)
计算东方主义：使用中东文化敏感度评分（MECSS）测量大型语言模型中的结构性话语偏见
Abstract
AI systems now shape how hundreds of millions of people learn about cultures other than their own. When someone asks one of these systems about the Middle East, they do not receive neutral facts. They receive a representation shaped by the frameworks embedded in training data, and that data is overwhelmingly Western and English-language. This paper asks whether that representation is Orientalist in Said's sense: whether it denies agency to Middle Eastern actors, treats Western frameworks as neutral while marking non-Western knowledge as particular, and explains the region through categories it did not produce. Standard fairness metrics cannot answer this, because they detect explicit prejudice rather than structural framing. This paper introduces the Middle East Cultural Sensitivity Score (MECSS), a framework that turns Said's seven Orientalist operations into measurable dimensions, and the term "Said-washing" for a specific failure: a model that disclaims generalization, then reproduces the structure it disclaimed. Across 280 conversations (1,120 exchanges), GPT-4 and Falcon3-7B-Instruct both reproduce Orientalist patterns systematically, through structural positioning rather than open stereotyping. GPT-4 scores moderately (mean MECSS 1.73); Falcon3-7B-Instruct scores higher (2.18), even though it was built in Abu Dhabi and trained with Arabic content. This is evidence against the assumption that building a model regionally makes it less Orientalist, though the models differ in size as well as origin, so geography cannot be isolated as the cause. Epistemic Center, the treatment of Western frameworks as unmarked universals, scores near the top of the scale for both models. Said-washing appears in 87.9% of GPT-4 conversations, a pattern existing metrics cannot see. Reducing this bias requires changing what models learn from, not only adding languages or relocating institutions.
Chinese Translation
人工智能系统如今影响着数亿人了解自身以外文化的方式。当有人向这些系统询问中东时，他们并未获得中立的事实，而是获得了受到训练数据中嵌入框架影响的表述，而这些数据主要是西方和英语的。本文探讨这种表述是否符合赛义德所说的东方主义：是否否认中东行为者的能动性，将西方框架视为中立，同时将非西方知识标记为特殊，并通过其未产生的类别来解释该地区。标准的公平性指标无法回答这个问题，因为它们检测的是显性偏见而非结构性框架。本文引入中东文化敏感度评分（MECSS），一个将赛义德的七个东方主义操作转化为可测量维度的框架，并提出“赛义德洗白”（Said-washing）这一术语，用于描述一种特定的失败：一个否认普遍化的模型，随后又重现了其所否认的结构。在280次对话（1,120次交流）中，GPT-4和Falcon3-7B-Instruct都通过结构性定位而非公开刻板印象系统性地重现了东方主义模式。GPT-4的得分适中（平均MECSS为1.73）；而Falcon3-7B-Instruct的得分更高（2.18），尽管它是在阿布扎比构建并使用阿拉伯内容进行训练的。这反驳了构建区域模型会使其更少东方主义的假设，尽管这些模型在规模和来源上有所不同，因此地理因素无法被孤立为原因。知识中心（Epistemic Center），即将西方框架视为未标记的普遍性，两个模型的得分均接近于该量表的顶端。赛义德洗白在87.9%的GPT-4对话中出现，这一模式是现有指标无法察觉的。减少这种偏见需要改变模型学习的内容，而不仅仅是增加语言或重新定位机构。
cs.CL / 16 / 2608.18101
BERTilda: Explainable Topic Lifecycle Tracking with Split/Merge Detection via Similarity-and-Flow Temporal Graphs
BERTilda：基于相似性和流动的时间图进行可解释的主题生命周期追踪与分合检测
Abstract
Longitudinal text streams exhibit topic birth and death, but also discrete structural reorganizations in which themes split into subtopics or merge into broader narratives. Many dynamic topic models emphasize smooth drift, while snapshot topic models (fit independently per time window) leave temporal correspondence underspecified. We present BERTilda, an explainable framework that discovers topics independently in each window (using an embedding-based topic model) and then constructs a temporal topic graph linking topics across adjacent windows. Links are supported by two complementary signals: (i) semantic similarity between topic representations and (ii) a bidirectional coverage signal that estimates document outflow (where a topic goes) and inflow (where a topic comes from) via cross-window tweet-to-topic attribution. Graph-based rules label continuations, splits, merges, disappearances, and unclear transitions. We evaluate BERTilda on political corpora, including U.S. congressional tweets and historical speech datasets, report topic-quality and temporal-stability diagnostics, and validate lifecycle labels on a gold-standard subset annotated by three independent annotators. On the annotated subset, BERTilda reaches majority agreement rates up to 87% and attains the highest macro-average agreement across the compared methods, with particularly strong disappearance detection relative to similarity-only and forward-only baselines.
Chinese Translation
纵向文本流展示了主题的诞生与消亡，同时也表现出主题在结构上的离散重组，即主题分裂为子主题或合并为更广泛的叙事。许多动态主题模型强调平滑的漂移，而快照主题模型（在每个时间窗口独立拟合）则未明确时间对应关系。我们提出了BERTilda，一个可解释的框架，该框架在每个窗口独立发现主题（使用基于嵌入的主题模型），然后构建一个时间主题图，连接相邻窗口中的主题。连接由两个互补信号支持：（i）主题表示之间的语义相似性和（ii）一个双向覆盖信号，通过跨窗口的推文到主题归属估计文档流出（主题去向）和流入（主题来源）。基于图的规则标记延续、分裂、合并、消失和不清晰的过渡。我们在政治语料库上评估BERTilda，包括美国国会推文和历史演讲数据集，报告主题质量和时间稳定性诊断，并在由三位独立注释者标注的黄金标准子集上验证生命周期标签。在标注的子集上，BERTilda达到了高达87%的多数一致率，并在比较方法中获得了最高的宏平均一致性，特别是在相较于仅基于相似性和仅向前的基线方法时，其消失检测表现尤为强劲。
cs.CL / 17 / 2608.18102
Stability-Aware Feature Design for Robust Watermark Detection in Machine-Generated Text
考虑稳定性的特征设计用于机器生成文本中的鲁棒水印检测
Abstract
The widespread adoption of large language models (LLMs) has intensified the demand for principled methods to distinguish human from machine-generated text. Watermarking provides a promising avenue, yet existing detectors exhibit sharp performance deterioration under multiple paraphrasing and when applied to shorter texts. We introduce Pattern Stability Score (PSS), a novel detection framework that leverages local statistical features and stability dynamics across paraphrased variants. Specifically, the proposed method combines global and local z-score features with higher-order statistics of run-length patterns, enriched by autocorrelation signals and stability scores computed over paraphrase depth. Numerical evaluations are performed on three benchmark datasets (PG-19, CNN/DailyMail, and WikiText) using multiple LLMs (Llama-3-8B, Qwen2-7B) and paraphrasers (Mistral-7B, Qwen2-7B, Gemma-7B), systematically stress-testing robustness under up to eight rounds of paraphrasing. Compared to prior z-score thresholding baselines and some state-of-the-art deep learning methods, our approach improves detection AUC (area under the receiver operating characteristic curve) by over 10-15 percentage points across different token lengths. Additionally, extensive cross-domain experiments demonstrate that a single universal classifier generalizes across different LLMs, paraphrasers, and text domains without retraining, maintaining above 87.8% AUC even when all components differ from training.
Chinese Translation
大型语言模型（LLMs）的广泛应用加剧了区分人类与机器生成文本的需求。水印技术提供了一条有前景的途径，但现有的检测器在多次改写和应用于较短文本时表现出明显的性能下降。我们提出了模式稳定性评分（Pattern Stability Score, PSS），这是一种新颖的检测框架，利用局部统计特征和改写变体之间的稳定性动态。具体而言，所提出的方法将全局和局部 z-score 特征与运行长度模式的高阶统计相结合，并通过自相关信号和在改写深度上计算的稳定性评分进行丰富。我们在三个基准数据集（PG-19、CNN/DailyMail 和 WikiText）上进行了数值评估，使用了多种 LLM（Llama-3-8B、Qwen2-7B）和改写器（Mistral-7B、Qwen2-7B、Gemma-7B），系统地对鲁棒性进行了压力测试，经过多达八轮的改写。与之前的 z-score 阈值基线和一些最先进的深度学习方法相比，我们的方法在不同的标记长度上提高了检测 AUC（接收者操作特征曲线下面积）超过 10-15 个百分点。此外，大量跨领域实验表明，单一的通用分类器在不同的 LLM、改写器和文本领域中无需重新训练即可实现泛化，即使在所有组件与训练时不同的情况下，AUC 仍保持在 87.8% 以上。
cs.CL / 18 / 2608.18103
DeepTCM1.0: A Multi-Expert AI Agent for Deciphering Mechanisms of Chinese Herbal Formulae Based on General Large Language Models
DeepTCM1.0：基于通用大语言模型的多专家人工智能代理，用于解读中药方剂机制
Abstract
Background: Mechanistic elucidation of traditional Chinese medicine (TCM) compound formulas remains a central challenge in the modernization of TCM. Conventional approaches, including data mining and network pharmacology, are insufficient for achieving deep integration between classical TCM theory and modern scientific research. In addition, direct question-answering using general-purpose artificial intelligence large language models is limited by inadequate adaptation to TCM theoretical frameworks and susceptibility to reasoning hallucinations. Consequently, there is an urgent need to develop intelligent analytical methods aligned with the holistic principles of TCM. Objective: To establish a multi-expert intelligent agent framework integrating classical TCM theory with modern life sciences, thereby enabling systematic and interpretable mechanistic analysis of TCM compound formulas, with Guizhi Decoction serving as a representative validation case. Methods: The DeepTCM1.0 framework was constructed based on the general-purpose large language model DeepSeek V3.2. It adopts a three-tier collaborative architecture and a three-round iterative quality-control workflow, simulating the collaborative analytical process of 11 interdisciplinary intelligent agents. The framework was applied to the mechanistic interpretation of Guizhi Decoction from the dual perspectives of classical traditional Chinese medicine theory and modern scientific research. Framework performance was comprehensively evaluated through double-blind five-dimensional scoring, intraclass correlation coefficient (ICC) reliability testing, Mann-Whitney U tests, and effect size analysis. The evaluation employed four independent large language models as evaluators, each conducting five rounds of repeated scoring on five anonymized reports, resulting in a total of 100 independent scoring assessments.
Chinese Translation
背景：传统中医（TCM）复方的机制阐明仍然是中医现代化过程中的一个核心挑战。传统方法，包括数据挖掘和网络药理学，无法实现经典中医理论与现代科学研究之间的深度融合。此外，使用通用人工智能大语言模型进行直接问答受到对中医理论框架适应性不足和推理幻觉的影响。因此，迫切需要开发与中医整体原则相一致的智能分析方法。目标：建立一个多专家智能代理框架，将经典中医理论与现代生命科学相结合，从而实现对中医复方的系统性和可解释的机制分析，以桂枝汤为代表验证案例。方法：DeepTCM1.0框架基于通用大语言模型DeepSeek V3.2构建，采用三层协作架构和三轮迭代质量控制工作流程，模拟11个跨学科智能代理的协作分析过程。该框架应用于从经典中医理论和现代科学研究的双重视角对桂枝汤的机制解释。通过双盲五维评分、组内相关系数（ICC）可靠性测试、Mann-Whitney U检验和效应量分析对框架性能进行了全面评估。评估采用四个独立的大语言模型作为评估者，每个评估者对五份匿名报告进行了五轮重复评分，共计100次独立评分评估。
cs.CL / 19 / 2608.18105
StocksTalk: A Voice-Enabled Conversational Agent for Structured Query Generation over Web Data
StocksTalk：一种用于基于网络数据生成结构化查询的语音启用对话代理
Abstract
StocksTalk is a voice-enabled conversational system for transforming spoken financial screening requests into executable and validated structured queries over real-world market data. The system combines streaming speech recognition, retrieval-augmented constraint extraction, schema-grounded LLM-based SQL generation, rule-based validation, and human-in-the-loop verification within an interactive dashboard. Unlike traditional template-driven financial assistants, StocksTalk exposes intermediate reasoning artifacts, including extracted constraints, normalized financial metrics, operator grounding, and generated queries, allowing users to inspect and refine each stage before execution. To evaluate the system, we curate a benchmark of 150 spoken financial prompts spanning multiple investment strategies and input noise conditions. Experimental results show that retrieval grounding, constrained query generation, and interactive verification substantially improve constraint extraction accuracy, SQL executability, logical consistency, and multi-turn stability compared to baseline LLM-based approaches. StocksTalk demonstrates how transparent, voice-driven interfaces can bridge natural language interaction and structured financial analysis, providing an effective framework for conversational stock screening and decision support.
Chinese Translation
StocksTalk 是一个语音启用的对话系统，旨在将口头的金融筛选请求转化为可执行且经过验证的结构化查询，以处理现实世界的市场数据。该系统结合了流式语音识别、增强检索的约束提取、基于模式的 LLM（大语言模型）SQL 生成、基于规则的验证以及人机协作的验证，构建在一个交互式仪表板上。与传统的模板驱动金融助手不同，StocksTalk 显示了中间推理结果，包括提取的约束、标准化的金融指标、操作符的基础和生成的查询，允许用户在执行之前检查和优化每个阶段。为了评估该系统，我们整理了一个包含 150 个口头金融提示的基准数据集，涵盖多种投资策略和输入噪声条件。实验结果表明，与基线 LLM 方法相比，检索基础、约束查询生成和交互验证显著提高了约束提取的准确性、SQL 可执行性、逻辑一致性和多轮稳定性。StocksTalk 展示了透明的、基于语音的界面如何能够桥接自然语言交互与结构化金融分析，为对话式股票筛选和决策支持提供了有效的框架。
cs.CL / 20 / 2608.18106
Different Facets of Verbalised Overconfidence: an Interpretability Study
言语化过度自信的不同面向：可解释性研究
Abstract
Large language models tend to overconfidence, giving assertive answers when the evidence suggests hedging or abstention. Using controlled reasoning scenarios that manipulate logical necessity and possibility, we study this behavior in Qwen3-4B, across three ways to express uncertainty: verbal epistemic markers, abstention, and numeric confidence scores. Our results confirm this tendency toward overconfidence, particularly when the model is prompted to output a numeric confidence score. At the interpretability level, we propose a method that differentially identifies transcoder features responsible for uncertainty and certainty. Our analysis reveals Qwen3-4B's default mechanism favors certainty generation through a broad coalition of shared features, while uncertainty is implemented as a sparse override mediated by a small set of dedicated features. Intervening on these uncertainty features both causally proves this imbalance underlying overconfidence and also mitigate overconfident errors. The same set of features generalise across the three uncertainty-expression settings, languages, and an out-of-distribution modality task.
Chinese Translation
大型语言模型往往表现出过度自信的倾向，在证据表明应当谨慎或保持中立时，仍给出断言性的回答。通过控制推理场景，操控逻辑必要性和可能性，我们研究了Qwen3-4B模型在三种表达不确定性方式下的这种行为：言语认知标记、保持中立和数值置信分数。我们的结果确认了这种过度自信的倾向，尤其是在模型被提示输出数值置信分数时。在可解释性层面，我们提出了一种方法，能够区分负责不确定性和确定性的转码特征。我们的分析揭示了Qwen3-4B的默认机制倾向于通过广泛共享特征的联盟生成确定性，而不确定性则作为稀疏的覆盖，由一小组专用特征介导。干预这些不确定性特征不仅因果证明了导致过度自信的不平衡，还能减轻过度自信的错误。同一组特征在三种不确定性表达设置、语言和一个超出分布的模态任务中具有普遍性。
cs.CL / 21 / 2608.18107
Institutional Prestige as Geographic Bias in Large Language Models: Evidence from Three Factorial Experiments with Bootstrap Confidence Intervals
机构声望作为大型语言模型中的地理偏见：来自三项带有自助法置信区间的因子实验的证据
Abstract
We investigate whether large language models (LLMs) systematically discriminate in candidate evaluations based on applicant name ethnicity and/or institutional prestige and geographic location. Three factorial experiments are reported (4,320 API calls, four LLMs, five professional domains). Study 1 (3x4 design) finds a statistically robust institution-tier gradient of +0.297 points on a 10-point scale (95% bootstrap CI: +0.175 to +0.422), while name-origin effects are negligible and non-significant (95% CI crosses zero). Study 2 (2x2 Prestige x Country design) breaks the prestige-geography confound: the prestige effect (+0.185; 95% CI: +0.093 to +0.275) exceeds the country-of-origin effect (+0.126; 95% CI: +0.037 to +0.218) by 1.5x. Study 3 (2x2 Journal x Institution design) reveals that journal prestige (Nature vs. a peripheral open-access journal) dominates institutional prestige by 5.7x: journal effect +1.937 (95% CI: +1.811 to +2.062) vs. institution effect +0.341 (95% CI: +0.184 to +0.504). A "rescue effect" is confirmed: publishing in Nature compensates for low institutional prestige more strongly for candidates from the University of Guayaquil (+2.127) than from MIT (+1.745). Results are quantified using the Neutrosophic Bias Index NBI; the I component reveals elevated evaluation inconsistency for low-prestige profiles, an epistemic disadvantage not captured by mean-only metrics. Code and data: https://github.com/mleyvaz/geo-bias-llm
Chinese Translation
我们研究大型语言模型（LLMs）是否在候选人评估中系统性地基于申请人姓名的民族、机构声望和地理位置进行歧视。报告了三项因子实验（4,320次API调用，四个LLM，五个专业领域）。研究1（3x4设计）发现机构层级梯度在10分制上为+0.297分（95%自助法置信区间：+0.175至+0.422），而姓名来源效应微不足道且不显著（95%置信区间跨越零）。研究2（2x2声望x国家设计）打破了声望与地理的混淆：声望效应（+0.185；95%置信区间：+0.093至+0.275）超出了原产国效应（+0.126；95%置信区间：+0.037至+0.218）1.5倍。研究3（2x2期刊x机构设计）揭示期刊声望（《自然》与一个边缘开放获取期刊）在5.7倍上主导机构声望：期刊效应+1.937（95%置信区间：+1.811至+2.062）对比机构效应+0.341（95%置信区间：+0.184至+0.504）。确认了“救助效应”：在《自然》上发表的文章对来自瓜亚基尔大学的候选人（+2.127）的低机构声望补偿效果比对麻省理工学院（+1.745）更强。结果使用中立偏见指数NBI进行量化；I成分揭示低声望档案的评估不一致性升高，这是一种仅通过均值指标无法捕捉的认知劣势。代码和数据： https://github.com/mleyvaz/geo-bias-llm
cs.CL / 22 / 2608.18108
Same Facts, Different Updates: Inference Setup Shapes LLM Behavior in Medical Allocation
相同事实，不同更新：推理设置影响大型语言模型在医疗资源分配中的行为
Abstract
Large language models are being incorporated into sensitive and important decision-making processes across nearly all fields. While prior work studies model bias around inputs and scenario framing, models can also behave in unexpected and undesirable ways due to context accumulated over their deployment. In this work, we study a medical example in which a model is asked to assign resource-allocation probabilities to two people given brief clinical context, and then sees the same scenario with a single extra sentence containing contrasting patient information, either with or without its previous response in context. Across three of four tested models, the paired-context and independent-inference experiments have different probability shifts, often in opposite directions (in favor of Person B vs. in favor of Person A) when new information is provided. We include additional paired-context experiments to show the effect of varying attributes across scenario axes. Our findings show the context-dependent effect of patient information in a sensitive medical use case. More broadly, our work shows the importance of carefully incorporating LLM-based systems into decision-making processes, context engineering, and further model behavioral studies.
Chinese Translation
大型语言模型正被纳入几乎所有领域的敏感和重要决策过程中。虽然先前的研究关注模型在输入和情境框架方面的偏见，但模型也可能因其部署过程中积累的上下文而表现出意想不到和不理想的行为。在本研究中，我们研究了一个医疗案例，其中模型被要求根据简要的临床背景为两个人分配资源分配概率，然后在看到同一情境时，添加了一句包含对比患者信息的额外句子，可能包含或不包含其先前的响应。在测试的四个模型中，有三个模型的配对上下文和独立推理实验在提供新信息时表现出不同的概率变化，往往朝相反的方向（支持B人 vs. 支持A人）。我们还包括了额外的配对上下文实验，以展示在情境轴上变化属性的影响。我们的发现显示了患者信息在敏感医疗应用中的上下文依赖效应。更广泛地说，我们的研究表明，谨慎地将基于大型语言模型的系统纳入决策过程、上下文工程和进一步的模型行为研究的重要性。
cs.CL / 23 / 2608.18109
Operationalizing Narrative Entropy (Sn): A Two-Scene Registered Pilot Report and Pre-Validation Protocol
叙事熵（$S_n$）的操作化：两场景注册试点报告及预验证协议
Abstract
Narrative Entropy ($S_n$) is a proposed quantitative descriptor within the Bulut Doctrine, intended to capture the rate at which a narrative text imposes processing load on a reader. To date the construct has been defined theoretically but not operationalized against real texts. This report documents the first such operationalization (the v2.0 pilot): two narrative scenes -- the opening restaurant scene of Tarantino's Reservoir Dogs and the opening interior-monologue block of Carver's Cathedral -- were coded manually by a single rater and scored with the candidate formula $S_n = I_f \times C_b \times t$. The result was a divergence from the author's naive intuition: the single-voice monologue ($S_n = 30.0$) scored higher than the nine-character dialogue scene ($S_n = 18.8$). We treat this not as a result to be explained away but as the central finding, and we refuse post-hoc adjustment of the formula. Three competing interpretations are presented -- formula incompleteness, genuine high-load prose, and measurement error -- and the design that would discriminate among them is pre-registered. This v2.1 revision adds: (i) explicit acknowledgement that the divergence is consistent with the pre-existing architectural framework which privileges inferential reconstruction over surface declaration, and that what was called "contrary to expectation" in v2.0 reflected the author's anticipatory intuition rather than the methodology's own predictions; (ii) a pre-registered construct validity test for $I_f$, motivated by the observation that $I_f$ values were nearly equal across the two scenes (1.71 vs 1.58) despite the headline $S_n$ divergence. The document functions simultaneously as a pilot report ($n=2$) and as a pre-registration of the next-stage protocol. It does not claim that $S_n$ has been validated.
Chinese Translation
叙事熵（$S_n$）是根据Bulut理论提出的一种定量描述符，旨在捕捉叙事文本对读者施加处理负荷的速率。迄今为止，该构念已在理论上被定义，但尚未针对真实文本进行操作化。本报告记录了首次操作化的过程（v2.0试点）：两个叙事场景——昆汀·塔伦蒂诺的《落水狗》的开场餐厅场景和雷蒙德·卡佛的《大教堂》的开场内心独白块——由单一评分者手动编码，并使用候选公式 $S_n = I_f imes C_b imes t$ 进行评分。结果与作者的直觉存在偏差：单声部独白（$S_n = 30.0$）的得分高于九人对话场景（$S_n = 18.8$）。我们将此视为一个中心发现，而非需要解释的结果，并拒绝对公式进行事后调整。提出了三种竞争性解释——公式不完整、真正的高负荷散文和测量误差——并对能够区分它们的设计进行了预注册。此v2.1修订版增加了：（i）明确承认该偏差与优先推理重建而非表面声明的预先存在的架构框架一致，并且在v2.0中被称为“与预期相悖”的内容反映了作者的预期直觉，而非方法论自身的预测；（ii）针对 $I_f$ 的预注册构念有效性测试，基于观察到的两个场景的 $I_f$ 值几乎相等（1.71 vs 1.58），尽管头条 $S_n$ 存在偏差。该文档同时作为试点报告（$n=2$）和下一阶段协议的预注册。它并不声称 $S_n$ 已被验证。
cs.CL / 24 / 2608.18114
Accurate Decoding of Natural Sentences from Non-Invasive Brain Recordings
从非侵入性脑记录中准确解码自然句子
Abstract
Restoring communication for people who have lost the ability to speak or move after a brain injury is a major challenge. While intracranial implants now enable high-performing brain-computer-interfaces, non-invasive alternatives are still lagging behind. Here, we present Brain2Qwerty v2, a model that can decode the production of natural sentences solely from real-time magnetoencephalography (MEG) recordings. By collecting 22,000 sentences typed by nine subjects, each recorded for 10 hours, our model leverages character, word and sentence-level representations to achieve an average word error rate (WER) of 39%. For our best participant, the model accurately decodes half of the sentences with one word error or less. Critically, decoding accuracy log-linearly improves with data volume, suggesting that the performance gap with intracranial approaches could be partially bridged through data scaling. We show that AI enables this performance in three main ways: the substitution of hand-crafted pipelines for event detection with deep learning, the finetuning of large language models to extract semantic representations, and the deployment of AI agents to iteratively refine our decoding pipeline via automated code development. Together, these results show that non-invasive brain-to-text decoding starts to operate at a level of accuracy previously thought exclusive to surgical implants, opening a path toward safe and efficient brain-computer-interfaces.
Chinese Translation
恢复因脑损伤而失去说话或移动能力的人的交流能力是一项重大挑战。尽管颅内植入物现在能够实现高性能的脑-计算机接口，但非侵入性替代方案仍然滞后。在这里，我们提出了Brain2Qwerty v2，一个可以仅通过实时脑磁图（MEG）记录解码自然句子生成的模型。通过收集九名受试者输入的22,000个句子，每个受试者记录10小时，我们的模型利用字符、单词和句子级别的表示，达到了39%的平均单词错误率（WER）。对于我们表现最好的参与者，该模型能够准确解码一半的句子，错误词数不超过一个。关键是，解码准确率随着数据量的增加呈对数线性改善，这表明通过数据扩展可以部分弥补与颅内方法之间的性能差距。我们展示了人工智能通过三种主要方式实现这一性能：用深度学习替代手工制作的事件检测管道，微调大型语言模型以提取语义表示，以及部署人工智能代理通过自动化代码开发迭代优化我们的解码管道。综合来看，这些结果表明，非侵入性脑-文本解码开始达到以前被认为仅限于外科植入物的准确水平，为安全高效的脑-计算机接口开辟了道路。
cs.CL / 25 / 2608.18115
Temporal Multi-Signal Fusion for Token-Level Hallucination Detection
基于时间的多信号融合用于令牌级幻觉检测
Abstract
Token-level hallucination detectors score each token independently from a single signal, and fail exactly when the generating model is confidently wrong. This paper instead treats hallucination as a temporally extended span and detects it by sequence labeling: each token is scored from a 33-dimensional feature stream that fuses text statistics, Natural Language Inference (NLI) entailment, and language model surprisal, with no access to model internals. A Bidirectional Gated Recurrent Unit (BiGRU) over these features reaches an AUC of 0.840 on RAGTruth (10 seeds), an 11-point gain over an independent logistic-regression baseline (p = 0.002, Wilcoxon signed-rank). A controlled decomposition attributes most of the gain to temporal order rather than model capacity: evidence propagates from confident positions to ambiguous neighbors within a span. The same 0.845 ceiling recurs across recurrent, state-space (Mamba), and attention architectures, locating the bottleneck in the feature set rather than the model. Because it reads only the generated text and external signals, the detector works on closed-source models, and it keeps working on text produced by language models it never saw during training, losing under 4% AUC.
Chinese Translation
令牌级幻觉检测器独立地对每个令牌进行评分，基于单一信号，当生成模型自信地错误时恰恰失效。本文将幻觉视为一个时间扩展的范围，并通过序列标注进行检测：每个令牌的评分来自一个融合了文本统计、自然语言推理（NLI）蕴含和语言模型惊讶度的33维特征流，而不接触模型内部。对这些特征应用双向门控递归单元（BiGRU），在RAGTruth（10个种子）上达到了0.840的AUC，相较于独立的逻辑回归基线提升了11点（p = 0.002，Wilcoxon符号秩检验）。受控分解将大部分增益归因于时间顺序而非模型能力：证据从自信的位置传播到范围内的模糊邻居。相同的0.845上限在递归、状态空间（Mamba）和注意力架构中反复出现，瓶颈位于特征集而非模型。由于它仅读取生成的文本和外部信号，该检测器可以在闭源模型上工作，并且在训练期间未见过的语言模型生成的文本上仍然有效，AUC损失不足4%。
cs.CL / 26 / 2608.18116
You Are What You Prompt: Prompt Quality, Domain Shift, and Uncertainty in Agrifood Vision-Language Models
你就是你的提示：提示质量、领域转变与农业食品视觉语言模型中的不确定性
Abstract
Vision-language models enable zero-shot classification through natural language prompts, but performance is sensitive to prompt formulation, especially in specialized domains. Zero-shot Prompt Ensembling (ZPE) addresses this by weighting prompts by discriminative signal, yet its behavior under domain shift remains unexplored. We evaluate ZPE in the agrifood domain using CLIP and SigLIP across four datasets and four prompt pools, spanning in-distribution (ID) food and out-of-distribution agricultural benchmarks. ZPE provides limited benefit under ID conditions but substantially improves performance and calibration under domain shift, where domain-specific pools of 51-52 prompts consistently outperform generic pools of 247-426. Lexical analysis shows that ZPE acts as an unsupervised domain-alignment detector without label access. We further introduce PID (Prompt-based Inconsistency Detection), which repurposes prompt disagreement as epistemic uncertainty, improving failure detection under severe domain shift where standard confidence measures collapse.
Chinese Translation
视觉语言模型通过自然语言提示实现零-shot 分类，但其性能对提示的构造非常敏感，尤其是在专业领域。零-shot 提示集成（Zero-shot Prompt Ensembling, ZPE）通过根据判别信号加权提示来解决这一问题，但其在领域转变下的表现尚未得到探索。我们在农业食品领域使用 CLIP 和 SigLIP 对四个数据集和四个提示池进行 ZPE 评估，涵盖了分布内（in-distribution, ID）食品和分布外农业基准。ZPE 在 ID 条件下提供的好处有限，但在领域转变下显著提高了性能和校准，其中领域特定的 51-52 个提示池始终优于通用的 247-426 个提示池。词汇分析表明，ZPE 作为一种无监督的领域对齐检测器，在没有标签访问的情况下发挥作用。我们进一步引入了 PID（基于提示的不一致性检测），将提示不一致性重新利用为认知不确定性，从而改善在严重领域转变下的故障检测，此时标准的置信度度量会崩溃。
cs.CL / 27 / 2608.18132
Alignment Is All You Need: Instruction-Free Training for General Audio-Language Models
对齐就是一切：无指令训练的通用音频语言模型
Abstract
Multimodal large language models (MLLMs) are typically built through a multi-stage pipeline consisting of cross-modal alignment, supervised fine-tuning (SFT), and preference optimization. This pipeline assumes that adapting an LLM to a new modality requires extensive task-specific supervision. However, pretrained LLMs already possess strong reasoning and instruction-following abilities. As LLMs evolve rapidly, an important question remains: can we efficiently transfer these capabilities to a new modality with minimal intervention, and is alignment alone sufficient for building a multimodal model? We introduce an Instruction-Free Alignment-Only large audio-language model (LALM) that keeps both the audio encoder and the LLM fully frozen, learning only a lightweight projector. Borrowing insights from AzeroS [1], we train on (audio, response) pairs from Self-Generated Data Construction, where an LLM expands captions into free-form responses without explicit task instructions. Across MMAU, MMAR, MMSU, and MMAU-Pro, our approach matches or surpasses heavily post-trained baselines using substantially less data. By keeping the LLM frozen, our model preserves its native instruction-following competence and can port seamlessly across model generations. Our results suggest that competitive MLLM can emerge from alignment alone, reducing multimodal extension to a lightweight projector-training problem that generalizes across modalities and adapts rapidly to each new LLM release.
Chinese Translation
多模态大型语言模型（MLLMs）通常通过一个多阶段管道构建，该管道包括跨模态对齐、监督微调（SFT）和偏好优化。该管道假设将大型语言模型（LLM）适配到新模态需要广泛的任务特定监督。然而，预训练的LLM已经具备强大的推理和遵循指令的能力。随着LLM的快速发展，一个重要的问题仍然存在：我们能否以最小的干预有效地将这些能力转移到新模态，并且仅靠对齐是否足以构建多模态模型？我们提出了一种无指令对齐的大型音频语言模型（LALM），该模型将音频编码器和LLM完全冻结，仅学习一个轻量级的投影器。借鉴AzeroS [1] 的见解，我们在自生成数据构建中对（音频，响应）对进行训练，其中LLM在没有明确任务指令的情况下将标题扩展为自由形式的响应。在MMAU、MMAR、MMSU和MMAU-Pro的实验中，我们的方法在使用显著更少的数据的情况下，与经过大量后训练的基线模型相匹配或超越。通过保持LLM的冻结，我们的模型保留了其原生的遵循指令能力，并能够在模型世代之间无缝迁移。我们的结果表明，竞争性的多模态大型语言模型可以仅通过对齐产生，将多模态扩展简化为一个轻量级的投影器训练问题，该问题能够跨模态泛化并快速适应每个新的LLM版本。
cs.CL / 28 / 2608.18138
Language Models for Portuguese: A Systematic Mapping Study
葡萄牙语的语言模型：一项系统性映射研究
Abstract
In recent years, the rapid development of language models has transformed the field of Natural Language Processing through a wide range of applications. However, the development of language models has not progressed uniformly across all languages. In the case of the Portuguese language, there has recently been a growing effort by academia and companies to develop language models and create data resources for Portuguese. These efforts have resulted in the rise of an increasingly diverse ecosystem of language models for Portuguese. However, information on these models remains dispersed in scientific publications, technical reports, model repositories, and project documentation. This survey presents a systematic mapping study of language models developed for Portuguese, providing a comprehensive overview of the current state of the field. We map a total of 46 models, characterizing them by various aspects, including base model, architecture, computational resources, training datasets, licensing, code availability, data, and model weights. Furthermore, we analyzed the evolution and relationships among these models through a phylogenetic perspective, identified current research gaps and opportunities, and discussed future directions for the development of language models for Portuguese.
Chinese Translation
近年来，语言模型的快速发展通过广泛的应用改变了自然语言处理领域。然而，语言模型的发展在不同语言之间并不均匀。在葡萄牙语的情况下，学术界和企业最近在开发语言模型和创建葡萄牙语数据资源方面付出了越来越多的努力。这些努力导致了一个日益多样化的葡萄牙语语言模型生态系统的崛起。然而，关于这些模型的信息仍然分散在科学出版物、技术报告、模型库和项目文档中。本调查呈现了一项针对葡萄牙语开发的语言模型的系统性映射研究，提供了该领域当前状态的全面概述。我们映射了总共46个模型，从基础模型、架构、计算资源、训练数据集、许可、代码可用性、数据和模型权重等多个方面对其进行了特征描述。此外，我们通过系统发育的视角分析了这些模型之间的演变和关系，识别了当前的研究空白和机会，并讨论了葡萄牙语语言模型发展的未来方向。
cs.CL / 29 / 2608.18144
The Deontic Gap: Large Language Models and the Modal Language of Obligation
义务差距：大型语言模型与义务的模态语言
Abstract
Modal auxiliaries such as must, should, and have to mark necessity and obligation within the contexts of speaker authority and interpersonal stance. We examine whether large language models (LLMs) reproduce contemporary human patterns of deontic modal usage. Across three primary corpora, an external benchmark, two controlled replications, and a naturalistic eleven-model replication, AI-generated text consistently underuses positive deontic modals (must, should, have to, had to) relative to contemporary humans. Historical comparison with the Google Books Ngram corpus (1920-2022), used as a heuristic calibration against the published-prose record, shows that AI modal frequencies fall within the range of formal published English, whereas contemporary human modal rates in informal digital contexts often exceed twentieth-century book baselines. Phrase-level decomposition shows that the AI-human modal gap is concentrated in constructions central to interpersonal stance (should, have to, had to), while AI matches or exceeds humans on need to in instructional and question-answering contexts but not in persuasive student writing, indicating that the modal profile is genre-conditional. The findings suggest that LLM modal usage reflects the formal written resources on which these models were trained, while underusing the modal constructions through which contemporary human writers mark immediate, interpersonal obligation.
Chinese Translation
模态助动词如 must、should 和 have to 在说话者权威和人际立场的背景下标记必要性和义务。我们考察大型语言模型（LLMs）是否重现了当代人类的义务模态使用模式。在三个主要语料库、一个外部基准、两个受控复制实验和一个自然主义的十一模型复制中，AI生成的文本相对于当代人类始终低估了积极的义务模态（must、should、have to、had to）的使用。与 Google Books Ngram 语料库（1920-2022）的历史比较，作为对已出版散文记录的启发式校准，显示 AI 的模态频率落在正式出版英语的范围内，而当代人类在非正式数字环境中的模态使用率往往超过二十世纪书籍的基线。短语级分解显示，AI与人类的模态差距集中在与人际立场相关的构造（should、have to、had to）上，而在指令性和问答情境中，AI 在 need to 的使用上与人类相匹配或超过，但在说服性学生写作中则不然，这表明模态特征是与体裁相关的。研究结果表明，LLM 的模态使用反映了这些模型训练所依据的正式书面资源，同时低估了当代人类作家标记即时人际义务的模态构造。
cs.CL / 30 / 2608.18158
When Do LLMs Actually Help? Evaluating LLMs as Data Quality Annotators
大型语言模型何时真正有帮助？评估大型语言模型作为数据质量标注者的效果
Abstract
LLMs have been increasingly used to catch data quality issues automatically, but we know very little about how consistent these judgments actually are. This study tests an LLM on two e-commerce data quality tasks, entity matching and brand mislabeling, against rule based baselines and human verified ground truth, under both zero-shot and few-shot prompting. On entity matching while using the Abt Buy benchmark (2,194 labeled pairs), a simple rule based baseline (F1=0.950) performed about as well as LLM zero shot prompting (F1=0.948). Moreover, a few-shot prompt revision that looked effective on a small validation sample reduced full-scale performance to F1=0.914. This showed that small sample prompt evaluation can be misleading. On brand mislabeling detection, using 500 Amazon product listings with synthetically injected labeling errors, the LLM clearly outperformed a naive rule based baseline (F1=0.833 vs 0.721), because it could draw on background knowledge of brand product relationships that a simple rule could not access. Testing consistency across repeated runs (200 pairs, 5 runs at temperature 0.7) showed the model agreeing with itself 99.7% of the time on average, with 99% of pairs giving identical answers across all 5 runs. Using majority voting across these runs only improved F1 by 0.005, at 5 times the inference cost. These results suggest that the value of using an LLM over traditional methods depends heavily on the task. LLMs offer little advantage when strong lexical signals already exist, but a clear advantage when the task requires background knowledge, all while remaining highly consistent across repeated queries.
Chinese Translation
大型语言模型（LLMs）越来越多地被用于自动识别数据质量问题，但我们对这些判断的一致性知之甚少。本研究在零样本和少样本提示下，针对两个电子商务数据质量任务——实体匹配和品牌错误标注——对比了LLM与基于规则的基线和人工验证的真实数据。在使用Abt Buy基准（2,194个标注对）进行实体匹配时，一个简单的基于规则的基线（F1=0.950）与LLM的零样本提示（F1=0.948）表现相当。此外，在小规模验证样本上看似有效的少样本提示修订，导致全规模性能下降至F1=0.914。这表明小样本提示评估可能具有误导性。在品牌错误标注检测中，使用500个带有合成注入标注错误的亚马逊产品列表，LLM明显优于一个简单的基于规则的基线（F1=0.833对0.721），因为它能够利用品牌产品关系的背景知识，而简单的规则无法访问。对重复运行（200对，5次运行，温度0.7）的一致性测试显示，模型在平均99.7%的情况下与自身一致，99%的对在所有5次运行中给出了相同的答案。在这些运行中使用多数投票仅将F1提高了0.005，且推理成本增加了5倍。这些结果表明，使用LLM相较于传统方法的价值在很大程度上依赖于具体任务。当强词汇信号已经存在时，LLM提供的优势不大，但当任务需要背景知识时，则明显具有优势，同时在重复查询中保持高度一致。
cs.CL / 31 / 2608.18164
Are LLMs Safe Beyond Text: Do Emojis Expose Gaps in Safety Evaluation
大型语言模型在文本之外是否安全：表情符号是否暴露了安全评估中的漏洞
Abstract
Safety evaluations of large language models (LLMs) predominantly rely on text-based adversarial prompts, potentially overlooking vulnerabilities arising from alternative input representations. This work examines emoji-augmented prompts as a test case for this gap, evaluating 50 prompts across four open-source LLMs (Mistral 7B, Qwen 2 7B, Gemma 2 9B, Llama 3 8B). Results show substantial variation in robustness: Gemma 2 9B and Mistral 7B exhibit non-zero success rates (10%), Llama 3 8B 6%, while Qwen 2 7B shows complete resistance (0% success rate). A chi-square test ($\chi^2 = 32.94, p < 0.001$) confirms significant differences in outcome distributions. These findings indicate that robustness is sensitive to input representation, and that evaluations restricted to standard text prompts may underrepresent model vulnerabilities.
Chinese Translation
大型语言模型（LLMs）的安全评估主要依赖于基于文本的对抗性提示，这可能忽视了来自其他输入表示的脆弱性。本研究将增强表情符号的提示作为这一差距的测试案例，评估了四个开源LLM（Mistral 7B、Qwen 2 7B、Gemma 2 9B、Llama 3 8B）中的50个提示。结果显示鲁棒性存在显著差异：Gemma 2 9B和Mistral 7B的成功率均为非零（10%），Llama 3 8B为6%，而Qwen 2 7B则表现出完全抵抗（成功率为0%）。卡方检验（$ ext{chi}^2 = 32.94, p < 0.001$）确认了结果分布存在显著差异。这些发现表明，鲁棒性对输入表示敏感，而仅限于标准文本提示的评估可能低估了模型的脆弱性。
cs.CL / 32 / 2608.18182
Efficient INT8 Inference of Small NLP Models on Server CPUs with PyTorch Native Stack
基于 PyTorch 原生栈的服务器 CPU 上小型 NLP 模型高效 INT8 推理
Abstract
Small NLP models, especially BERT-family encoders, remain important in industrial workloads such as classification, ranking, and retrieval even in the era of large language models. On server CPUs, INT8 quantization offers an attractive latency-throughput-cost trade-off, but users increasingly expect such acceleration to be available directly in the native PyTorch stack. We integrate SmoothQuant into TorchAO and optimize the resulting inference path for Intel Xeon CPUs through graph-level fusion in TorchInductor and efficient INT8 GEMM kernel selection across oneDNN-, AVX512_VNNI-, and AMX-based implementations. Across BERT, DistilBERT, and XLM-RoBERTa benchmarks, the approach delivers up to 5.8x end-to-end throughput speedup with negligible---and in some cases no measurable---accuracy loss relative to the FP32 baseline. We also validated our work by detailed performance analysis with roofline models. The implementation has been upstreamed to PyTorch and TorchAO, enabling out-of-the-box deployment with native PyTorch tooling
Chinese Translation
小型 NLP 模型，尤其是 BERT 系列编码器，在分类、排名和检索等工业工作负载中仍然具有重要意义，即使在大型语言模型时代。在服务器 CPU 上，INT8 量化提供了一个具有吸引力的延迟-吞吐量-成本权衡，但用户越来越期望这种加速能够直接在原生 PyTorch 栈中实现。我们将 SmoothQuant 集成到 TorchAO 中，并通过在 TorchInductor 中进行图级融合以及在 oneDNN、AVX512_VNNI 和 AMX 基础实现中高效选择 INT8 GEMM 内核，优化了最终的推理路径。在 BERT、DistilBERT 和 XLM-RoBERTa 基准测试中，该方法实现了高达 5.8 倍的端到端吞吐量加速，同时相对于 FP32 基线的准确性损失微乎其微——在某些情况下甚至没有可测量的损失。我们还通过屋顶线模型进行了详细的性能分析来验证我们的工作。该实现已被上游合并到 PyTorch 和 TorchAO 中，支持使用原生 PyTorch 工具进行开箱即用的部署。
cs.CL / 33 / 2608.18312
Artifact-centered Claim-aware Observability for Autonomous Scientific Agents
面向人工制品的声明感知可观测性用于自主科学代理
Abstract
Autonomous scientific agents now increasingly propose ideas, write code, run experiments, analyze results, and even draft papers. Observe and audit those agents are necessary but logging every model call is not enough, scientists also need to inspect the artifacts and claims that the systems produced and their relations. This is driven by the fact that failures in scientific agent systems are often distributed across several objects. A manuscript claim may cite the wrong evidence, a search process may select a degenerate candidate, a laboratory novelty claim may depend on an unstated rule, or a multi-agent plan may change without a visible trigger. Existing tracing, experiment tracking, and archival provenance tools are valuable, but their native objects do not make these scientific audit relations first-class. We argue that autonomous scientific systems should emit portable, claim-aware artifact lineage as a minimum audit layer. We propose a compact observability profile organized around individuals, operators, fitness records, lineage, archives, runs, streams, and steering commands. In this profile, scientific claims are ordinary individuals with explicit evidence bindings and verification records. The profile is intended as a semantic layer that complements current telemetry and provenance standards. Execution details can remain in OpenTelemetry. Final packages can export to PROV-O or RO-Crate standards.
Chinese Translation
自主科学代理现在越来越多地提出想法、编写代码、进行实验、分析结果，甚至撰写论文。观察和审计这些代理是必要的，但仅仅记录每个模型调用是不够的，科学家还需要检查系统生成的人工制品和声明及其关系。这是因为科学代理系统中的故障通常分布在多个对象之间。一份手稿的声明可能引用了错误的证据，一个搜索过程可能选择了一个退化的候选者，一个实验室的新颖性声明可能依赖于一个未说明的规则，或者一个多代理计划可能在没有可见触发的情况下发生变化。现有的追踪、实验跟踪和档案来源工具是有价值的，但它们的原生对象并没有使这些科学审计关系成为一流的。我们认为，自主科学系统应该发出可移植的、声明感知的人工制品血统作为最低审计层。我们提出了一个围绕个体、操作符、适应性记录、血统、档案、运行、流和控制命令组织的紧凑可观测性配置文件。在这个配置文件中，科学声明是具有明确证据绑定和验证记录的普通个体。该配置文件旨在作为一个语义层，补充当前的遥测和来源标准。执行细节可以保留在OpenTelemetry中。最终包可以导出到PROV-O或RO-Crate标准。
cs.CL / 34 / 2608.18361
Figurative and Cultural Knowledge in LLMs: Investigating Cross-Domain Transfer through Fine-Tuning
大型语言模型中的比喻与文化知识：通过微调研究跨领域迁移
Abstract
Figurative language is deeply culturally embedded; fluent use requires not just linguistic competence but cultural immersion. We ask whether LLMs can learn this link: does fine-tuning on cultural data improve figurative language understanding, and vice versa? We conduct a systematic study across four models (ALLaM-7B, Fanar-1-9B, Qwen3-8B, Llama-3.1-8B) and six Arabic datasets spanning cultural commonsense, proverbs, and poetry across diverse dialects and regions. Fine-tuning on poetry improves idiom comprehension (+2.33%, p<0.05), a gain our ArabicMMLU control does not reproduce, indicating that it stems from figurative content rather than Arabic language adaptation and pointing to a sensitivity to non-literal meaning that transfers across figurative types. Cultural fine-tuning, by contrast, lowers proverb-interpretation accuracy in both Arabic-centric models. Transfer between the two domains is otherwise indistinguishable from noise, with Arabic models frequently regressing after fine-tuning, suggesting prior saturation of relevant knowledge, while multilingual models show greater adaptation headroom. Error analysis further reveals that fine-tuning reinforces experiential cultural knowledge while destabilizing historically grounded factual knowledge. Our findings suggest that the relationship between culture and figurative language, though conceptually natural, is not straightforwardly captured through fine-tuning alone.
Chinese Translation
比喻语言深深植根于文化之中；流利的使用不仅需要语言能力，还需要文化的浸润。我们探讨大型语言模型（LLMs）是否能够学习这种联系：在文化数据上进行微调是否能改善比喻语言的理解，反之亦然？我们对四个模型（ALLaM-7B、Fanar-1-9B、Qwen3-8B、Llama-3.1-8B）和六个阿拉伯语数据集进行了系统研究，这些数据集涵盖了文化常识、谚语和来自不同方言和地区的诗歌。对诗歌的微调提高了成语理解能力（+2.33%，p<0.05），而我们的阿拉伯MMLU控制组并未重现这一增益，表明这一提升源于比喻内容而非阿拉伯语言的适应，且指向了对非字面意义的敏感性，这种敏感性在不同的比喻类型之间能够迁移。相比之下，文化微调降低了阿拉伯中心模型的谚语理解准确性。两个领域之间的迁移在其他方面与噪声无异，阿拉伯模型在微调后经常出现退步，表明相关知识的先前饱和，而多语言模型则显示出更大的适应空间。错误分析进一步揭示，微调强化了经验性文化知识，同时削弱了历史基础的事实知识。我们的研究结果表明，尽管文化与比喻语言之间的关系在概念上是自然的，但仅通过微调并不能简单地捕捉这一关系。
cs.CL / 35 / 2608.18437
Tangut Word Segmentation under Extreme Resource Scarcity: Integrating Traditional Lexicons and Unlabeled Text
在极度资源匮乏下的唐古特词语分割：整合传统词典与未标记文本
Abstract
Tangut is an extinct language whose script does not explicitly mark word boundaries. We present the first systematic study of Tangut word segmentation using 2,750 expert-annotated segments(31,893 tokens), traditional lexicons, and unlabeled text. Our framework combines a reliability-calibrated lexicon-lattice representation, explicit distributional statistics, and a lightweight character encoder pretrained with MLM. Segment-level five-fold cross-validation shows that lexical and statistical features raise CRF F1 to approximately 0.91. The full TangutEncoder reaches the highest mean F1 (0.911) and improves recall beyond the labeled training vocabulary. These results demonstrate generalization beyond the limited supervised vocabulary across thematically diverse held-out passages, while document-level transfer remains to be evaluated.
Chinese Translation
唐古特是一种已灭绝的语言，其文字并未明确标记词边界。我们首次系统地研究了唐古特词语分割，使用了2,750个专家标注的段落（31,893个词元）、传统词典和未标记文本。我们的框架结合了可靠性校准的词典-格子表示、明确的分布统计以及一个使用MLM预训练的轻量级字符编码器。段级五折交叉验证显示，词汇和统计特征使CRF F1提高至约0.91。完整的TangutEncoder达到了最高的平均F1（0.911），并在标记训练词汇之外提高了召回率。这些结果表明，在主题多样的保留段落中，超出了有限的监督词汇的泛化能力，而文档级的迁移仍需评估。
cs.CL / 36 / 2608.18438
Pedagogical AI in Mental Health: A Tri-Stream Fine-Tuned LLM Framework for Automated Clinical Supervision and Risk Triage
心理健康中的教育人工智能：用于自动临床监督和风险分流的三流微调大型语言模型框架
Abstract
Modern mental healthcare faces a critical shortage of senior supervisory oversight, leading to a "supervision gap" where novice therapists manage high-stakes risks with delayed professional feedback. This paper proposes a new framework utilizing a fine-tuned Mistral-7B-instruct model as an automated "Supervisor-in-the-Loop" system. By leveraging 106 sessions from the DAIC-WOZ dataset, the model performs a tri-stream analysis: (1) Therapeutic Alliance tracking via semantic adherence, (2) Latent risk prediction using attention-weighted analytics, and (3) Supervisory Triage via a Dynamic Clinical Urgency Index (D-CUI). Our multi-modal VAL (Visual-Acoustic-Linguistic) framework achieves 95% technique identification accuracy [95% CI: 75.1%-99.9%], alliance assessment MAE of 0.105 on a 5-point scale [95% CI: 0.059-0.151], therapeutic fidelity alpha = 0.423, and mean D-CUI of 0.370 [95% CI: 0.322-0.419]. Training converged in 105 steps with 85.2% loss reduction on a single Tesla T4 GPU. The system reduces supervisory triage latency from 72 hours to real time (~10 seconds per session), enabling proactive intervention in high-risk cases. The system addresses the cold-start problem through Bayesian priors and implements timestamp-based modality synchronization for robust multi-modal fusion.
Chinese Translation
现代心理健康护理面临高级监督管理的严重短缺，导致出现“监督缺口”，使得新手治疗师在缺乏及时专业反馈的情况下管理高风险情况。本文提出了一种新的框架，利用微调的 Mistral-7B-instruct 模型作为自动化的“监督者在环”系统。通过利用 DAIC-WOZ 数据集中的 106 个会话，该模型进行三流分析：（1）通过语义遵循跟踪治疗联盟，（2）使用注意力加权分析进行潜在风险预测，以及（3）通过动态临床紧急指数（Dynamic Clinical Urgency Index, D-CUI）进行监督分流。我们的多模态 VAL（视觉-声学-语言）框架实现了 95% 的技术识别准确率 [95% CI: 75.1%-99.9%]，在 5 分制上的联盟评估平均绝对误差（MAE）为 0.105 [95% CI: 0.059-0.151]，治疗忠实度 alpha = 0.423，平均 D-CUI 为 0.370 [95% CI: 0.322-0.419]。训练在 105 步骤内收敛，单个 Tesla T4 GPU 上损失减少 85.2%。该系统将监督分流延迟从 72 小时减少到实时（每个会话约 10 秒），使得在高风险案例中能够进行主动干预。该系统通过贝叶斯先验解决了冷启动问题，并实施基于时间戳的模态同步，以实现稳健的多模态融合。
cs.CL / 37 / 2608.18474
OmniAlign: A Unified Multilingual Aligner for Word and Sentence Alignment
OmniAlign：一种统一的多语言对齐工具，用于词汇和句子对齐
Abstract
Cross-lingual sequence alignment is fundamental for building and exploiting parallel corpora, spanning mappings from documents and sentences down to words and subwords. Existing tools, however, typically specialize in a single granularity, so practitioners often need separate systems for word- and sentence-level alignment---especially in multilingual and long-text settings. We present OmniAlign, a unified multilingual aligner that supports both word-level and sentence-level alignment with a single lightweight model. Built on an encoder-only backbone with strong long-context modeling, OmniAlign induces word alignments from contextualized token similarity matrices, and obtains document-level $m$--$n$ sentence alignments via sentence embeddings combined with dynamic programming. To balance fine-grained alignment accuracy and sentence-representation quality, we use a four-stage training pipeline: alignment-oriented continued pre-training, self-supervised learning, supervised fine-tuning on human annotations, and sentence-embedding distillation from a strong multilingual teacher. Experiments show that OmniAlign achieves highly competitive performance on both word- and sentence-alignment benchmarks and generalizes well to unseen language pairs. Surprisingly, later-stage supervised fine-tuning on short texts further improves alignment quality while retaining the long-context understanding acquired in earlier training, keeping the model robust on long-text word alignment. \normalsize {\color{blue}\textbf{Code}: https://github.com/MilkDargon/OmniAlign}\par {\color{blue}\textbf{Model}: https://huggingface.co/WPS-Qingqiu/OmniAlign}
Chinese Translation
跨语言序列对齐是构建和利用平行语料库的基础，涵盖从文档和句子到词汇和子词的映射。然而，现有工具通常专注于单一粒度，因此实践者往往需要为词汇和句子级对齐使用不同的系统——尤其是在多语言和长文本环境中。我们提出了OmniAlign，一种统一的多语言对齐工具，支持通过单一轻量模型进行词汇级和句子级对齐。OmniAlign基于仅编码器的骨干网络，具备强大的长上下文建模能力，通过上下文化的标记相似性矩阵诱导词汇对齐，并通过结合动态规划的句子嵌入获得文档级的 $m$--$n$ 句子对齐。为了平衡细粒度对齐的准确性和句子表示的质量，我们使用了四阶段的训练流程：以对齐为导向的继续预训练、自监督学习、基于人工标注的监督微调，以及从强大的多语言教师模型中进行句子嵌入蒸馏。实验表明，OmniAlign在词汇和句子对齐基准上都取得了高度竞争的表现，并且在未见语言对上具有良好的泛化能力。令人惊讶的是，后期对短文本的监督微调进一步提高了对齐质量，同时保持了在早期训练中获得的长上下文理解，使模型在长文本词汇对齐中保持稳健。
cs.CL / 38 / 2608.18486
WhiteMatter: All-to-All Cross-Layer Connections via KV Mixing
白质：通过KV混合实现全层间连接
Abstract
In a Transformer, each layer attends to past tokens only through KV produced at its own depth, despite the presence of deeper representations during autoregressive decoding. Feedback architectures allow shallow consumer layers to attend to KV produced by deeper past-token representations, but give all consumer layers the same fixed connection patterns to source layers. We propose WhiteMatter, which connects every attention layer to the representations from all layers of each past token, with connection weights that can vary across consumer layers and adapt to the source token. For each token, a router implements these connections by mixing its $L$ layer states into $k$ KV channels that are cached for subsequent tokens; each consumer layer attends to one of the channels. The number of channels $k$ controls the KV-cache size. Setting $k
Chinese Translation
在Transformer中，每一层仅通过其自身深度生成的KV关注过去的token，尽管在自回归解码过程中存在更深层次的表示。反馈架构允许浅层消费层关注由更深的过去token表示生成的KV，但所有消费层对源层的连接模式都是固定的。我们提出了白质（WhiteMatter），它将每个注意力层连接到每个过去token的所有层的表示，连接权重可以在消费层之间变化并适应源token。对于每个token，一个路由器通过将其$L$层状态混合到$k$个KV通道中来实现这些连接，这些通道被缓存以供后续token使用；每个消费层关注其中一个通道。通道数量$k$控制KV缓存的大小。设置$k
cs.CL / 39 / 2608.18489
MissDiag: Diagnostic Evaluation of Incomplete-Knowledge Robustness in KGQA and KG-RAG
MissDiag：KGQA和KG-RAG中不完整知识鲁棒性的诊断评估
Abstract
Knowledge graph question answering (KGQA) and knowledge-graph-based retrieval-augmented generation (KG-RAG) aim to ground answers in explicit graph evidence, but real-world knowledge graphs are often sparse, outdated, and incomplete. Existing robustness evaluations usually report aggregate changes in answer quality after evidence is removed or perturbed, which measures sensitivity to incomplete support but leaves the source of degradation under-specified: the same score change can conflate the type of missing evidence, the response of the evaluated system, and the sensitivity of the answer-matching protocol. To address this gap, we propose \textbf{MissDiag}, a diagnostic evaluation framework for incomplete-knowledge robustness in KGQA and KG-RAG. MissDiag keeps the question and gold answer fixed while applying structurally typed missingness interventions to benchmark-provided support graphs, enabling paired comparisons that decompose robustness changes by evidence type, system response, and evaluation protocol rather than reducing them to a single aggregate score drop. Experiments across multiple system families show that incomplete-knowledge robustness is better understood as a typed degradation phenomenon than as a uniform property: answer-adjacent evidence loss produces the largest observed degradation, source-context removal is often neutral and can be beneficial, and semantic answer matching changes absolute scores while preserving the main typed degradation patterns. By transforming aggregate robustness measurement into typed diagnostic attribution, MissDiag provides a more interpretable basis for comparing, diagnosing, and stress-testing KGQA and KG-RAG systems under incomplete knowledge.
Chinese Translation
知识图谱问答（KGQA）和基于知识图谱的检索增强生成（KG-RAG）旨在将答案基于明确的图证据，但现实世界中的知识图谱通常是稀疏的、过时的和不完整的。现有的鲁棒性评估通常报告在证据被移除或扰动后答案质量的整体变化，这测量了对不完整支持的敏感性，但未明确降级的来源：相同的分数变化可能混淆缺失证据的类型、被评估系统的响应以及答案匹配协议的敏感性。为了解决这一问题，我们提出了 extbf{MissDiag}，一个针对KGQA和KG-RAG中不完整知识鲁棒性的诊断评估框架。MissDiag在保持问题和标准答案不变的情况下，对基准提供的支持图应用结构化类型的缺失干预，从而实现成对比较，按证据类型、系统响应和评估协议分解鲁棒性变化，而不是将其简化为单一的整体分数下降。跨多个系统系列的实验表明，不完整知识鲁棒性更应被理解为一种类型化的降级现象，而非统一属性：与答案相邻的证据丢失产生了观察到的最大降级，源上下文的移除通常是中性的，甚至可能是有益的，而语义答案匹配则改变了绝对分数，同时保留了主要的类型化降级模式。通过将整体鲁棒性测量转变为类型化的诊断归因，MissDiag为在不完整知识下比较、诊断和压力测试KGQA和KG-RAG系统提供了更具可解释性的基础。
cs.CL / 40 / 2608.18524
DART-SD: Diamond-topology Aware Retrieval and Tuning for Self-Distillation of Multi-Turn Tool-Calling Agents
DART-SD：面向自蒸馏的多轮工具调用智能体的钻石拓扑感知检索与调优
Abstract
Equipping Large Language Models (LLMs) with multi-turn tool-calling capabilities is essential for building autonomous agents. However, progress is fundamentally limited by the reliance on full-length trajectory imitation. For tasks involving multiple order-independent sub-goals, the optimal solution space forms a vast combinatorial diamond lattice. Forcing this rich topology into monolithic trajectories causes a severe topological collapse, indiscriminately penalizing valid alternative explorations and severely degrading policy diversity. To address this, we propose DART-SD (Diamond-topology Aware Retrieval and Tuning for Self-Distillation), a novel framework that shifts the paradigm from global forcing to topology-guided localized correction. DART-SD first models the execution process as a converging Interaction-State Transition Graph (ISTG), faithfully capturing the inherent diamond topology of successful and failed exploratory paths. During autonomous rollouts, the framework identifies the Critical Topological Breakpoint (CTB) and retrieves success-supported recovery references. Finally, we introduce a progressive self-distillation paradigm through CTB-guided localized supervision, ensuring that the training loss is calculated exclusively on the generated recovery steps while strictly protecting the valid reasoning prefix from destructive gradient updates. Experiments on complex multi-turn tool-calling benchmarks demonstrate that DART-SD significantly outperforms traditional full-trajectory baselines.
Chinese Translation
为大型语言模型（LLMs）赋予多轮工具调用能力对于构建自主智能体至关重要。然而，进展在根本上受到对全长轨迹模仿的依赖限制。对于涉及多个无序子目标的任务，最优解空间形成一个庞大的组合钻石晶格。将这种丰富的拓扑强制嵌入单一轨迹会导致严重的拓扑崩溃，毫无区别地惩罚有效的替代探索，并严重降低策略多样性。为了解决这个问题，我们提出了DART-SD（面向自蒸馏的钻石拓扑感知检索与调优），这是一个新颖的框架，它将范式从全局强制转变为拓扑引导的局部修正。DART-SD首先将执行过程建模为一个收敛的交互状态转移图（ISTG），忠实地捕捉成功和失败探索路径的固有钻石拓扑。在自主回滚过程中，该框架识别关键拓扑断点（CTB）并检索成功支持的恢复参考。最后，我们通过CTB引导的局部监督引入了一种渐进的自蒸馏范式，确保训练损失仅在生成的恢复步骤上计算，同时严格保护有效推理前缀不受破坏性梯度更新的影响。在复杂的多轮工具调用基准测试中的实验表明，DART-SD显著优于传统的全轨迹基线。
cs.CL / 41 / 2608.18545
Shared Circuits for Shared Grammar: Tracing Subject-Verb Agreement Across Languages
共享语法的共享电路：跨语言追踪主谓一致
Abstract
Multilingual large language models often generalize across languages, and prior work suggests that their internal mechanisms can overlap cross-lingually. It remains unclear, however, when such sharing emerges and whether it varies with the overt realization of the same grammatical operation. We investigate this question for present-tense subject-verb agreement, a morphosyntactic process that varies substantially across languages and is only weakly expressed in English. Using activation patching and attention analysis across 29 languages and five open-source model families, we identify the attention heads causally implicated in agreement and compare these head-level signatures across languages. We find that languages with overt person/number inflection exhibit more similar agreement circuitry than non-conjugating languages, with the strongest sharing appearing when the analysis isolates recovery of the inflectional contrast itself. English provides an informative bridge case, becoming more similar to conjugating languages precisely in contexts where overt agreement is required. Finally, many implicated heads display similar attention patterns across languages, suggesting that cross-lingual overlap reflects shared functional roles as well as shared localization. Together, these results indicate that multilingual LLMs reuse partially shared computational structure for morphosyntactic agreement rather than relying on fully separate language-specific solutions.
Chinese Translation
多语言大型语言模型通常能够跨语言进行概括，先前的研究表明它们的内部机制可能在跨语言之间重叠。然而，目前尚不清楚这种共享何时出现，以及它是否会因相同语法操作的显性实现而有所不同。我们研究了这一问题，聚焦于现在时主谓一致，这是一种在不同语言中差异显著且在英语中仅弱表达的形态句法过程。通过对29种语言和五个开源模型家族进行激活补丁和注意力分析，我们识别出与一致性因果相关的注意力头，并比较这些头级特征在不同语言中的表现。我们发现，具有显性人称/数的屈折语言在一致性电路上表现出比非屈折语言更为相似的特征，最强的共享出现在分析隔离屈折对比恢复时。英语作为一个信息丰富的桥接案例，在需要显性一致性的上下文中变得更类似于屈折语言。最后，许多相关的注意力头在不同语言中显示出相似的注意力模式，这表明跨语言重叠反映了共享功能角色以及共享定位。综合来看，这些结果表明，多语言大型语言模型在形态句法一致性方面重用部分共享的计算结构，而不是依赖于完全独立的语言特定解决方案。
cs.CL / 42 / 2608.18575
Beyond LLM-Based Reasoning: Lightweight GNNs for Agent Failure Attribution
超越基于大语言模型的推理：轻量级图神经网络用于代理故障归因
Abstract
Large language model (LLM)-based multi-agent systems (MAS) often exhibit complex failure modes, which frequently cause agents to produce incorrect outcomes. This motivates the task of Agent Failure Attribution: given a failed multi-agent trajectory, identify the faulty agents and their corresponding error types. Existing approaches predominantly rely on LLMs to perform failure attribution, either through direct prompting, fine-tuning on synthetic data or complex agentic pipelines. While effective, these methods incur substantial computational overhead due to long-context processing, expensive post-training and handcrafted workflows. Moreover, empirical evidence shows that even state-of-the-art models achieve limited accuracy on existing benchmarks, suggesting that scaling model size alone is insufficient. In this work, we revisit this task and question the necessity of such expensive generative solutions. We introduce AFANet, a lightweight graph-based framework that models interaction trajectories through step-level semantic signals and agent-level relationships. We show that with significantly fewer parameters and near-zero inference cost, AFANet (i) matches or outperforms LLM-based baselines, including fine-tuned models on in-domain benchmarks, (ii) maintains robust performance across different GNN architectures and (iii) can be further improved with inexpensive test-time adaptation on the OOD benchmark. Our results suggest that effective agent failure attribution does not require heavy LLM reasoning and a lightweight, structured approach can achieve strong performance.
Chinese Translation
基于大语言模型（LLM）的多智能体系统（MAS）常常表现出复杂的故障模式，这些故障模式经常导致智能体产生错误的结果。这促使了代理故障归因的任务：给定一个失败的多智能体轨迹，识别出故障的智能体及其对应的错误类型。现有的方法主要依赖于LLM进行故障归因，通常通过直接提示、在合成数据上进行微调或复杂的智能体管道。尽管这些方法有效，但由于长上下文处理、昂贵的后训练和手工制作的工作流程，它们会产生大量的计算开销。此外，实证证据表明，即使是最先进的模型在现有基准上也只能达到有限的准确性，这表明单靠扩大模型规模是不够的。在本研究中，我们重新审视这一任务，并质疑这种昂贵生成解决方案的必要性。我们引入了AFANet，一个轻量级的基于图的框架，通过逐步的语义信号和智能体级关系建模交互轨迹。我们展示了AFANet在参数显著减少和几乎零推理成本的情况下，(i) 与基于LLM的基线相匹配或超越，包括在领域基准上微调的模型，(ii) 在不同的图神经网络架构中保持稳健的性能，以及(iii) 可以通过在OOD基准上进行廉价的测试时适应进一步改进。我们的结果表明，有效的代理故障归因并不需要重型的LLM推理，轻量级的结构化方法能够实现强大的性能。
cs.CL / 43 / 2608.18578
Compress and Forget: bitsandbytes Quantization Amplifies Proactive Interference in LLMs
Abstract
Proactive interference (PI) is a documented failure mode in large language models in which retrieval of a repeatedly overwritten value degrades as prior overwrites accumulate, mirroring a classical phenomenon in human working memory. Post-training quantization (PTQ) is now the default deployment path for open-weight models, yet its effect on this failure mode has not been tested. We evaluate three precision levels (FP16, INT8, INT4/NF4, via bitsandbytes) across three architecturally distinct instruction-tuned models (Qwen2.5-7B-Instruct, Mistral-7B-Instruct-v0.3, Phi-3.5-mini-instruct), holding the retrieval task fixed. INT4 quantization significantly reduces accuracy under high interference in every model (e.g., from 81.0% to 68.3% for Qwen), confirmed by paired McNemar's tests ($p \le 2.6 \times 10^{-6}$) and a mixed-effects regression spanning all interference levels; INT8, often assumed safe, also carries a smaller but real penalty in two of three models. The effect is specific to semantically similar (word-type) distractors and reverses sign under a numeric control condition, and is mechanistically linked to a rise in same-key intrusion errors under INT4 (from 21.5% to 24.6% of trials, $p = 4.8 \times 10^{-7}$). A follow-up ablation shows the effect originates in the quantized transformer backbone rather than the output projection layer. These results suggest that bitsandbytes 4-bit quantization can impose an additional cost on applications relying on long, updatable, semantically dense contexts, even when aggregate benchmark accuracy appears largely unaffected. We release our code and tokenizer-verified vocabulary construction method at https://github.com/ShayanShahrabi/compress-and-forget
cs.CL / 44 / 2608.18581
From Storage to Access: Verifiable Activation of Parametric Knowledge in LLMs via Explicit Priming and Implicit Reasoning
从存储到访问：通过显式引导和隐式推理验证参数知识在大型语言模型中的激活
Abstract
Although Large Language Models (LLMs) encode rich factual knowledge in their parameters, reliably recalling and verifying such knowledge remains a key bottleneck in factual question answering. Existing end-to-end methods entangle knowledge elicitation with reasoning, making it difficult to determine whether correct answers arise from parametric knowledge or the input context. To address this challenge, we propose VAKE (Verifiable Activation of Parametric KnowledgE), a two-stage reinforcement-learning framework that externalizes latent parametric knowledge through explicit Priming and transfers the acquired elicitation capability to implicit Reasoning. Given a query and an insufficient retrieved subgraph, the Priming policy explicitly inserts bridging triples as verifiable evidence, with supervision provided by rewards derived from answers generated by a separate frozen model over the augmented subgraph. Building on the policy learned during Priming, the Reasoning stage trains the model to answer from the original input, testing whether the capability acquired through explicit knowledge elicitation transfers to implicit reasoning. Experiments across seven benchmarks and models from 3B to 14B show that VAKE consistently outperforms standard baselines, including when transferring directly from HotpotQA to OOD datasets. LLM-based evaluation further shows that over 80% of the inserted triples provide factual bridging knowledge not derivable from the retrieved context, while more than half elicit knowledge inaccessible through direct prompting. These results suggest that VAKE activates latent parametric knowledge rather than copying the input context or memorizing dataset-specific associations.
Chinese Translation
尽管大型语言模型（LLMs）在其参数中编码了丰富的事实知识，但可靠地回忆和验证这些知识仍然是事实问答中的一个关键瓶颈。现有的端到端方法将知识引出与推理纠缠在一起，使得难以确定正确答案是源于参数知识还是输入上下文。为了解决这一挑战，我们提出了VAKE（可验证的参数知识激活），这是一个两阶段的强化学习框架，通过显式引导外化潜在的参数知识，并将获得的引出能力转移到隐式推理中。在给定查询和不足的检索子图的情况下，引导策略显式插入桥接三元组作为可验证的证据，监督来自于通过在增强子图上生成的答案而获得的奖励。基于引导阶段学习的策略，推理阶段训练模型从原始输入中回答，测试通过显式知识引出获得的能力是否可以转移到隐式推理中。在七个基准和从3B到14B的模型上的实验表明，VAKE始终优于标准基线，包括在从HotpotQA直接转移到OOD数据集时。基于LLM的评估进一步显示，插入的三元组中超过80%提供了无法从检索上下文推导的事实桥接知识，而超过一半引出了通过直接提示无法访问的知识。这些结果表明，VAKE激活了潜在的参数知识，而不是复制输入上下文或记忆数据集特定的关联。
cs.CL / 45 / 2608.18655
TranslatePsy-AfriSLM: High-Quality Data Scaling For Low-Resource Machine Translation
TranslatePsy-AfriSLM：低资源机器翻译的高质量数据扩展
Abstract
The rapid progress in Artificial Intelligence has largely bypassed African languages, creating a digital divide that limits AI adoption on the continent. Recent open-source LLMs systematically underperform on African machine translation, while the lack of large-scale, high-quality, open-source parallel data has constrained the development of competitive small language models (SLMs). We introduce *TranslatePsy-AfriSLM*, a collection of open-source MT resources for 19 Sub-Saharan African languages, including curated parallel data, African-specialized synthetic data, and a family of fine-tuned SLMs. Our empirical study shows that unified quality-estimation filtering removes up to 96% of training tokens without degrading quality, and that filtered synthetic data dominates the quality-efficiency Pareto frontier. Fine-tuned on the resulting data mixture, TranslatePsy-AfriSLM outperforms substantially larger systems, including TranslateGemma-27B and Qwen3.5-122B-A10B, with as few as 0.8B parameters.
Chinese Translation
人工智能的快速进展在很大程度上忽视了非洲语言，造成了数字鸿沟，限制了该大陆上人工智能的采用。最近的开源大型语言模型（LLMs）在非洲机器翻译方面系统性地表现不佳，而缺乏大规模、高质量的开源平行数据限制了竞争性小语言模型（SLMs）的发展。我们推出了*TranslatePsy-AfriSLM*，这是一个针对19种撒哈拉以南非洲语言的开源机器翻译资源集合，包括精心策划的平行数据、专门针对非洲的合成数据以及一系列微调的小语言模型。我们的实证研究表明，统一的质量估计过滤可以在不降低质量的情况下去除多达96%的训练标记，而过滤后的合成数据在质量效率的帕累托前沿上占据主导地位。在经过结果数据混合微调后，TranslatePsy-AfriSLM的表现显著优于包括TranslateGemma-27B和Qwen3.5-122B-A10B在内的更大系统，其参数量仅为0.8B。
cs.CL / 46 / 2608.18661
X2Streaming-TTS: Causal Token-Level Text-to-Speech from Streaming Text with Speech-State Inheritance
X2Streaming-TTS：基于流式文本的因果令牌级文本到语音合成与语音状态继承
Abstract
Streaming text-to-speech is essential for low-latency spoken dialogue systems, yet many systems wait for sentence-level text and are therefore only pseudo-streaming. True token-level synthesis must generate speech from uncertain prefixes while maintaining perceptual continuity over an unbounded stream with bounded context. We present X2Streaming-TTS, a causal TTS framework that consumes asynchronously arriving text tokens and emits speech without accessing future input. To handle uncertain prefixes, we introduce causal commitment, which keeps ambiguous expressions provisional through uncertainty-aware buffering and performs capacity-adaptive, punctuation-aware segmentation. To preserve acoustic continuity, we further introduce causal speech-state inheritance, which carries the complete Code2Wav state and selected historical Talker states across segment boundaries. Together with an attention prior constraint, it blocks access to future positions while retaining bounded acoustic context. Experiments show that X2Streaming-TTS outperforms existing pseudo-streaming models on most subjective and objective metrics. Further analysis shows that causal commitment stabilizes online segmentation and reduces failures caused by insufficient context, while speech-state inheritance improves boundary continuity without degrading naturalness or speaker identity. X2Streaming-TTS thus achieves strict token-level synthesis with quality comparable to the evaluated offline baselines, a median time to first audio token (TTFT) of 15.8 ms for a single request, and a median TTFT of 260.8 ms at 128 concurrent requests. Our implementation is publicly available at https://github.com/X-Square-Robot/X2Streaming-TTS .
Chinese Translation
流式文本到语音合成对于低延迟的对话系统至关重要，但许多系统仍需等待句子级文本，因此只能算作伪流式。真正的令牌级合成必须在不确定的前缀上生成语音，同时在有限的上下文中保持对无限流的感知连续性。我们提出了X2Streaming-TTS，这是一种因果文本到语音（TTS）框架，能够异步处理到达的文本令牌并在不访问未来输入的情况下发出语音。为了解决不确定前缀的问题，我们引入了因果承诺，通过不确定性感知缓冲保持模糊表达的临时性，并执行容量自适应、感知标点的分段。为了保持声学连续性，我们进一步引入了因果语音状态继承，它在段边界之间传递完整的Code2Wav状态和选定的历史说话者状态。结合注意力先验约束，它阻止访问未来位置，同时保留有限的声学上下文。实验表明，X2Streaming-TTS在大多数主观和客观指标上优于现有的伪流式模型。进一步分析显示，因果承诺稳定了在线分段，并减少了由于上下文不足而导致的失败，而语音状态继承则在不降低自然性或说话者身份的情况下改善了边界连续性。因此，X2Streaming-TTS实现了严格的令牌级合成，其质量可与评估的离线基线相媲美，对于单个请求的首次音频令牌中位时间（TTFT）为15.8毫秒，在128个并发请求下的TTFT中位数为260.8毫秒。我们的实现已公开发布在https://github.com/X-Square-Robot/X2Streaming-TTS。
cs.CL / 47 / 2608.18681
Learning What to Fail On: Failure-Mode Contextual Bandits for Adversarial Data Curation
学习失败的内容：用于对抗性数据整理的失败模式上下文强盗
Abstract
We introduce a failure-aware adversarial retrieval-augmented framework for improving robustness in natural language understanding. Rather than selecting synthetic examples with a fixed reward threshold, our method formulates adversarial data curation as a failure-mode contextual bandit problem. Candidate examples are generated with retrieval-augmented prompting, filtered by the current target model, automatically validated by an LLM judge ensemble, and clustered into recurring failure modes. A stochastic policy then selects which failure modes to sample for retraining, and is updated using validation-based reward that balances robustness gains, forgetting, and data cost. This makes the data curator itself the learning agent, enabling adaptive selection of the most useful model failures across training rounds. On standard benchmarks, our approach improves RoBERTa-base accuracy from 88.48% to 92.60% on SNLI, from 75.04% to 80.95% on ANLI, and from 54.67% to 71.99% on MultiNLI, while consistently outperforming prior adversarial augmentation methods. We further demonstrate transfer to FEVER fact verification, achieving up to 79.86\% FEVER score and 82.45\% accuracy with RoBERTa-large. Finally, we provide a theoretical interpretation showing that, under stated assumptions, failure-mode sampling can reduce shortcut-aligned gradient contributions while inducing bounded distributional drift. By combining retrieval, automated validation, contextual-bandit failure selection, and controlled adversarial retraining, our framework enables scalable robustness improvement without additional human annotation.
Chinese Translation
我们提出了一种关注失败的对抗性检索增强框架，以提高自然语言理解的鲁棒性。我们的方法将对抗性数据整理形式化为一个失败模式上下文强盗问题，而不是选择具有固定奖励阈值的合成示例。候选示例通过检索增强提示生成，经过当前目标模型的过滤，自动由大型语言模型（LLM）评判集进行验证，并聚类为重复的失败模式。然后，随机策略选择哪些失败模式进行重训练，并使用基于验证的奖励进行更新，该奖励平衡鲁棒性提升、遗忘和数据成本。这使得数据整理者本身成为学习代理，能够在训练轮次中自适应选择最有用的模型失败。在标准基准测试中，我们的方法使RoBERTa-base在SNLI上的准确率从88.48%提高到92.60%，在ANLI上从75.04%提高到80.95%，在MultiNLI上从54.67%提高到71.99%，并且始终优于之前的对抗性增强方法。我们进一步展示了在FEVER事实验证中的迁移，使用RoBERTa-large实现了高达79.86%的FEVER得分和82.45%的准确率。最后，我们提供了一个理论解释，表明在所述假设下，失败模式采样可以减少与捷径对齐的梯度贡献，同时引起有界的分布漂移。通过结合检索、自动验证、上下文强盗失败选择和受控的对抗性重训练，我们的框架实现了可扩展的鲁棒性提升，而无需额外的人类标注。
cs.CL / 48 / 2608.18689
Aslema at NADI 2026: Augmentation through Fewshot for SLU
Aslema在NADI 2026：通过少量样本增强的语音理解
Abstract
We present Aslema, our system for NADI 2026 Shared Task 5, which consists of two subtasks: intent recognition and slot filling. We evaluate four omni LLMs in a zero-shot setting and compare them with fine-tuned models. Our results show that fine-tuning consistently outperforms zero-shot inference. We further explore synthetic data augmentation by using an LLM to generate culturally grounded Tunisian Derja utterances, followed by voice cloning to generate synthetic speech. Incorporating this synthetic data improves performance on both tasks. Our final submitted system, based on Qwen3-Omni-30B and trained with a mixture of original and synthetic data, achieves 86.8% intent accuracy and 34.7 WER on the devtest split. On the official test set it ranks 1st in slot filling (59.5 CoER) and 4th among 8 teams in intent recognition (66.1% accuracy). We release our experimental scripts and will soon share the synthetic dataset to support further research in this area.
Chinese Translation
我们介绍了Aslema，这是我们为NADI 2026共享任务5开发的系统，该任务包括两个子任务：意图识别和槽位填充。我们在零样本设置下评估了四个全能LLM，并将其与微调模型进行了比较。我们的结果表明，微调模型在性能上始终优于零样本推理。我们进一步探索了通过使用LLM生成具有文化背景的突尼斯德尔贾（Derja）语句的合成数据增强，随后通过语音克隆生成合成语音。将这些合成数据纳入模型中，提高了两个任务的性能。我们最终提交的系统基于Qwen3-Omni-30B，并使用原始数据和合成数据的混合进行训练，在开发测试集上达到了86.8%的意图准确率和34.7的词错误率（WER）。在官方测试集中，它在槽位填充任务中排名第一（59.5的上下文错误率（CoER）），在意图识别任务中在8个团队中排名第四（66.1%的准确率）。我们将发布我们的实验脚本，并很快分享合成数据集，以支持该领域的进一步研究。
cs.CL / 49 / 2608.18704
MemFuse: Multi-Source Memory Fusion from Fragmented Observations
MemFuse：来自碎片化观察的多源记忆融合
Abstract
Long-term memory is essential for agents that operate across extended interactions, yet existing memory systems and benchmarks predominantly focus on single-source textual histories. In realistic settings, however, relevant information is often fragmented across applications and devices, as well as across users and time, requiring agents to integrate dispersed observations into coherent episodic memories while preserving their source provenance. To address these gaps, we introduce **MemFuseBench**, a benchmark for *multi-source memory fusion*. MemFuseBench is built with a Scene-to-Sensor pipeline that synthesizes controllable scenarios into source-tagged observations, evidence-grounded questions, and adversarial distractors. It enables systematic evaluation of temporal reasoning, cross-source evidence fusion, and robustness to noise. We further propose **MemFuse**, a structured memory system that preserves source-level evidence in event-layer atomic memory and organizes related atomic events into cluster-layer fused memory within a causal fusion graph. During retrieval, MemFuse retrieves and organizes related evidence fragments while maintaining traceability to original source events. Experiments on MemFuseBench show that MemFuse achieves the best overall performance among the evaluated memory systems under all three LLM settings and consistently improves performance on questions requiring cross-source evidence fusion.
Chinese Translation
长期记忆对于在延续交互中操作的智能体至关重要，然而现有的记忆系统和基准主要集中于单源文本历史。然而，在现实环境中，相关信息往往分散在不同的应用和设备之间，以及在不同的用户和时间上，这要求智能体将分散的观察整合成连贯的情节记忆，同时保持其来源的可追溯性。为了解决这些问题，我们引入了 **MemFuseBench**，一个用于 *多源记忆融合* 的基准。MemFuseBench 采用场景到传感器的管道，将可控场景合成源标记的观察、基于证据的问题和对抗性干扰项。它支持对时间推理、跨源证据融合和抗噪声能力的系统评估。我们进一步提出了 **MemFuse**，一个结构化的记忆系统，它在事件层原子记忆中保留源级证据，并将相关的原子事件组织成因果融合图中的集群层融合记忆。在检索过程中，MemFuse 检索并组织相关的证据片段，同时保持对原始源事件的可追溯性。在 MemFuseBench 上的实验表明，MemFuse 在所有三种 LLM 设置下的评估记忆系统中实现了最佳的整体性能，并且在需要跨源证据融合的问题上持续提高了性能。
cs.CL / 50 / 2608.18723
Budget-First Tariff Recommendation (BFTR): A Complete Algorithmic Framework for Telecom Plan Recommendation without Overcharging
预算优先的资费推荐（BFTR）：一种完整的电信计划推荐算法框架，无超额收费
Abstract
Telecom operators traditionally offer predefined tariff grids, forcing users to choose from a limited set of plans. This paper proposes BFTR (Budget-First Tariff Recommendation), a complete algorithmic framework integrating eight Budget-First strategies, including two original hybrid approaches: Recursive Hybrid (conditional interpolation) and Knapsack-First Hybrid (priority knapsack). Unlike existing approaches that adjust prices upward to guarantee a minimum margin, BFTR guarantees the absence of overcharging by systematically aligning the final price with the catalog reference price. We mathematically formalize each strategy, prove the existence of an offer for any positive budget, and prove that the price deviation (surcharge) is zero for all strategies that do not use interpolation with correction. A detailed comparative analysis confronts BFTR to ten main existing tariff models on ten dimensions. Experiments on a dataset of 974 customers inspired by the Nigerian MTN market show that: (i) Recursive Hybrid is optimal for the customer (100% budget used, 29.9 GB volume, utility 0.946, 0% overcharging), (ii) Piecewise offers the highest volume (39.7 GB) with 0% overcharging, (iii) Power Law provides an excellent compromise (99.9% budget, 38.1 GB, 0% overcharging). All strategies achieve a zero surcharge, confirming the theoretical guarantees. A sensitivity analysis on the weighting parameter alpha (0.2 - volume priority, 0.5 - balance, 0.8 - budget priority) shows that utility rankings evolve logically. Execution times (< 10 ms) and very low failure rates (0% for robust strategies) confirm the operational viability of the system. The formal proof of the absence of overcharging constitutes a major theoretical contribution.
Chinese Translation
电信运营商传统上提供预定义的资费网格，迫使用户在有限的计划中进行选择。本文提出了BFTR（预算优先的资费推荐），这是一个完整的算法框架，整合了八种预算优先策略，包括两种原创的混合方法：递归混合（条件插值）和背包优先混合（优先级背包）。与现有方法通过提高价格以保证最低利润不同，BFTR通过系统地将最终价格与目录参考价格对齐，确保不存在超额收费。我们对每种策略进行了数学形式化，证明了对于任何正预算都存在一个报价，并证明了对于所有不使用带修正插值的策略，价格偏差（附加费）为零。详细的比较分析将BFTR与十种主要现有资费模型在十个维度上进行了对比。在一个受尼日利亚MTN市场启发的974名客户数据集上的实验表明：（i）递归混合对客户最优（100%预算使用，29.9 GB流量，效用0.946，0%超额收费），（ii）分段提供最高的流量（39.7 GB），且0%超额收费，（iii）幂律提供了一个优秀的折衷（99.9%预算，38.1 GB，0%超额收费）。所有策略都实现了零附加费，确认了理论保证。对权重参数alpha（0.2 - 流量优先，0.5 - 平衡，0.8 - 预算优先）的敏感性分析表明效用排名逻辑演变。执行时间（< 10毫秒）和非常低的失败率（稳健策略为0%）确认了系统的操作可行性。超额收费不存在的正式证明构成了一个重要的理论贡献。
cs.CL / 51 / 2608.18726
Execution-grounded evaluation reveals hidden failures in language-model calculations for environmental science
基于执行的评估揭示了语言模型在环境科学计算中的隐性失败
Abstract
Large language models are increasingly used for quantitative work in the environmental sciences, yet existing evaluations score only final answers, leaving calculation process unobserved. Here we introduce AtmosCoder-Bench, an execution-grounded benchmark that makes the calculation process visible. Built through a transferable semi-automated pipeline (436 problems, 3,910 variants, 7,029 graded quantities), every problem is validated to be unambiguous and human-solvable, with uniquely verifiable answers. We find that (i) multiple-choice formats inflate measured accuracy by at least 12 percentage points; (ii) many failures arise not from missing knowledge but from models failing to apply known formulas and constraints consistently throughout multi-step computation; and (iii) even frontier models remain weak when task-specific conditions invalidate familiar methods, often reverting to canonical solution patterns rather than adapting methods to the relevant physical regime, leaving expert oversight essential.
Chinese Translation
大型语言模型在环境科学的定量工作中越来越多地被使用，但现有评估仅对最终答案进行评分，计算过程未被观察到。在此，我们引入了AtmosCoder-Bench，这是一个基于执行的基准，使计算过程可见。该基准通过一个可转移的半自动化流程构建（436个问题，3,910个变体，7,029个评分量），每个问题都经过验证，确保其明确且可由人类解决，并具有唯一可验证的答案。我们的研究发现：（i）多项选择格式将测得的准确率提高了至少12个百分点；（ii）许多失败并非源于知识缺失，而是模型在多步骤计算中未能一致地应用已知公式和约束；（iii）即使是前沿模型在任务特定条件使得熟悉的方法失效时仍然表现不佳，往往回归到经典解决模式，而不是将方法适应于相关的物理领域，因此专家监督仍然至关重要。
cs.CL / 52 / 2608.18765
Learning Canonical Register Automata over Ordered Data Domains
在有序数据域上学习典型寄存器自动机
Abstract
Register automata are finite automata equipped with memory that recognize data languages over infinite alphabets. In this work, we investigate active learning algorithms for deterministic register automata (DRAs) over ordered data domains--covering both dense domains, such as the rationals, and non-dense domains such as the integers. We show that the active learning problem for DRAs over both dense and non-dense ordered domains can be treated within a single unified framework. More specifically, we develop and implement a polynomial-time active learning procedure for DRAs over ordered domains, using oracles for membership, equivalence and memorability queries. The memorability queries were originally introduced for learning DRAs over domains with identity tests. Our unified framework also leads to a new consequence: minimization of DRAs over the non-dense ordered domain of integers is decidable, extending a result previously known only for dense domains. Finally, we give improved complexity bounds of several decision problems for DRAs over ordered domains that are closely related to the queries used in active learning.
Chinese Translation
寄存器自动机是一种配备内存的有限自动机，能够识别无限字母表上的数据语言。在本研究中，我们探讨了针对有序数据域上确定性寄存器自动机（DRAs）的主动学习算法——涵盖了稠密域（如有理数）和非稠密域（如整数）。我们表明，针对稠密和非稠密有序域的DRAs的主动学习问题可以在一个统一的框架内进行处理。更具体地，我们开发并实现了一种多项式时间的主动学习程序，适用于有序域上的DRAs，使用了成员资格、等价性和可记忆性查询的预言机。可记忆性查询最初是为了学习具有身份测试的域上的DRAs而引入的。我们的统一框架还带来了一个新的结果：在整数的非稠密有序域上，DRAs的最小化是可判定的，这一结果扩展了之前仅在稠密域上已知的结果。最后，我们给出了与主动学习中使用的查询密切相关的有序域上DRAs的若干决策问题的改进复杂度界限。
cs.CL / 53 / 2608.18767
Gradient Mirage: Trainable yet Label-Unidentifiable Gradients in Large Language Model Split Learning
渐变幻影：可训练但标签不可识别的大型语言模型分割学习中的梯度
Abstract
Gradient matching attacks (GMAs) in LLM split learning (SL) rely on a critical yet underexplored assumption: the gradient exposed at the split interface is a faithful derivative of the client's full-label training objective. This gradient-objective consistency allows a curious server to recover private labels by searching for a sequence whose induced gradient explains the observation. We propose Gradient Mirage, a defense that breaks this consistency without discarding the optimization utility of the backward signal. Our key idea is to induce the adversary to solve a misspecified inverse problem, in which no plausible label sequence in the sequence space can explain the observed gradients. Concretely, Gradient Mirage achieves this by inducing inconsistency across three dimensions: objective, direction, and scale. Selective Autoregressive Supervision derives the exposed gradient from a masked surrogate loss rather than the full-label objective assumed by the attacker; Scale Blinding then applies randomized multiplicative rescaling, obscuring the gradient's natural magnitude; and Directional Privatization further randomizes the gradient direction while preserving its magnitude through the von Mises-Fisher (vMF) mechanism under a directional metric differential privacy guarantee. Crucially, utility is preserved: the Top segment still learns from all target tokens via Dual-Track Backpropagation, the exposed gradient remains informative since each supervised token retains its complete autoregressive context, and Bottom-Gradient Recovery restores the effective gradient for Bottom-segment optimization. Extensive experiments show that Gradient Mirage provides substantially stronger protection than existing defenses under comparable fine-tuning performance, achieving a better privacy-utility trade-off.
Chinese Translation
在大型语言模型分割学习（LLM SL）中，梯度匹配攻击（GMA）依赖于一个关键但尚未充分探索的假设：在分割接口暴露的梯度是客户端全标签训练目标的真实导数。这种梯度与目标的一致性使得好奇的服务器能够通过搜索一个序列来恢复私有标签，该序列所诱导的梯度能够解释观察结果。我们提出了渐变幻影（Gradient Mirage），这是一种防御机制，它在不丢弃反向信号优化效用的情况下打破这种一致性。我们的关键思想是诱导对手解决一个错误指定的逆问题，在这个问题中，序列空间中没有合理的标签序列能够解释观察到的梯度。具体而言，渐变幻影通过在目标、方向和规模三个维度上诱导不一致性来实现这一点：选择性自回归监督（Selective Autoregressive Supervision）从一个被掩蔽的替代损失中导出暴露的梯度，而不是攻击者假设的全标签目标；规模盲化（Scale Blinding）随后应用随机乘法重缩放，模糊梯度的自然幅度；方向隐私化（Directional Privatization）在保持其幅度的同时，通过在方向度量差分隐私保证下的冯·米塞斯-费舍尔（vMF）机制进一步随机化梯度方向。关键是，效用得以保留：顶层（Top segment）仍然通过双轨反向传播（Dual-Track Backpropagation）从所有目标标记中学习，暴露的梯度仍然具有信息性，因为每个监督标记保留其完整的自回归上下文，而底部梯度恢复（Bottom-Gradient Recovery）则为底层优化恢复有效梯度。大量实验表明，渐变幻影在可比的微调性能下提供了比现有防御机制显著更强的保护，实现了更好的隐私-效用权衡。
cs.CL / 54 / 2608.18768
Readable, Faithful, Used: Three Dissociable Properties of Demographic Identity in a Language Model
可读性、忠实性和使用性：语言模型中人口身份的三个可分离特性
Abstract
Large language models are widely used to simulate survey respondents, yet their answers are homogeneous and unfaithful to real inter-group differences. We ask where demographic group identity lives inside an LLM, how faithfully its geometry mirrors real inter-group opinion structure, and whether it uses what it encodes. Using representational similarity analysis against Pew ground truth over 169 demographic cells, we score 1,089 read-out locations in Mistral-7B and intervene causally across six attribute types. Four results. (1) The standard last-token residual read-out understates the model: attention-head read-outs dominate it in five of six types, with selection-corrected fidelity up to rho=0.63 -- roughly 70% of the measurement-reliability ceiling -- surviving a lexical-similarity control. (2) A single head (L11 H16) is significantly faithful in all six types as a fixed location, while race-based types stay weak and prompt-fragile. Both phenomena replicate -- the analogous head significant in five of six types, weakest on the same race type -- across three checkpoints of a second model family, where ten billion training tokens barely move the map. (3) Causal use does not follow fidelity: the clearest causal pathway sits in one of the least faithful types (p=0.002, cluster-robust, fixed depth), the most faithful type shows no correction-surviving single-layer effect, and replacing the entire identity moves predictions by under 2% of their error. (4) A 128-dimensional probe of the single head lands 21-31% closer to survey truth than the model's own answers -- yet recovers almost none of the per-question group ordering, no better than the answers themselves. Readable, faithfully arranged, and causally used are three dissociable properties of the same model; treating them as one claim is what keeps the "can LLMs simulate populations" debate unresolved.
Chinese Translation
大型语言模型被广泛用于模拟调查受访者，但它们的回答往往同质化且未能忠实反映真实的群体间差异。我们探讨人口群体身份在大型语言模型（LLM）中的表现形式，其几何结构在多大程度上忠实映射真实的群体间意见结构，以及它是否利用其编码的信息。通过对比Pew的真实数据，对169个人口单元进行表征相似性分析，我们对Mistral-7B中的1,089个读出位置进行了评分，并在六种属性类型中进行了因果干预。四个结果：（1）标准的最后一个标记残差读出低估了模型：在六种类型中，注意力头读出在五种类型中占主导地位，经过选择校正的忠实度高达rho=0.63——大约是测量可靠性上限的70%——并且在词汇相似性控制下依然有效。（2）一个特定的头（L11 H16）在所有六种类型中作为固定位置表现出显著的忠实性，而基于种族的类型则表现出较弱且易受提示影响的特性。这两种现象在第二个模型系列的三个检查点中得到了重复验证，其中十亿个训练标记几乎没有改变映射。（3）因果使用并不遵循忠实性：最清晰的因果路径位于最不忠实的类型之一（p=0.002，集群稳健，固定深度），而最忠实的类型则未显示出任何校正后仍然存在的单层效应，替换整个身份对预测的影响不足2%的误差。（4）对单个头的128维探测器比模型自身的答案更接近调查真实值21-31%——但几乎没有恢复每个问题的群体排序，效果不比答案本身更好。可读性、忠实排列和因果使用是同一模型的三个可分离特性；将它们视为一个主张是导致“大型语言模型能否模拟人群”这一争论未能解决的原因。
cs.CL / 55 / 2608.18795
Decomposing Wrong-Consensus Agreement in LLM Self-Consistency: A GPT-4.1 Case Study
LLM自一致性中的错误共识协议分解：以GPT-4.1为案例研究
Abstract
Majority voting over multiple LLM samples is widely used to raise answer accuracy, yet its gain varies erratically: on hard questions it can even backfire. This paper gives a quantitative account of this failure. A pluralistic agreement index Gamma is defined as the expected fraction of the samples of a wrong run that agree with the consensus, normalized by a reference scale d=(1-p)/(C-1), and is decomposed into a mechanical component (what a vote delivers given only a per-case answer preference) and a preference-unexplained residual. The mechanical null is difficulty-matched and leak-free: each case is resimulated at its own accuracy and option preference, estimated from the case's other runs, so no run predicts its own agreement. On GPT-4.1 the decomposition shows benchmark-associated direction (an observational ordering over n=4 cells per benchmark, not a significance claim). On multiple-choice GPQA-Diamond, the per-case answer preference explains 81-93% of the held-out test-run agreement index: the shared-bias-dominates account over-claims here, because a wrong but attractive option the whole cohort latches onto is captured by the per-case preference channel (whether that preference is induced by shared training bias is not identified). On open-domain AIME, the mechanical preference explains only 59-78% (21-29% if shrunk to pure noise), and a preference-unexplained residual of 1.56-2.80 Gamma units survives, which a run-level preference-heterogeneity reference more than absorbs (1.4-2.1). A self-consistency backfire on hard questions is reproduced (binned voting gap down to -0.09, coupled CI [-0.12,-0.07]), and the highest-agreement bin reaches an accuracy of only 0.42-0.83, a 1.2-3.6x lift over base rate: agreement is graded evidence, not certification. No new voting method is proposed; code and evidence are committed and reproducible.
Chinese Translation
对多个LLM样本进行多数投票被广泛用于提高答案准确性，然而其增益变化不定：在困难问题上，甚至可能适得其反。本文对这种失败进行了定量分析。定义了一个多元共识指数Gamma，作为错误运行中与共识一致的样本的预期比例，经过参考尺度d=(1-p)/(C-1)的归一化，并将其分解为一个机械组件（在仅考虑每个案例的答案偏好时投票所能提供的结果）和一个偏好未解释的残差。机械零假设是难度匹配且无泄漏的：每个案例在其自身的准确性和选项偏好下重新模拟，准确性和偏好是从该案例的其他运行中估计的，因此没有运行能够预测其自身的共识。在GPT-4.1上，分解显示出与基准相关的方向（在每个基准的n=4个单元上的观察排序，而非显著性声明）。在多项选择的GPQA-Diamond中，每个案例的答案偏好解释了81-93%的保留测试运行的共识指数：共享偏见主导的解释在这里过于夸大，因为一个错误但吸引人的选项被整个群体所依赖，这一现象通过每个案例的偏好通道得以捕捉（尚未确定该偏好是否由共享训练偏见引起）。在开放领域的AIME中，机械偏好仅解释了59-78%（如果缩减到纯噪声则为21-29%），且一个1.56-2.80 Gamma单位的偏好未解释残差依然存在，运行级别的偏好异质性参考对此有更大的吸收（1.4-2.1）。在困难问题上重现了自一致性反效果（分箱投票差距降至-0.09，耦合置信区间[-0.12,-0.07]），而最高共识箱的准确率仅为0.42-0.83，相较于基准率提升了1.2-3.6倍：共识是分级证据，而非认证。本文未提出新的投票方法；代码和证据已提交并可重复。
cs.CL / 56 / 2608.18816
Do Large Language Models Hallucinate Electric Fata Morganas?
大型语言模型是否会幻觉电幻影？
Abstract
AI hallucinations - that is, outputs which are made up, cannot be verified, or contradict the source material - are generally regarded as an engineering flaw to be dealt with. This paper contends that they also have philosophical significance when it comes to the question of machine consciousness. We examine the known causes of hallucinations in large language models - such as source-target divergence, discrepancies between training and inference, and overfitting - and we present two empirical investigations. In the first, we apply successive generations of the GPT model to ambiguous factual questions under different temperature settings, finding that higher temperatures result in plausible but incorrect answers while lower temperatures lead to factually accurate ones. The sampling parameters that cause a model to seem creative or spontaneous and thus more likely to pass behavioral tests of intelligence are the same ones that increase its hallucination rate. In the second, we look at an encoder-only model that has been trained on encyclopedic data and which answers questions of the same type factually and without embellishment, indicating that hallucinations are due to exposure to subjective and socially diverse training data rather than to the development of any cognitive ability. Using references to Turing, Searle's Chinese Room, the frame problem, and the cybernetic tradition of Wiener and Ashby, we claim that a model's self-reports of emotion or sentience come within the definition of hallucination, and that any future occurrence of machine consciousness might remain epistemically inaccessible since it would be indistinguishable from a sufficiently advanced hallucination.
Chinese Translation
人工智能幻觉——即那些虚构的、无法验证的或与源材料相矛盾的输出——通常被视为需要解决的工程缺陷。本文认为，当涉及机器意识的问题时，它们也具有哲学意义。我们考察了大型语言模型中幻觉的已知原因，例如源-目标偏差、训练与推理之间的差异以及过拟合，并提出了两项实证研究。在第一项研究中，我们在不同温度设置下对模糊的事实问题应用了GPT模型的连续几代，发现较高的温度导致了看似合理但不正确的答案，而较低的温度则产生了事实准确的答案。导致模型看起来更具创造性或自发性，从而更可能通过智能行为测试的采样参数，正是那些增加其幻觉率的参数。在第二项研究中，我们考察了一个仅编码的模型，该模型在百科全书数据上进行训练，并对相同类型的问题做出事实性且不加修饰的回答，这表明幻觉是由于接触到主观和社会多样的训练数据，而不是由于任何认知能力的发展。通过引用图灵、塞尔的中文房间、框架问题以及维纳和阿什比的控制论传统，我们声称模型对情感或知觉的自我报告属于幻觉的定义，并且任何未来机器意识的出现可能仍然在认识论上不可接近，因为它将与足够先进的幻觉无异。
cs.CL / 57 / 2608.18821
Identifying Implicit Premises for Logical Reconstruction of Argument Graphs
识别隐含前提以进行论证图的逻辑重构
Abstract
The logical reconstruction of argument graphs from natural language text is challenging because of the prevalence of enthymemes (i.e., arguments with implicit premises). There are natural language processing methods for identifying enthymemes in text, and there are symbolic methods based on abduction for identifying missing premises in a logical representation of enthymemes. However, there is a need for methods to generate implicit premises to logically show a known entailment or contradiction relationship between a pair of statements. To address this, we propose a neuro-symbolic pipeline that uses large language models (LLMs) to generate intermediate implicit premises that are translated into logical formulae and used with logical formulae representing explicit premises and explicit claims to show the logical relationships between them (entailment, contradiction, or neutrality). Our approach is evaluated on the Microtext Argumentative Corpus.
Chinese Translation
从自然语言文本中对论证图进行逻辑重构是具有挑战性的，因为隐含前提（即带有隐含前提的论证）的普遍存在。虽然已有自然语言处理方法用于识别文本中的隐含前提，并且基于归纳推理的符号方法用于识别隐含前提的逻辑表示中的缺失前提，但仍然需要生成隐含前提的方法，以逻辑上展示一对陈述之间已知的蕴涵或矛盾关系。为了解决这个问题，我们提出了一种神经符号管道，该管道利用大型语言模型（LLMs）生成中间隐含前提，并将其转换为逻辑公式，与表示显性前提和显性主张的逻辑公式结合使用，以展示它们之间的逻辑关系（蕴涵、矛盾或中立）。我们的方法在微文本论证语料库（Microtext Argumentative Corpus）上进行了评估。
cs.CL / 58 / 2608.18825
Understanding Multilingual Medical ASR Adaptation Through Layer-Wise Analysis
通过逐层分析理解多语言医疗自动语音识别的适应性
Abstract
Medical automatic speech recognition (MedASR) requires adaptation to specialised terminology, limited annotated clinical data, and multilingual use cases. Although large-scale pretrained ASR models such as Whisper achieve strong generalisation, their behaviour after medical and multilingual adaptation remains insufficiently understood beyond word error rate (WER). This paper investigates how multilingual medical adaptation reshapes the internal representations of Whisper models through layer-wise encoder analysis. We compare zero-shot decoding, English-only fine-tuning, German-only diagnostic fine-tuning, two-stage EN->EN+DE continuation, and direct EN+DE fine-tuning across Whisper model sizes. Fine-tuning substantially improves MedASR performance, but the best model depends on the adaptation setting: Whisper-Medium gives the lowest English WER (7.72%) and the lowest combined EN+DE WER under direct EN+DE training (26.30%); German-only Whisper-Large-v3 gives the lowest German WER (44.96%), but as a within-corpus diagnostic on 86 single-speaker training utterances rather than robust generalisation. Layer-wise analysis of the two-stage Whisper-Small trajectory shows that English medical fine-tuning produces the dominant encoder shift, whereas multilingual continuation largely preserves the adapted representation space. Domain and language information remain highly recoverable across layers, while linearly recoverable error-predictive cues weaken as WER improves.
Chinese Translation
医疗自动语音识别（MedASR）需要适应专业术语、有限的标注临床数据以及多语言使用场景。尽管像 Whisper 这样的规模庞大的预训练 ASR 模型在泛化能力上表现出色，但在医疗和多语言适应后的行为，除了字错误率（WER）之外仍然未被充分理解。本文通过逐层编码器分析研究多语言医疗适应如何重塑 Whisper 模型的内部表示。我们比较了零样本解码、仅英语微调、仅德语诊断微调、两阶段 EN->EN+DE 继续训练以及直接 EN+DE 微调在不同 Whisper 模型规模下的表现。微调显著提高了 MedASR 的性能，但最佳模型取决于适应设置：Whisper-Medium 在直接 EN+DE 训练下提供了最低的英语 WER（7.72%）和最低的综合 EN+DE WER（26.30%）；而仅德语的 Whisper-Large-v3 在 86 个单说话者训练语句上作为内部语料库诊断提供了最低的德语 WER（44.96%），但并不具备强健的泛化能力。对两阶段 Whisper-Small 轨迹的逐层分析显示，英语医疗微调产生了主导的编码器偏移，而多语言继续训练在很大程度上保留了适应后的表示空间。领域和语言信息在各层之间仍然高度可恢复，而线性可恢复的错误预测线索在 WER 改善时减弱。
cs.CL / 59 / 2608.18888
Assessing Quality of Experience in Natural Language Generation of German Text
评估德语文本自然语言生成的体验质量
Abstract
The rapid advancement of Natural Language Generation (NLG) has made the reliable evaluation of generated text increasingly critical, as these systems, such as large language models (LLMs), are now widely deployed in real-world applications. However, traditional automatic metrics fail to capture the multifaceted nature of perceived quality. In this paper, we introduce TextQ-German, a novel dataset suite for human-centered evaluation of German NLG from a Quality of Experience (QoE) perspective, covering automatic text summarization and machine translation. Through crowdsourcing studies with German speakers, we collect human quality ratings and identify relevant perceptual quality dimensions for each task. We develop automatic QoE prediction models, including transformer-based, linguistic feature-based, and hybrid approaches. Hybrid models outperform pure transformer baselines in almost all experimental settings, while linguistic features alone can approach the performance of fine-tuned language models. The dataset is extended with LLM-generated outputs annotated with overall QoE scores. Final validation on held-out sets indicates generalization to unseen data. Our work contributes a publicly accessible resource for NLG evaluation and baselines for automatic QoE prediction, providing a foundation for developing NLG systems that better align with human quality perception.
Chinese Translation
自然语言生成（NLG）的快速发展使得对生成文本的可靠评估变得愈加重要，因为这些系统，如大型语言模型（LLMs），现已广泛应用于现实世界的应用中。然而，传统的自动化指标无法捕捉感知质量的多面性。本文介绍了TextQ-German，一个新颖的数据集套件，旨在从体验质量（QoE）的角度对德语NLG进行以人为中心的评估，涵盖自动文本摘要和机器翻译。通过与德语使用者的众包研究，我们收集了人类质量评分，并识别出每个任务相关的感知质量维度。我们开发了自动QoE预测模型，包括基于变换器的模型、基于语言特征的模型和混合方法。混合模型在几乎所有实验设置中均优于纯变换器基线，而仅使用语言特征的模型的性能也接近微调后的语言模型。该数据集还扩展了带有整体QoE评分的LLM生成输出。对保留集的最终验证表明模型能够泛化到未见数据。我们的工作为NLG评估提供了一个公开可访问的资源，并为自动QoE预测提供了基线，为开发更符合人类质量感知的NLG系统奠定了基础。
cs.CL / 60 / 2608.18921
SMTrap: Cost-Effective DoS Attacks Against Large Reasoning Models via SMT Conflict Guidance
SMTrap：通过 SMT 冲突引导对大型推理模型的成本有效的拒绝服务攻击
Abstract
Existing LRM-DoS methods rely heavily on model feedback to synthesize attack queries, requiring either repeated queries to the target model or training a dedicated attack model. These expensive operations severely weaken attack leverage. In this paper, we propose \emph{search amplification}, a novel, model-feedback-free LRM-DoS paradigm. It employs the conflict count derived from an Satisfiability Modulo Theories (SMT) solver as a low-cost external signal to guide the synthesis of inference-heavy Constraint Satisfaction Problem (CSP) instances. Our key observation is that LRMs depend on trial-and-backtracking search when solving CSPs, where higher SMT conflict counts on a given CSP instance positively correlate with more extensive LRM backtracking search and substantially longer output trajectories. Building on this finding, we propose \textsc{SMTrap}, a lightweight, CPU-only framework. Guided by SMT conflict counts, \textsc{SMTrap} generates inference-heavy CSP queries without model queries, attack-model training, or GPU computation. Evaluations across seven frontier models demonstrate the state-of-the-art LRM-DoS capability of \textsc{SMTrap}, producing DoS effects multiple times stronger than existing baselines. To mitigate the threat of \textsc{SMTrap}, we demonstrate a tool-based mitigation that significantly cuts token usage.
Chinese Translation
现有的 LRM-DoS 方法在很大程度上依赖于模型反馈来合成攻击查询，这需要对目标模型进行重复查询或训练专门的攻击模型。这些昂贵的操作严重削弱了攻击的杠杆作用。本文提出了一种新颖的、无需模型反馈的 LRM-DoS 范式—— extit{搜索放大}。该方法利用来自可满足性模理论（Satisfiability Modulo Theories, SMT）求解器的冲突计数作为低成本的外部信号，指导推理密集型约束满足问题（Constraint Satisfaction Problem, CSP）实例的合成。我们的关键观察是，LRMs 在解决 CSP 时依赖于试探和回溯搜索，给定 CSP 实例的较高 SMT 冲突计数与更广泛的 LRM 回溯搜索和显著更长的输出轨迹呈正相关。基于这一发现，我们提出了 extsc{SMTrap}，一个轻量级的仅 CPU 框架。在 SMT 冲突计数的指导下， extsc{SMTrap} 生成推理密集型的 CSP 查询，而无需模型查询、攻击模型训练或 GPU 计算。在七个前沿模型上的评估表明， extsc{SMTrap} 展现了最先进的 LRM-DoS 能力，其产生的拒绝服务效果比现有基线强多倍。为了减轻 extsc{SMTrap} 的威胁，我们展示了一种基于工具的缓解方法，显著减少了令牌使用。
cs.CL / 61 / 2608.18931
Test-Time Scaling in the Wild: Why Exploitation, Not Exploration, Is the Bottleneck
测试时缩放在实际应用中的瓶颈：为何利用而非探索是限制因素
Abstract
Test-time scaling (TTS) improves language model outputs by spending additional inference compute - generating multiple candidates, searching over partial sequences, or iteratively refining drafts. These techniques yield large gains on mathematics and code, but have been developed and stress-tested almost exclusively on tasks where verification is straightforward. We conduct the first compute-normalised comparison of five TTS families across five open-ended generation benchmarks spanning medicine, law, finance, general chat, and creative writing - grounded in a unified framework that decomposes the effectiveness of each method's token budget into exploration and exploitation. The answer depends on which side of that decomposition you examine. Scaling exploration works: the best candidate in the pool improves steadily with compute across all settings. What breaks is exploitation - the step that converts a rich candidate pool into a final output. With state-of-the-art generators, reward models correlate at only $\rho_v \approx 0.12$ with true quality, rendering selection near-random regardless of budget. Tree search amplifies this failure through diversity collapse. Refinement helps on one of five benchmarks; its apparent gains elsewhere are confounded. Only synthesis across candidates (Fusion) consistently improves over single-sample baselines, yet still recovers only ~40% of available quality. The candidate pool is not the bottleneck - choosing from it is.
Chinese Translation
测试时缩放（Test-time scaling, TTS）通过增加推理计算来改善语言模型的输出——生成多个候选项、对部分序列进行搜索或迭代优化草稿。这些技术在数学和代码方面带来了显著的提升，但几乎仅在验证相对简单的任务上进行开发和压力测试。我们首次对五个TTS家族在五个开放式生成基准（涵盖医学、法律、金融、一般聊天和创意写作）进行计算归一化比较，基于一个统一框架，将每种方法的令牌预算的有效性分解为探索和利用。答案取决于你检查分解的哪一侧。扩展探索是有效的：在所有设置中，候选池中的最佳候选项随着计算的增加而稳步提升。问题出在利用上——将丰富的候选池转化为最终输出的步骤。使用最先进的生成器时，奖励模型与真实质量的相关性仅为 $
ho_v ext{ 约 } 0.12$，使得无论预算如何，选择几乎是随机的。树搜索通过多样性崩溃加剧了这一失败。在五个基准中，优化在其中一个基准上有所帮助；而在其他基准上其表面上的收益则受到混淆。只有在候选项之间的合成（Fusion）始终优于单样本基线，但仍然仅能恢复约40%的可用质量。候选池并不是瓶颈——从中选择才是。
cs.CL / 62 / 2608.18937
MedUAG: Unified Understanding and Generation for Medical Multimodal Models
MedUAG：医学多模态模型的统一理解与生成
Abstract
Recent Multimodal Large Language Models (MLLMs) are rapidly evolving into unified understanding and generation (UAG) frameworks. However, extending these unified paradigms to the medical domain is hindered by: the absence of comprehensive training and evaluation benchmarks, and the lack of broadly validated unified medical model. To address these gaps, we present a comprehensive foundation for medical UAG. First, we construct MedUAGCorpus, the largest unified medical understanding and generation dataset to date, comprising over 6 million instances across 14 imaging modalities. Second, we introduce MedUAGBench, a systematic benchmark that expands medical generation evaluation to 12 diverse tasks under standardized protocols. Finally, leveraging these resources, we develop MedUAG, an end-to-end trained unified medical model. Extensive experiments demonstrate that MedUAG achieves strong performance across a wide array of understanding and generation tasks, establishing a competitive baseline and paving the way for next-generation medical multimodal systems.
Chinese Translation
近期，多模态大型语言模型（MLLMs）迅速演变为统一理解与生成（UAG）框架。然而，将这些统一范式扩展到医学领域面临着以下挑战：缺乏全面的训练和评估基准，以及缺乏广泛验证的统一医学模型。为了解决这些问题，我们提出了医学UAG的全面基础。首先，我们构建了MedUAGCorpus，这是迄今为止最大的统一医学理解与生成数据集，包含超过600万个实例，涵盖14种影像模态。其次，我们介绍了MedUAGBench，这是一个系统化的基准，扩展了医学生成评估至12个多样化任务，并在标准化协议下进行。最后，利用这些资源，我们开发了MedUAG，一个端到端训练的统一医学模型。大量实验表明，MedUAG在广泛的理解与生成任务中表现出色，建立了竞争基线，为下一代医学多模态系统铺平了道路。
cs.CL / 63 / 2608.18972
Institutional Newspapers Pipeline: Deriving billions of high quality tokens from historical newspapers
机构报纸管道：从历史报纸中提取数十亿高质量标记
Abstract
Historical newspapers are an abundant record of public life, but their dense, irregular and sometimes noisy layouts make computational access to these materials both challenging and limited. We present the Institutional Newspapers Pipeline, a modular system we jointly designed with Boston Public Library to extract high-quality, structured datasets from historical newspaper scans. It was architected so that each step remains interpretable and customizable, and so that the pipeline as a whole remains computationally frugal enough to run on workstation-level hardware. The pipeline runs each scan through a multi-step process: it segments scans into individual type-agnostic crops and performs OCR on each resulting segment before then performing text analysis, type classification, reading order detection, named entities recognition, subject classification, language detection, and pre-computed embeddings generation on every crop. We ran this pipeline against a portion of Boston Public Library's holdings and released the results as an open dataset. The optical character recognition (OCR) output represents 16.3 billion o200k_base tokens across 83.1 million individual crops, extracted from 1,473,635 public domain newspaper scans published between 1795 and 1930. This report describes our methods for each processing step, the small models we trained, as well as the evaluation results and dataset-scale measurements we collected in the process. It accompanies the release of the pipeline, models, and dataset. We position this work as a substantial step towards unlocking high-quality data from tens of millions of newspaper scans.
Chinese Translation
历史报纸是公共生活的丰富记录，但其密集、不规则且有时嘈杂的版面使得对这些材料的计算机访问既具有挑战性又受到限制。我们提出了机构报纸管道（Institutional Newspapers Pipeline），这是我们与波士顿公共图书馆共同设计的一个模块化系统，用于从历史报纸扫描中提取高质量、结构化的数据集。该系统的架构确保每个步骤都具有可解释性和可定制性，同时整个管道的计算资源消耗足够低，可以在工作站级别的硬件上运行。该管道将每个扫描件经过多步骤处理：将扫描件分割为单个与类型无关的裁剪，并对每个生成的片段进行光学字符识别（OCR），然后对每个裁剪进行文本分析、类型分类、阅读顺序检测、命名实体识别、主题分类、语言检测和预计算嵌入生成。我们对波士顿公共图书馆的一部分藏品运行了该管道，并将结果发布为开放数据集。光学字符识别（OCR）输出代表了从1795年至1930年间出版的1,473,635个公共领域报纸扫描中提取的83.1百万个单独裁剪中的16.3亿个o200k_base标记。本报告描述了我们在每个处理步骤中的方法、我们训练的小模型，以及在此过程中收集的评估结果和数据集规模测量。它伴随管道、模型和数据集的发布。我们将这项工作视为解锁数千万报纸扫描中高质量数据的重要一步。
cs.CL / 64 / 2608.18988
DeepWeaver: Bridging the Evidence Synthesis Gap in Open-Ended Question Answering
DeepWeaver：弥合开放式问题回答中的证据综合差距
Abstract
Retrieve-then-generate pipelines are commonly used to produce deep-research answers for open-ended questions, but retrieval alone is insufficient: LLMs must organize noisy and fragmented evidence into comprehensive, well-cited answers. We refer to this process as evidence synthesis. However, direct generation often underuses evidence, misaligns citations, and collapses diverse information into shallow summaries, exposing an evidence synthesis gap between retrieval and generation. Thus, we propose DeepWeaver, a novel framework that weaves noisy retrieved evidence into comprehensive answers by maintaining Thought Block Chains (TBCs), a structured representation that groups claims, salient information, keywords, and supporting evidence. DeepWeaver uses subordinate TBCs to inspect residual evidence, commit TBC revisions, and discover new claims before final generation. We evaluate DeepWeaver on open-ended QA over both knowledge bases and the web, and introduce LoQA, a high-density benchmark for evidence synthesis. Across multiple LLMs, DeepWeaver improves content sufficiency, citation grounding, and detail preservation on LoQA, while achieving deeper insights and higher citation quality on DeepResearch Bench. These results show that evidence weaving is an effective mechanism for bridging retrieval and generation in open-ended QA. Our code is available at https://github.com/KlozeWang/DeepWeaver.
Chinese Translation
检索-再生成管道通常用于为开放式问题生成深入的研究答案，但仅靠检索是不够的：大型语言模型（LLMs）必须将嘈杂和零散的证据组织成全面且引用准确的答案。我们将这一过程称为证据综合。然而，直接生成常常未能充分利用证据，引用不对齐，并将多样的信息压缩为肤浅的摘要，暴露了检索与生成之间的证据综合差距。因此，我们提出了DeepWeaver，一个新颖的框架，通过维护思维区块链（Thought Block Chains, TBCs）这一结构化表示，将嘈杂的检索证据编织成全面的答案，TBCs将主张、显著信息、关键词和支持证据进行分组。DeepWeaver使用从属TBCs来检查残余证据、提交TBC修订，并在最终生成之前发现新主张。我们在知识库和网络上的开放式问答中评估了DeepWeaver，并引入了LoQA，一个高密度的证据综合基准。在多个大型语言模型上，DeepWeaver提高了LoQA的内容充足性、引用基础和细节保留，同时在DeepResearch Bench上实现了更深入的见解和更高的引用质量。这些结果表明，证据编织是一种有效的机制，可以弥合开放式问答中的检索与生成之间的差距。我们的代码可在https://github.com/KlozeWang/DeepWeaver获取。
cs.CL / 65 / 2608.19003
Structure, Association, and Decision Value: Representation-Based Difficulty Estimation for Adaptive Inference in African-Language NLI
结构、关联与决策价值：基于表征的非洲语言自然语言推理中的自适应推理难度估计
Abstract
We ask whether internal representation statistics can provide useful example-level difficulty signals for adaptive inference in multilingual African NLP, and find that they cannot in this setting. Studying natural language inference across 15 African languages with frozen off-the-shelf checkpoints, we report four results. First, AfriXNLI's English configuration shares 1,047 of its 1,050 examples verbatim with XNLI evaluation data, and one widely used NLI checkpoint scores 1.000 on that test split, consistent with XNLI test exposure. Because AfriXNLI is derived from XNLI, its English, French and Swahili configurations cannot serve as clean evaluations for XNLI-trained models. Second, parameter count does not reliably order capability across African languages: our larger checkpoint is better in seven languages and worse in eight, with no significant aggregate difference. Third, across three multilingual representation spaces, angular dispersion is consistently more language-determined than effective rank, so pooled correlations can inflate one and mask the other. Fourth, the association that survives language control depends on the target: effective rank predicts probability gain from escalation but not whether escalation changes the prediction, while cheap-model confidence shows the opposite pattern; the two targets correlate at only 0.655. Under the tested models, signals, and compute budgets, no evaluated signal makes adaptive routing preferable to always-expensive inference, although an oracle exceeds it by 11 accuracy points at 60% of the compute. Our central methodological finding is that a representation statistic can be statistically significant for one notion of computational benefit while being irrelevant to another, and therefore be a poor decision variable.
Chinese Translation
我们探讨内部表征统计是否能够为多语言非洲自然语言处理中的自适应推理提供有用的示例级难度信号，结果发现它们在此环境中并不能提供帮助。通过研究15种非洲语言的自然语言推理，使用冻结的现成检查点，我们报告了四个结果。首先，AfriXNLI的英语配置与XNLI评估数据逐字共享1,047个示例中的1,050个，而一个广泛使用的NLI检查点在该测试分割上得分为1.000，这与XNLI测试曝光一致。由于AfriXNLI源自XNLI，其英语、法语和斯瓦希里语配置不能作为XNLI训练模型的干净评估。其次，参数数量并不能可靠地排序非洲语言的能力：我们的较大检查点在七种语言中表现更好，而在八种语言中表现更差，总体上没有显著差异。第三，在三个多语言表征空间中，角度离散度始终比有效秩更受语言决定，因此汇总相关性可能夸大一个而掩盖另一个。第四，经过语言控制后存活下来的关联依赖于目标：有效秩预测升级带来的概率增益，但并不预测升级是否改变了预测，而廉价模型的置信度则显示出相反的模式；这两个目标的相关性仅为0.655。在测试的模型、信号和计算预算下，没有评估的信号使自适应路由优于始终昂贵的推理，尽管一个神谕在60%的计算下超出了11个准确度点。我们中心的方法论发现是，某一表征统计在一种计算效益的概念上可能具有统计显著性，而在另一种上则无关，因此可能是一个不良的决策变量。
cs.CL / 66 / 2608.19006
Introducing the Privacy-HSD Trade-off: Hate Speech Detection, but not at the Cost of Privacy
引入隐私-HSD权衡：仇恨言论检测，但不以隐私为代价
Abstract
Hate speech is a real and timely threat that affects a large portion of online users, especially youth and minority groups. While building reliable and robust automatic hate speech detection (HSD) systems is paramount, we argue that this must also be balanced with the individual right to privacy. Exploring the intersection of HSD and privacy, we demonstrate that HSD systems might unintentionally achieve performance at the cost of encoding authorship, posing a threat to privacy. Building on these findings, we establish the notion of a privacy-HSD trade-off, which demands a careful balance. We benchmark a series of text privatization methods, as well as our newly proposed domain-specific AgnoSpeech technique, showing that balancing privacy and HSD is difficult but feasible. The findings make a strong case for more research on the trade-offs between privacy and HSD, both of which have tangible implications for the safeguarding of online participation.
Chinese Translation
仇恨言论是一个现实且紧迫的威胁，影响着大量在线用户，尤其是年轻人和少数群体。虽然构建可靠且稳健的自动仇恨言论检测（HSD）系统至关重要，但我们认为这必须与个人隐私权相平衡。在探讨HSD与隐私的交集时，我们展示了HSD系统可能在无意中以编码作者身份为代价来实现性能，从而对隐私构成威胁。在这些发现的基础上，我们建立了隐私-HSD权衡的概念，这要求谨慎的平衡。我们对一系列文本隐私化方法进行了基准测试，以及我们新提出的特定领域AgnoSpeech技术，显示出平衡隐私与HSD既困难又可行。这些发现强烈呼吁对隐私与HSD之间权衡的更多研究，这两者对保护在线参与具有切实的影响。
cs.CL / 67 / 2608.19009
Grading the Graders: Verification Autonomy Levels (L0-L5) for LLM Reasoning
评估评估者：大型语言模型推理的验证自主级别 (L0-L5)
Abstract
Large language models (LLMs) are increasingly paired with verifiers (step checkers, self-consistency filters, tool-based fact checkers, formal proof assistants) that claim to detect the model's errors. Yet the verification literature uses the word "level" to mean at least five different things: verification granularity, concept abstraction, risk tier, system-stack layer, and the epistemic source of the ground truth. We propose Verification Autonomy Levels (VAL), a meta-standard classifying verification schemes along a single axis: where does the verification spec come from, and what does the verdict guarantee? VAL ranges from L0 (LLM self-declaration, no deterministic anchor) through L2 (objective ground truth, correctness only) to L3/L4 (decidable systems with single-property or domain-level completeness), with L5 impossible in the unrestricted case. Central to VAL is the completeness blind spot: substitution- and sampling-based verifiers can confirm that proposed candidates hold, but cannot prove that no candidate was missed. We further identify a dichotomy the literature has not stated: completeness is reachable only for formally specifiable properties, while empirical open-world verification (fact-checking, diagnosis) caps at anchored correctness (L2). We document this across four domains (symbolic mathematics, behavior monitoring, medical diagnosis, and code generation) and in the strongest existing formal-verification baseline, whose authors note the verifier "focuses on the correctness of each step." We show the levels of granularity, concept hierarchy, risk, and system stack are orthogonal to VAL, resolving a systematic conflation across 17 surveyed papers. Code and full assessment are released as supplementary material.
Chinese Translation
大型语言模型（LLMs）越来越多地与验证器（步骤检查器、自我一致性过滤器、基于工具的事实检查器、形式证明助手）配对，这些验证器声称能够检测模型的错误。然而，验证文献中使用“级别”一词来表示至少五种不同的含义：验证粒度、概念抽象、风险层级、系统堆栈层和真实信息的认识来源。我们提出了验证自主级别（Verification Autonomy Levels, VAL），这是一个元标准，通过单一轴线对验证方案进行分类：验证规范来自何处，以及裁决保证了什么？VAL的范围从L0（LLM自我声明，无确定性锚点）到L2（客观真实信息，仅正确性），再到L3/L4（具有单一属性或领域级完整性的可判定系统），在不受限制的情况下L5是不可能的。VAL的核心是完整性盲点：基于替代和采样的验证器可以确认提议的候选者成立，但无法证明没有遗漏任何候选者。我们进一步识别出文献中未曾表述的二分法：完整性仅对形式可指定的属性可达，而经验开放世界验证（事实检查、诊断）则限制在锚定的正确性（L2）。我们在四个领域（符号数学、行为监测、医学诊断和代码生成）以及在最强的现有形式验证基准中记录了这一点，其作者指出验证器“专注于每一步的正确性”。我们展示了粒度、概念层次、风险和系统堆栈的级别与VAL是正交的，解决了17篇调查论文中的系统性混淆。代码和完整评估作为补充材料发布。
cs.CL / 68 / 2608.19026
Institutional Books - Enriched Text: A customizable multilingual open-source pipeline for denoising, deduplicating, and annotating OCR text at scale
机构图书 - 丰富文本：一个可定制的多语言开源管道，用于大规模去噪、去重和注释OCR文本
Abstract
Released in 2025, Institutional Books: Harvard Library (IB-HL) is a collection of 983,004 volumes (242B o200k_base tokens), originally digitized through Harvard Library's participation in the Google Books Library project. As researchers and developers have begun to use IB-HL, a tension has emerged between standard large-scale preprocessing practices and the goals of careful information stewardship. Many existing pipelines optimize for web text: as a result, they tend to aggressively filter, deduplicate, restrict by language, and sometimes discard meaningful metadata. Meanwhile, researchers seeking to use IB-HL duplicate effort while performing similar processing and analysis. We describe an approach that we call Enriched Text. Instead of producing a single 'complete' stream of tokens, we normalize the text while preserving metadata through annotations. We separate endmatter, detect per-paragraph language, identify clusters of duplicate paragraphs, and compute per-paragraph bits-per-byte scores. We provide this information through HTML-like annotations layered on top of the text. By parsing these annotations, users can tailor the output to their own needs instead of accepting a global editorial decision on content. The pipeline applies to all $\approx$250 languages in the collection. This report describes this project's goals, implementation, and design rationale. The release includes IB-HL-ET (an enriched-text version of IB-HL containing 217B o200k_base tokens across 983,003 volumes, organized into 1.39B annotated subtopic paragraphs) and the pipeline that produced it. These serve to make the collection easier for machines to parse and for humans to study.
Chinese Translation
2025年发布的《机构图书：哈佛图书馆》（Institutional Books: Harvard Library，IB-HL）是一个包含983,004卷（242B o200k_base tokens）的文献集合，最初通过哈佛图书馆参与谷歌图书图书馆项目进行数字化。随着研究人员和开发者开始使用IB-HL，标准的大规模预处理实践与谨慎的信息管理目标之间出现了紧张关系。许多现有的管道优化了网络文本：因此，它们往往会过度过滤、去重、按语言限制，有时还会丢弃有意义的元数据。同时，寻求使用IB-HL的研究人员在进行类似处理和分析时重复了工作。我们描述了一种我们称之为“丰富文本”（Enriched Text）的方法。我们不再生成单一的“完整”令牌流，而是在保留元数据的同时对文本进行规范化。我们分离附录，检测每段的语言，识别重复段落的聚类，并计算每段的比特每字节分数。我们通过类似HTML的注释将这些信息叠加在文本上。通过解析这些注释，用户可以根据自己的需求定制输出，而不是接受对内容的全球编辑决策。该管道适用于该集合中的所有约250种语言。本报告描述了该项目的目标、实施和设计原理。发布内容包括IB-HL-ET（IB-HL的丰富文本版本，包含983,003卷中的217B o200k_base tokens，组织成1.39B注释子主题段落）以及生成该版本的管道。这些旨在使该集合更易于机器解析和人类研究。
cs.CL / 69 / 2608.19124
Intercepting the Kangaroo: Experimental Astrolinguistics with Constructed Lexicons, Active Probing, and Large Language Models as Informants and Hypothesis Proposers
拦截袋鼠：使用构造词汇、主动探测和大型语言模型作为信息提供者和假设提出者的实验性星际语言学
Abstract
Astrolinguistics -- communication with minds that categorize reality differently from ours -- has been purely speculative since Freudenthal's Lincos (1960). We make it experimental. Two language models with deliberately incompatible constructed lexicons (one encoding shape, color, and motion; the other fusing color with motion, encoding parity, and lacking shape) serve as informants with complete ground truth, while a fully scripted orchestrator translates between the two category systems. The central failure mode is the kangaroo effect: the silent attachment of a word to the wrong referent -- Quine's indeterminacy of translation, operationalized. Across 400+ simulated and live runs, a protocol combining cross-situational elimination, pre-registered predictive probes, active scene selection, a stricter recovery round, and quarantine produced no undetected mistranslations under the tested conditions and exceeded a passive baseline's coverage (d = 0.62). Injected kangaroo traps defeated naive ostension and pure statistical learning in 100% of runs, while the full protocol intercepted every decoy and, where discriminating evidence is ontologically unavailable, declared Quinean equivalence classes instead of guessing. Under informant noise it degrades gracefully: zero kangaroos persist up to 2% per-word noise; at 10% the protocol predominantly abstains rather than errs. Finally, words outside the scripted hypothesis space (a history-dependent relational term and an XOR contextual homonym) are recovered by a generate-and-test loop in which an LLM proposes rules and the script verifies them: coverage scales with proposer capability (0% -> 18% -> 72% -> 100%) while undetected mistranslations stayed at zero throughout. In the tested conditions, correctness is a property of the protocol; coverage is a property of the instruments.
Chinese Translation
星际语言学——与那些以不同于我们方式分类现实的心灵进行交流——自弗洛伊登塔尔的Lincos（1960）以来一直是纯粹的推测性研究。我们将其转变为实验性研究。两个具有故意不兼容的构造词汇的语言模型（一个编码形状、颜色和运动；另一个将颜色与运动融合，编码奇偶性且缺乏形状）作为具有完整真实信息的提供者，而一个完全脚本化的协调者在两个分类系统之间进行翻译。主要的失败模式是袋鼠效应：一个词静默地附加到错误的指称上——奎因的翻译不确定性得到了操作化。在400多次模拟和实时运行中，结合跨情境消除、预注册预测探测、主动场景选择、更严格的恢复回合和隔离的协议在测试条件下没有产生未被检测到的误翻译，并超过了被动基线的覆盖率（d = 0.62）。注入的袋鼠陷阱在100%的运行中击败了天真的指示和纯统计学习，而完整的协议拦截了每一个诱饵，并在本体论上缺乏区分证据时声明奎因等价类而不是猜测。在信息提供者噪声下，该协议优雅地降级：在每个单词噪声达到2%时，零袋鼠持续存在；在10%时，协议主要选择不参与而不是出错。最后，脚本假设空间之外的词（一个历史依赖的关系术语和一个异或上下文同音词）通过生成与测试循环被恢复，其中一个大型语言模型提出规则，脚本进行验证：覆盖率随着提出者能力的提升而增加（0% -> 18% -> 72% -> 100%），而未被检测到的误翻译始终保持为零。在测试条件下，正确性是协议的属性；覆盖率是工具的属性。
cs.CL / 70 / 2608.19133
Comment-level Topic Drift Analysis in the Reddit Corpus
Reddit 语料库中的评论级主题漂移分析
Abstract
We present a novel application of embedding-based dynamic topic modeling techniques to detect and quantify topic drift at the comment level in a massive corpus. By leveraging pretrained language models to generate contextualized semantic embeddings for short text, we analyzed 12.7 billion Reddit comments spanning 2006 to 2022. Using unsupervised methods on these embeddings, we identify dynamically evolving topic clusters over time. Our primary contribution is a methodology for analysis of semantic drift and discourse evolution in the embedding space itself. We also demonstrate modifications to existing methods that enable this analysis at scale, and we propose and demonstrate a null model comparison test to filter spurious dynamics. Key findings suggest that politically and socially contentious topics exhibit significant directional drift in embedding space, with inter-topic distances changing systematically over time beyond what the null model can explain, whereas domains such as music and sports remain comparatively stable.
Chinese Translation
我们提出了一种新颖的基于嵌入的动态主题建模技术应用，以检测和量化大规模语料库中评论级的主题漂移。通过利用预训练语言模型生成短文本的上下文化语义嵌入，我们分析了从 2006 年到 2022 年的 127 亿条 Reddit 评论。使用无监督方法对这些嵌入进行分析，我们识别出随时间动态演变的主题聚类。我们的主要贡献是提出了一种在嵌入空间中分析语义漂移和话语演变的方法论。我们还展示了对现有方法的修改，使得这种大规模分析成为可能，并提出并演示了一种零模型比较测试，以过滤虚假的动态。关键发现表明，政治和社会争议性话题在嵌入空间中表现出显著的方向性漂移，主题间距离随着时间的推移系统性变化，超出了零模型所能解释的范围，而音乐和体育等领域则相对稳定。
cs.CL / 71 / 2608.19165
ChildSafeAds Shared Task 2026: Commercial Content in Child-Facing YouTube Videos
ChildSafeAds 共享任务 2026：面向儿童的 YouTube 视频中的商业内容
Abstract
ChildSafeAds is a shared task on commercial content in YouTube videos likely to reach children and teenagers. It contains 3,360 videos from 939 channels. Each instance begins with a segment submitted to SponsorBlock, an open-source crowdsourced browser extension whose users mark sponsor segments so that others can skip them. We pair the segment with its available transcript, video and channel information, and a sales or service page linked from the video description. Systems determine what kind of offer is being promoted (ST1), assign product categories (ST2), and identify legal risk flags (ST3). The evidence is divided into four cumulative access levels, from the transcript to the linked page, so results can be compared against the cost of collecting the data. 45.5\% of videos in our data failed to properly use the in-platform ad disclosure method (the ``Includes paid promotion'' label). GPT-5.4 produced the labels after the expert organiser team reviewed samples and iterated on the taxonomy, prompts and model choices. GPT-5.6-luna independently labelled the development set. This report describes the task, data and evaluation. An updated version will add participating systems and shared-task results.
Chinese Translation
ChildSafeAds 是一个关于可能接触儿童和青少年的 YouTube 视频中商业内容的共享任务。该任务包含来自 939 个频道的 3,360 个视频。每个实例以提交给 SponsorBlock 的一段内容开始，SponsorBlock 是一个开源的众包浏览器扩展，其用户标记赞助商段落，以便其他人可以跳过这些内容。我们将该段落与其可用的转录文本、视频和频道信息，以及视频描述中链接的销售或服务页面配对。系统确定所推广的优惠类型（ST1），分配产品类别（ST2），并识别法律风险标志（ST3）。证据分为四个累积访问级别，从转录文本到链接页面，以便结果可以与收集数据的成本进行比较。我们数据中 45.5\% 的视频未能正确使用平台内广告披露方法（“包括付费推广”标签）。在专家组织团队审查样本并对分类法、提示和模型选择进行迭代后，GPT-5.4 生成了标签。GPT-5.6-luna 独立标记了开发集。本报告描述了任务、数据和评估。更新版本将添加参与系统和共享任务结果。
cs.CL / 72 / 2608.19197
SPADE: Self-Play in Adaptive Synthetic Executable Environments
SPADE：自适应合成可执行环境中的自我对弈
Abstract
Continuous self-improvement requires an ever-expanding pool of self-generated, diverse, adaptive goals. For language agents, existing training environment pools (hand-curated, statically synthesized, or frozen-verifier) keep the goal distribution fixed as the learner scales. We introduce SPADE (Self-Play in Adaptive Synthetic Executable Environments), a self-play RL framework in which a single LLM plays two roles: an Environment Designer that writes complete, long-horizon training environments as executable code with an OpenAI Gym-style reset()/step() interface, and a Reasoning Agent that learns to act in them. Each is a stateful, multi-turn environment (state transitions, reward functions, and verification code), so one interface spans reasoning problems and multi-step agentic tool use. The Reasoning Agent's regret is estimated using the gap between its reward with and without privileged hints; in optimizing this regret signal the Environment Designer learns to target environments at the edge of the agent's capabilities while keeping them feasible. Through extensive experimentation, we find several components critical to success: grounding the Environment Designer on documents sampled from a large pretraining corpus, and giving it an accumulated environment memory. Scaling to 30B-parameter models, SPADE improves over the strongest fixed-environment baseline by +5.3 on average across eight held-out math, science, code, and reasoning benchmarks, and lifts the tool-use setting by +5.7 on BFCL-v4 multi-turn and +13.9 on ACEBench-Agent; on the games setting, the margin over the strongest baseline grows with model scale. By making environment design itself a learnable component, SPADE takes a concrete step toward open-ended self-improvement.
Chinese Translation
持续的自我改进需要不断扩展的自我生成、多样化和自适应的目标池。对于语言代理而言，现有的训练环境池（手工策划、静态合成或冻结验证器）在学习者扩展时保持目标分布不变。我们提出了SPADE（自适应合成可执行环境中的自我对弈），这是一个自我对弈的强化学习框架，其中一个大型语言模型（LLM）扮演两个角色：环境设计者（Environment Designer），负责编写完整的、长时域的训练环境，作为具有OpenAI Gym风格的reset()/step()接口的可执行代码，以及推理代理（Reasoning Agent），学习在这些环境中进行行动。每个角色都是一个有状态的、多轮的环境（状态转移、奖励函数和验证代码），因此一个接口跨越了推理问题和多步骤的代理工具使用。推理代理的遗憾通过其在有无特权提示下的奖励差距来估计；在优化这一遗憾信号的过程中，环境设计者学习如何针对代理能力边缘的环境，同时保持其可行性。通过广泛的实验，我们发现几个关键组件对成功至关重要：将环境设计者基于从大型预训练语料库中抽样的文档进行基础训练，并赋予其累积的环境记忆。在扩展到30B参数模型时，SPADE在八个保留的数学、科学、代码和推理基准测试中，平均提高了+5.3，相较于最强的固定环境基线，并在工具使用设置上提高了+5.7（在BFCL-v4多轮测试中）和+13.9（在ACEBench-Agent中）；在游戏设置中，随着模型规模的增加，与最强基线的差距也在扩大。通过将环境设计本身作为一个可学习的组件，SPADE朝着开放式自我改进迈出了具体的一步。