cs.RO / 1 / 2608.23575
Pattern-Derived Visual Swarm Games: Multi-Scale Drone-Vision States for Interception and Sustainability Audits
基于模式的视觉群体游戏：用于拦截和可持续性审计的多尺度无人机视觉状态
Abstract
We convert drone-vision annotation streams into virtual swarm-game states without controlling physical drones. VisDrone and UAVSwarm metadata are compressed into a Bloom representation; deterministic probes produce bounded capability vectors, image-space formations, finite zero-sum payoffs, and human-readable visual overlays. The audit scales from $6\times 6$ to $32\times 32$ finite games and adds a repeated Markov layer with stock, fatigue, adaptation, exposure, stress, budget, data-growth, model-improvement, and entropy-budget state variables. Local screen tuning raises robust screen security from $0.526$ to $0.593$, and the $32\times 32$ tuned screen reaches value $0.616$. A field readout audit shows that fixed-pixel rasters do not improve monotonically: $128\times 128$ accuracy is $67.2\%$ and hotspot error is $0.136$. The diagnosed error is shrinking image-plane bandwidth. A finite empirical-risk encoder over scale-normalized Gaussian bandwidths selects a scale-normalized encoder with $\lambda=1.50$, reaching $77.6\%$ accuracy at $128\times 128$ and reducing joint loss by $0.185$. A server-side audit checks $16{,}777{,}216$ target-localization states, and a 32-round repeated-game audit over $16{,}777{,}216$ trajectories selects a budget-adaptive policy with value $0.461$.
Chinese Translation
我们将无人机视觉标注流转换为虚拟群体游戏状态，而无需控制物理无人机。VisDrone和UAVSwarm元数据被压缩为布隆表示；确定性探测产生有界能力向量、图像空间构型、有限零和收益和人类可读的视觉叠加。审计的规模从$6 imes 6$到$32 imes 32$的有限游戏，并增加了一个重复的马尔可夫层，包含库存、疲劳、适应、暴露、压力、预算、数据增长、模型改进和熵预算状态变量。局部屏幕调优将稳健的屏幕安全性从$0.526$提高到$0.593$，而$32 imes 32$调优后的屏幕达到值$0.616$。现场读出审计显示，固定像素栅格并未单调改善：$128 imes 128$的准确率为$67.2\%$，热点误差为$0.136$。诊断出的误差是图像平面带宽的缩小。一个有限的经验风险编码器在尺度归一化的高斯带宽上选择一个尺度归一化的编码器，$ ext{λ}=1.50$，在$128 imes 128$时达到$77.6\\%$的准确率，并将联合损失减少$0.185$。服务器端审计检查$16{,}777{,}216$个目标定位状态，而在$16{,}777{,}216$条轨迹上的32轮重复游戏审计选择了一个预算自适应策略，值为$0.461$。
cs.RO / 2 / 2608.23629
Macro-Operator Generation and Predicate Selection for TAMP Operator Learning
宏操作生成与谓词选择用于任务与运动规划操作学习
Abstract
Creating symbolic operators by hand is one of the main bottlenecks in deploying Task and Motion Planning systems (TAMP). Recent works show that these operators can instead be learned directly from demonstration data. Existing methods, however, typically learn each action in isolation and cannot capture the recurring multi-step structure of manipulation tasks, so the search becomes intractable on long sequential tasks. A further inefficiency arises in the symbolic state: every provided predicate is evaluated at every search node, even when it never appears in any learned operator. We present a system that addresses both problems together. Its central component is the automatic generation of macro-operators, composite actions that compress a recurring sequence of individual actions into a single planning step. Our system discovers causally linked action pairs directly from the training data, where one action produces exactly the condition that the next one requires, and turns each pair into a new operator. Alongside this, our system prunes every predicate that no learned operator references, which shrinks the symbolic state evaluated at each search node. Together, these changes shorten the effective planning horizon, and the benefit they bring grows with the length of the task. Across four TAMP domains, our method reaches up to a 4.6x planning speedup compared to the baseline method, namely Learning Operators for TAMP. More importantly, it solves a long sequential task that the baseline cannot solve. Macro-operator discovery thus not only accelerates planning but, in certain domains, determines solvability in practice.
Chinese Translation
手动创建符号操作是部署任务与运动规划系统（TAMP）中的主要瓶颈之一。近期的研究表明，这些操作可以直接从演示数据中学习。然而，现有的方法通常孤立地学习每个动作，无法捕捉到操作任务中反复出现的多步骤结构，因此在长序列任务上的搜索变得不可行。此外，在符号状态中还存在进一步的低效：每个提供的谓词在每个搜索节点上都被评估，即使它在任何学习到的操作中从未出现。我们提出了一个系统，旨在同时解决这两个问题。其核心组件是宏操作的自动生成，这是一种复合动作，将一系列重复的单个动作压缩为一个规划步骤。我们的系统直接从训练数据中发现因果关联的动作对，其中一个动作恰好产生下一个动作所需的条件，并将每对动作转化为一个新的操作。同时，我们的系统修剪掉所有未被学习操作引用的谓词，从而缩小在每个搜索节点上评估的符号状态。通过这些变化，规划的有效时间范围缩短，其带来的好处随着任务长度的增加而增长。在四个TAMP领域中，我们的方法与基线方法（即TAMP的操作学习）相比，实现了最高4.6倍的规划加速。更重要的是，它解决了基线无法解决的长序列任务。因此，宏操作的发现不仅加速了规划，而且在某些领域中，实际上决定了可解性。
cs.RO / 3 / 2608.23650
Concept-Guided Exploration: Building Persistent, Actionable Scene Graphs
概念引导的探索：构建持久的可操作场景图
Abstract
The perception of 3D space by mobile robots is rapidly moving from flat metric grid representations to hybrid metric-semantic graphs built from human-interpretable concepts. While most approaches first build metric maps and then add semantic layers, we explore an alternative, concept-first architecture in which spatial understanding emerges from asynchronous concept agents that directly instantiate and manage semantic entities. Our robot employs two spatial concepts (room and door), implemented as autonomous processes within a cognitive distributed architecture. These concept agents cooperatively build a shared scene graph representation of indoor layouts through active exploration and incremental validation. The key architectural principle is hierarchical constraint propagation: Room instantiation provides geometric and semantic priors to guide and support door detection within wall boundaries. The resulting structure is maintained by a complementary functional principle based on prediction-matching loops. This approach is designed to yield an actionable, human-interpretable spatial representation without relying on any pre-existing global metric map, supporting scalable operation and persistent, task-relevant understanding in structured indoor environments.
Chinese Translation
移动机器人对三维空间的感知正迅速从平面度量网格表示转向基于人类可解释概念构建的混合度量-语义图。虽然大多数方法首先构建度量地图，然后添加语义层，但我们探索了一种替代的概念优先架构，其中空间理解源于异步概念代理，这些代理直接实例化和管理语义实体。我们的机器人采用了两个空间概念（房间和门），这些概念作为认知分布式架构中的自主过程实现。这些概念代理通过主动探索和增量验证，共同构建室内布局的共享场景图表示。关键的架构原则是层次约束传播：房间实例化提供几何和语义先验，以指导和支持在墙体边界内的门检测。所得到的结构由基于预测-匹配循环的互补功能原则维护。这种方法旨在生成可操作的、人类可解释的空间表示，而不依赖于任何现有的全局度量地图，支持在结构化室内环境中的可扩展操作和持久的、与任务相关的理解。
cs.RO / 4 / 2608.23831
Learning to Act While Waiting: RL Finetuning of Generalist Robot Policies Under Inference Latency
在等待中学习行动：在推理延迟下对通用机器人策略进行强化学习微调
Abstract
While reinforcement learning (RL) allows generalist robot policies to continually improve during deployment, the large model size of modern generalist policies, such as VLAs, poses a fundamental obstacle to effective RL improvement. In particular, their severe inference latency---which can lead to pauses or jerky movements---can alter the effective environment dynamics and, if not correctly accounted for, break the Markov assumption that RL relies on, causing standard RL algorithms to fail completely. In this work, we introduce a latency-aware framework, Asynchronous RL with Intermediate Information (ARLI), that enables RL-based improvement of generalist policies under inference delays. Our framework builds on asynchronous inference approaches, which interleave action generation with execution to hide latency, and addresses its incompatibility with RL by providing a low-latency RL policy design that maximizes reactivity within the inference window through two contributions: state augmentations that restore near-Markovian structure by incorporating committed actions and a mid-inference observation. We evaluate our approach across simulated and real-world manipulation tasks, and find that it enables effective finetuning under inference delays where standard RL fails entirely, even matching or exceeding the performance of standard RL in idealized no-latency settings.
Chinese Translation
虽然强化学习（RL）允许通用机器人策略在部署过程中不断改进，但现代通用策略（如 VLA）的庞大模型规模对有效的 RL 改进构成了根本障碍。特别是，它们严重的推理延迟——可能导致暂停或抖动的动作——可以改变有效的环境动态，并且如果没有正确考虑，会破坏 RL 所依赖的马尔可夫假设，从而导致标准 RL 算法完全失效。在本研究中，我们提出了一种考虑延迟的框架，即带有中间信息的异步强化学习（ARLI），该框架使得在推理延迟下基于 RL 的通用策略改进成为可能。我们的框架建立在异步推理方法之上，这些方法将动作生成与执行交错进行，以隐藏延迟，并通过提供一种低延迟的 RL 策略设计来解决其与 RL 的不兼容性，该设计通过两项贡献最大化推理窗口内的反应性：状态增强，通过结合已承诺的动作和中间推理观察来恢复近马尔可夫结构。我们在模拟和真实世界的操作任务中评估了我们的方法，发现它能够在标准 RL 完全失效的推理延迟下实现有效的微调，甚至在理想的无延迟环境中匹配或超越标准 RL 的性能。
cs.RO / 5 / 2608.23839
Resilience Matters for Embodied Agents System: New Metrics, Systematic Evaluation, and Optimization
韧性对具身智能体系统的重要性：新指标、系统评估与优化
Abstract
Embodied Agents System (EAS) are increasingly deployed in open-world physical domains, where reliability directly dictates deployment quality and human-agent trust. However, existing evaluations rely on outcome-centric metrics as success rate or safety scores that collapse diverse execution trajectories into coarse scores, obscuring the dynamic processes underlying agent behavior. Therefore, they ignore a critical property of EAS -- which we define as the Resilience -- that reflects how EASs recover, stabilize, and extend under perturbations and across iterative updates. The lack of resilience is particularly critical in open-world environments due to continuous unexpected disruptions, thus directly affecting the quality of EAS deployment. To address this problem, we gain insight from the resilience-engineering concepts to EAS groundings and propose a novel resilience evaluation framework that can be flexibly applied to any EAS. Specifically, we define the first comprehensive resilience metrics suite for EASs system that exposes Rebound, Stability, and Graceful Extensibility across embodied tasks execution, providing a practical grounding for EAS resilience analysis. We further implement the resilience evaluation layer that transforms execution process into assessments for diagnosis and optimization. Across 400 household tasks with 10 EAS, we reveal the process-level distinction hidden by outcome metrics, including recovery cost differences among successful episodes ($\Delta C_{rec}=25.2$), increased instability and task-family degradation. Metrics-guided optimizations reduce recovery cost and increase stability, graceful extensibility completion, showing the diagnostic effect of resilience evaluation. Our results reveal a trade-off among resilience characteristics, suggesting that a resilient EAS construction should be configured according to deployment-specific requirements.
Chinese Translation
具身智能体系统（EAS）越来越多地应用于开放世界物理领域，其中可靠性直接决定了部署质量和人机信任。然而，现有评估依赖于以结果为中心的指标，如成功率或安全评分，这些指标将多样的执行轨迹简化为粗略的评分，掩盖了智能体行为背后的动态过程。因此，它们忽视了EAS的一个关键属性——我们将其定义为韧性——反映了EAS在扰动和迭代更新中如何恢复、稳定和扩展。由于持续的意外干扰，韧性的缺乏在开放世界环境中尤为关键，因此直接影响EAS的部署质量。为了解决这一问题，我们借鉴韧性工程的概念，提出了一种新颖的韧性评估框架，可以灵活应用于任何EAS。具体而言，我们定义了首个全面的EAS韧性指标套件，揭示了具身任务执行中的反弹、稳定性和优雅扩展性，为EAS韧性分析提供了实用的基础。我们进一步实现了韧性评估层，将执行过程转化为诊断和优化的评估。在对400个家庭任务和10个EAS的研究中，我们揭示了结果指标所掩盖的过程级别区别，包括成功情节之间的恢复成本差异（$ ext{Δ}C_{rec}=25.2$）、不稳定性增加和任务家族退化。基于指标的优化减少了恢复成本并提高了稳定性和优雅扩展完成度，显示了韧性评估的诊断效果。我们的结果揭示了韧性特征之间的权衡，建议韧性EAS的构建应根据特定部署需求进行配置。
cs.RO / 6 / 2608.23863
DreamLedger: Execution-Settled Credit Files for World-Model Imagination in Robot Decision Loops
梦账本：机器人决策循环中的执行结算信用文件用于世界模型想象
Abstract
Robots are beginning to act on world-model predictions, yet reliability is still expressed through instantaneous, model-internal signals. DreamLedger instead treats reliability as a persistent deployment object: an execution-settled credit file recording how often consumed predictions are borne out, indexed by operating condition, region, and prediction horizon, and consulted before each use. Each consumed prediction is registered as a claim; attributable outcomes are settled against arriving reality at zero labeling cost, an attribution stage excludes measurement-contaminated outcomes, and a settlement-supervised head complements sparse bins. The resulting credit gates consumption: low-credit predictions shorten the dependent horizon or trigger additional observation; every reliance event remains auditable via dependency tickets and replayable logs. We evaluate DreamLedger in three simulated domains (indoor flight, tabletop manipulation, 2D navigation), via mounts on unmodified DreamerV3, TD-MPC2, and V-JEPA 2-AC, and on a real Franka manipulator. Claim failure is dose-monotone in all 12 held-out condition-horizon cells. Credit-gated planning reduces burned imagination (consumed claims that later fail to redeem) by 62% (95% CI 43-81%) versus blind consumption, with equal success and comparable collision rates. At matched risk targets, persistent books cut verification probes from 1.00 to 0.36/episode in manipulation, at success 0.94 versus 0.98; settlement-grounded calibration retains moderate, seed-consistent operating points unlike raw instantaneous gates. The same trust layer operates across decoder-, latent-, and token-space interfaces, including V-JEPA 2-AC settled on real robot frames. On hardware, settlement remains operational under real sensing and contact noise, a deployment failure loop is re-priced online, and all 1,062 registered spends replay from the audit logs.
Chinese Translation
机器人开始根据世界模型的预测进行行动，但其可靠性仍然通过瞬时的模型内部信号来表达。梦账本则将可靠性视为一个持久的部署对象：一个执行结算的信用文件，记录消费预测的兑现频率，并按操作条件、区域和预测时间范围进行索引，在每次使用前进行查询。每个消费的预测被注册为一个索赔；可归因的结果与实际到达的现实进行结算，且没有标记成本，归因阶段排除了受测量污染的结果，结算监督头补充了稀疏的箱子。最终的信用限制了消费：低信用预测缩短了依赖时间范围或触发额外观察；每个依赖事件都可以通过依赖票据和可重放日志进行审计。我们在三个模拟领域（室内飞行、桌面操作、二维导航）中评估梦账本，通过未修改的DreamerV3、TD-MPC2和V-JEPA 2-AC进行挂载，并在真实的Franka操纵器上进行测试。在所有12个保留的条件-时间范围单元中，索赔失败呈剂量单调。在与盲目消费相比，信用限制规划将消耗的想象（后期未能兑现的消费索赔）减少了62%（95%置信区间43-81%），同时成功率和碰撞率相当。在匹配风险目标下，持久的账本将验证探测从每集1.00减少到0.36，在操作成功率为0.94对比0.98；基于结算的校准保持了适度的、种子一致的操作点，而不是原始的瞬时限制。相同的信任层在解码器、潜在空间和令牌空间接口中运行，包括在真实机器人框架上结算的V-JEPA 2-AC。在硬件上，结算在真实传感和接触噪声下仍然可操作，部署失败循环在线重新定价，所有1,062个注册支出均可从审计日志中重放。
cs.RO / 7 / 2608.23887
Interpreting Control Latents for System Identification via Conditional Flow Matching
通过条件流匹配解释系统识别的控制潜变量
Abstract
Latent-conditioned adaptive policies can control robots across changing dynamics, but their learned latents remain internal representations of the policy rather than physical models that can be inspected, rolled out, or used by other control modules. This limits closed-loop analysis, diagnosis, and further improvement of a fixed policy. A direct mapping from latent to physical parameters is also under-specified, because multiple systems can induce similar closed-loop behavior. We therefore decode each operational latent into a distribution of quadrotor models using conditional flow matching. The decoded distribution enables two downstream uses without modifying the policy: online predictive tuning of a high-level controller around the fixed low-level policy, and robustness analysis under specified disturbances. Under perturbed actuator dynamics, decoded-model predictive tuning reduces position tracking RMSE by $23\%$ and heading RMSE by $45\%$ relative to fixed gains. Under Gaussian force disturbances, decoded-model ensembles closely predict the lateral tracking-error evolution. Together, these results show that control latents can be converted into physical model ensembles for tuning, robustness analysis, and diagnosis of frozen adaptive policies.
Chinese Translation
潜变量条件自适应策略能够在变化的动态中控制机器人，但其学习到的潜变量仍然是策略的内部表示，而不是可以检查、展开或被其他控制模块使用的物理模型。这限制了固定策略的闭环分析、诊断和进一步改进。潜变量到物理参数的直接映射也存在不确定性，因为多个系统可以产生相似的闭环行为。因此，我们使用条件流匹配将每个操作潜变量解码为四旋翼模型的分布。解码后的分布在不修改策略的情况下支持两种下游应用：围绕固定低级策略的高层控制器的在线预测调优，以及在指定干扰下的鲁棒性分析。在扰动的执行器动态下，解码模型的预测调优相较于固定增益将位置跟踪均方根误差（RMSE）降低了23%，航向RMSE降低了45%。在高斯力干扰下，解码模型集能够准确预测横向跟踪误差的演变。这些结果表明，控制潜变量可以转换为物理模型集，用于调优、鲁棒性分析和冻结自适应策略的诊断。
cs.RO / 8 / 2608.23972
Safety-aware Model Predictive Path Integral Control with Signal Temporal Logic
基于安全意识的模型预测路径积分控制与信号时序逻辑
Abstract
Safety-aware motion planning remains a challenge in robotics, especially when missions are time-critical and are under complex specifications. In this paper, we propose safety-aware-stl-mppi, a computationally efficient sampling-based receding-horizon planning framework designed to promote satisfaction of constraints expressed in Signal Temporal Logic (STL). Our approach encodes discrete-time STL formulas into candidate time-varying control barrier functions (CBF), which are integrated into a model predictive path integral (MPPI) controller. Our method inherits the benefits of low computational cost from an efficiently parallelizable sampling based planner and utilizes CBF for constraints expressed in STL. We compare against several MPPI baselines using four artificial Mars Rover planning case studies with a diverse environment and cost setups, where we show our method consistently achieving high safety and efficiency. We show a quadcopter planning experiment with NVIDIA Isaac Lab.
Chinese Translation
安全意识的运动规划在机器人技术中仍然是一个挑战，尤其是在任务时间紧迫且具有复杂规范的情况下。本文提出了一种计算效率高的基于采样的回退视野规划框架——安全意识的 STL-MPPI，旨在促进满足以信号时序逻辑（STL）表达的约束。我们的方法将离散时间的 STL 公式编码为候选的时变控制屏障函数（CBF），并将其集成到模型预测路径积分（MPPI）控制器中。我们的方法继承了高效可并行化的基于采样的规划器的低计算成本的优势，并利用 CBF 来处理以 STL 表达的约束。我们在四个不同环境和成本设置的人工火星车规划案例研究中，与多个 MPPI 基线进行比较，结果表明我们的方法在安全性和效率上始终表现优异。我们还展示了一个使用 NVIDIA Isaac Lab 的四旋翼规划实验。
cs.RO / 9 / 2608.23983
Sensorless damage-safe grasping
无传感器的安全抓取
Abstract
Robotic fruit harvesting must hold produce securely without bruising it, yet compression stiffness varies several-fold with ripeness within a single species, so no fixed grip force spans the range. Rather than tune force, we bound deformation: a controller closes the gripper until the object's estimated compression strain reaches a user-specified limit $\varepsilon$, using only the encoder position and motor-effort signal on every servo gripper---no tactile or force-torque sensor. Dividing an effort-based contact force by a lower bound on object stiffness makes the stop provably conservative---true compression stays at or below $\varepsilon$---for any $\varepsilon$ above a contact-detection strain floor we identify and quantify: robust detection itself spends compression, linearly in closing speed, making speed an explicit throughput--gentleness knob. Unlike a hand-tuned force threshold, $\varepsilon$ is a certified, size-scaling, operator-interpretable damage limit, and a ready safe-action parameter for learned grasping policies. In MuJoCo simulation over a realistic fruit-stiffness range, under a sensor-noise model calibrated to the real servo, the controller holds $\ge 98\,\%$ grasp at $0\,\%$ damage across all medium-to-firm stiffnesses for the entire certified $\varepsilon$ range, which neither fixed-force baseline attains; on stiffness-graded 3D-printed TPU cubes it matches baseline grasp success at roughly half the grip force and cuts soft-object damage from $100\,\%$ to $40\,\%$.
Chinese Translation
机器人水果采摘必须牢牢抓住果实而不造成损伤，但在同一物种内，压缩刚度随着成熟度的变化而变化数倍，因此没有固定的抓取力能够涵盖这一范围。我们并不调整力，而是限制变形：控制器在物体的估计压缩应变达到用户指定的极限 $ ext{ε}$ 之前，关闭夹具，仅使用每个伺服夹具的编码器位置和电机努力信号——不使用触觉或力-扭矩传感器。将基于努力的接触力除以物体刚度的下限，使得停止的行为在理论上是保守的——真实的压缩保持在或低于 $ ext{ε}$——对于我们识别和量化的任何高于接触检测应变下限的 $ ext{ε}$：鲁棒检测本身会消耗压缩，线性依赖于闭合速度，使得速度成为一个明确的通量-温和性调节器。与手动调整的力阈值不同，$ ext{ε}$ 是一个经过认证的、规模可扩展的、操作员可解释的损伤限制，也是学习抓取策略的一个现成的安全操作参数。在 MuJoCo 模拟中，针对一个现实的水果刚度范围，在一个经过校准的传感器噪声模型下，控制器在整个认证的 $ ext{ε}$ 范围内，在所有中等到坚硬的刚度下保持 $ ext{≥ 98 ext{ extperthousand}}$ 的抓取成功率，且损伤为 $0 ext{ ext{ extperthousand}}$，而固定力基线无法达到；在刚度分级的 3D 打印 TPU 立方体上，它在大约一半的抓取力下与基线抓取成功率相匹配，并将软物体的损伤从 $100 ext{ ext{ extperthousand}}$ 降低到 $40 ext{ ext{ extperthousand}}$。
cs.RO / 10 / 2608.23994
Bridging Teacher Expectations and Robot Learning via Coupling Dynamics
通过耦合动态弥合教师期望与机器人学习之间的差距
Abstract
Human-robot teaching focuses on enabling nontechnical experts to customize robots according to their needs after deployment. With recent advances in machine learning, human-robot teaching is no longer confined to offline learning where the data gathering step from a human teacher is separated from when the robot learns. Instead, more recent approaches for human-robot teaching focus on coupling human teaching with robot learning. This coupling impacts the structure, timing, and content of the teaching and learning interaction. However, it is currently unclear how such coupling dynamics affect humanrobot teaching effectiveness and human perceptions towards the teaching process. Informed by human learning theories, in this paper we propose a new scale for classifying human-robot teaching interactions according to coupling dynamics present between the human teacher and robot learner. We apply this scale to a subset of the human-robot teaching literature to identify how coupling dynamics and human teacher mental model mismatches with the ground truth robot learning system affect teaching effectiveness and human perceptions towards the teaching process
Chinese Translation
人机教学的重点在于使非技术专家能够在机器人部署后根据其需求定制机器人。随着机器学习的最新进展，人机教学不再局限于离线学习，即人类教师的数据收集步骤与机器人学习的过程是分开的。相反，最近的人机教学方法更关注将人类教学与机器人学习耦合在一起。这种耦合影响了教学与学习互动的结构、时机和内容。然而，目前尚不清楚这种耦合动态如何影响人机教学的有效性以及人类对教学过程的感知。基于人类学习理论，本文提出了一种新的分类尺度，用于根据人类教师与机器人学习者之间存在的耦合动态对人机教学互动进行分类。我们将该尺度应用于人机教学文献的一个子集，以识别耦合动态及人类教师的心理模型与真实机器人学习系统之间的不匹配如何影响教学有效性和人类对教学过程的感知。
cs.RO / 11 / 2608.24026
NeurRAFT: Robot Motion Planning via Anchor-Level Flow Matching with Clearance-Aware Preference Tuning
NeurRAFT：通过锚级流匹配与清晰度感知偏好调节的机器人运动规划
Abstract
Recent end-to-end neural motion planners generate trajectories from raw sensor observations, avoiding the privileged geometric models required by classical planners. However, collision-free planning in cluttered environments remains challenging. We present NeurRAFT, a generative planning framework based on anchor-level flow matching and clearance-aware preference tuning. Unlike prior neural planners that model dense waypoint sequences and spend capacity on redundant local details and smoothness, NeurRAFT operates on compact anchor waypoints. We train the planner using a Jacobian-weighted loss that accounts for the task-space impact of each anchor. At inference, the anchors are generated in two integration steps, followed by cubic-spline interpolation to recover a smooth, full-resolution trajectory. Since imitation learning from positive demonstrations cannot distinguish collision-free from near-collision trajectories, collision-prone behaviors persist at test time. Rather than relying on post-hoc corrections, we directly reshape the pretrained planner's distribution toward safer solutions without augmenting inference. Specifically, Direct Preference Optimization shifts probability mass toward trajectories with larger obstacle clearance, with the resulting improvement directly absorbed into the planner parameters. Experiments show substantial improvements over state-of-the-art planners, while real-world experiments demonstrate zero-shot transfer to a Franka robot under noisy and partially occluded depth observations. Video results available at https://neurraft.github.io/.
Chinese Translation
最近的端到端神经运动规划器能够从原始传感器观测中生成轨迹，避免了传统规划器所需的特权几何模型。然而，在杂乱环境中实现无碰撞规划仍然具有挑战性。我们提出了NeurRAFT，这是一种基于锚级流匹配和清晰度感知偏好调节的生成规划框架。与之前的神经规划器建模密集的航点序列并在冗余的局部细节和平滑性上消耗容量不同，NeurRAFT在紧凑的锚航点上进行操作。我们使用考虑每个锚点在任务空间影响的雅可比加权损失来训练规划器。在推理时，锚点通过两个集成步骤生成，随后通过三次样条插值恢复平滑的全分辨率轨迹。由于从正示范中进行模仿学习无法区分无碰撞轨迹和接近碰撞轨迹，因此在测试时仍会出现易碰撞行为。我们并不依赖事后修正，而是直接重新塑造预训练规划器的分布，以朝向更安全的解决方案，而无需增加推理负担。具体而言，直接偏好优化将概率质量转移到具有更大障碍物间隙的轨迹上，所产生的改进直接吸收到规划器参数中。实验表明，相较于最先进的规划器有显著改进，而现实世界实验则展示了在噪声和部分遮挡的深度观测下对Franka机器人进行零-shot迁移的能力。视频结果可在 https://neurraft.github.io/ 查看。
cs.RO / 12 / 2608.24039
Design-to-Plan: A Large Language Model-Based Multi-Agent Framework for Manufacturing Process Planning from 3D CAD Models and 2D Engineering Drawings
设计到计划：基于大型语言模型的多智能体框架用于从3D CAD模型和2D工程图进行制造过程规划
Abstract
Manufacturing process planning transforms heterogeneous design information into coherent manufacturing decisions. However, existing approaches focus on isolated subtasks, such as feature recognition, drawing interpretation, or tool selection, and struggle to support the full reasoning chain from design artifacts to process plans. This is critical when planning must interpret 3D CAD models, 2D engineering drawings, materials, and domain-specific rules. To address this gap, this paper presents Design-to-Plan, a large language model (LLM)-based multi-agent framework for end-to-end manufacturing process planning. An orchestrator coordinates specialized agents for 3D feature recognition, 2D drawing analysis, 2D-3D context fusion, knowledge retrieval, process sequencing, tool selection, and report generation. Rather than using LLMs as standalone text generators, the framework deploys them as reasoning agents that interact with deterministic modules and knowledge sources to produce consistent and traceable decisions. In this hybrid design, deterministic modules and specialized agents extract structured information from CAD and drawing inputs, while LLM agents perform context-aware reasoning, retrieve manufacturing rules, resolve conflicts, and generate planning outputs. The framework is evaluated using 300 benchmark cases across three downstream ReAct-enabled agents, plus separate evaluations of CAD feature recognition, drawing analysis, and 2D-3D context fusion. The parallel architecture achieves 100% success across downstream agents, Tool F1 scores of 95.9%-97.6%, 90% source detection accuracy in conflict analysis, and a 60%-68% reduction in token usage for key planning tasks. Results show that structured LLM-based multi-agent coordination can bridge design representations and manufacturing knowledge, enabling scalable, efficient, and traceable design-to-plan automation.
Chinese Translation
制造过程规划将异构设计信息转化为连贯的制造决策。然而，现有的方法主要集中在孤立的子任务上，如特征识别、图纸解读或工具选择，难以支持从设计工件到过程计划的完整推理链。当规划需要解释3D CAD模型、2D工程图、材料和特定领域规则时，这一点尤为关键。为了解决这一问题，本文提出了Design-to-Plan，一个基于大型语言模型（LLM）的端到端制造过程规划的多智能体框架。一个协调者负责协调专门的智能体进行3D特征识别、2D图纸分析、2D-3D上下文融合、知识检索、过程排序、工具选择和报告生成。该框架并非将LLM作为独立的文本生成器，而是将其作为推理智能体，与确定性模块和知识源进行交互，以生成一致且可追溯的决策。在这种混合设计中，确定性模块和专门智能体从CAD和图纸输入中提取结构化信息，而LLM智能体则执行上下文感知推理、检索制造规则、解决冲突并生成规划输出。该框架通过300个基准案例进行评估，涵盖三个下游的ReAct启用智能体，并对CAD特征识别、图纸分析和2D-3D上下文融合进行了单独评估。并行架构在下游智能体中实现了100%的成功率，工具F1得分为95.9%-97.6%，冲突分析中的源检测准确率为90%，关键规划任务的令牌使用量减少了60%-68%。结果表明，基于结构化LLM的多智能体协调能够桥接设计表示与制造知识，实现可扩展、高效且可追溯的设计到计划自动化。
cs.RO / 13 / 2608.24042
Hierarchical Skill Retrieval for Data-Efficient Adaptation of Vision-Language-Action Models
用于数据高效适应视觉-语言-动作模型的层次技能检索
Abstract
While Vision-Language-Action (VLA) models pretrained on large-scale robot datasets provide a strong foundation for robot manipulation, their performance can degrade when adapted to new tasks with limited task-specific demonstrations. Retrieval offers a practical way to reuse existing demonstrations for data-efficient adaptation, but existing methods often rely on visual similarity, state-action representations, or task-level language matching. These approaches may overlook the hierarchical structure of long-horizon manipulation tasks, where complete task matches are rare but reusable skills are often abundant. To address this challenge, we propose Hierarchical Skill Retrieval (HSR), a retrieval framework for data-efficient VLA adaptation. Specifically, HSR first decomposes a target task into candidate skill sequences. It evaluates each plan based on both semantic plausibility and skill reliability estimated from the prior dataset. The selected decomposition is then used for hybrid retrieval. This combines subtask-level language retrieval with behavior-feature reranking to identify demonstrations that are both semantically relevant and compatible with the target task. Finally, we adapt the policy through a two-stage pretraining and finetuning pipeline, which separates general skill acquisition from task-specific adaptation. Experiments on the LIBERO benchmark and several real-world robot manipulation tasks show that HSR improves the average success rate by 10.3% and 21.3% over the strongest baseline, respectively. These results demonstrate the effectiveness of structured skill-level retrieval for data-efficient VLA adaptation. Videos and code are available at https://hoar012.github.io/HSR-Project.
Chinese Translation
虽然在大规模机器人数据集上预训练的视觉-语言-动作（VLA）模型为机器人操作提供了坚实的基础，但在适应具有有限任务特定演示的新任务时，其性能可能会下降。检索提供了一种实用的方法来重用现有演示以实现数据高效的适应，但现有方法通常依赖于视觉相似性、状态-动作表示或任务级语言匹配。这些方法可能忽视了长时间操作任务的层次结构，在这些任务中，完整的任务匹配是稀有的，但可重用的技能往往是丰富的。为了解决这一挑战，我们提出了层次技能检索（Hierarchical Skill Retrieval, HSR），这是一个用于数据高效VLA适应的检索框架。具体而言，HSR首先将目标任务分解为候选技能序列。它根据语义合理性和从先前数据集中估计的技能可靠性评估每个计划。所选的分解随后用于混合检索。这结合了子任务级语言检索与行为特征重排序，以识别在语义上相关且与目标任务兼容的演示。最后，我们通过一个两阶段的预训练和微调流程来适应策略，该流程将一般技能获取与任务特定适应分开。在LIBERO基准和多个真实世界机器人操作任务上的实验表明，HSR分别提高了平均成功率10.3%和21.3%，相较于最强基线。这些结果证明了结构化技能级检索在数据高效VLA适应中的有效性。视频和代码可在 https://hoar012.github.io/HSR-Project 获取。
cs.RO / 14 / 2608.24094
SIREN-Bench: Behavior-Driven Generation and Evaluation of Emergency-Vehicle Interactions
SIREN-Bench：基于行为的紧急车辆交互生成与评估
Abstract
Emergency vehicles (EMVs) can reorganize surrounding traffic as civilian vehicles brake, change lanes, or form rescue corridors in response to their passage. Evaluating these safety-critical interactions requires behavior-level control over both EMV privileges and civilian responses, together with consistent sensing and ground truth. Existing datasets and simulation benchmarks do not directly provide this combination. We present \textbf{SIREN}, a behavior-driven SUMO--CARLA co-simulation platform for generating EMV--civilian interactions. SIREN couples SUMO's network-level traffic evolution and behavior logic with CARLA's continuous vehicle control and synchronized onboard sensing; depending on the active behavior, the interaction is controlled by SUMO, CARLA, or jointly. We instantiate the platform as \textbf{SIREN-Bench-v1}, comprising seven parameterized interaction templates across emergency levels L1--L3 and three behavior families, with synchronized sensor observations and simulator-native annotations. We demonstrate the benchmark through three representative tasks: 3D object detection, trajectory prediction, and vision-language risk understanding. Evaluations of nine trajectory predictors, four LiDAR-based detectors, and five vision-language models reveal behavior-dependent failure modes. Traffic-clearance interactions are hardest for detection, privileged intersection traversal is hardest for prediction, and no learned predictor outperforms the constant-velocity reference on average. Vision-language models perform substantially better on normal traffic than on near-miss and collision events. These results demonstrate the value of behavior-centered benchmarking and establish SIREN as an extensible data-generation and evaluation platform for autonomous-driving and transportation safety research.
Chinese Translation
紧急车辆（EMVs）可以在其通过时重新组织周围交通，导致民用车辆刹车、变道或形成救援通道。评估这些安全关键的交互需要对EMV特权和民用响应进行行为级控制，同时还需一致的感知和真实数据。现有的数据集和仿真基准并未直接提供这种组合。我们提出了 extbf{SIREN}，一个基于行为的SUMO--CARLA协同仿真平台，用于生成EMV与民用车辆的交互。SIREN将SUMO的网络级交通演变和行为逻辑与CARLA的连续车辆控制和同步车载感知相结合；根据活跃的行为，交互由SUMO、CARLA或两者共同控制。我们将该平台实例化为 extbf{SIREN-Bench-v1}，包含七个参数化的交互模板，涵盖紧急级别L1--L3和三种行为类别，配有同步的传感器观测和仿真原生注释。我们通过三个代表性任务展示该基准：3D物体检测、轨迹预测和视觉-语言风险理解。对九个轨迹预测器、四个基于LiDAR的检测器和五个视觉-语言模型的评估揭示了依赖行为的失败模式。交通清理交互对检测而言是最困难的，特权交叉口通行对预测而言是最困难的，且没有任何学习的预测器在平均上超越恒速参考。视觉-语言模型在正常交通情况下的表现明显优于近失误和碰撞事件。这些结果展示了以行为为中心的基准测试的价值，并确立了SIREN作为一个可扩展的数据生成和评估平台，用于自动驾驶和交通安全研究。
cs.RO / 15 / 2608.24101
TrAct: Bridging Robot Control and Visual Prediction with Visual Tracks
TrAct：通过视觉轨迹连接机器人控制与视觉预测
Abstract
Robot actions are inherently embodiment-specific and only weakly aligned with image-space visual changes, limiting their effectiveness as conditioning signals for robot world models. In contrast, visual tracks provide an embodiment-agnostic representation of how task-relevant points move through a scene, offering dense image-space guidance for accurate and spatially precise future video prediction. Building on this observation, we propose TrAct, a world-model-based robot decision-making framework that uses visual tracks as an intermediate interface between control and prediction. TrAct consists of three components: a Vision-Language-Action-and-Track model (VLAT) that jointly predicts candidate actions and corresponding visual tracks from the current observation and language instruction; a track-conditioned world model (TWM) that predicts future visual outcomes conditioned on the proposed tracks; and a vision-language reward model (VLAC) that scores the predicted outcomes. At inference time, VLAT generates candidate action-track pairs, TWM rolls out their visual consequences, and VLAC selects the track whose predicted outcome best satisfies the instruction; the action paired with the selected track is then executed by the robot. Experiments on the proposed LIBERO-INTEGRAL benchmark and real-world Franka manipulation show that TrAct improves success rates from 27% to 55% in simulation and from 49% to 76% on real-world tasks compared with the strong VLA baseline $\pi_{0.5}$. Furthermore, TWM consistently improves video prediction quality over the action-conditioned world model (AWM). These results demonstrate that visual tracks provide an effective shared interface between robot control and visual prediction, enabling more accurate world modeling and stronger robot generalization.
Chinese Translation
机器人动作本质上是特定于体现的，并且与图像空间中的视觉变化仅弱相关，这限制了它们作为机器人世界模型条件信号的有效性。相比之下，视觉轨迹提供了一种与体现无关的表示，展示了任务相关点如何在场景中移动，为准确且空间精确的未来视频预测提供了密集的图像空间指导。基于这一观察，我们提出了TrAct，这是一种基于世界模型的机器人决策框架，利用视觉轨迹作为控制与预测之间的中介接口。TrAct由三个组件组成：一个视觉-语言-动作-轨迹模型（VLAT），该模型从当前观察和语言指令中共同预测候选动作及其相应的视觉轨迹；一个轨迹条件世界模型（TWM），该模型根据提议的轨迹预测未来的视觉结果；以及一个视觉-语言奖励模型（VLAC），该模型对预测结果进行评分。在推理时，VLAT生成候选动作-轨迹对，TWM展开它们的视觉后果，VLAC选择预测结果最符合指令的轨迹；然后，机器人执行与所选轨迹配对的动作。在所提出的LIBERO-INTEGRAL基准和真实世界Franka操作的实验中，TrAct在模拟中的成功率从27%提高到55%，在真实世界任务中从49%提高到76%，相比于强大的VLA基线$ ext{π}_{0.5}$。此外，TWM在视频预测质量上始终优于动作条件世界模型（AWM）。这些结果表明，视觉轨迹提供了机器人控制与视觉预测之间有效的共享接口，从而实现更准确的世界建模和更强的机器人泛化能力。
cs.RO / 16 / 2608.24111
Trajectory-Level Continuous Action Representation for Robotic Manipulation
用于机器人操作的轨迹级连续动作表示
Abstract
We propose CAT, a trajectory-level continuous action representation framework for robotic manipulation. Existing visuomotor systems often entangle action representation with control frequency or rely on fixed temporal parameterizations. This leads to representational redundancy at high sampling rates and limits the modeling of critical motion. CAT instead encodes action trajectories within a fixed real-time interval into a set of continuous latent tokens. To ensure temporal consistency across varying control frequencies, we further incorporate a frequency-aware positional encoding that establishs a shared temporal coordinate system. Trajectory-level regularization further stabilizes the latent representation. This approach prevents representation growth with timestep density and avoids reliance on predefined temporal parameterizations. Extensive system-level evaluations on LIBERO, MimicGen, and real-world long-horizon manipulation tasks demonstrate that CAT-based policies consistently outperform both competitive VQ-based and continuous visuomotor baselines under matched training settings. Across various model backbones and control frequencies, CAT consistently improves success rates. These results highlight the advantages of trajectory-level continuous action modeling for scalable robotic manipulation across varying control rates.
Chinese Translation
我们提出了CAT，一个用于机器人操作的轨迹级连续动作表示框架。现有的视觉运动系统通常将动作表示与控制频率纠缠在一起，或依赖于固定的时间参数化。这导致在高采样率下的表示冗余，并限制了关键运动的建模。CAT则在固定的实时间隔内将动作轨迹编码为一组连续的潜在标记。为了确保在不同控制频率下的时间一致性，我们进一步结合了频率感知的位置编码，建立了一个共享的时间坐标系统。轨迹级正则化进一步稳定了潜在表示。这种方法防止了随着时间步密度的增加而导致的表示增长，并避免依赖于预定义的时间参数化。在LIBERO、MimicGen和现实世界的长时间操作任务上进行的广泛系统级评估表明，在匹配的训练设置下，基于CAT的策略在性能上始终优于竞争的基于VQ和连续视觉运动的基线。在各种模型骨干和控制频率下，CAT始终提高成功率。这些结果突显了轨迹级连续动作建模在不同控制速率下可扩展的机器人操作中的优势。
cs.RO / 17 / 2608.24115
PonderPounce: A Pretrained MLLM as an Episode Context Engine for Robot Control
PonderPounce：作为机器人控制的情节上下文引擎的预训练多模态大语言模型
Abstract
Multimodal large language models (MLLMs) can integrate long visual histories, reason under partial observability, and infer behavior from a few examples. Yet vision-language-action (VLA) models generally inherit pretrained representations without using this contextual capacity as episode memory. Memory-dependent policies address this gap through purpose-built history mechanisms. PonderPounce instead reuses an MLLM's native causal context as robot memory. Ponder, a System2 MLLM, accumulates episode observations, demonstrations, and prior cognition in its native causal context and can generate subgoal text and demonstration reasoning for internal use. Pounce, a System1 VLA, receives the current observation, instruction, and proprioception directly; through the Ponder--Pounce interface, it asynchronously receives only the newest continuous cognition token and its age. Both are jointly trained end to end without a purpose-built memory module or separate bridge pretraining. Optimized serving achieves p50 latencies of 78ms for cognition refresh and 25ms for action-model invocation, supporting 20Hz action playback. On RoboMME with base-scale training data, PonderPounce reaches 60.83% with 9B and 50.04% with 0.8B under the same Pounce architecture and interface, versus 44.51% for FrameSamp+Modul and 17.93% for the current-observation \pi_{0.5}. With 9x data, it reaches 75.54% versus 57.88% for FrameSamp+Modul. On RoboCasa-DC, the same interface learns from action supervision alone and reaches 12.5% versus 11.6% for the strongest published demonstration-conditioned baseline, falling to 8.6% when cognition is replaced by a learned null state.
Chinese Translation
多模态大语言模型（MLLMs）能够整合长时间的视觉历史，在部分可观测性下进行推理，并从少量示例中推断行为。然而，视觉-语言-动作（VLA）模型通常继承预训练的表示，而未利用这种上下文能力作为情节记忆。依赖记忆的策略通过专门构建的历史机制来解决这一问题。PonderPounce 则重用 MLLM 的原生因果上下文作为机器人记忆。Ponder 是一个 System2 MLLM，能够在其原生因果上下文中累积情节观察、示范和先前认知，并可以生成用于内部使用的子目标文本和示范推理。Pounce 是一个 System1 VLA，直接接收当前观察、指令和本体感知；通过 Ponder-Pounce 接口，它异步接收最新的连续认知令牌及其年龄。两者在没有专门构建的记忆模块或单独的桥接预训练的情况下进行端到端联合训练。优化后的服务实现了 78 毫秒的认知刷新 p50 延迟和 25 毫秒的动作模型调用延迟，支持 20Hz 的动作播放。在使用基础规模训练数据的 RoboMME 上，PonderPounce 在相同的 Pounce 架构和接口下达到了 60.83%（9B）和 50.04%（0.8B）的成绩，而 FrameSamp+Modul 的成绩为 44.51%，当前观察 C0_{0.5} 的成绩为 17.93%。在使用 9 倍数据时，其成绩达到了 75.54%，而 FrameSamp+Modul 的成绩为 57.88%。在 RoboCasa-DC 上，相同的接口仅通过动作监督进行学习，达到了 12.5%的成绩，而最强的已发布示范条件基线为 11.6%，当认知被替换为学习的空状态时，成绩降至 8.6%。
cs.RO / 18 / 2608.24155
Coverage Planning for Robotic Tooth Preparation in Densely Constrained Environments
在高度受限环境中进行机器人牙齿预备的覆盖规划
Abstract
Tooth preparation refers to the controlled removal of tooth structure to create an optimal substrate for fixed restorations and is a core procedure in restorative dentistry. Automating this task is particularly challenging for robots because the dental bur must operate within a densely constrained intraoral workspace, where even sub-millimeter deviations can compromise outcomes or damage adjacent structures. This paper presents a novel robotic system for autonomous full-crown tooth preparation. The proposed framework includes: 1) an anatomy-aware toolpath planning algorithm that conforms precisely to a technician-designed preparation model while protecting adjacent teeth, and 2) a clearance-oriented end-effector yaw assignment strategy that allows intraoral access while reducing the risk of soft-tissue interference. Together, these features enable the robot to accurately mill the irregular tooth surface with an average geometric deviation of 0.117 mm (RMSE), achieving both restoration quality and clinical safety. A series of simulations and phantom-head experiments validate the system's feasibility and effectiveness.
Chinese Translation
牙齿预备是指对牙齿结构进行控制性去除，以创造一个适合固定修复体的最佳基底，这是修复牙科中的核心程序。对于机器人来说，自动化这一任务尤其具有挑战性，因为牙科钻头必须在一个高度受限的口腔内工作空间内操作，甚至亚毫米级的偏差都可能影响结果或损坏邻近结构。本文提出了一种新型的自主全冠牙齿预备机器人系统。所提出的框架包括：1）一种解剖学感知的工具路径规划算法，该算法精确符合技术人员设计的预备模型，同时保护邻近牙齿；2）一种以间隙为导向的末端执行器偏航分配策略，允许口腔内的访问，同时降低软组织干扰的风险。这些特性使得机器人能够以平均几何偏差0.117毫米（均方根误差）精确铣削不规则的牙齿表面，实现修复质量和临床安全性的双重目标。一系列仿真和假头实验验证了该系统的可行性和有效性。
cs.RO / 19 / 2608.24162
Robust Slip Detection and Material Classification via Spatiotemporal Transformers on a Uniformly-Illuminated Visuo-Tactile Sensor
基于均匀照明视触觉传感器的时空变换器实现的鲁棒滑动检测与材料分类
Abstract
Tactile sensing is central to robotic manipulation, among which slip detection stands out as a quintessential and critical task. However, existing slip datasets are predominantly limited to binary classification, lacking fine-grained directional perception. To address this limitation, we propose a visuo-tactile sensor featuring customized uniform RGB illumination, alongside a unified perception framework. At the hardware level, the sensor achieves high-precision, sub-millimeter depth reconstruction. Based on this capability, we collect a multi-task visuo-tactile dataset encompassing 15 objects, synchronously generating depth information for each data sample. Algorithmically, we design a dual-head TimeSformer network to process dynamic spatiotemporal slip. On unseen objects, this network achieves robust accuracies of 95.5% and 91.5% for 3-class contact state prediction and fine-grained 8-class slip direction classification, respectively. Furthermore, static tactile-based object class recognition utilizing a ResNet-50 backbone yields an outstanding accuracy of 98.8% across 15 categories. The proposed hardware-software framework provides high-fidelity feedback and a powerful multi-modal perception baseline for complex robotic manipulation.
Chinese Translation
触觉感知是机器人操作的核心，其中滑动检测作为一项典型且关键的任务尤为重要。然而，现有的滑动数据集主要局限于二分类，缺乏细粒度的方向感知。为了解决这一限制，我们提出了一种配备定制均匀RGB照明的视触觉传感器，并设计了统一的感知框架。在硬件层面，该传感器实现了高精度的亚毫米级深度重建。基于此能力，我们收集了包含15种物体的多任务视触觉数据集，同步生成每个样本的深度信息。在算法方面，我们设计了一个双头TimeSformer网络以处理动态时空滑动信息。在未见过的物体上，该网络在三分类接触状态预测和细粒度八分类滑动方向识别任务中，分别达到了95.5%和91.5%的鲁棒准确率。此外，基于静态触觉的物体类别识别采用ResNet-50骨干网络，在15个类别上实现了98.8%的优异准确率。所提出的软硬件框架为复杂机器人操作提供了高保真反馈和强大的多模态感知基线。
cs.RO / 20 / 2608.24199
NVIDIA Cosmos-H-Dreams: Real-Time Generative Physics Simulation for Surgical Robotics
NVIDIA Cosmos-H-Dreams：用于外科机器人手术的实时生成物理模拟
Abstract
Generative simulation for surgical robotics still lacks real-time interaction. Physical-robot experiments, often involving animal or cadaver labs, are time-consuming, costly, and difficult to reproduce, while classical simulators struggle to capture photorealistic appearance and deformable-tissue dynamics. We address this gap with Cosmos-H-Dreams, an integrated real-time surgical world-model system combining an action-conditioned generative model, a teacher-to-student distillation recipe, and a deployment stack built on the NVIDIA FlashDreams streaming-inference library. Starting from Cosmos-H-Surgical-Simulator, a multi-embodiment action-conditioned surgical video world model fine-tuned on the large-scale Open-H-Embodiment corpus, we post-train this checkpoint on embodiment- and procedure-specific data. By distilling the resulting bidirectional teacher into a causal, few-step student with Self Forcing, we turn a passive video generator into a controllable surgical simulator that streams at $\sim$160 inference FPS on a single NVIDIA RTX PRO 6000 Blackwell workstation GPU. Crucially, Cosmos-H-Dreams is controller-agnostic: any interface that emits a stream of robot kinematics can drive it. We demonstrate live control through a browser keyboard over WebRTC, a Meta Quest headset over WebXR, a commercial surgical robot console such as CMR Surgical's Versius, and learned policies operating in closed loop. To our knowledge, this is the first interactive surgical world model supporting live human and policy control. Human operators and policies alike can act inside the synthesized world and observe the consequences in real time. We release Cosmos-H-Dreams as an open surgical simulation system, providing a common foundation for surgical education, scalable synthetic data generation, and future intraoperative decision support.
Chinese Translation
外科机器人手术的生成模拟仍然缺乏实时交互。物理机器人实验通常涉及动物或尸体实验室，耗时、成本高且难以重复，而经典模拟器在捕捉照片级真实外观和可变形组织动态方面存在困难。我们通过Cosmos-H-Dreams填补了这一空白，这是一个集成的实时外科世界模型系统，结合了基于动作的生成模型、教师到学生的蒸馏方法，以及基于NVIDIA FlashDreams流推理库构建的部署栈。从Cosmos-H-Surgical-Simulator开始，这是一个在大规模Open-H-Embodiment语料库上微调的多体现基于动作的外科视频世界模型，我们在特定于体现和程序的数据上对该检查点进行了后训练。通过将生成的双向教师蒸馏为一个因果的、少步的学生模型，并使用自我强迫（Self Forcing），我们将一个被动的视频生成器转变为一个可控的外科模拟器，该模拟器在单个NVIDIA RTX PRO 6000 Blackwell工作站GPU上以约160帧每秒的速度进行推理。至关重要的是，Cosmos-H-Dreams是控制器无关的：任何发出机器人运动学流的接口都可以驱动它。我们通过WebRTC的浏览器键盘、WebXR的Meta Quest头显、CMR Surgical公司的Versius等商业外科机器人控制台，以及在闭环中操作的学习策略展示了实时控制。据我们所知，这是第一个支持实时人类和策略控制的交互式外科世界模型。人类操作者和策略都可以在合成的世界中采取行动，并实时观察后果。我们将Cosmos-H-Dreams作为一个开放的外科模拟系统发布，为外科教育、可扩展的合成数据生成和未来的术中决策支持提供了共同基础。
cs.RO / 21 / 2608.24217
CARO: Contact-Agnostic Residual Observation for Zero-Shot Robust Quadruped Locomotion
CARO：一种无接触残差观测框架用于零-shot稳健四足运动
Abstract
We propose CARO, a contact-agnostic residual observation framework for policy adaptation. CARO embeds a fixed-base Euler--Lagrange model into the reinforcement learning control loop and constructs a torque-level residual observation without requiring torque sensors, explicit contact estimation, or vision-based measurements of the floating-base position and linear velocity. A disturbance observer extracts a structured signal representing dynamics mismatch, while the policy learns to exploit this feedback for online adaptation. CARO is trained under the same terrain, command, and domain-randomization conditions as the nominal policy, without specialized disturbance curricula or additional adaptation supervision. Nevertheless, it achieves substantially improved zero-shot robustness in simulation and sim-to-real transfer tasks involving out-of-distribution payloads, center-of-mass shifts, terrain geometries, abrupt dynamics changes, and elevated-platform landings.
Chinese Translation
我们提出了CARO，一种用于策略适应的无接触残差观测框架。CARO将固定基Euler-Lagrange模型嵌入到强化学习控制循环中，并构建了一个扭矩级别的残差观测，无需扭矩传感器、显式接触估计或基于视觉的浮动基位置和线性速度测量。扰动观测器提取一个结构化信号，表示动态不匹配，而策略学习利用这一反馈进行在线适应。CARO在与名义策略相同的地形、指令和领域随机化条件下进行训练，无需专门的扰动课程或额外的适应监督。尽管如此，它在模拟和从模拟到现实的转移任务中，在涉及分布外负载、质心偏移、地形几何、突变动态变化和高平台着陆的情况下，显著提高了零-shot稳健性。
cs.RO / 22 / 2608.24242
A Durable Vision-Based Tactile Fingertip for Robotic Manipulation
一种耐用的基于视觉的触觉指尖用于机器人操作
Abstract
Currently available commercial vision-based tactile sensors provide rich contact information but remain vulnerable to abrasion and repeated concentrated loading, limiting their use in demanding robotic applications. This work presents a durable tactile fingertip comprising a soft silicone gel with a nonpigmented, textured, thin thermoplastic-polyurethane protective film and a replaceable sensing cartridge. Durability was evaluated using two accelerated laboratory procedures: a rotating-drum sanding test and a repetitive probe test applying 39.2 N (4.0 kgf) at 45 cycles per minute. Under the defined sanding conditions, the developed sensor reached the protective-film rupture endpoint after approximately 2-3 hours. During repetitive probe testing, all nine developed sensors remained functionally usable when testing was discontinued: seven after 5 days, one after 6 days, and one after 8 days. Commercial GelSight Mini and DIGIT specimens exhibited initial surface-film rupture after approximately 24-30 seconds of sanding and 25-35 minutes of repetitive loading. Damage to the developed sensor progressed gradually and produced little interference with tactile imaging at the test endpoints. These observations establish durability improvements of more than two orders of magnitude under the defined accelerated conditions. Combining increased durability, gradual degradation, and rapid cartridge replacement offers a practical approach to maintainable vision-based tactile sensing for demanding robotic applications.
Chinese Translation
目前可用的商业视觉触觉传感器提供丰富的接触信息，但仍然容易受到磨损和重复集中加载的影响，限制了它们在高要求机器人应用中的使用。本研究提出了一种耐用的触觉指尖，采用软硅胶材料，配以无色素、纹理化的薄热塑性聚氨酯保护膜和可更换的传感器组件。通过两种加速实验室程序评估了其耐用性：旋转鼓砂测试和施加39.2 N（4.0 kgf）重复探针测试，频率为每分钟45次。在定义的砂磨条件下，所开发的传感器在大约2-3小时后达到了保护膜破裂的终点。在重复探针测试中，当测试停止时，所有九个开发的传感器仍然保持功能可用：七个在5天后，一个在6天后，另一个在8天后。商业GelSight Mini和DIGIT样本在大约24-30秒的砂磨和25-35分钟的重复加载后表现出初始表面膜破裂。所开发传感器的损伤逐渐加重，并在测试终点对触觉成像产生了很小的干扰。这些观察结果在定义的加速条件下证明了耐用性提高了两个数量级以上。结合增强的耐用性、渐进的退化和快速的组件更换，为高要求的机器人应用提供了一种可维护的基于视觉的触觉传感的实用方法。
cs.RO / 23 / 2608.24485
NeuralParker: A Reinforcement Learning Planner for Irregular Parking Environments
NeuralParker：一种用于不规则停车环境的强化学习规划器
Abstract
Automated parking commonly assumes marked slots and short approach maneuvers. Delivery and service vehicles, however, may need to reach an operator-specified pose in an irregular bounded environment from a distant start. Existing learning-based parking planners often rely on local observations, which can restrict long-range route reasoning. To address this problem, we present NeuralParker, a reinforcement learning-based hybrid planner for arbitrary-pose parking. NeuralParker encodes full-environment obstacle and boundary geometry in a target-relative vertex representation, allowing the policy to retain route-defining context throughout the approach. It further couples a learned curvature--length arc policy with an in-loop terminal ensemble that selects from diverse cubic Hermite connections using a curvature-regularized cost. We also establish factorial and long-range route-choice benchmarks to evaluate planning success and trajectory quality. Experiments on these benchmarks show that NeuralParker achieves higher planning success and better overall trajectory quality than the evaluated baselines, while ablation studies support the benefits of the target-relative global representation and terminal ensemble. Finally, a real-vehicle evaluation confirms that the planner transfers effectively to real delivery-vehicle perception at a working parking site, planning successfully at low computational cost.
Chinese Translation
自动停车通常假设存在标记的停车位和短距离的接近机动。然而，送货和服务车辆可能需要从远处起始位置到达操作员指定的姿态，在不规则的边界环境中。现有的基于学习的停车规划器往往依赖于局部观测，这限制了长距离路径推理。为了解决这个问题，我们提出了NeuralParker，一种基于强化学习的混合规划器，能够处理任意姿态的停车。NeuralParker在目标相对顶点表示中编码了全环境的障碍物和边界几何信息，使得策略在接近过程中能够保持路径定义的上下文。它进一步结合了学习到的曲率-长度弧策略和一个循环终端集成，该集成使用曲率正则化成本从多样的三次Hermite连接中进行选择。我们还建立了因子和长距离路径选择基准，以评估规划成功率和轨迹质量。在这些基准上的实验表明，NeuralParker在规划成功率和整体轨迹质量上均优于评估的基线，同时消融研究支持了目标相对全局表示和终端集成的优势。最后，真实车辆评估确认该规划器能够有效转移到实际送货车辆的感知，在工作停车场成功规划，且计算成本低。
cs.RO / 24 / 2608.24525
RoG-DAgger: Rollout-Guided Post-Training for End-to-End Driving
RoG-DAgger：基于回滚指导的端到端驾驶后训练
Abstract
Recent end-to-end driving systems demonstrate strong performance on closed-loop benchmarks, yet are still predominantly trained on fixed expert-collected data using open-loop imitation learning. This training-inference mismatch leaves the policy vulnerable in policy-induced states, where accumulated errors can lead to safety-critical failures. A promising post-training approach to overcome this issue is Dataset Aggregation (DAgger), which gathers expert demonstrations in policy-induced states and subsequently fine-tunes the policy on the resulting aggregated dataset. Existing driving DAgger pipelines, however, face three challenges: i) the expert is restricted to a limited trajectory-and-speed solution space, ii) takeover may occur too early or too late relative to impending failures, and iii) privileged expert decisions may rely on information unavailable to the student. To address this, we introduce RoG-DAgger, a post-training framework that uses short-horizon kinematic rollouts to construct high-quality expert demonstrations in safety-critical states. Specifically, RoG-DAgger expands the expert's trajectory-and-speed solution space and evaluates candidate plans through rollout to construct preventive supervision. Moreover, it uses rollout solvability to time the takeover near the estimated point of no return. Lastly, it aligns the expert's field of view with that of the student to provide student-compatible supervision. Across in-distribution (including long-horizon) and out-of-distribution evaluations, RoG-DAgger improves the end-to-end model SimLingo by 5.3 driving-score points and 6.2 percentage points in success rate on Bench2Drive, doubles its driving score from 22 to 44 on Longest6 v2, and improves out-of-distribution success rate from 55\% to 66\% on Fail2Drive.
Chinese Translation
近期的端到端驾驶系统在闭环基准测试中表现出强大的性能，但仍然主要依赖于使用开放式模仿学习在固定专家收集的数据上进行训练。这种训练与推理的不匹配使得策略在政策引发的状态下变得脆弱，累积的错误可能导致安全关键的失败。一种有前景的后训练方法是数据集聚合（Dataset Aggregation, DAgger），它在政策引发的状态中收集专家演示，并随后在结果聚合的数据集上微调策略。然而，现有的驾驶DAgger流程面临三个挑战：i）专家受限于有限的轨迹和速度解空间，ii）接管可能相对于即将发生的失败发生得过早或过晚，iii）特权专家的决策可能依赖于学生无法获得的信息。为了解决这些问题，我们提出了RoG-DAgger，一个后训练框架，利用短时间范围的运动学回滚在安全关键状态中构建高质量的专家演示。具体而言，RoG-DAgger扩展了专家的轨迹和速度解空间，并通过回滚评估候选计划以构建预防性监督。此外，它利用回滚可解性来在估计的不可逆转点附近时机接管。最后，它将专家的视野与学生的视野对齐，以提供与学生兼容的监督。在分布内（包括长时间范围）和分布外的评估中，RoG-DAgger使端到端模型SimLingo的驾驶得分提高了5.3分，成功率提高了6.2个百分点，在Bench2Drive上，驾驶得分从22提高到44，在Longest6 v2上，成功率从55\%提高到66\%.
cs.RO / 25 / 2608.24572
Fiber Optic Sensing Glove for High Performance Dexterous Manipulation Capture
用于高性能灵巧操作捕捉的光纤传感手套
Abstract
Capturing hand pose during dexterous manipulation remains difficult: vision-based methods degrade under occlusion and challenging lighting, while sensorized gloves, though occlusion-free, are prone to drift and magnetic interference and rarely match motion-capture accuracy. We introduce a fiber optic sensing glove for full hand pose tracking that targets these failure modes, using multi-core shape-sensing fibers that capture each fiber's full 3D shape rather than curvature alone. A novel pipeline registers each reconstructed fiber shape to a common hand reference frame, and a new inverse-kinematics solver reconstructs full hand pose at 60 Hz using curve constraints. Benchmarked on a 2-hour dataset of dexterous object manipulation tasks across 5 subjects, the glove achieves 7.2 mm mean fingertip position error against motion capture ground truth, reduced to 4.9 mm by a one-time factory calibration of the fiber routing hub that transfers across users and sessions. These capabilities enable high-fidelity data capture and bimanual virtual teleoperation - both essential to advancing the robotics field.
Chinese Translation
在灵巧操作过程中捕捉手部姿态仍然具有挑战性：基于视觉的方法在遮挡和复杂光照条件下效果不佳，而传感手套虽然避免了遮挡，但容易受到漂移和磁干扰的影响，且很少能达到运动捕捉的精度。我们提出了一种光纤传感手套，用于全手部姿态跟踪，旨在解决这些失败模式，采用多核形状传感光纤，捕捉每根光纤的完整三维形状，而不仅仅是曲率。一个新颖的处理流程将每个重建的光纤形状注册到一个共同的手部参考框架中，而一个新的逆运动学求解器使用曲线约束以60 Hz的频率重建完整的手部姿态。在对5名受试者进行的2小时灵巧物体操作任务的数据集上进行基准测试，该手套在运动捕捉的真实值下实现了7.2毫米的平均指尖位置误差，通过一次性工厂校准光纤路由中心，用户和会话间的误差降低至4.9毫米。这些能力使得高保真数据捕捉和双手虚拟遥操作成为可能，这对于推动机器人技术领域的发展至关重要。
cs.RO / 26 / 2608.24603
Gripper-aware Vision Language Action Models
考虑夹具的视觉语言动作模型
Abstract
Vision language action models (VLAs) have advanced general purpose robotic grasping and manipulation by enabling robots to interpret visual observations and natural language instructions to generate executable action sequences. However, existing VLAs often implicitly assume gripper invariance, despite grasping strategies being inherently embodiment-dependent. Different gripper types, such as parallel-jaw and suction, usually require distinct interaction strategies to achieve the same grasping objective. Moreover, current datasets for VLAs predominantly rely on parallel-jaw grippers, limiting gripper-aware learning. To address this gap, we introduce MiGA, a multi-gripper-aware dataset spanning five distinct gripper types across multiple robots with 103,000 demonstrations, explicitly capturing strategy divergence under shared task objectives. We further propose GVLA, which combines a new multi-gripper tokenizer with adapter-based policy routing. Our new gripper encoding induces structured embedding information that balances parameter sharing and strategy differentiation, while layer-wise probing confirms meaningful gripper-conditioned representations for VLAs. Intensive experiments in both simulation and real-world robots show that our GVLA outperforms the current baselines across evaluated settings. Our method also improves zero-shot generalization or few-shot adaptation to new objects or unseen tasks, and enable more efficient gripper adaptation.
Chinese Translation
视觉语言动作模型（VLAs）通过使机器人能够解读视觉观察和自然语言指令，从而生成可执行的动作序列，推动了通用机器人抓取和操作的进展。然而，现有的VLA通常隐含地假设夹具的不变性，尽管抓取策略本质上依赖于具体的实现。不同类型的夹具，如平行夹爪和吸盘，通常需要不同的交互策略以实现相同的抓取目标。此外，当前的VLA数据集主要依赖于平行夹爪，这限制了对夹具感知学习的探索。为了解决这一问题，我们引入了MiGA，一个涵盖五种不同夹具类型的多夹具感知数据集，涉及多个机器人，共计103,000个演示，明确捕捉在共享任务目标下的策略差异。我们进一步提出了GVLA，它结合了一种新的多夹具分词器和基于适配器的策略路由。我们新的夹具编码引入了结构化的嵌入信息，平衡了参数共享和策略差异，同时逐层探测确认了对VLA有意义的夹具条件表示。在模拟和真实机器人中的大量实验表明，我们的GVLA在评估设置中超越了当前的基准。我们的方法还改善了对新物体或未见任务的零样本泛化或少样本适应能力，并实现了更高效的夹具适应。
cs.RO / 27 / 2608.24618
VIP: Variation-based Iterative-learning Planning for Robotic Navigation
VIP：基于变异的迭代学习规划用于机器人导航
Abstract
Over the past decade, autonomous robotic systems have been increasingly deployed in applications such as surveying, search and rescue, and last-mile delivery. These applications require robots to generate safe and efficient motion plans in large, complex, and obstacle-dense environments, often under limited onboard computing resources. However, conventional planning methods commonly rely on finite-dimensional trajectory parameterization or increasingly long prediction horizons, leading to rapidly growing computational costs, particularly in multi-robot scenarios. This paper presents a novel variation-based iterative-learning planning (VIP) framework for efficient motion planning of both single robots and robotic swarms. Instead of optimizing a large number of discrete trajectory variables, VIP directly updates the planning command as a continuous function in an infinite-dimensional function space. The same variation-based update can be implemented in a model-in-the-loop manner for offline planning or in a robot-in-the-loop manner between online physical executions. By avoiding the computational burden associated with horizon expansion and high-dimensional trajectory discretization, VIP maintains a per-iteration computational complexity of $\mathcal{O}(n)$, where $n$ denotes the number of spatial discretization points. Extensive simulations and real-world experiments demonstrate that the proposed framework can efficiently generate and iteratively improve motion plans for different planning objectives, robotic platforms, and swarm configurations, highlighting its effectiveness, computational efficiency, and scalability as a general planning methodology.
Chinese Translation
在过去十年中，自主机器人系统在测绘、搜索与救援以及最后一公里配送等应用中得到了越来越广泛的部署。这些应用要求机器人在大型、复杂且障碍物密集的环境中生成安全高效的运动规划，通常在有限的机载计算资源下进行。然而，传统的规划方法通常依赖于有限维的轨迹参数化或越来越长的预测时间范围，导致计算成本迅速增长，特别是在多机器人场景中。本文提出了一种新颖的基于变异的迭代学习规划（VIP）框架，用于单个机器人和机器人群体的高效运动规划。VIP并不是优化大量离散轨迹变量，而是直接在无限维函数空间中将规划命令更新为连续函数。相同的基于变异的更新可以以模型在环（model-in-the-loop）方式实现离线规划，或以机器人在环（robot-in-the-loop）方式在在线物理执行之间进行。通过避免与时间范围扩展和高维轨迹离散化相关的计算负担，VIP保持每次迭代的计算复杂度为$ ext{O}(n)$，其中$n$表示空间离散化点的数量。大量的仿真和实际实验表明，所提出的框架能够高效生成并迭代改进不同规划目标、机器人平台和群体配置的运动规划，突显了其作为通用规划方法的有效性、计算效率和可扩展性。
cs.RO / 28 / 2608.24714
GaussianWAM: Distilling Geometry and Semantics from 3D Gaussian Fields into World-Action Models
GaussianWAM：从3D高斯场中提取几何和语义以构建世界-动作模型
Abstract
World-Action Models (WAMs) jointly learn future visual prediction and action generation, using video dynamics as a representation-learning signal for robotic manipulation. However, their video latents are primarily optimized for visual prediction and are not explicitly encouraged to preserve cross-view geometric structure or spatially localized, object-relevant semantics. We propose \textbf{GaussianWAM}, a training-time representation-enhancement framework that organizes geometric and semantic supervision through a 3D Gaussian field. Given synchronized multi-view observations, frozen geometry and vision foundation models provide depth, camera parameters, and dense semantic features. GaussianWAM binds these heterogeneous signals to shared Gaussian primitives and renders spatially aligned semantic, depth, and coverage targets, which are distilled into the current-observation representations of the WAM. All teacher models, Gaussian components, and auxiliary prediction heads are removed after training, leaving the original WAM inference path without additional modules or forward computation. On LIBERO-Plus, GaussianWAM improves FastWAM from 52.05\% to 71.29\% and Cosmos Policy from 71.52\% to 77.30\%. Direct CLIP and VGGT distillation already establishes a strong FastWAM baseline of 69.37\%, while Gaussian-field unification further improves it to 71.29\%, supporting the benefit of spatially organizing heterogeneous teacher signals. GaussianWAM also improves performance on standard LIBERO and shows positive transfer trends on RoboTwin and real-world manipulation. These results suggest that training-time Gaussian distillation provides a practical way to inject geometry- and semantics-related supervision into WAM representations without changing their deployment architecture.
Chinese Translation
世界-动作模型（WAMs）共同学习未来的视觉预测和动作生成，利用视频动态作为机器人操控的表示学习信号。然而，它们的视频潜变量主要针对视觉预测进行优化，并未明确鼓励保留跨视角的几何结构或空间局部的、与物体相关的语义。我们提出了 extbf{GaussianWAM}，一种在训练时增强表示的框架，通过3D高斯场组织几何和语义的监督。给定同步的多视角观测，冻结的几何和视觉基础模型提供深度、相机参数和密集的语义特征。GaussianWAM将这些异构信号绑定到共享的高斯原语，并渲染空间对齐的语义、深度和覆盖目标，这些目标被提炼到WAM的当前观测表示中。所有教师模型、高斯组件和辅助预测头在训练后被移除，保留原始的WAM推理路径，而无需额外的模块或前向计算。在LIBERO-Plus上，GaussianWAM将FastWAM的性能从52.05\%提升至71.29\%，将Cosmos Policy的性能从71.52\%提升至77.30\%。直接的CLIP和VGGT蒸馏已经建立了69.37\%的强FastWAM基线，而高斯场统一进一步将其提升至71.29\%，支持空间组织异构教师信号的益处。GaussianWAM还提高了标准LIBERO上的性能，并在RoboTwin和现实世界操控中显示出积极的迁移趋势。这些结果表明，训练时的高斯蒸馏为在WAM表示中注入与几何和语义相关的监督提供了一种实用的方法，而无需改变其部署架构。
cs.RO / 29 / 2608.24724
Fiber Bragg Grating Whiskers for Bioinspired Hydrodynamic Perception on Underwater Robots
用于水下机器人生物启发水动力感知的光纤布拉格光栅须状结构
Abstract
Harbor seals track hydrodynamic trails with their vibrissae, enabling passive perception of moving targets in dark or turbid water. Inspired by this capability, we present compact fiber Bragg grating (FBG) whiskers for underwater robots. Like seal whiskers, they have a non-uniform taper and elliptical cross-section. Controlled towing experiments show a monotonic relative-flow response from 0.1 to 0.6 m/s, a strong reduction of self-induced oscillation relative to a cylindrical baseline, and a pronounced dependence on angle of attack. Experiments with a pitching foil show that the whiskers can detect the characteristic vortices shed by a stationary or moving source, detectable several seconds after the source has passed. Using this information, a single front-mounted whisker enabled a small underwater robot to distinguish between continuing straight and executing a turn, selecting the correct branch in 17 of 20 trials (85.0%) from whisker signals alone. These results connect bioinspired hydrodynamic sensing to robot action and suggest the utility of whiskers for tracking underwater objects.
Chinese Translation
海港海豹利用其触须追踪水动力轨迹，从而在黑暗或浑浊的水中实现对移动目标的被动感知。受到这种能力的启发，我们提出了用于水下机器人的紧凑型光纤布拉格光栅（FBG）须状结构。与海豹的触须类似，它们具有非均匀锥度和椭圆截面。受控拖曳实验表明，在0.1至0.6 m/s的范围内，须状结构表现出单调的相对流动响应，相较于圆柱基线，自激振荡显著减少，并且对攻角有明显依赖。与一个摆动翼的实验表明，须状结构能够检测到由静止或移动源产生的特征涡流，这些涡流在源经过数秒后仍可被探测。利用这些信息，一个前置的须状结构使得一台小型水下机器人能够区分继续直行和执行转弯，仅凭须状信号在20次试验中正确选择了17次（85.0%）。这些结果将生物启发的水动力感知与机器人动作连接起来，并暗示了须状结构在追踪水下物体中的潜在应用。
cs.RO / 30 / 2608.24741
One-Shot Learning from Demonstration of Contact-Rich Robotic Manipulation by Identifying Physical Interactions
通过识别物理交互实现接触丰富的机器人操作的一次性学习
Abstract
Learning from Demonstration (LfD) allows robots to learn manipulation tasks directly from humans, thereby supporting the versatile application of robots. Most LfD methods do not explicitly model the physical interactions between a robot and its environment, such as the making and breaking of contact, while these are crucial during manipulation tasks. Because the same basic physical interactions recur often, they can be a basis for robust, generalizable, and adaptive task reproduction. We propose an LfD method that explicitly uses what physical interactions take place where and when. Using that information, a hybrid position-force controller tracks demonstrated trajectories until contact-based transition conditions from the demonstrations are met. We evaluate our method in real robot experiments consisting of opening doors and locks, bolt picking and screwing, dislodging, and surface contouring. We show that explicitly modeling physical interactions benefits LfD in four ways. First, by allowing reproduction of complex, sequential, and contact-rich manipulation tasks using only a single demonstration and no prior knowledge of the task. Second, by facilitating robustness to unknown geometric variations in the environment. Third, by facilitating generalization when geometric variations are known. Fourth, by facilitating online adaptation using geometric information explored during task reproduction. We discuss how robustness, generalization, and adaptivity can be explicitly implemented, which is generally lacking in the LfD literature. Thereby, our work aims to close a gap in interpretable few-shot LfD of robotic manipulation.
Chinese Translation
从示范学习（LfD）使机器人能够直接从人类学习操作任务，从而支持机器人多样化的应用。大多数LfD方法并未明确建模机器人与其环境之间的物理交互，例如接触的建立和断开，而这些在操作任务中至关重要。由于相同的基本物理交互经常重复出现，它们可以成为稳健、可泛化和自适应任务再现的基础。我们提出了一种LfD方法，明确使用物理交互发生的时间和地点。利用这些信息，混合位置-力控制器跟踪示范轨迹，直到满足示范中的基于接触的过渡条件。我们在实际机器人实验中评估了我们的方法，这些实验包括开门和锁、螺栓挑选和拧紧、物体松动和表面轮廓处理。我们展示了明确建模物理交互对LfD有四方面的好处。首先，允许仅通过一次示范而无需任务的先验知识来再现复杂、顺序和接触丰富的操作任务。其次，增强了对环境中未知几何变化的鲁棒性。第三，促进了在已知几何变化时的泛化。第四，利用在任务再现过程中探索的几何信息促进在线适应。我们讨论了如何明确实施鲁棒性、泛化和适应性，而这在LfD文献中通常缺乏。因此，我们的工作旨在填补机器人操作可解释的少量示范LfD中的空白。
cs.RO / 31 / 2608.24882
Latent Action as Intention Enables Efficient Future Imagination for World Action Models
潜在行动作为意图使世界行动模型的未来想象更加高效
Abstract
World action models (WAMs) improve robot control by modeling how observations evolve, but generating future observations at test time incurs substantial latency. Fast-WAM removes this process for efficiency; however, our matched implementations show lower generalization for Fast-WAM than for future-aware alternatives, especially with scarce robot demonstrations and in out-of-distribution scenarios. To bridge this gap, we introduce **LAWA**, a WAM architecture that uses compact latent actions as an operational representation of future intentions, enabling efficient test-time future imagination without generating future observations. Specifically, a discrete tokenizer enhanced by action-free pre-training produces manipulation-centric codebook targets. LAWA jointly denoises a continuous latent state anchored to these targets with executable action chunks while omitting the future-video branch at inference. On RoboCasa, LAWA achieves state-of-the-art average success rates of 65.6% and 80.8% in the few-shot and full data settings, improving over the matched Fast-WAM baseline by 9.6 and 4.5 points, respectively. It also preserves the performance level of the matched Joint-WAM variant while requiring 42.9% lower inference latency. LAWA also demonstrates competitive zero-shot robustness on LIBERO-Plus and superior performance on real-world tasks. These results show that future imagination need not be discarded: retaining it with compact latent actions yields an effective trade-off among performance, generalization, and latency. Code and models will be released.
Chinese Translation
世界行动模型（WAMs）通过建模观察如何演变来改善机器人控制，但在测试时生成未来观察会导致显著的延迟。Fast-WAM消除了这一过程以提高效率；然而，我们的匹配实现显示，Fast-WAM的泛化能力低于未来感知的替代方案，尤其是在稀缺的机器人演示和分布外场景下。为了解决这一问题，我们引入了**LAWA**，一种WAM架构，它使用紧凑的潜在行动作为未来意图的操作表示，从而在不生成未来观察的情况下实现高效的测试时未来想象。具体而言，通过无行动预训练增强的离散标记器生成以操控为中心的代码本目标。LAWA联合去噪与这些目标锚定的连续潜在状态，并使用可执行的行动块，同时在推理时省略未来视频分支。在RoboCasa上，LAWA在少样本和全数据设置中分别实现了65.6%和80.8%的最新平均成功率，分别比匹配的Fast-WAM基线提高了9.6和4.5个百分点。它还保持了与匹配的Joint-WAM变体相同的性能水平，同时推理延迟降低了42.9%。LAWA在LIBERO-Plus上也展示了竞争性的零-shot鲁棒性，并在现实世界任务中表现优越。这些结果表明，未来想象不必被抛弃：通过紧凑的潜在行动保留未来想象在性能、泛化和延迟之间实现了有效的权衡。代码和模型将被发布。
cs.RO / 32 / 2608.24885
Do Robotic World Models Really Follow Actions? Diagnosing and Aligning Action-Conditioned Generation for Policy Learning
机器人世界模型真的遵循动作吗？诊断和对齐基于动作的生成以进行策略学习
Abstract
Action-conditioned world models are increasingly used as learned simulators for policy evaluation and improvement, yet their effectiveness rests on an unverified assumption: generated futures faithfully reflect arbitrary valid actions. Existing benchmarks are typically confined to expert demonstrations, leaving off-expert action following inadequately evaluated. To address this gap, we introduce WorldEcho, which probes action following over a broader action distribution using visual integrity and SE(3) trajectory alignment. Our diagnosis shows that current world models reasonably execute expert actions but struggle with diverse off-expert trajectories, either ignoring the commanded actions or producing visually invalid rollouts. We further propose WorldSync, which strengthens action following along three complementary axes: distributional coverage, representational grounding, and intervention-effect alignment. It broadens the training distribution over action consequences, grounds intermediate video representations in action-induced robot dynamics through an Action-Forcing Expert, and aligns predicted changes under action interventions with the corresponding changes in ground-truth futures. Experiments on RoboTwin benchmarks and real-robot tasks show that WorldSync improves WorldEcho metrics and serves as a more reliable simulator for iterative policy improvement, enabling policies to achieve higher success rates.
Chinese Translation
基于动作的世界模型越来越多地被用作策略评估和改进的学习模拟器，但它们的有效性建立在一个未经验证的假设上：生成的未来能够忠实地反映任意有效的动作。现有的基准通常局限于专家演示，导致非专家动作的跟随评估不足。为了解决这一问题，我们引入了WorldEcho，它通过视觉完整性和SE(3)轨迹对齐，探测更广泛动作分布下的动作跟随情况。我们的诊断显示，当前的世界模型能够合理地执行专家动作，但在处理多样的非专家轨迹时表现不佳，要么忽略指令动作，要么产生视觉上无效的结果。我们进一步提出了WorldSync，它在三个互补的方面增强了动作跟随：分布覆盖、表示基础和干预效果对齐。它扩展了基于动作后果的训练分布，通过一个动作强制专家将中间视频表示与动作引起的机器人动态相结合，并将动作干预下的预测变化与真实未来的相应变化对齐。在RoboTwin基准和真实机器人任务上的实验表明，WorldSync改善了WorldEcho的指标，并作为一个更可靠的模拟器用于迭代策略改进，使策略能够实现更高的成功率。
cs.CV / 1 / 2608.23593
Fidelity Preference, Not Demographic Preference: A Pixel-Level Attribute-Sensitivity Audit of Image Aesthetic/Preference Scorers
忠诚偏好，而非人口统计偏好：图像美学/偏好评分者的像素级属性敏感性审计
Abstract
Text-to-image systems use learned aesthetic scorers to filter training data and guide generation, but whether these scores encode demographic attributes as objective quality is unclear. We audit four scorers (LAION-Aesthetics, PickScore, ImageReward, HPSv2) using pixel-level interventions on skin tone and body type in synthetic and real images. Our key finding is that along skin-lightness, the dominant effect is fidelity preference: unaltered images score highest, and perturbations in either direction are penalized (inverted-U). Placebo arms show this penalty is not an artifact of the skin operator, as applying the same CIELAB L* shift to non-skin regions yields similar penalty magnitudes. However, the penalty is operator-dependent and holds for all operators only for LAION-Aes. Critically, audits on synthetic images alone are misleading: LAION-Aes shows strong preference for darker skin on synthetic faces, but on 1470 real faces the preference reverses and becomes much smaller, and amplification becomes non-significant. Across scorers, synthetic results do not transfer -- reversing for LAION-Aes and HPSv2, attenuating for PickScore. We contribute a reproducible benchmark with artifact control and synthetic/real cross-validation, and an auditability criterion for pixel-level causal isolation (valid for skin tone, not for body type due to deformation). Population-stratified analysis shows fidelity-penalty asymmetry is not robust across groups after FDR correction except for HPSv2. Our findings show naive synthetic audits misjudge bias direction and magnitude, and only within-image causal isolation on real data can distinguish true demographic bias from fidelity preference.
Chinese Translation
文本到图像系统使用学习到的美学评分器来过滤训练数据并指导生成，但这些评分是否将人口统计属性编码为客观质量尚不清楚。我们对四个评分器（LAION-Aesthetics、PickScore、ImageReward、HPSv2）进行了审计，采用对合成和真实图像的肤色和体型进行像素级干预。我们的主要发现是，在肤色明亮度方面，主导效应是忠诚偏好：未改变的图像得分最高，任何方向的扰动都会受到惩罚（倒U型）。安慰剂组显示这种惩罚并不是肤色操作符的伪影，因为对非肤色区域施加相同的CIELAB L*偏移会产生类似的惩罚幅度。然而，惩罚是操作符依赖的，并且仅对LAION-Aes在所有操作符中成立。关键是，仅对合成图像的审计是误导性的：LAION-Aes对合成面孔表现出对深色肤色的强烈偏好，但在1470个真实面孔上，偏好反转并变得更小，放大效应变得不显著。在评分器之间，合成结果并不转移——对于LAION-Aes和HPSv2反转，对于PickScore减弱。我们贡献了一个可重复的基准，具有伪影控制和合成/真实交叉验证，以及一个像素级因果隔离的可审计性标准（对肤色有效，但由于变形对体型无效）。人口分层分析显示，忠诚惩罚不对称在FDR校正后在各组之间并不稳健，除了HPSv2。我们的发现表明，天真的合成审计错误判断了偏见的方向和幅度，只有在真实数据中进行图像内因果隔离才能区分真正的人口统计偏见和忠诚偏好。
cs.CV / 2 / 2608.23634
The Blending Ratio Is Not Where the Performance Is: Diagnosing Prototype Blending for Few-Shot Adaptation of Vision-Language Models
混合比例并非性能所在：诊断视觉-语言模型的少样本适应原型混合
Abstract
Many few-shot adaptation methods for vision-language models classify with a convex combination of the zero-shot text prototype and the mean of the K labelled image features, with a single blending ratio routinely tuned on held-out labels, often on the test set itself. We ask what the family's own bias-variance justification invites: what is the right ratio, can it be estimated without validation data, and is finding it where the performance is? First, the ratio minimising prototype mean-squared error has a closed form whose support-set plug-in is exactly a positive-part James-Stein coefficient shrinking towards the text prototype. Across 4,800 cells (ten datasets, five backbones including SigLIP, five shot counts, five seeds, four prompt tiers) this theoretically optimal ratio is a reliable estimate of the wrong quantity: on the 950 primary-tier cells where it is defined it trails a test-set-oracle ratio by 8.5 points. It saturates near 1, discarding the text prior for a nearest-class-mean classifier, because 78% of the text-image prototype distance it treats as bias is a class-independent offset that the arg max largely cancels. We prove the mechanism and bound its share of the damage at 26% by a counterfactual. Second, leave-one-out on the support set alone sets a ratio landing within 0.9 points of the oracle blend, so it is estimable without validation data. Third, validation-free linear probes beat even the oracle-tuned blend: CLAP by +1.9 points and LP++ by +1.5 on average, and at K >= 4 all four validation-free baselines sit above the oracle, the linear probes by margins excluding zero. These results locate the ceiling in the model class, not the hyperparameter: the ratio can be set near-optimally for free, and it is still not where the performance is. Code, cached features, per-cell records: https://huggingface.co/datasets/Liangzhi-Li/clipbench-blending
Chinese Translation
许多视觉-语言模型的少样本适应方法通过零样本文本原型与 K 个标记图像特征均值的凸组合进行分类，通常在保留标签上调节单一混合比例，往往是在测试集上。我们提出了该方法家族自身的偏差-方差解释所引发的问题：什么是正确的比例，是否可以在没有验证数据的情况下进行估计，以及找到它是否就是性能所在？首先，最小化原型均方误差的比例有一个封闭形式，其支持集插值正好是一个正部分的 James-Stein 系数，向文本原型收缩。在 4800 个单元（十个数据集、五个骨干网络，包括 SigLIP、五个样本数量、五个种子、四个提示层级）中，这个理论上的最优比例是错误量的可靠估计：在 950 个主要层级单元中，它落后于测试集神谕比例 8.5 个点。它在接近 1 时饱和，抛弃了文本先验，变成了最近类均值分类器，因为它视为偏差的文本-图像原型距离中有 78% 是一个类无关的偏移量，arg max 在很大程度上抵消了这一点。我们通过反事实证明了这一机制，并将其造成的损害限制在 26%。其次，仅在支持集上进行的留一法设置的比例与神谕混合相差不超过 0.9 个点，因此可以在没有验证数据的情况下进行估计。第三，无验证的线性探测器甚至超越了神谕调优的混合：CLAP 平均提高 1.9 个点，LP++ 平均提高 1.5 个点，并且在 K >= 4 时，所有四个无验证基线均高于神谕，线性探测器的边际不包括零。这些结果定位了模型类别的上限，而非超参数：该比例可以在接近最优的情况下免费设置，但它仍然不是性能所在。代码、缓存特征、每个单元记录： https://huggingface.co/datasets/Liangzhi-Li/clipbench-blending
cs.CV / 3 / 2608.23636
Cross-Generation Optimization of YOLOv26, YOLOv11, and YOLOv8 for Fine-Grained Small-Object Detection and Instance Segmentation in Complex Orchards
YOLOv26、YOLOv11 和 YOLOv8 在复杂果园中细粒度小物体检测与实例分割的跨代优化
Abstract
Small-object detection and instance segmentation remain challenging in orchard environments because of green-on-green similarity, occlusion, and limited pixel representation of fine fruit anatomy. This study presents a cross-generation benchmark of Ultralytics YOLOv8, YOLOv11, and YOLOv26 for detecting and segmenting apple fruitlet, calyx, and peduncle structures for robotic orchard perception. Five model scales (n, s, m, l, and x) were evaluated under conventional 640 x 640 and small-object focused 960 x 960 training configurations, yielding 30 experiments. Increasing model capacity did not consistently improve accuracy. YOLOv11s-960 achieved the highest observed mask mAP@50:95 (0.402) and box mAP@50:95 (0.426), while YOLOv26s-960 achieved comparable values of 0.397 and 0.425 with only 10.37 M parameters and 34.1 GFLOPs. Peduncle remained the most challenging class. Overall, compact-to-moderate YOLO models with small-object-focused training provided favorable accuracy efficiency trade-offs, establishing a practical benchmark for fine-grained agricultural robotics and orchard perception. Github Link: https://github.com/rnjnspkt/Optimizing-and-Comparing-Ultralytics-YOLOv26-YOLOv11-and-YOLOv8-for-Small-Object-Detection-and-Seg
Chinese Translation
小物体检测和实例分割在果园环境中仍然面临挑战，主要由于绿色与绿色之间的相似性、遮挡以及细小果实解剖结构的像素表示有限。本研究呈现了 Ultralytics YOLOv8、YOLOv11 和 YOLOv26 的跨代基准，用于检测和分割苹果果实、萼片和花梗结构，以支持机器人果园感知。评估了五种模型规模（n、s、m、l 和 x）在常规的 640 x 640 和专注于小物体的 960 x 960 训练配置下，共进行了 30 次实验。增加模型容量并未始终提高准确性。YOLOv11s-960 达到了最高的观察到的掩膜 mAP@50:95（0.402）和框 mAP@50:95（0.426），而 YOLOv26s-960 则以仅 10.37 M 参数和 34.1 GFLOPs 达到了可比的值 0.397 和 0.425。花梗仍然是最具挑战性的类别。总体而言，采用小物体专注训练的紧凑到中等规模的 YOLO 模型提供了良好的准确性与效率的权衡，为细粒度农业机器人和果园感知建立了实用的基准。Github 链接：https://github.com/rnjnspkt/Optimizing-and-Comparing-Ultralytics-YOLOv26-YOLOv11-and-YOLOv8-for-Small-Object-Detection-and-Seg
cs.CV / 4 / 2608.23664
Scaling Reinforcement Learning for Diffusion Models via Velocity Matching
通过速度匹配扩展强化学习以适应扩散模型
Abstract
Reward fine-tuning is becoming an important tool for adapting diffusion models to human preferences and task-specific objectives, but existing methods largely inherit policy-gradient machinery from large language models. Unlike autoregressive models, diffusion models do not provide tractable likelihoods for generated samples. As a result, current approaches either construct trajectory likelihoods from stochastic denoising transitions or approximate endpoint likelihoods with evidence lower bound, introducing additional computation and algorithmic complexity. We demonstrate that this likelihood-based machinery is not necessary for effective diffusion reward fine-tuning. We propose reward-based velocity matching (RVM), a simple trajectory-free update that acts directly on the velocity field. RVM reinforces directions associated with high-reward generations, suppresses those with low reward, and involves an optional anchor term controlling drift from a reference velocity. Notably, it provides a general framework that recovers recent fine-tuning methods, including RAM and DiffusionNFT, as special cases. Across various large-scale diffusion models reward fine-tuning tasks, RVM is competitive with or outperforms trajectory-based policy-gradient methods under substantially reduced training cost. We further find that, once the velocity update is simplified, the particular loss variant matters less than reward and anchor design. For video generation, standard preference rewards can favor visually clean but nearly static outputs; introducing a new dynamic-tracking reward that substantially improve motions while improving overall VBench performance. These results suggest that scalable reward fine-tuning for diffusion models is better posed in the native velocity representation than as likelihood-based policy optimization.
Chinese Translation
奖励微调正成为将扩散模型适应于人类偏好和特定任务目标的重要工具，但现有方法在很大程度上继承了来自大型语言模型的策略梯度机制。与自回归模型不同，扩散模型并不提供生成样本的可处理似然性。因此，当前的方法要么从随机去噪转变构建轨迹似然，要么用证据下界近似终点似然，导致额外的计算和算法复杂性。我们证明，这种基于似然的机制对于有效的扩散奖励微调并不是必要的。我们提出了基于奖励的速度匹配（RVM），这是一种简单的无轨迹更新，直接作用于速度场。RVM 强化与高奖励生成相关的方向，抑制低奖励的方向，并涉及一个可选的锚定项以控制与参考速度的漂移。值得注意的是，它提供了一个通用框架，可以将最近的微调方法（包括 RAM 和 DiffusionNFT）作为特例恢复。在各种大规模扩散模型的奖励微调任务中，RVM 在显著降低训练成本的情况下与基于轨迹的策略梯度方法具有竞争力或表现更好。我们进一步发现，一旦速度更新被简化，特定的损失变体对奖励和锚定设计的重要性降低。对于视频生成，标准的偏好奖励可能偏向于视觉上干净但几乎静止的输出；引入一种新的动态跟踪奖励，显著改善运动，同时提高整体 VBench 性能。这些结果表明，对于扩散模型，可扩展的奖励微调在本地速度表示中比基于似然的策略优化更为适合。
cs.CV / 5 / 2608.23720
Platonic Representation Hypothesis on World Models
关于世界模型的柏拉图表征假说
Abstract
World models have demonstrated significant potential for perceiving and simulating complex environments. Despite their strong performance, the fundamental nature of their learned representations remains poorly understood. In this paper, we investigate the Platonic Representation Hypothesis within this domain by proposing the Predictive Consistency Assumption: we posit that the optimization of a shared state transition objective acts as a selective pressure that encourages heterogeneous models to converge toward a shared latent structure. Through systematic experiments with the DINO World Model (DINO-WM), in which we vary visual encoders to create heterogeneous models, we find that capable world models evolve toward geometrically similar internal structures. Moreover, via model stitching, we show that the internal features of one world model can be mapped to another with limited performance degradation, providing evidence of functional compatibility. Our findings suggest that the pursuit of predictive consistency can promote shared, transition-compatible latent structure across world models.
Chinese Translation
世界模型在感知和模拟复杂环境方面展现出显著潜力。尽管它们的性能强劲，但其学习表征的基本性质仍然不甚明了。本文通过提出预测一致性假设，探讨了该领域中的柏拉图表征假说：我们假设，共享状态转移目标的优化作为一种选择压力，促使异质模型趋向于共享的潜在结构。通过对 DINO 世界模型（DINO-WM）进行系统实验，我们改变视觉编码器以创建异质模型，发现有能力的世界模型向几何相似的内部结构演化。此外，通过模型拼接，我们展示了一个世界模型的内部特征可以映射到另一个模型上，且性能下降有限，提供了功能兼容性的证据。我们的研究结果表明，追求预测一致性可以促进世界模型之间共享的、兼容状态转移的潜在结构。
cs.CV / 6 / 2608.23723
DriftAD: Visually-Guided Text Drift for Few-Shot Industrial Anomaly Detection
DriftAD：用于少样本工业异常检测的视觉引导文本漂移
Abstract
Few-shot anomaly detection (FSAD) has recently benefited from vision-language models such as CLIP, which enable anomaly de?tection by aligning visual features with text descriptions of normal and abnormal states. However, existing methods typically rely on static text prompts that are applied uniformly across the entire feature hierarchy and spatial dimensions. This rigid global-to-local matching fails to capture the highly localized and scale-dependent physical variations of industrial defects. To address this, we propose DriftAD, a FSAD framework built on three key modules. First, an Anomaly Signal Amplification (ASA) module enhances subtle defect signals through spatial and frequency branches before text-visual matching. Second, Visually-Guided Text Drift (VGTD) dynamically transforms frozen CLIP text embeddings, steering them into layer?wise, spatially-adaptive anomaly descriptors conditioned on local visual context at each encoder depth. Third, Drift-Guided Spatial Gating (DGSG) uses the drifted abnormal descriptor as a spatial probe to selectively enhance anomaly-relevant visual features. Addi?tionally, a drift separation loss prevents representational collapse of the drifted descriptors, and a gate supervision loss enforces spatially discriminative gating in DGSG. Extensive experiments on MVTec?AD and VisA demonstrate state-of-the-art performance across all 1-, 2-, and 4-shot settings on both image-level and pixel-level metrics. Code is available at https://github.com/wenyang001/DriftAD.
Chinese Translation
少样本异常检测（FSAD）最近受益于视觉-语言模型，如CLIP，这些模型通过将视觉特征与正常和异常状态的文本描述对齐来实现异常检测。然而，现有方法通常依赖于在整个特征层次和空间维度上均匀应用的静态文本提示。这种刚性的全局到局部匹配未能捕捉工业缺陷的高度局部化和尺度依赖的物理变化。为了解决这个问题，我们提出了DriftAD，一个基于三个关键模块构建的FSAD框架。首先，异常信号放大（ASA）模块通过空间和频率分支在文本-视觉匹配之前增强微妙的缺陷信号。其次，视觉引导文本漂移（VGTD）动态转换冻结的CLIP文本嵌入，将其引导为基于每个编码器深度的局部视觉上下文条件的层级、空间自适应异常描述符。第三，漂移引导空间门控（DGSG）使用漂移的异常描述符作为空间探针，选择性地增强与异常相关的视觉特征。此外，漂移分离损失防止漂移描述符的表示崩溃，而门控监督损失则在DGSG中强制执行空间上具有区分性的门控。在MVTec-AD和VisA上的大量实验表明，在图像级和像素级指标的1-shot、2-shot和4-shot设置中均实现了最先进的性能。代码可在https://github.com/wenyang001/DriftAD获取。
cs.CV / 7 / 2608.23728
Velocity-coupled Representation Refinement for Satellite Orbit Prediction
速度耦合表示精炼用于卫星轨道预测
Abstract
Satellite orbit prediction, which aims to forecast future orbital trajectories from historical observations, is important for collision warning and safe space operations. With advances in time-series forecasting, learning-based methods have emerged as a promising solution for satellite prediction. In orbital dynamics, a satellite state is typically described by position and velocity, where position characterizes trajectory geometry and velocity reflects its instantaneous direction and rate of change. However, most existing methods mainly focus on temporal dependencies within position sequences while rarely exploiting the intrinsic coupling between position and velocity, which is essential for modeling satellite motion. To this end, we propose OrbitNet, a velocity-aware representation learning method for accurate satellite orbit prediction. It lifts conventional position-sequence forecasting to a position-velocity coupled representation learning paradigm by exploiting relationships among satellite state variables. Specifically, we develop a velocity-coupled representation refinement strategy to enhance positional representations through cross-variable interactions between position and velocity. We further introduce orbital segment modeling, which partitions historical trajectories into temporal segments and performs segment-level temporal learning to capture local motion variations and long-range evolution patterns. Extensive experiments show that OrbitNet outperforms large time-series foundation models and representative general forecasting methods under both in-domain evaluation on Starlink and zero-shot evaluation across six unseen satellite constellations. We expect this work to encourage further exploration of satellite-aware representation learning for trajectory time-series forecasting.
Chinese Translation
卫星轨道预测旨在根据历史观测数据预测未来的轨道轨迹，对于碰撞预警和安全的太空操作至关重要。随着时间序列预测技术的发展，基于学习的方法已成为卫星预测的有希望的解决方案。在轨道动力学中，卫星状态通常由位置和速度描述，其中位置表征轨迹几何形状，速度反映其瞬时方向和变化速率。然而，现有的大多数方法主要关注位置序列中的时间依赖性，鲜有利用位置和速度之间的内在耦合关系，而这对于建模卫星运动是至关重要的。为此，我们提出了OrbitNet，一种关注速度的表示学习方法，用于准确的卫星轨道预测。它通过利用卫星状态变量之间的关系，将传统的位置序列预测提升到位置-速度耦合表示学习范式。具体而言，我们开发了一种速度耦合表示精炼策略，通过位置和速度之间的交叉变量交互来增强位置表示。我们进一步引入轨道段建模，将历史轨迹划分为时间段，并进行段级时间学习，以捕捉局部运动变化和长程演化模式。大量实验表明，OrbitNet在Starlink的领域内评估和六个未见卫星星座的零样本评估中，均优于大型时间序列基础模型和代表性的通用预测方法。我们期望这项工作能够鼓励进一步探索卫星感知的表示学习，以用于轨迹时间序列预测。
cs.CV / 8 / 2608.23730
More Motion Is Not Always Better Motion: Corpus Composition Governs Whether Augmentation Helps SMPL-Based Parkinsonian Gait Severity Estimation
更多的运动并不总是更好的运动：语料库组成决定了增强是否有助于基于SMPL的帕金森步态严重性估计
Abstract
We grade MDS-UPDRS gait severity from SMPL motion using three frozen MotionAGFormer encoders as featurizers, reaching macro-F1 0.58 on a hidden, multi-site test set. Because the system's members differ only in their lifting corpus, evaluating encoders singly on that test set isolates what that corpus contributes. Six pools drawn from one inertial dataset, varying only in which walking tasks they include, score between 0.32 and 0.53, and just one of them beats the 0.51 of an encoder given no outside motion at all. What separates them is not how much data they hold but whether they carry a contrast in walking speed, the variation this representation appears to depend : a further pool adding a third collection site at fixed task composition does worse still. The same rule explains why exact synthetic motion and monocularly reconstructed web video both fail to help. Modifying the learned representation itself, rather than the corpus behind it, cost every variant that attempted it.
Chinese Translation
我们使用三个冻结的MotionAGFormer编码器作为特征提取器，从SMPL运动中对MDS-UPDRS步态严重性进行评分，在一个隐藏的多地点测试集上达到了宏观F1值0.58。由于系统的成员仅在其提升语料库上有所不同，因此在该测试集上单独评估编码器可以隔离出该语料库的贡献。六个来自同一惯性数据集的池，唯一的区别在于它们包含的步态任务，得分在0.32到0.53之间，只有一个池的表现超过了在没有外部运动的情况下编码器的0.51。它们之间的区别不在于数据量的多少，而在于它们是否包含步态速度的对比，这种表示似乎依赖于这种变化：一个在固定任务组成下增加第三个收集地点的池表现更差。同样的规则也解释了为什么精确的合成运动和单目重建的网络视频都未能提供帮助。修改学习到的表示本身，而不是其背后的语料库，使得每一个尝试这样做的变体都遭遇了失败。
cs.CV / 9 / 2608.23746
CRISP: Calibration-Aware Visual State Space Duality for Remote Sensing Semantic Segmentation
CRISP：面向遥感语义分割的校准感知视觉状态空间对偶
Abstract
State space models, especially Visual State Space Duality (VSSD), have emerged as efficient linear-time alternatives to Transformers for dense visual tasks. However, we observe that VSSD compresses spatial context into a global aggregation that suppresses high-frequency responses, causing excessive boundary smoothing in remote sensing semantic segmentation. To address this, we propose CRISP, a calibration framework with two components. Its core, the Duality Calibration Operator (DCO), restores local contrast and boundary responses through residual injection and frequency calibration within the VSSD backbone, without altering its linear complexity. To retain the recovered detail, an Orthogonal Multi-Prototype (OMP) head assigns multiple orthogonally constrained prototypes per class to model large intra-class variance. Extensive experiments on Potsdam, Vaihingen, and LoveDA show that, with approximately 30M parameters, CRISP achieves consistent gains in mean F1 (mF) and mIoU while remaining competitive with state-of-the-art methods. Code is available at https://github.com/crazylifeha/CRISP.
Chinese Translation
状态空间模型，特别是视觉状态空间对偶（Visual State Space Duality, VSSD），已成为密集视觉任务中高效的线性时间替代方案。然而，我们观察到VSSD将空间上下文压缩为全局聚合，这抑制了高频响应，导致遥感语义分割中边界过度平滑。为了解决这个问题，我们提出了CRISP，一个包含两个组件的校准框架。其核心是对偶校准算子（Duality Calibration Operator, DCO），通过残差注入和频率校准在VSSD主干中恢复局部对比度和边界响应，而不改变其线性复杂度。为了保留恢复的细节，一个正交多原型（Orthogonal Multi-Prototype, OMP）头为每个类别分配多个正交约束原型，以建模较大的类内方差。在Potsdam、Vaihingen和LoveDA上的大量实验表明，CRISP在约3000万参数的情况下，在平均F1（mF）和平均交并比（mIoU）上实现了一致的提升，同时与最先进的方法保持竞争力。代码可在https://github.com/crazylifeha/CRISP获取。
cs.CV / 10 / 2608.23752
Too much of a good thing -- when knowledge distillation promotes overfitting, and how to avoid it
好事多磨——当知识蒸馏促进过拟合时，以及如何避免它
Abstract
The growing size of Convolutional Neural Networks has led to increasingly large and costly models. Knowledge Distillation (KD) addresses this by transferring knowledge from a large network (teacher) to a small one (student), also reducing the training data required. KD is traditionally applied only at the network's final output. However, its behaviour when applied at intermediate network layers has received little attention. This raises the question of whether intermediate block-wise KD, which provides supervision throughout the network, could offer an advantage under specific conditions, such as few instances per class, which is common in fine-grained datasets. This work proposes a student design based on simple, homogeneous blocks mirroring those of the teacher, distilling knowledge between corresponding blocks. Across eleven datasets, we show that on classic datasets, distilling only the last block is sufficient -- and often best--, whereas fine-grained, data-scarce settings benefit substantially from intermediate supervision, with even a single additional distillation point narrowing the gap considerably. We further study how this supervision should be guided, exploring configurations of varying granularity and informed by an explainability analysis based on attention maps, Centered Kernel Alignment, and Grad-CAM, alongside the impact of teacher and student fine-tuning strategies. This work shows that intermediate block-wise distillation, guided appropriately, is key to building compact data-efficient models without sacrificing accuracy.
Chinese Translation
卷积神经网络的规模不断扩大，导致模型变得越来越庞大且成本高昂。知识蒸馏（Knowledge Distillation, KD）通过将知识从大型网络（教师）转移到小型网络（学生）来解决这一问题，同时减少所需的训练数据。KD 传统上仅应用于网络的最终输出。然而，当应用于中间网络层时，其行为却鲜有关注。这引发了一个问题，即中间块级 KD 是否能在特定条件下（例如每类样本较少的情况，这在细粒度数据集中很常见）提供优势。本研究提出了一种基于简单、同质块的学生设计，镜像教师的结构，在相应的块之间进行知识蒸馏。在十一种数据集中，我们展示了在经典数据集上，仅蒸馏最后一个块就足够了——并且通常是最佳选择，而在细粒度、数据稀缺的环境中，中间监督显著受益，甚至一个额外的蒸馏点也能显著缩小差距。我们进一步研究了这种监督应如何引导，探索不同粒度的配置，并基于注意力图、中心核对齐（Centered Kernel Alignment）和 Grad-CAM 的可解释性分析，研究教师和学生微调策略的影响。本研究表明，适当引导的中间块级蒸馏是构建紧凑且数据高效模型的关键，而不牺牲准确性。
cs.CV / 11 / 2608.23790
Primate vision reveals a missing principle for robust dynamic AI
灵长类视觉揭示鲁棒动态人工智能的缺失原则
Abstract
How does an intelligent visual system combine what objects look like with how they move while remaining robust as appearance changes? We addressed this question by comparing human perception and neural activity in macaque inferior temporal cortex with representations from image- and video-based neural networks spanning recognition, segmentation, optic-flow processing and predictive world modeling. Temporal integration improved object representations, but most video recognition models generalized poorly when appearance was disrupted while motion structure was preserved. Humans and macaque IT remained robust. Notably, predictive world models combined strong cross-appearance generalization with the closest correspondence to IT, outperforming other video-modeling approaches in neural fidelity. Yet no model reproduced the cortical transformation from early appearance-dominated responses toward later appearance-invariant motion coding. These results identify progressive integration of motion into object representations as a principle of robust dynamic vision and implicate predictive learning as a promising route toward realizing this computation in artificial systems.
Chinese Translation
一个智能视觉系统如何在外观变化的情况下，将物体的形态与运动方式结合起来，同时保持鲁棒性？我们通过比较人类感知和猕猴下颞皮层（macaque inferior temporal cortex）的神经活动，与涵盖识别、分割、光流处理及预测世界建模的基于图像和视频的神经网络表示，来探讨这一问题。时间整合提升了物体表示，但大多数视频识别模型在外观被破坏而运动结构保持时泛化能力较差。人类和猕猴IT区域则保持鲁棒性。值得注意的是，预测世界模型（predictive world models）不仅具备强大的跨外观泛化能力，还与IT区域表现出最接近的对应关系，其神经保真度优于其他视频建模方法。然而，没有任何模型能够再现从早期以外观为主导的响应向后期外观不变的运动编码的皮层转变。这些结果表明，将运动逐步整合进物体表示是实现鲁棒动态视觉的关键原则，并暗示预测学习是实现该计算的人工系统中的有前景路径。
cs.CV / 12 / 2608.23799
Restoring Without Forgetting: Continual Learning Across Image Degradations
无遗忘恢复：跨图像退化的持续学习
Abstract
Recent progress in image restoration has converged on all-in-one architectures that jointly handle multiple degradations within a single network. These methods are effective on static benchmarks but target a closed-world setting that assumes simultaneous access to every target degradation at training time. In practice, degradations are encountered sequentially as field-deployed systems progressively face new environmental conditions, and historical training data is often unavailable due to privacy or storage constraints. Accommodating a new degradation then requires either retraining on the union of all prior data, which is often costly or infeasible, or fine-tuning, which causes catastrophic forgetting. We formulate multi-degradation image restoration as a continual domain-incremental learning problem, in which degradations arrive incrementally and prior data is unavailable. Our proposed Restoring without Forgetting (RwF) framework learns a lightweight adapter for each new degradation, eliminating forgetting by construction at a fraction of the cost of dedicated per-domain networks. To isolate degradation learning from dataset variation, we construct a benchmark spanning five degradation domains under shared image content. At test time, an unsupervised routing mechanism identifies the appropriate restoration path for unknown inputs without requiring domain labels. Across the five-domain sequence, RwF improves final average PSNR over naive sequential fine-tuning by 15.25 dB and 11.83 dB on the Restormer and NAFNet backbones, respectively. The framework transfers to eleven canonical real-degradation benchmarks (3,465 images) at 89.5% routing accuracy with only a +0.94 dB oracle PSNR gap, establishing, to our knowledge, the first systematic baseline for continual multi-degradation image restoration.
Chinese Translation
近期图像恢复的进展集中在一种全能架构上，该架构在单一网络中共同处理多种退化。这些方法在静态基准测试中表现有效，但目标是一个封闭世界的设置，假设在训练时能够同时访问每种目标退化。在实际应用中，退化是逐步出现的，因为现场部署的系统逐渐面临新的环境条件，并且由于隐私或存储限制，历史训练数据往往不可用。因此，适应新的退化需要在所有先前数据的并集上重新训练，这通常成本高昂或不可行，或者进行微调，这会导致灾难性遗忘。我们将多退化图像恢复表述为一个持续的领域增量学习问题，其中退化逐步到达且先前数据不可用。我们提出的无遗忘恢复（Restoring without Forgetting, RwF）框架为每种新的退化学习一个轻量级适配器，通过构建消除遗忘，成本仅为专用每领域网络的一小部分。为了将退化学习与数据集变异隔离，我们构建了一个基准，涵盖五个退化领域，并共享图像内容。在测试时，一种无监督路由机制识别未知输入的适当恢复路径，而无需领域标签。在五个领域序列中，RwF在Restormer和NAFNet主干上分别比简单的顺序微调提高了15.25 dB和11.83 dB的最终平均PSNR。该框架在十一项经典真实退化基准（3,465张图像）上以89.5%的路由准确率转移，仅有+0.94 dB的oracle PSNR差距，建立了我们所知的第一个系统性基准，用于持续多退化图像恢复。
cs.CV / 13 / 2608.23803
LUCAID: Agentic Multimodal AI for Lung Cancer Precision Pathology
LUCAID：用于肺癌精准病理的智能多模态人工智能
Eich, Marie-Lisa, Standvoss, Kai, Milbich, Timo, Möllers, Alexander, Hägele, Miriam, Anders, Philipp, Tharun, Lars, Kontradiuk, Hanna, Kons, Sebastian, Aldoj, Nader, Adigüzel, Recepcan, Narai, Adam, Hönig, Lukas, Striebel, Jonathan, Yang, Binru, Dragomir, Mihnea P., Sextro, Marvin, Keyl, Philipp, Jurmeister, Philipp, Krupar, Rosemarie, Ramberger, Evelyn, Wells, James, Ribbat-Idel, Julika, Kunft, Andreas, Shuaib, Hussam, Grohé, Christian, Büttner, Reinhard, Horst, David, Müller, Klaus-Robert, Ruff, Lukas, Alber, Maximilian, Klauschen, Frederick, Schallenberg, Simon
Abstract
Lung cancer tissue diagnostics is complex, as therapy decisions in precision oncology rely on the integration of histomorphological, immunohistochemical, and molecular features. Yet pathological assessment remains largely visual and semi-quantitative and shows interobserver variability, while existing artificial intelligence (AI) tools cover only selected tasks, rarely reach generalizable expert-level performance, and lack prospective clinical validation. To address these challenges, we developed and clinically validated LUCAID, an agentic AI system for precision lung cancer pathology. An integrative agent couples diagnostic reasoning with nine modules that cover the full routine workflow, from quality control, tumor detection and segmentation, histological subtyping, tumor microenvironment profiling, tumor cellularity quantification, and predictive biomarker scoring (PD-L1, MET, TROP-2) to automated structured report generation. LUCAID enables users to interactively query the module outputs and generate reports that contextualize the results. Against large-scale expert ground-truth annotations, the analysis modules achieved F1 scores of 0.82-0.95. In prospective clinical validation, LUCAID reached 93.0% concordance with an expert-panel adjudicated reference standard across clinically actionable decisions, compared with 68.3-81.1% for five experienced thoracic pathologists.
Chinese Translation
肺癌组织诊断复杂，因为精准肿瘤学中的治疗决策依赖于组织形态学、免疫组化和分子特征的整合。然而，病理评估仍然主要是视觉和半定量的，并且存在观察者间的变异性，而现有的人工智能（AI）工具仅覆盖特定任务，鲜有达到可推广的专家级表现，并且缺乏前瞻性的临床验证。为了解决这些挑战，我们开发并临床验证了LUCAID，一个用于精准肺癌病理的智能AI系统。该集成智能体将诊断推理与九个模块相结合，覆盖从质量控制、肿瘤检测和分割、组织学亚型分类、肿瘤微环境分析、肿瘤细胞密度量化到预测性生物标志物评分（PD-L1、MET、TROP-2）及自动化结构化报告生成的完整常规工作流程。LUCAID使用户能够交互式查询模块输出并生成上下文化的报告。在大规模专家真实标注的对比中，分析模块的F1分数达到了0.82-0.95。在前瞻性临床验证中，LUCAID在临床可操作决策上与专家小组裁定的参考标准达到了93.0%的一致性，而五位经验丰富的胸部病理学家的表现为68.3-81.1%。
cs.CV / 14 / 2608.23836
Predicting Radiologist Expertise from 3D Gaze Patterns During CT Interpretation
基于3D注视模式预测放射科医生的专业水平
Abstract
Accurate interpretation of volumetric CT requires efficient navigation of 3D image volumes and attention to diagnostically relevant regions. While eye-tracking has been widely studied in 2D medical imaging, its use for expertise assessment in CT settings remains limited. We propose a gaze-informed transformer framework for expertise classification in thoracic CT. Using a DINOv2 backbone, radiologist fixation patterns are integrated into volumetric feature learning through (1) a learnable log-space bias in self-attention and (2) gaze-weighted pooling of patch embeddings. We trained and evaluated our approach on 182 CT reading sessions from five radiologists with varying levels of experience. On a held-out test set, the model achieves an ROC-AUC of 0.91 and F1 score of 0.86, outperforming adapted methods. These findings suggest that incorporating visual search behavior into transformers may support objective, process-based expertise assessment in radiology. Code is available via https://github.com/leiluk1/GazeToSkill.
Chinese Translation
准确解读体积CT需要有效导航3D图像体积并关注诊断相关区域。尽管眼动追踪在2D医学影像中得到了广泛研究，但在CT环境中用于专业水平评估的应用仍然有限。我们提出了一种基于注视信息的变换器框架，用于胸部CT的专业水平分类。通过使用DINOv2骨干网络，放射科医生的注视模式通过（1）自注意力中的可学习对数空间偏置和（2）基于注视权重的补丁嵌入池化整合到体积特征学习中。我们在五位具有不同经验水平的放射科医生的182个CT阅读会话上训练和评估了我们的方法。在一个保留的测试集上，该模型达到了0.91的ROC-AUC和0.86的F1分数，优于改编的方法。这些发现表明，将视觉搜索行为纳入变换器可能有助于支持放射学中基于过程的客观专业水平评估。代码可通过https://github.com/leiluk1/GazeToSkill获取。
cs.CV / 15 / 2608.23838
Infant Care Video Dataset for Classification of Interventions Using Transformers
用于干预分类的婴儿护理视频数据集：基于变换器的研究
Abstract
Healthcare documentation in the neonatal intensive care unit (NICU) presents significant challenges, with nurses spending approximately 25\% of their time on record-keeping, while up to 60\% of interventions remain undocumented. Motivated by the need to detect interventions from video automatically, we present the Infant Care Video Dataset (ICVD), a collection of 4,144 videos spanning 12 simulated intervention classes designed for developing automated documentation systems. Our manikin-based approach systematically varies conditions, such as camera angle and clinician skin tone, while ensuring privacy compliance. Using video transformer architectures (TimeSformer and MotionFormer), we establish strong baseline performance (93.97\% and 93.17\% top-1 accuracy) among the 12 infant care classes. Our ablation study comparing temporal models with a framewise approach (23.17\% accuracy) demonstrates a 70.80\% performance gap, validating the need for temporal modeling. The ICVD provides a foundation for developing automated documentation systems to reduce clinical burden in neonatal care environments and improve existing practices.
Chinese Translation
新生儿重症监护病房（NICU）的医疗文档记录面临重大挑战，护士约25%的时间用于记录，而高达60%的干预措施未被记录。为了自动检测视频中的干预措施，我们提出了婴儿护理视频数据集（ICVD），该数据集包含4,144个视频，涵盖12个模拟干预类别，旨在开发自动文档记录系统。我们基于模型的研究系统地改变条件，如摄像机角度和临床医生的肤色，同时确保隐私合规性。使用视频变换器架构（TimeSformer和MotionFormer），我们在12个婴儿护理类别中建立了强大的基线性能（93.97%和93.17%的top-1准确率）。我们的消融研究比较了时间模型与逐帧方法（23.17%的准确率），显示出70.80%的性能差距，验证了时间建模的必要性。ICVD为开发自动文档记录系统提供了基础，旨在减轻新生儿护理环境中的临床负担并改善现有实践。
cs.CV / 16 / 2608.23845
Object Counting Across Modalities: Taxonomies, Benchmarks, Applications, and Open Challenges
跨模态物体计数：分类法、基准、应用与开放挑战
Abstract
Object-counting methods have rapidly shifted from class-specific density regression to open-vocabulary, foundation-model-backed counters. These methods now enumerate instances from various visual and textual prompts. While this shift marks major conceptual progress, our survey argues that claims of universal generality have outpaced the evaluative infrastructure. Most progress metrics rely on a few saturated benchmarks that models exploit for statistical regularities. Newly introduced diagnostic datasets reveal systematic failures in semantic grounding, temporal identity, and spatial reasoning with occlusion. To address these failures, we introduce a five-axis taxonomy (modality, mechanism, prompting, supervision level, and generalization setting). We use this taxonomy to audit the literature across application domains, including microscopy, remote sensing, crowd counting, and agriculture. This formalizes prevailing challenges into six structural contradictions. From these, we propose a roadmap for compositional scene understanding, active counting agents, and unified multimodal evaluation protocols. The main imperative is to build a robust evaluation infrastructure to distinguish open-world generalization from benchmark-specific optimization, rather than simple incremental engineering.
Chinese Translation
物体计数方法迅速从特定类别的密度回归转向开放词汇、基础模型支持的计数器。这些方法现在能够从各种视觉和文本提示中枚举实例。尽管这一转变标志着重大的概念进步，但我们的调查认为，关于普遍性的一些主张已经超出了评估基础设施的能力。大多数进展指标依赖于少数饱和的基准，这些基准被模型利用以寻找统计规律。新近引入的诊断数据集揭示了在语义基础、时间身份和空间推理（尤其是遮挡情况下）的系统性失败。为了解决这些失败，我们提出了一个五轴分类法（模态、机制、提示、监督水平和泛化设置）。我们利用这一分类法对包括显微镜、遥感、人口计数和农业等应用领域的文献进行审计。这一方法将当前面临的挑战形式化为六个结构性矛盾。基于这些矛盾，我们提出了一条关于组合场景理解、主动计数代理和统一多模态评估协议的路线图。主要的任务是建立一个强大的评估基础设施，以区分开放世界泛化与基准特定优化，而不是简单的增量工程。
cs.CV / 17 / 2608.23850
DDMS: Discriminative Distillation of Multi-view Foundational Features into Single-view Models
DDMS：多视角基础特征向单视角模型的区分蒸馏
Abstract
Foundational visual features such as DINO have played a critical role across modern computer vision, and have recently become key components in multi-view feed-forward geometry estimators. In this work, we demonstrate that by re-distilling these multi-view models---their internal knowledge of 3D geometry---into a single-view estimator, we can obtain enhanced 3D consistent foundational features. Our key idea is to construct a multi-view teacher by fusing pretrained 2D foundation features with multi-view geometric features, and refining the fused representation with a discriminative ranking objective. Through our discriminative distillation framework, we enforce the learned features to be both 3D consistent and locally distinctive, while keeping them aligned with the feature space of the original foundation model to preserve the semantic structure of the pretrained representation. Consistency and local discriminability are critical for 3D computer vision problems such as forming semantic and geometric correspondences across images. To demonstrate the effectiveness of our method, we perform comprehensive experiments spanning multiple angles: direct feature analysis, dense prediction transfer, and explicit 3D lifting and rendering. Across these evaluations, our method consistently produces stronger 3D-aware foundation features that improve multi-view consistency and local discriminability while preserving the semantic transferability of the original representation.
Chinese Translation
基础视觉特征如 DINO 在现代计算机视觉中发挥了关键作用，最近已成为多视角前馈几何估计器的关键组成部分。在本研究中，我们展示了通过将这些多视角模型——它们对 3D 几何的内部知识——重新蒸馏到单视角估计器中，我们可以获得增强的 3D 一致基础特征。我们的关键思想是通过将预训练的 2D 基础特征与多视角几何特征融合，构建一个多视角教师，并通过区分排名目标来精炼融合表示。通过我们的区分蒸馏框架，我们强制学习到的特征在保持与原始基础模型特征空间对齐的同时，既要 3D 一致又要局部独特，以保留预训练表示的语义结构。一致性和局部可区分性对于 3D 计算机视觉问题至关重要，例如在图像之间形成语义和几何对应关系。为了证明我们方法的有效性，我们进行了全面的实验，涵盖多个角度：直接特征分析、密集预测转移以及显式 3D 提升和渲染。在这些评估中，我们的方法始终产生更强的 3D 感知基础特征，改善了多视角一致性和局部可区分性，同时保留了原始表示的语义可迁移性。
cs.CV / 18 / 2608.23853
LUX: A Lesion-Aware Graph-Conditioned Visual - Language Architecture for Explainable Endoscopic Captioning
LUX：一种考虑病变的图条件视觉-语言架构，用于可解释的内窥镜图像字幕生成
Abstract
The interpretation of endoscopic imagery in ulcerative colitis is complex and subjective, with variability in human assessment and subtle mucosal inflammation. Although deep learning has advanced automated analysis, most vision-language models rely on global visual embeddings that overlook the localized and relational nature of pathological evidence, limiting clinical reliability and interpretability. We introduce LUX (Lesion-aware Unified eXplainable captioning), a graph-conditioned vision-language architecture for explainable endoscopic image captioning. LUX constructs a lesion-centric scene graph from Grad-CAM and CBAM activation maps, representing pathological regions as nodes and encoding their spatial and clinical relationships. These graph embeddings are integrated into the cross-attention layers of a T5 decoder, enabling generated words to attend to specific lesion nodes rather than only to global image features. This provides direct alignment between linguistic content and pathological evidence, supporting token-level interpretability and relational reasoning. LUX outperforms strong baseline and state-of-the-art medical captioning models across BLEU, METEOR, ROUGE-L, and CIDEr, with particularly strong gains in CIDEr. It also reduces hallucinated clinical findings and improves lesion-level grounding through stronger correspondence between generated tokens and localized pathological regions.
Chinese Translation
对溃疡性结肠炎内窥镜图像的解读复杂且主观，存在人类评估的变异性以及微妙的黏膜炎症。尽管深度学习推动了自动化分析的发展，但大多数视觉-语言模型依赖于全球视觉嵌入，忽视了病理证据的局部和关系特性，从而限制了临床的可靠性和可解释性。我们提出了LUX（Lesion-aware Unified eXplainable captioning），一种用于可解释的内窥镜图像字幕生成的图条件视觉-语言架构。LUX从Grad-CAM和CBAM激活图构建了一个以病变为中心的场景图，将病理区域表示为节点，并编码它们的空间和临床关系。这些图嵌入被整合到T5解码器的交叉注意力层中，使生成的词能够关注特定的病变节点，而不仅仅是全局图像特征。这提供了语言内容与病理证据之间的直接对齐，支持令牌级的可解释性和关系推理。LUX在BLEU、METEOR、ROUGE-L和CIDEr等指标上超越了强基线和最先进的医学字幕生成模型，尤其在CIDEr上取得了显著提升。它还减少了虚构的临床发现，并通过生成的令牌与局部病理区域之间更强的对应关系改善了病变级别的定位。
cs.CV / 19 / 2608.23864
AffineTok: Semantic Affine Consistency for Diffusion-Friendly Visual Tokenizer
AffineTok：用于扩散友好的视觉标记器的语义仿射一致性
Abstract
Visual tokenizers increasingly inject semantic supervision into latent spaces to make downstream diffusion easier. Yet how these semantics should be organized to facilitate denoising remains underexplored. In this paper, we define the semantic recovery objective: the denoising process should recover the semantic content of the clean image from noisy latent, and a good tokenizer should make it easier. Existing approaches train a projector to predict the semantics directly from the noisy latent. We argue that this predicts the average of clean-image semantics, whereas what really needs to be aligned is the semantics of averaged clean latents. More importantly, we demonstrate that the semantic recovery error orthogonally decomposes into the error of the optimal semantic prediction directly from the noisy latent and the error between these two predictions. We therefore identify their consistency as the missing requirement and call it Semantic Affine Consistency (SAC). To examine whether this overlooked requirement is closely related to downstream generation, we introduce M_SAC, a tokenizer-side proxy for SAC. Across the evaluated tokenizers and diffusion model scales, M_SAC closely tracks generation quality, reaching a Pearson correlation of 0.960 with SiT-XL gFID, thereby motivating SAC-guided tokenizer training. We then introduce AffineTok, which promotes SAC through two complementary, training-only components. Global Semantic Coordination Token (GSCT) coordinates the semantic organization of clean latents, keeping semantic averaging meaningful, while Posterior-Mean Semantic Alignment (PMSA) predicts posterior-mean latents from noisy inputs and supervises their semantics. On ImageNet 256, compared with the baseline, AffineTok reduces gFID by 26% at 20 epochs and, with continued training, achieves a new state-of-the-art gFID of 1.21 without classifier-free guidance and 1.10 with guidance.
Chinese Translation
视觉标记器越来越多地将语义监督注入潜在空间，以便于下游扩散。然而，如何组织这些语义以促进去噪仍然未得到充分探讨。在本文中，我们定义了语义恢复目标：去噪过程应从噪声潜在中恢复干净图像的语义内容，而一个好的标记器应使这一过程更为容易。现有的方法训练一个投影器直接从噪声潜在中预测语义。我们认为，这种方法预测的是干净图像语义的平均值，而真正需要对齐的是平均干净潜在的语义。更重要的是，我们证明语义恢复误差正交分解为直接从噪声潜在中进行的最佳语义预测误差和这两种预测之间的误差。因此，我们将它们的一致性识别为缺失的要求，并称之为语义仿射一致性（Semantic Affine Consistency, SAC）。为了检验这一被忽视的要求是否与下游生成密切相关，我们引入了 M_SAC，作为 SAC 的标记器侧代理。在评估的标记器和扩散模型规模中，M_SAC 与生成质量密切相关，达到与 SiT-XL gFID 的皮尔逊相关系数为 0.960，从而推动了基于 SAC 的标记器训练。随后，我们介绍了 AffineTok，它通过两个互补的仅训练组件促进 SAC。全局语义协调标记（Global Semantic Coordination Token, GSCT）协调干净潜在的语义组织，使语义平均保持有意义，而后验均值语义对齐（Posterior-Mean Semantic Alignment, PMSA）则从噪声输入中预测后验均值潜在，并监督其语义。在 ImageNet 256 上，与基线相比，AffineTok 在 20 个训练周期内将 gFID 降低了 26%，并且在持续训练下，达到了不使用无分类器引导的 1.21 的新的最先进的 gFID，以及使用引导的 1.10。
cs.CV / 20 / 2608.23869
Gen2Physics: Grounding Generated 3D Meshes in Physics via Multi-View Material Decomposition
Gen2Physics：通过多视角材料分解将生成的3D网格与物理相结合
Abstract
While state-of-the-art generative models produce high-fidelity 3D meshes, these outputs lack the physical properties required for interactive simulation, gaming, or robotics. We introduce Gen2Physics, a unified and automated framework that grounds generated meshes in physics by automatically decomposing them into their constituent material components. Unlike prior approaches, which focus on volumetric representations incompatible with standard physics engines, Gen2Physics operates directly on meshes to produce immediately simulation-ready assets. Our pipeline integrates a fine-tuned Vision Transformer for dense material segmentation, a robust 2D-to-3D consistency projection, and a Vision-Language Model (VLM) guided refinement that leverages contextual reasoning to assign physical properties and infer internal geometry (solid vs. hollow). By converting surface patches into volumes with distinct densities, our method enables physically plausible dynamic simulations. Experimental results on the ABO-500 and PartNet-Material benchmarks demonstrate that Gen2Physics more than doubles the material segmentation accuracy of prior physics-grounding pipelines (15.6 to 48.3 mIoU), while matching the mass-estimation accuracy of volumetric methods and being the only approach to output watertight per-material sub-meshes.
Chinese Translation
尽管最先进的生成模型能够生成高保真的3D网格，但这些输出缺乏用于交互式模拟、游戏或机器人所需的物理属性。我们提出了Gen2Physics，一个统一且自动化的框架，通过自动将生成的网格分解为其组成材料组件，将其与物理相结合。与以往关注于与标准物理引擎不兼容的体积表示的方法不同，Gen2Physics直接在网格上操作，以生成立即可用于模拟的资产。我们的流程集成了经过微调的视觉变换器（Vision Transformer）用于密集材料分割，一个强健的2D到3D一致性投影，以及一个视觉-语言模型（Vision-Language Model, VLM）引导的细化过程，利用上下文推理来分配物理属性并推断内部几何形状（实心与空心）。通过将表面贴片转换为具有不同密度的体积，我们的方法使得物理上合理的动态模拟成为可能。在ABO-500和PartNet-Material基准上的实验结果表明，Gen2Physics的材料分割准确率超过了先前物理基础管道的两倍（从15.6提升至48.3 mIoU），同时与体积方法的质量估计准确率相匹配，并且是唯一能够输出密闭的每种材料子网格的方法。
cs.CV / 21 / 2608.23880
LG-GER: Language-Guided Group Emotion Recognition via Multimodal Evidence Distillation
LG-GER：通过多模态证据蒸馏的语言引导群体情感识别
Abstract
Inferring the collective emotional state of a group of people from a single image, a task known as group emotion recognition (GER), requires integrating spatially distributed cues such as faces, poses, interactions, and scene context. Current methods rely on detector-driven multi-stream pipelines. These are trained with only image-level supervision that lacks guidance on which regions matter or how strongly each contributes. We propose LG-GER, a language-guided distillation framework that uses a multimodal large language model (MLLM) to generate dense, spatially grounded evidence, i.e., bounding boxes paired with emotion signals and confidence scores, for the training images. This structured evidence is distilled into a single vision-language model (VLM) backbone through four complementary losses: classification, region-text grounding, spatial emotion, and spatial confidence regression. At inference, LG-GER requires no detectors, no MLLM, and no multi-stream fusion, making GER practical for real-time and resource-constrained deployment. LG-GER has been evaluated on two benchmark GER datasets (GroupEmoW and GAF~3.0) and achieves competitive or superior results compared to state-of-the-art methods that require detection and multi-stream processing at inference.
Chinese Translation
从单幅图像推断一组人的集体情感状态的任务被称为群体情感识别（GER），这需要整合空间分布的线索，如面孔、姿势、互动和场景上下文。目前的方法依赖于基于检测器的多流管道，这些方法仅通过图像级监督进行训练，缺乏对重要区域的指导或各区域对情感的贡献程度。我们提出了LG-GER，一种语言引导的蒸馏框架，利用多模态大语言模型（MLLM）生成密集的、空间基础的证据，即与情感信号和置信度分数配对的边界框，用于训练图像。这种结构化证据通过四种互补损失（分类、区域-文本对齐、空间情感和空间置信度回归）蒸馏到一个单一的视觉-语言模型（VLM）主干中。在推理阶段，LG-GER不需要检测器、不需要MLLM，也不需要多流融合，使得GER在实时和资源受限的部署中变得实用。LG-GER在两个基准GER数据集（GroupEmoW和GAF~3.0）上进行了评估，并且与需要检测和多流处理的最先进方法相比，取得了具有竞争力或更优的结果。
cs.CV / 22 / 2608.23903
Continual Visual Learning under Evolving Semantic Concept Shift
在不断演变的语义概念转变下的持续视觉学习
Abstract
Visual foundation models are commonly adapted under the assumption that the appearance of incoming data may change while the semantic meaning of the prediction task remains fixed. In long-lived visual systems, however, taxonomies, policies, and concept definitions can themselves evolve, causing the same visual evidence to require a different interpretation. We study this setting as evolving semantic concept shift and introduce SemReWrite, a framework for selectively updating obsolete visual--semantic mappings while preserving knowledge that remains valid. SemReWrite represents changes between old and revised semantic specifications, combines semantic discrepancy with sparse revised supervision to localize affected visual regions, and uses an input-dependent low-rank rewriting mechanism together with structured semantic memory, preservation, and obsolete-decision suppression. We further introduce EvoShift-Bench, spanning ImageNet, iNaturalist, CUB-200-2011, and DomainNet, with semantic transitions including class split, merge, boundary revision, insertion, partial redefinition, recurrence, and mixed semantic--appearance shift. To explicitly evaluate selective semantic revision, we introduce Rewrite Accuracy (RA) and Preservation Accuracy (PA) for affected and unaffected regions, respectively, Obsolete Retention (OR) for measuring residual outdated semantic associations, and the Selective Revision Score (SRS), which jointly summarizes rewriting and preservation performance. Experiments show that SemReWrite achieves a stronger balance between learning revised semantics and retaining unaffected knowledge than prompt replacement, conventional fine-tuning, parameter-efficient adaptation, and continual-learning strategies.
Chinese Translation
视觉基础模型通常是在假设输入数据的外观可能变化而预测任务的语义意义保持不变的情况下进行适应。然而，在长期存在的视觉系统中，分类法、政策和概念定义本身可能会演变，导致相同的视觉证据需要不同的解释。我们将这种情境研究为不断演变的语义概念转变，并引入SemReWrite，一个选择性更新过时视觉-语义映射的框架，同时保留仍然有效的知识。SemReWrite表示旧的和修订的语义规范之间的变化，将语义差异与稀疏的修订监督结合，以定位受影响的视觉区域，并使用依赖输入的低秩重写机制以及结构化的语义记忆、保留和过时决策抑制。我们进一步介绍了EvoShift-Bench，涵盖ImageNet、iNaturalist、CUB-200-2011和DomainNet，包含的语义转变包括类别分裂、合并、边界修订、插入、部分重新定义、复发和混合语义-外观转变。为了明确评估选择性语义修订，我们引入了重写准确率（Rewrite Accuracy，RA）和保留准确率（Preservation Accuracy，PA），分别用于受影响和未受影响区域，过时保留（Obsolete Retention，OR）用于测量残余的过时语义关联，以及选择性修订得分（Selective Revision Score，SRS），它共同总结了重写和保留的性能。实验表明，SemReWrite在学习修订语义和保留未受影响知识之间实现了比提示替换、传统微调、参数高效适应和持续学习策略更强的平衡。
cs.CV / 23 / 2608.23921
HAP: Head-Adaptive Visual Token Pruning via Cross-Modal Alignment
HAP：通过跨模态对齐的头部自适应视觉标记剪枝
Abstract
Recent Vision-Language Models encode high-resolution images into long visual token sequences, incurring prohibitive prefill costs. To compress them, existing methods score each visual token by averaging text-to-visual attention uniformly across all heads, which assumes every head matches the query. However, our empirical analysis shows that misaligned heads dominate the average, amplifying background tokens and drowning out fine-grained cues. To address this, we propose PAQ (Prompt-Grounded Attention Quality), a metric quantifying how well each head aligns the prompt with image regions. Built on PAQ, our pruning proceeds in three stages. Given a target FLOPs budget, we first partition the transformer layers into groups and allocate a visual token budget to each. Within each group, we then aggregate per-head attention maps via PAQ-weighted softmax into a group-level matrix. Finally, we score visual tokens by this matrix's magnitude and retain the allocated budget per group. By weighting heads with PAQ, our method scores tokens by attention signals that more faithfully reflect prompt relevance, rather than diluting them through uniform averaging. Across 18 benchmarks, our method delivers state-of-the-art trade-offs. Specifically, on LLaVA-1.5-7B (9 tasks), retaining only \textbf{5.6\%} tokens preserves \textbf{99.1\%} of the original performance, surpassing the strongest baseline AutoPrune by 4.2 points. Code is available in https://github.com/baokou-fw2/HAP.
Chinese Translation
近期的视觉-语言模型将高分辨率图像编码为长视觉标记序列，导致高昂的预填充成本。为了压缩这些序列，现有方法通过在所有头部均匀平均文本到视觉的注意力来对每个视觉标记进行评分，这假设每个头部都与查询匹配。然而，我们的实证分析表明，未对齐的头部主导了平均值，放大了背景标记并淹没了细粒度线索。为了解决这个问题，我们提出了PAQ（Prompt-Grounded Attention Quality），一个量化每个头部如何将提示与图像区域对齐的指标。在PAQ的基础上，我们的剪枝过程分为三个阶段。给定目标FLOPs预算，我们首先将变换器层划分为组，并为每组分配视觉标记预算。在每组内，我们通过PAQ加权的softmax将每个头部的注意力图聚合为组级矩阵。最后，我们通过该矩阵的大小对视觉标记进行评分，并保留每组的分配预算。通过使用PAQ对头部进行加权，我们的方法根据更真实地反映提示相关性的注意力信号对标记进行评分，而不是通过均匀平均来稀释它们。在18个基准测试中，我们的方法提供了最先进的权衡。具体而言，在LLaVA-1.5-7B（9个任务）上，仅保留 extbf{5.6\%}的标记即可保留 extbf{99.1\\%}的原始性能，超过最强基线AutoPrune 4.2分。代码可在https://github.com/baokou-fw2/HAP获取。
cs.CV / 24 / 2608.23923
ROI-Gated SAHI: Content-Adaptive Slicing-Based Inference for Efficient Object Detection
ROI门控SAHI：基于内容自适应切片的高效目标检测推理
Abstract
Slicing-Aided Hyper Inference (SAHI) improves small object detection in high-resolution images but often spends substantial compute on background tiles. We propose region-of-interest (ROI)-Gated SAHI, an inference-time framework that introduces a lightweight proposer to localize foreground regions and restrict sliced refinement to informative areas. We evaluate the framework in two settings. On the COCO128 full split dataset comprising 128 images, static ROI-gating is slower on average than Full SAHI, achieving a speed ratio of 0.88, and yields a lower
[email protected] of 0.6602 compared with 0.7569 for Full SAHI. A simple adaptive routing policy with $\tau =$ 0.4 educes the mean latency, achieving a slight gain of 1.02$\times$ over Full SAHI. On a three-image sparse-to-dense case study, ROI-gating achieves speedups ranging from 0.96$\times$ to 6.90$\times$ with a mean speedup of 3.41$\times$. These results show that ROI-gating is most beneficial in sparse scenes and requires policy-based routing for robust average behavior.
Chinese Translation
切片辅助超推理（SAHI）在高分辨率图像中改善了小物体检测，但通常在背景切片上消耗大量计算资源。我们提出了区域兴趣（ROI）门控SAHI，这是一种推理时框架，引入了轻量级提议器以定位前景区域，并将切片细化限制在信息丰富的区域。我们在两个设置中评估该框架。在包含128张图像的COCO128完整拆分数据集上，静态ROI门控的平均速度比完整SAHI慢，速度比为0.88，且其
[email protected]为0.6602，低于完整SAHI的0.7569。一个简单的自适应路由策略（$ au = 0.4$）降低了平均延迟，相较于完整SAHI实现了1.02$ imes$的轻微增益。在一个三图像稀疏到密集的案例研究中，ROI门控的加速范围从0.96$ imes$到6.90$ imes$，平均加速为3.41$ imes$。这些结果表明，ROI门控在稀疏场景中最为有利，并且需要基于策略的路由以实现稳健的平均表现。
cs.CV / 25 / 2608.23927
GlanceWAM: Sparse Test-Time Imagination for World-Action Models
GlanceWAM：世界-动作模型的稀疏测试时想象
Abstract
Video generative models provide rich physical priors for robot learning, yet existing world-action models (WAMs) face a fundamental trade-off: synchronous video generation at control rate is latency-prohibitive, while abandoning test-time visual imagination sacrifices task success. We show that visual imagination achieves both real-time inference and superior success rates when generated asynchronously off the critical path and consumed directly in latent space. We introduce GlanceWAM, which decouples imagination from control within a single video DiT: an asynchronous proposer glances ahead on a slow clock to imagine a single lookahead frame seconds into the future in the background, while an action head decodes action chunks at control rate (48 ms) purely in latent space without blocking. Enabled by a non-interfering attention mask that isolates video representations and staleness-robust horizon training that accommodates asynchronous lookahead aging, GlanceWAM breaks the speed-success dilemma. Trained purely on demonstrations, it attains 72.2% on the 24-task RoboCasa kitchen benchmark (surpassing synchronous Cosmos Policy at 67.1% and imagination-free co-training at 64.4%) and 99.0% on LIBERO, executing at 48 ms per chunk on an NVIDIA A100 GPU (24x faster than synchronous baselines). Code is available at https://github.com/linhanwang/GlanceWAM.
Chinese Translation
视频生成模型为机器人学习提供了丰富的物理先验，但现有的世界-动作模型（WAMs）面临一个基本的权衡：以控制速率同步生成视频会导致延迟问题，而放弃测试时的视觉想象则会牺牲任务成功率。我们展示了当视觉想象在关键路径之外异步生成并直接在潜在空间中消费时，能够实现实时推理和更高的成功率。我们引入了GlanceWAM，它在单个视频DiT中将想象与控制解耦：一个异步提议者在慢速时钟上提前观察，以在后台想象未来几秒的单帧前瞻图像，而一个动作头则在潜在空间中以控制速率（48毫秒）解码动作块，且不造成阻塞。得益于一种不干扰的注意力掩码，该掩码隔离了视频表示，以及适应异步前瞻老化的抗陈旧性训练，GlanceWAM打破了速度与成功率之间的困境。它仅通过示范进行训练，在24任务的RoboCasa厨房基准测试中达到了72.2%的成功率（超过了同步的Cosmos Policy的67.1%和无想象的共同训练的64.4%），在LIBERO上达到了99.0%的成功率，在NVIDIA A100 GPU上以每块48毫秒的速度执行（比同步基线快24倍）。代码可在https://github.com/linhanwang/GlanceWAM获取。
cs.CV / 26 / 2608.23928
RefineRank: Joint Box Refinement and Ranking for Surgical Spatio-Temporal Grounding
RefineRank：手术时空定位的联合框精炼与排序
Abstract
Surgical spatio-temporal grounding (STG) requires locating, at each video time specified by a procedural question, the object that the question asks about. Existing approaches face a trade-off: vision language models understand the question context but produce imprecise coordinates, whereas open-set detectors provide localized candidate boxes whose confidence does not reflect which box answers the question. We introduce RefineRank, which closes this gap at the candidate-box level. A compact trainable module, RefineNet, combines the language and regional features of a frozen medical vision language model with the proposals of a frozen open-set detector: it predicts a bounded coordinate correction and a quality score for every candidate box, and a fixed decoding rule returns the original or refined box with the highest score. On the MedVidBench Official Rankings (Verified), RefineRank records 0.421 STG mIoU, the highest displayed STG score, while its global multi-metric rank is 11. In a controlled evaluation on separate training and evaluation videos, coordinate correction raises the candidate oracle upper bound from 0.6772 to 0.7302, and ranking the joint pool of original and refined candidates by their RefineNet scores improves STG mIoU from 0.2719 to 0.4534, whereas separately trained selectors over the same pool reach at most 0.4186. These results show that a small box-level module can reconcile question understanding with precise localization without retraining either backbone. Code is available at [https://github.com/linzhe001/RefineRank](https://github.com/linzhe001/RefineRank).
Chinese Translation
手术时空定位（STG）要求在每个由程序性问题指定的视频时间点上，定位问题所询问的对象。现有方法面临权衡：视觉语言模型理解问题上下文，但产生不精确的坐标，而开放集检测器提供的局部候选框的置信度并不能反映哪个框能够回答问题。我们提出了RefineRank，旨在在候选框层面弥补这一差距。一个紧凑的可训练模块RefineNet，将冻结的医学视觉语言模型的语言和区域特征与冻结的开放集检测器的提议相结合：它为每个候选框预测一个有界的坐标修正和一个质量评分，并通过固定的解码规则返回得分最高的原始框或精炼框。在MedVidBench官方排名（已验证）中，RefineRank记录了0.421的STG mIoU，显示出最高的STG得分，而其全球多指标排名为11。在对分开的训练和评估视频进行的控制评估中，坐标修正将候选的oracle上限从0.6772提高到0.7302，而通过RefineNet评分对原始和精炼候选的联合池进行排序，使STG mIoU从0.2719提高到0.4534，而在同一池上单独训练的选择器最多只能达到0.4186。这些结果表明，一个小型的框级模块可以在不重新训练任何主干的情况下，将问题理解与精确定位相结合。代码可在[https://github.com/linzhe001/RefineRank](https://github.com/linzhe001/RefineRank)获取。
cs.CV / 27 / 2608.23930
SceneReGen: Generative Reconstruction of 3D Scenes from a Single Image
SceneReGen：基于单幅图像的3D场景生成重建
Abstract
Single-image 3D scene reconstruction must complete partially observed objects and place them coherently in a shared observation-aligned scene frame. Object-level generative priors offer strong completion ability, but their centered, scale-normalized outputs are typically expressed in an object frame, creating a fundamental representation gap between object generation and scene reconstruction. We introduce SceneReGen, a generative reconstruction framework that reinterprets scene reconstruction as the generation and assembly of complete object assets in a shared observation-aligned scene frame. SceneReGen addresses the generation-reconstruction gap through selective pose factorization: each object's observed orientation is encoded directly in the generated mesh, while translation and scale are estimated from instance-level and global scene evidence. Given a scene image and instance masks, a geometry encoder extracts dense cues; learnable shape queries condition a pretrained DiT-based 3D generator to produce complete meshes in their observed orientations, while position queries fuse object and scene features to assemble them in the shared frame. On the 3D-FUTURE evaluation subset, SceneReGen achieves the best scene-level CD, scene-level F-Score, and 3D bounding-box IoU among the evaluated methods, ties the best object-level CD, and ranks second in object-level F-Score. Qualitative outputs in autonomous-driving and embodied-AI scenes further illustrate the potential of asset-centric reconstruction beyond indoor furniture.
Chinese Translation
单幅图像的3D场景重建必须完成部分观察到的物体，并将它们一致地放置在共享的观察对齐场景框架中。物体级生成先验提供了强大的补全能力，但它们的中心化、尺度归一化输出通常以物体框架表示，从而在物体生成与场景重建之间产生了根本的表示差距。我们提出了SceneReGen，一个生成重建框架，将场景重建重新解释为在共享的观察对齐场景框架中生成和组装完整的物体资产。SceneReGen通过选择性姿态因式分解来解决生成与重建之间的差距：每个物体的观察方向直接编码在生成的网格中，而平移和尺度则从实例级和全局场景证据中估计。给定场景图像和实例掩码，几何编码器提取密集线索；可学习的形状查询条件化一个预训练的基于DiT的3D生成器，以产生在其观察方向上的完整网格，而位置查询则融合物体和场景特征，以在共享框架中组装它们。在3D-FUTURE评估子集上，SceneReGen在评估方法中实现了最佳的场景级CD、场景级F-Score和3D边界框IoU，物体级CD与最佳结果持平，并在物体级F-Score中排名第二。在自动驾驶和具身人工智能场景中的定性输出进一步展示了以资产为中心的重建在室内家具之外的潜力。
cs.CV / 28 / 2608.23943
Luce: Relightable Gaussians for 3D Asset Generation
Luce：可重光照的高斯模型用于3D资产生成
Abstract
High-fidelity image-to-3D generation requires a 3D representation that captures both geometry and appearance. To support relighting and integration into standard rendering pipelines, the representation should include physically based rendering (PBR) modalities such as albedo, metallic-roughness, and surface normals. We propose Luce, a 3D representation that unifies geometry and PBR materials within a voxelized multimodal Gaussian cloud, using dedicated Gaussian primitives for each modality. A variational autoencoder compresses this representation into a unified material-aware latent space. A rectified-flow transformer generates this latent from a single image, conditioned on multi-layer features from a pretrained image encoder that preserve both semantic context and fine spatial detail. The latent then decodes into relightable PBR Gaussians and an optional textured mesh with a tangent-space normal map. On Toys4K, Luce achieves state-of-the-art single-image-to-3D generation, improving FID by 28% over the strongest baseline. We further introduce a benchmark of AI-generated images, on which Luce improves the CLIP image-alignment score over the best baseline (0.8519 vs. 0.8299). Luce generates relightable, geometrically accurate, and materially faithful assets that preserve fine details such as text, logos, and inscriptions.
Chinese Translation
高保真图像到3D生成需要一种能够同时捕捉几何形状和外观的3D表示。为了支持重光照并集成到标准渲染管线中，该表示应包括物理基础渲染（PBR）模式，如反照率、金属粗糙度和表面法线。我们提出了Luce，这是一种将几何形状和PBR材料统一在体素化的多模态高斯云中的3D表示，使用专门的高斯原语来表示每种模式。变分自编码器将该表示压缩为统一的材料感知潜在空间。经过校正的流变换器从单幅图像生成该潜在表示，条件是来自预训练图像编码器的多层特征，这些特征保留了语义上下文和细致的空间细节。然后，该潜在表示解码为可重光照的PBR高斯模型和一个可选的带切线空间法线贴图的纹理网格。在Toys4K数据集上，Luce实现了最先进的单图像到3D生成，相比最强基线提高了28%的FID。我们进一步引入了一个AI生成图像的基准测试，Luce在该基准上提高了CLIP图像对齐得分，相比最佳基线（0.8519 vs. 0.8299）。Luce生成的资产在几何上准确且材料上忠实，能够保留文本、标志和铭文等细节。
cs.CV / 29 / 2608.23974
Boot-and-Feedback Framework for Generalist-Expert Model Collaboration in Breast Ultrasound Diagnosis
乳腺超声诊断中通用专家模型协作的引导与反馈框架
Abstract
Breast ultrasound (BUS) is widely used for breast cancer diagnosis yet remains operator-dependent. While deep learning shows promise, ensuring diagnostic reliability and interpretability is challenging. Recent Multimodal Large Language Models (MLLMs) often generate spurious descriptions due to limited domain knowledge, which mislead downstream expert models and compromise clinical validity. To address these challenges, we propose the Boot-and-Feedback (BooF) model collaboration framework for synergistic MLLM-expert interaction. Specifically, in the Boot Stage, the MLLM is guided by the BI-RADS lexicon and preliminary benign-malignant vision-expert predictions, enabling it to transfer general reasoning to BUS analysis while avoiding hallucinations. Subsequently, the Feedback Stage integrates these descriptions with visual features via a lightweight Attention-Gated Cross-Modality Fusion Module. This allows the expert to leverage textual feedback while adaptively filtering noise. Extensive experiments on multiple BUS datasets demonstrate that BooF substantially outperforms state-of-the-art methods in terms of diagnostic accuracy and interpretability.
Chinese Translation
乳腺超声（BUS）广泛应用于乳腺癌诊断，但仍然依赖于操作人员。尽管深度学习展现出潜力，但确保诊断的可靠性和可解释性仍然具有挑战性。近期的多模态大型语言模型（MLLMs）由于领域知识有限，常常生成虚假的描述，这会误导下游专家模型并影响临床有效性。为了解决这些问题，我们提出了引导与反馈（BooF）模型协作框架，以实现多模态大型语言模型与专家之间的协同互动。具体而言，在引导阶段，MLLM在BI-RADS词汇和初步良恶性视觉专家预测的指导下，能够将一般推理转移到乳腺超声分析中，同时避免幻觉。随后，反馈阶段通过轻量级注意力门控跨模态融合模块将这些描述与视觉特征结合。这使得专家能够利用文本反馈，同时自适应地过滤噪声。在多个乳腺超声数据集上的大量实验表明，BooF在诊断准确性和可解释性方面显著优于最先进的方法。
cs.CV / 30 / 2608.23984
Source-Face Authenticity Detection for 3D Gaussian Heads Reconstructed from a Single Portrait: A Benchmark and Dedicated Detector
基于单幅肖像重建的3D高斯头源面孔真实性检测：基准与专用检测器
Abstract
Recent advances in single-image 3D Gaussian head reconstruction have enabled highly realistic and freely renderable digital heads from a single portrait. However, reconstruction and rendering can weaken the forgery traces in the source portrait, making the resulting 3D face difficult to classify whether its underlying face is real or fake, and thereby posing risks to identity authentication and face privacy. To study this problem, we introduce the first large-scale benchmark for this task by collecting real portraits and fake portraits from multiple sources and evaluate representative existing detectors on this benchmark, revealing their lack of explicit mechanisms for retaining fine-grained information and maintaining feature consistency across rendered views. To directly address these two limitations, we propose a detector trained with a two-stage strategy. In Stage I, masked autoencoding encourages the visual backbone to retain the fine-grained appearance information required for local reconstruction, while multi-view contrastive learning enforces feature consistency across rendered views of the same head. Since CLS tokens at different depths exhibit complementary spatial attention patterns, Stage II freezes the adapted backbone and concatenates low-, middle-, and high-level CLS tokens for classification. Experiments show that our method achieves the highest accuracy and ranks first across all reported metrics among the evaluated detectors.
Chinese Translation
最近在单幅图像3D高斯头重建方面的进展，使得从单幅肖像生成高度逼真且可自由渲染的数字头部成为可能。然而，重建和渲染可能会削弱源肖像中的伪造痕迹，使得生成的3D面孔难以判断其底层面孔是真实的还是伪造的，从而对身份认证和面孔隐私构成风险。为了解决这一问题，我们通过收集来自多个来源的真实肖像和伪造肖像，首次引入了这一任务的大规模基准，并在该基准上评估了代表性的现有检测器，揭示了它们在保留细粒度信息和维护渲染视图间特征一致性方面缺乏明确机制。为了直接解决这两个限制，我们提出了一种采用两阶段策略训练的检测器。在第一阶段，掩码自编码鼓励视觉主干保留局部重建所需的细粒度外观信息，而多视图对比学习则强制同一头部的渲染视图之间保持特征一致性。由于不同深度的CLS（Class Token）标记表现出互补的空间注意模式，第二阶段冻结适应后的主干，并将低、中、高层的CLS标记进行拼接以进行分类。实验表明，我们的方法在所有评估的检测器中实现了最高的准确率，并在所有报告的指标中排名第一。
cs.CV / 31 / 2608.24010
Absorbing Gradient Conflicts: Modeling Semantic Variance via Kent Distributions for Cross-Modal Hashing
吸收梯度冲突：通过肯特分布建模语义方差以实现跨模态哈希
Abstract
Supervised proxy-based deep cross-modal hashing has become the dominant paradigm for large-scale retrieval. However, prevalent methods model class proxies as deterministic points in the embedding space. This rigid assumption causes severe gradient conflicts in multi-label scenarios, where gradient conflicts arising from label co-occurrence lead to severe gradient contention and optimization collapse. To resolve this, we propose Kent-based Distributional Proxy Hashing (KDPH), a novel framework that shifts proxy representation from static points to flexible anisotropic Kent distributions on the hypersphere. Unlike point proxies that must shift their positions to accommodate conflicting gradients, KDPH absorbs these conflicts by dynamically adjusting its directional variance. This allows the proxy to maintain a stable semantic mean direction while stretching to cover diverse label correlations. Furthermore, to ensure stable training of these geometric parameters, we derive a tailored loss function incorporating the Cayley transform to enforce strict orthogonality. To the best of our knowledge, KDPH is the first framework to successfully introduce the Kent distributions into cross-modal hashing. Experiments on three benchmark datasets demonstrate that KDPH mitigates proxy collapse and chaotic oscillation, significantly outperforms state-of-the-art methods. Code is available at https://github.com/Senmo996/KDPH-official-code.
Chinese Translation
基于监督代理的深度跨模态哈希已成为大规模检索的主流范式。然而，现有方法将类别代理建模为嵌入空间中的确定性点。这一僵化的假设在多标签场景中导致严重的梯度冲突，其中标签共现引发的梯度冲突导致严重的梯度争用和优化崩溃。为了解决这个问题，我们提出了基于肯特分布的代理哈希（Kent-based Distributional Proxy Hashing, KDPH），这是一个新颖的框架，它将代理表示从静态点转变为超球面上的灵活各向异性肯特分布。与必须移动位置以适应冲突梯度的点代理不同，KDPH通过动态调整其方向方差来吸收这些冲突。这使得代理能够保持稳定的语义均值方向，同时扩展以覆盖多样的标签相关性。此外，为了确保这些几何参数的稳定训练，我们推导出了一种定制的损失函数，结合了凯莱变换以强制执行严格的正交性。据我们所知，KDPH是首个成功将肯特分布引入跨模态哈希的框架。在三个基准数据集上的实验表明，KDPH缓解了代理崩溃和混乱振荡，显著优于最先进的方法。代码可在 https://github.com/Senmo996/KDPH-official-code 获取。
cs.CV / 32 / 2608.24020
IterCAD: Iterative Program Repair for CAD Code Generation from Orthographic Views
IterCAD：基于正投影视图的CAD代码生成的迭代程序修复
Abstract
Generating executable parametric CAD code from dimension-annotated orthographic drawings is a challenging task requiring geometric understanding, procedural reasoning, and precise numerical prediction. Existing vision-language approaches typically formulate this problem as one-shot generation, preventing the model from inspecting intermediate CAD results and correcting early mistakes, often leading to non-executable code or geometrically inconsistent outputs. In this paper, we propose IterCAD, an iterative framework that reformulates orthographic-view-to-CAD generation as a progressive program repair process. Instead of predicting the final CAD code in a single pass, IterCAD repeatedly analyzes the current CAD result, reasons about its discrepancy with the target views, and explicitly decides whether to REVISE the code or STOP the refinement process. To make iterative repair learnable, we further construct IterCAD-RS, a structured revise-or-stop supervision set containing both repairable intermediate CAD states and already-correct states, and develop a three-stage training strategy for initial generation, revision learning, and multi-turn RL optimization. By closing the loop between visual understanding, geometric verification, and code refinement, IterCAD progressively corrects structural and parametric errors. Experiments on CADExpert show that IterCAD consistently improves code executability and geometric fidelity over strong one-shot baselines.
Chinese Translation
从带有尺寸标注的正投影图生成可执行的参数化CAD代码是一项具有挑战性的任务，要求具备几何理解、过程推理和精确的数值预测。现有的视觉-语言方法通常将此问题表述为一次性生成，限制了模型检查中间CAD结果和纠正早期错误的能力，常常导致生成不可执行的代码或几何上不一致的输出。本文提出了IterCAD，一个将正投影视图到CAD生成重新表述为渐进程序修复过程的迭代框架。IterCAD不是在一次性预测最终的CAD代码，而是反复分析当前的CAD结果，推理其与目标视图的差异，并明确决定是修订代码还是停止修正过程。为了使迭代修复可学习，我们进一步构建了IterCAD-RS，一个包含可修复的中间CAD状态和已修正状态的结构化修订或停止监督集，并开发了一个三阶段的训练策略，涵盖初始生成、修订学习和多轮强化学习优化。通过在视觉理解、几何验证和代码修正之间闭合循环，IterCAD逐步纠正结构和参数错误。在CADExpert上的实验表明，IterCAD在代码可执行性和几何保真度上始终优于强大的单次生成基线。
cs.CV / 33 / 2608.24025
Low-Rank Velocity Fields as a Structural Prior for Unsupervised 4D Medical Image Interpolation
低秩速度场作为无监督4D医学图像插值的结构先验
Abstract
Endpoint-only unsupervised 4D medical image interpolation synthesizes intermediate volumes from sparsely sampled sequences with only the start and end volumes available for training; however, this weakly constrained setting often yields intermediates with unstable boundaries and non-physiological motion, limiting interpretability and downstream analysis. We propose low-rank velocity fields as a structural prior, constraining motion to a structured Tucker low-rank velocity field space that decomposes motion into globally shared spatial bases and a compact sample-specific core, thereby encouraging spatially correlated, anatomy-consistent deformation while suppressing voxel-wise high-frequency artifacts. To capture global coordination and local non-rigid details, we model motion in a coarse-to-fine multi-scale scheme and compose scale-wise deformations at inference to synthesize volumes at arbitrary times. We further provide a theoretical analysis showing that, under Tucker parameterization, low-rank parameters control the smoothness energy of the velocity field, explaining why low-rank modeling promotes smoother motion. Experiments on ACDC and 4D-Lung demonstrate state-of-the-art performance, remaining competitive with methods trained with intermediate-frame supervision, and producing intermediates with improved structural coherence and more stable anatomical contours.
Chinese Translation
仅使用端点的无监督4D医学图像插值从稀疏采样序列中合成中间体积，仅使用起始和结束体积进行训练；然而，这种弱约束的设置往往导致中间体积具有不稳定的边界和非生理运动，从而限制了解释性和后续分析。我们提出将低秩速度场作为结构先验，将运动约束在一个结构化的Tucker低秩速度场空间中，该空间将运动分解为全局共享的空间基和紧凑的样本特定核心，从而鼓励空间相关的、符合解剖学的变形，同时抑制体素级的高频伪影。为了捕捉全局协调和局部非刚性细节，我们在粗到细的多尺度方案中建模运动，并在推理时组合尺度级变形，以合成任意时间的体积。我们进一步提供理论分析，表明在Tucker参数化下，低秩参数控制速度场的平滑能量，解释了为什么低秩建模促进了更平滑的运动。在ACDC和4D-Lung上的实验表明，所提方法具有最先进的性能，与使用中间帧监督训练的方法相比仍具竞争力，并生成具有更好结构一致性和更稳定解剖轮廓的中间体积。
cs.CV / 34 / 2608.24027
Phase-Aligned Finite-Fourier Periodic Deformation for 4D Medical Image Interpolation
相位对齐的有限傅里叶周期变形用于4D医学图像插值
Abstract
4D medical image interpolation aims to recover missing volumes from sparsely observed time points and is important for dynamic anatomical analysis in applications such as cardiac MRI and thoracic CT, where motion is often repetitive or near-periodic over clinically relevant intervals. A key challenge is that this structure is not always encoded directly in deformation representations for interpolation. In addition, physiological motion is often non-uniform, so equal temporal intervals do not necessarily correspond to equal amounts of anatomical change. To address these issues, we formulate interpolation as learning a continuous deformation process with a phase-structured prior. Given two endpoint volumes, we parameterize a phase-conditioned velocity field with a finite Fourier basis, which embeds near-periodic motion patterns directly into the deformation space and supports continuous querying at arbitrary target times. We further introduce a phase-aligned temporal reparameterization that maps normalized within-interval time to a latent motion phase according to deformation variation intensity, thereby better modeling non-uniform motion progression. Intermediate volumes are then synthesized by continuously warping both endpoints, followed by bidirectional fusion and lightweight residual refinement. Experiments on ACDC and 4D-Lung show that the proposed method achieves state-of-the-art performance over existing baselines while producing anatomically plausible and coherent intermediate volumes from sparse observations.
Chinese Translation
4D医学图像插值旨在从稀疏观察的时间点恢复缺失的体积，对于心脏MRI和胸部CT等应用中的动态解剖分析至关重要，这些应用中运动往往在临床相关的时间间隔内是重复或近周期性的。一个关键挑战是，这种结构并不总是直接编码在插值的变形表示中。此外，生理运动通常是非均匀的，因此相等的时间间隔不一定对应于相等的解剖变化量。为了解决这些问题，我们将插值公式化为学习具有相位结构先验的连续变形过程。给定两个端点体积，我们用有限傅里叶基函数对相位条件的速度场进行参数化，将近周期性运动模式直接嵌入变形空间，并支持在任意目标时间的连续查询。我们进一步引入了一种相位对齐的时间重新参数化，将标准化的区间内时间映射到根据变形变化强度的潜在运动相位，从而更好地建模非均匀运动进程。然后，通过连续扭曲两个端点来合成中间体积，随后进行双向融合和轻量级残差精炼。在ACDC和4D-Lung上的实验表明，所提出的方法在现有基线之上实现了最先进的性能，同时从稀疏观察中生成了解剖上合理且连贯的中间体积。
cs.CV / 35 / 2608.24043
ConsensusTAS: Self-Supervised Temporal Action Segmentation for Long-Horizon Construction Videos
ConsensusTAS：用于长时间构建视频的自监督时序动作分割
Abstract
Recognizing sequential construction activities is important for collaborative human-robot work; for example, robots are able to understand workers' current and upcoming actions and provide timely tool delivery or physical support. However, despite extensive research on construction worker activity recognition, existing studies have been limited to classifying activity categories, such as climbing, lifting, and walking, instead of recognizing fine-grained activity transitions from long-horizon sequences. Addressing this problem is challenging because annotating action temporal boundaries in long construction videos is time-consuming. In this study, we propose ConsensusTAS, a label-free, self-supervised learning approach to segment continuous video streams into distinct activity phases by exploiting the internal consensus of candidate segmentations. We evaluated our algorithm on three public datasets, where it outperformed state-of-the-art methods, achieving an F1@10 of 73.08 on GTEA, an F1@10 of 64.33 on Breakfast, and an F1@50 of 33.50 on static-camera videos from Assembly101. We also tested it on real-world construction videos, where post-hoc evaluation showed that the model successfully recognized and segmented actions within the composite activity of bricklaying, such as spreading mortar on a brick, placing the brick, pressing, and aligning. Compared with other temporal action segmentation models that require computationally intensive large vision-language models, our method can run on a CPU, which provides practical value for video surveillance and human-robot collaboration on mobile robotic platforms.
Chinese Translation
识别连续的施工活动对于人机协作工作至关重要；例如，机器人能够理解工人当前和即将进行的动作，并提供及时的工具交付或物理支持。然而，尽管在建筑工人活动识别方面进行了广泛研究，现有研究仍然局限于对活动类别的分类，如攀爬、举起和行走，而未能识别长时间序列中的细粒度活动过渡。解决这个问题具有挑战性，因为在长时间的施工视频中标注动作的时间边界是耗时的。在本研究中，我们提出了ConsensusTAS，一种无标签的自监督学习方法，通过利用候选分割的内部共识，将连续的视频流分割为不同的活动阶段。我们在三个公共数据集上评估了我们的算法，结果超越了最先进的方法，在GTEA上达到了73.08的F1@10，在Breakfast上达到了64.33的F1@10，在Assembly101的静态摄像头视频上达到了33.50的F1@50。我们还在真实的施工视频上进行了测试，后期评估显示该模型成功识别并分割了砖砌复合活动中的动作，如在砖上涂抹砂浆、放置砖块、压实和对齐。与其他需要计算密集型大型视觉-语言模型的时序动作分割模型相比，我们的方法可以在CPU上运行，这为视频监控和移动机器人平台上的人机协作提供了实际价值。
cs.CV / 36 / 2608.24053
WeMM-Embedding: WeChat Multi-Modal Embedding Technical Report
WeMM-Embedding：微信多模态嵌入技术报告
Abstract
Universal multimodal embeddings are becoming a core component of modern AI systems, enabling heterogeneous content to be represented in a shared space for applications such as retrieval, recommendation, classification, and agentic systems. In this report, we present WeMM-Embedding, a family of universal multimodal embedding models supporting text, images, videos, visual documents, and arbitrarily interleaved multimodal inputs with flexible output dimensions. The family comprises 2B, 4B, and 9B variants and is trained in two stages: a large-scale multimodal alignment stage, followed by a refinement stage using curated data, fine-grained relevance supervision, and cross-scale knowledge transfer. Across extensive evaluations, WeMM-Embedding achieves leading performance on multiple public benchmarks. Notably, the 2B variant already surpasses the previously leading 8B open-source baseline on MMEB-v2, while the 9B variant further achieves a new state-of-the-art overall score of 80.6. WeMM-Embedding also demonstrates strong practical performance across WeChat applications, with substantial gains on a 26-task in-house benchmark and consistent improvements across 14 online A/B tests. It has been deployed at scale across recommendation and search applications, including WeChat Channels, Official Accounts, Moments, and e-commerce services. We have released the model weights and code to facilitate future research at https://github.com/Tencent/WeMM-Embedding.
Chinese Translation
通用多模态嵌入正成为现代人工智能系统的核心组成部分，使得异构内容能够在共享空间中表示，以便于检索、推荐、分类和智能系统等应用。在本报告中，我们提出了WeMM-Embedding，一个支持文本、图像、视频、视觉文档以及任意交错多模态输入的通用多模态嵌入模型系列，具有灵活的输出维度。该系列包括2B、4B和9B变体，并分为两个阶段进行训练：首先是大规模多模态对齐阶段，随后是使用精选数据、细粒度相关性监督和跨尺度知识转移的精细化阶段。在广泛的评估中，WeMM-Embedding在多个公共基准测试中取得了领先的性能。值得注意的是，2B变体已经超过了之前领先的8B开源基线在MMEB-v2上的表现，而9B变体则进一步实现了80.6的新状态-of-the-art整体得分。WeMM-Embedding在微信应用中也展现了强大的实际性能，在一个26项的内部基准测试中取得了显著提升，并在14个在线A/B测试中持续改善。该模型已在推荐和搜索应用中大规模部署，包括微信视频号、公众号、朋友圈和电子商务服务。我们已发布模型权重和代码，以促进未来的研究，网址为 https://github.com/Tencent/WeMM-Embedding。
cs.CV / 37 / 2608.24063
VisCache: Visual KV Cache Pruning for Efficient Vision Large Language Model Inference
VisCache：用于高效视觉大型语言模型推理的视觉KV缓存剪枝
Abstract
While Vision Large Language Models (VLLMs) have achieved remarkable success in multimodal reasoning, their long-context inference remains prohibitively expensive due to the massive computation and memory overhead of visual Key-Value (KV) caches. Existing KV compression methods often apply uniform pruning across visual tokens and layers, leading to substantial information loss and degraded performance.To address this challenge, we propose \textbf{VisCache}, a plug-and-play framework for coarse-to-fine \textbf{Vis}ual KV \textbf{Cache} pruning without training, which consists of two synergistic stages. First, a lightweight VLM filters temporal redundancy by selectively forwarding semantically informative keyframes. Second, we introduce {PruneKV}, a surgical KV compression algorithm tailored to the attention dynamics of VLLMs. Unlike rigid pruning strategies, PruneKV adopts a parabolic layer-wise budget allocation together with an asymmetric update mechanism that selectively prunes keys while fusing values, thereby preserving critical contextual information. Extensive experiments demonstrate that VisCache substantially improves inference efficiency, achieving up to {2.35$\times$ speedup} and significant memory reduction while maintaining competitive performance with only {19--28\%} KV cache retention. VisCache consistently outperforms existing baselines, establishing a new Pareto frontier between efficiency and performance for long-context VLLM inference. Code is available at https://github.com/Wlklk/VisCache
Chinese Translation
尽管视觉大型语言模型（VLLMs）在多模态推理方面取得了显著成功，但其长上下文推理仍然因视觉键值（KV）缓存的巨大计算和内存开销而成本高昂。现有的KV压缩方法通常在视觉标记和层之间应用均匀剪枝，导致信息损失和性能下降。为了解决这一挑战，我们提出了 extbf{VisCache}，一个无需训练的粗到细 extbf{Vis}ual KV extbf{Cache}剪枝的即插即用框架，包含两个协同阶段。首先，一个轻量级的VLM通过选择性地转发语义信息丰富的关键帧来过滤时间冗余。其次，我们引入了{PruneKV}，一种针对VLLMs注意力动态的外科KV压缩算法。与刚性剪枝策略不同，PruneKV采用抛物线层级预算分配以及不对称更新机制，选择性地剪枝键的同时融合值，从而保留关键的上下文信息。大量实验表明，VisCache显著提高了推理效率，达到了{2.35$ imes$加速}和显著的内存减少，同时在仅保留{19--28 ext{%}}的KV缓存的情况下保持了竞争力的性能。VisCache始终优于现有基线，建立了长上下文VLLM推理中效率与性能之间的新Pareto前沿。代码可在https://github.com/Wlklk/VisCache获取。
cs.CV / 38 / 2608.24068
Representation Learning in Diffusion and Flow-based Model: An Application Aspect
扩散与流模型中的表征学习：应用视角
Abstract
Diffusion models and flow-based models have recently become the dominant paradigms in generative modeling, largely due to their ability to learn rich, multi-level visual representations through large-scale training. This creates a bidirectional relationship between generative models and representation learning: improving representation learning enhances generation quality, while the learned representations can be leveraged for broader understanding tasks. This survey systematically explores this interplay with a focus on applications. We propose a three-tier progressive framework that organizes existing works from three perspectives: using representation learning to improve generative capabilities, exploiting generative models to extract representations for perception tasks, and ultimately moving toward general-purpose unified applications. We systematically categorize representative methods across a wide range of downstream tasks, including image classification, dense visual prediction, instance-level perception, and annotation-scarce scenarios. By providing a unified taxonomy and identifying key challenges, this survey aims to clarify the underlying logic of current research and suggest promising directions for future exploration. We hope this work can serve as a valuable reference for researchers interested in harnessing the representation power of generative models for applications beyond generation.
Chinese Translation
扩散模型和流模型最近已成为生成建模的主导范式，这在很大程度上得益于它们通过大规模训练学习丰富的多层次视觉表征的能力。这在生成模型与表征学习之间形成了双向关系：改善表征学习可以提升生成质量，而学习到的表征可以用于更广泛的理解任务。本调查系统地探讨了这种相互作用，重点关注应用方面。我们提出了一个三级渐进框架，从三个角度组织现有工作：利用表征学习提高生成能力，利用生成模型提取感知任务的表征，最终朝着通用统一应用的方向发展。我们系统地对代表性方法进行了分类，涵盖了广泛的下游任务，包括图像分类、密集视觉预测、实例级感知和标注稀缺场景。通过提供统一的分类法并识别关键挑战，本调查旨在阐明当前研究的基本逻辑，并提出未来探索的有希望方向。我们希望这项工作能为有兴趣利用生成模型的表征能力进行超越生成应用的研究者提供有价值的参考。
cs.CV / 39 / 2608.24093
Joint-Embedding Prediction of Masked Point Tubes for Self-Supervised Learning on 4D Point Cloud Videos
基于联合嵌入预测的掩膜点管自监督学习在4D点云视频中的应用
Abstract
Self-supervised representation learning for 4D point cloud videos is challenging because annotations are costly and reconstruction-based pretraining can overemphasize low-level geometric details. We propose a JEPA-style framework that learns from unlabeled spatiotemporal point clouds through latent point-tube prediction. Instead of reconstructing raw coordinates, the model masks spatiotemporal regions and predicts their target representations from visible context representations in feature space. To stabilize latent prediction, we incorporate Sketched Isotropic Gaussian Regularization, which encourages non-collapsed embeddings without relying on explicit reconstruction targets. This formulation aims to capture both spatial structure and temporal dynamics while keeping the pretraining objective aligned with downstream semantic recognition. Experiments on action and gesture recognition benchmarks show that the learned representations improve downstream fine-tuning, limited-label learning, and cross-dataset transfer. These results suggest that JEPA-style latent prediction is a promising alternative to reconstruction-centered pretraining for 4D point cloud videos.
Chinese Translation
4D点云视频的自监督表示学习面临挑战，因为标注成本高昂且基于重建的预训练可能过于强调低级几何细节。我们提出了一种JEPA（Joint-Embedding Prediction Approach）风格的框架，通过潜在点管预测从未标记的时空点云中学习。该模型不是重建原始坐标，而是掩盖时空区域，并从特征空间中的可见上下文表示预测其目标表示。为了稳定潜在预测，我们引入了素描各向同性高斯正则化（Sketched Isotropic Gaussian Regularization），该方法鼓励非压缩嵌入，而不依赖于显式重建目标。该公式旨在捕捉空间结构和时间动态，同时保持预训练目标与下游语义识别的一致性。在动作和手势识别基准上的实验表明，所学习的表示改善了下游微调、有限标签学习和跨数据集迁移。这些结果表明，JEPA风格的潜在预测是4D点云视频中重建中心预训练的有前景的替代方案。
cs.CV / 40 / 2608.24105
DRRG: A Discrete Diffusion Framework for Radiology Report Generation
DRRG：一种用于放射学报告生成的离散扩散框架
Abstract
Purpose: Automatic radiology report generation (RRG) has been widely explored to improve reporting accuracy and reduce radiologists' workload. Most existing methods rely on autoregressive (AR) frameworks that generate reports token by token and cannot revise earlier content, making them prone to error propagation and inconsistent with the iterative refinement process of radiological reporting. In contrast, discrete diffusion large language models (DLLMs) generate text through iterative denoising, naturally enabling report refinement. However, DLLMs have not been extensively investigated for RRG. In this study, we developed and evaluated a discrete diffusion framework for RRG that enables iterative refinement rather than conventional left-to-right autoregressive decoding. Materials and methods: We developed DRRG, a DLLM-based framework that formulates RRG as iterative masked-token denoising. DRRG incorporates a clinical-entities-aware complementary mask to improve token supervision coverage and emphasize clinically important entities, together with a concept-conditioning module that injects image-derived clinical concepts into visual representations. DRRG was trained and evaluated on MIMIC-CXR and CheXpert Plus. Results: On MIMIC-CXR, DRRG achieved BLEU-4 of 0.210, CheXpert-F1 of 0.549, RadGraph-F1 of 0.281, GREEN of 0.360, and RaTEScore of 0.604, outperforming the compared methods on most reported metrics, despite employing a substantially smaller LLM decoder. On CheXpert Plus, DRRG achieved the highest BLEU-4 (0.119) and CheXpert-F1 (0.347) among the compared methods. Conclusion: Discrete diffusion provides an effective alternative to autoregressive radiology report generation by enabling iterative, bidirectional report refinement. Incorporating clinically focused masking and image-derived concept conditioning improves report quality and clinical consistency.
Chinese Translation
目的：自动放射学报告生成（RRG）已被广泛研究，以提高报告的准确性并减轻放射科医生的工作负担。现有大多数方法依赖于自回归（AR）框架，逐个生成报告的标记，无法修订早期内容，这使得它们容易出现错误传播，并与放射学报告的迭代完善过程不一致。相比之下，离散扩散大型语言模型（DLLMs）通过迭代去噪生成文本，自然地支持报告的完善。然而，DLLMs在RRG中的应用尚未得到广泛研究。在本研究中，我们开发并评估了一种用于RRG的离散扩散框架，该框架支持迭代完善，而不是传统的从左到右的自回归解码。材料与方法：我们开发了DRRG，一个基于DLLM的框架，将RRG公式化为迭代掩码标记去噪。DRRG结合了临床实体感知的补充掩码，以提高标记监督覆盖率并强调临床重要实体，同时引入了一个概念条件模块，将图像衍生的临床概念注入视觉表示中。DRRG在MIMIC-CXR和CheXpert Plus上进行了训练和评估。结果：在MIMIC-CXR上，DRRG达到了BLEU-4为0.210，CheXpert-F1为0.549，RadGraph-F1为0.281，GREEN为0.360，以及RaTEScore为0.604，尽管使用了明显较小的LLM解码器，但在大多数报告的指标上超越了比较方法。在CheXpert Plus上，DRRG在比较方法中达到了最高的BLEU-4（0.119）和CheXpert-F1（0.347）。结论：离散扩散为自回归放射学报告生成提供了一种有效的替代方案，通过支持迭代的双向报告完善，结合临床聚焦的掩码和图像衍生的概念条件，提高了报告质量和临床一致性。
cs.CV / 41 / 2608.24107
MatReplace: A Reference-Free, Conditioning-Aligned Benchmark for Material Replacement in Interior Scenes
MatReplace：一种无参考、条件对齐的室内场景材料替换基准
Abstract
Material replacement is a common interior-design operation: changing the material of a selected surface while preserving its geometry, surroundings, and illumination. Despite its commercial relevance, no public benchmark isolates this task, and evaluating it is challenging. Reference-based metrics penalize valid outputs in this inherently one-to-many setting, favor the style of the reference generator, and cannot fairly compare editors that receive different forms of guidance. We introduce MatReplace, a reference-free benchmark that evaluates edits along four verifiable dimensions: local material correctness, global lighting harmony, outside preservation, and inside structure. It defines three tracks that vary one conditioning signal at a time: (A) instruction only, (B) instruction plus region mask, and (C) material reference image instead of instruction. Our results reveal a clear divide between naming and visually grounding materials. In Track A, leading closed-source editors achieve exemplar-level material rendering and surpass the exemplar anchor under our primary aggregate. In Track B, masks help only mask-compatible models with weak scene preservation, with task-paired, single-seed effects ranging from +0.137 to -0.090 across aligned model families. In Track C, reference-image conditioning degrades every family under both aggregates, by -0.031 to -0.508; in the worst cases, models repaint the reference image itself and perform worse than returning the input unchanged. Thus, named-material rendering is largely solved by the strongest closed editors on this distribution, but grounding materials from pixels remains an open challenge. Expert ratings validate our ranking (Kendall's tau = 0.68) and align with our aggregates more closely than GT-referenced or CLIP-based baselines.
Chinese Translation
材料替换是常见的室内设计操作：在保持几何形状、周围环境和照明的情况下，更改选定表面的材料。尽管这一操作在商业上具有重要意义，但目前没有公共基准专门针对这一任务，评估也颇具挑战性。基于参考的度量在这一固有的一对多设置中惩罚有效输出，偏向于参考生成器的风格，并且无法公平比较接受不同形式指导的编辑器。我们提出了MatReplace，这是一种无参考的基准，沿四个可验证维度评估编辑效果：局部材料正确性、全局光照和谐性、外部保留和内部结构。它定义了三个轨道，每次变化一个条件信号：（A）仅指令，（B）指令加区域掩码，以及（C）材料参考图像代替指令。我们的结果揭示了命名材料与视觉基础材料之间的明显差异。在轨道A中，领先的闭源编辑器实现了典范级的材料渲染，并在我们的主要综合指标下超越了典范锚点。在轨道B中，掩码仅对掩码兼容模型有帮助，且这些模型的场景保留能力较弱，任务配对的单种子效果在对齐模型家族中范围为+0.137到-0.090。在轨道C中，参考图像条件使每个家族在两个综合指标下均表现下降，降幅为-0.031到-0.508；在最糟糕的情况下，模型重新绘制参考图像本身，表现甚至不如返回未更改的输入。因此，命名材料渲染在这一分布上基本上被最强的闭源编辑器解决，但从像素中定位材料仍然是一个未解的挑战。专家评分验证了我们的排名（Kendall's tau = 0.68），并与我们的综合指标比GT参考或基于CLIP的基线更为一致。
cs.CV / 42 / 2608.24119
TransPhy: Visual In-Context Learning for Physically Grounded Image Editing
TransPhy：面向物理基础的图像编辑的视觉上下文学习
Abstract
Visual demonstrations provide a natural interface for specifying image transformations that are difficult to describe exhaustively with text. However, existing visual in-context learning (VICL) methods primarily focus on appearance-level relation transfer and provide limited support for physically grounded transformations, whose outcomes depend on material properties, geometry, object interactions, and environmental conditions. Given a source--target exemplar pair and a query image, physically grounded VICL requires a model to infer the demonstrated transformation, adapt its effects to the query-specific scene context, and preserve rule-irrelevant content. We introduce PhysVICL-74, comprising 74 physically grounded transformation rules and 5,240 source--target image pairs that form nearly 75K training and evaluation contexts. Its benchmark split separately evaluates novel-instance transfer and unseen-rule generalization. We further propose TransPhy, a framework that decomposes physically grounded VICL into physical-rule induction and transition-aligned rendering. TransPhy first predicts the demonstrated rule and an explicit query-specific target-state description, and then synthesizes the target image through token-wise mixture-of-experts adaptation, with expert routing guided by localized transition cues. Experiments show that TransPhy improves physical-rule adherence, query consistency, and unseen-rule generalization over existing visual in-context editing methods.
Chinese Translation
视觉演示为指定图像变换提供了一种自然的接口，这些变换难以用文本全面描述。然而，现有的视觉上下文学习（VICL）方法主要集中于外观层面的关系转移，对物理基础变换的支持有限，这些变换的结果依赖于材料属性、几何形状、物体交互和环境条件。给定一个源-目标示例对和一个查询图像，物理基础的VICL要求模型推断出演示的变换，将其效果适应于特定查询场景上下文，并保留与规则无关的内容。我们引入了PhysVICL-74，其中包含74条物理基础变换规则和5,240个源-目标图像对，形成近75K的训练和评估上下文。其基准划分分别评估新实例转移和未见规则的泛化。我们进一步提出了TransPhy，一个将物理基础的VICL分解为物理规则归纳和过渡对齐渲染的框架。TransPhy首先预测演示的规则和明确的查询特定目标状态描述，然后通过逐标记的专家混合适应合成目标图像，专家路由由局部过渡线索引导。实验表明，TransPhy在物理规则遵循、查询一致性和未见规则泛化方面优于现有的视觉上下文编辑方法。
cs.CV / 43 / 2608.24121
Graph-Supervised Hierarchical Clinical Alignment for Radiology Report Generation with Large Language Models
基于图的分层临床对齐用于大型语言模型的放射学报告生成
Abstract
Radiology report generation (RRG) has recently benefited from large language models, which substantially improve report fluency. However, clinically faithful generation remains challenging because current supervision is still imposed mostly at the report level. This creates a granularity mismatch: radiology reports are composed of disease-grounded findings, while existing methods are trained mainly with whole-report objectives. To address this problem, we propose Graph-Supervised Hierarchical Clinical Alignment, which reformulates image-report supervision as a hierarchical clinical alignment problem. Our method structures this alignment as a disease-conditioned process, where supervision is decomposed into two levels: Disease-Centric Alignment for fine-grained disease-specific correspondence, and Global Clinical Semantic Alignment for report-level semantic coherence. A clinical knowledge graph is used as a training-time-only structural prior that defines disease-specific supervision units and their clinical relationships, introducing no additional overhead at inference. Because standard contrastive alignment could produce false negatives when studies share overlapping pathologies, we combine instance-conditioned discriminative matching with disease-conditioned soft regularization, enabling fine-grained yet clinically consistent cross-modal representations. Experiments on MIMIC-CXR, IU-Xray, and COV-CTR show that our method consistently improves performance on both conventional and clinical metrics. Notably, our 3B model surpasses several prior systems with larger 7B/13B backbones, suggesting that improving supervision structure, rather than increasing model size, can be more effective for RRG.
Chinese Translation
放射学报告生成（RRG）最近受益于大型语言模型，这显著提高了报告的流畅性。然而，临床真实的生成仍然具有挑战性，因为目前的监督主要是在报告层面进行的。这造成了粒度不匹配：放射学报告由基于疾病的发现组成，而现有方法主要以整篇报告为目标进行训练。为了解决这个问题，我们提出了基于图的分层临床对齐，将图像-报告监督重新表述为一个分层的临床对齐问题。我们的方法将这种对齐结构化为一个基于疾病的过程，其中监督被分解为两个层次：以疾病为中心的对齐用于细粒度的疾病特定对应，以及全局临床语义对齐用于报告级别的语义一致性。临床知识图谱作为仅在训练时使用的结构先验，定义了疾病特定的监督单元及其临床关系，在推理时不会引入额外的开销。由于标准的对比对齐在研究共享重叠病理时可能产生假阴性，我们结合了基于实例的区分匹配与基于疾病的软正则化，使得跨模态表示既细粒度又临床一致。在MIMIC-CXR、IU-Xray和COV-CTR上的实验表明，我们的方法在传统和临床指标上均持续提高了性能。值得注意的是，我们的3B模型超越了几种具有更大7B/13B骨干网的先前系统，这表明改善监督结构而非单纯增加模型规模，对于放射学报告生成可能更为有效。
cs.CV / 44 / 2608.24130
Syn2RealTrack: Bridging the Gap Between Synthetic and Real-World Datasets for Online Multi-View Multi-Target Tracking
Syn2RealTrack：弥合合成数据集与真实世界数据集之间的差距，用于在线多视角多目标跟踪
Abstract
Multi-camera 3D perception systems for warehouse scenes are trained largely on synthetic data and evaluated on physically captured environments. The resulting synthetic-to-real gap, which corrupts ground-plane localization and cross-camera identity association, is usually treated as one deficiency for a single domain-adaptation module to absorb; we argue instead that it enters the pipeline at three separable points: the camera calibration, the object shape prior, and the assumption that the object census is known, each admitting a different local remedy. Our online pipeline, Syn2RealTrack, follows this decomposition: lens distortion is recovered from images alone under a calibration that provides none, detections are fused across views by a visibility-weighted part-based descriptor that abstains on occluded parts rather than guessing, person height is measured in closed form from calibration instead of copied from a synthetic prior, and a closed-world cardinality prior is paired with a causal filter that removes the phantom boxes the prior manufactures. The system therefore adapts by reallocating trust between geometry and appearance without retraining a feature extractor. On the AI City Challenge 2026 Track~1 evaluation server it reaches a 3D Higher Order Tracking Accuracy (HOTA) of 52.0118%. The code will be released at https://github.com/SKKUAutoLab/aic26_mc3dp
Chinese Translation
用于仓库场景的多摄像头三维感知系统主要在合成数据上进行训练，并在物理捕获的环境中进行评估。由此产生的合成到真实的差距会破坏地面定位和跨摄像头身份关联，通常被视为单一领域适应模块需要解决的缺陷；我们则认为这一问题在三个可分离的环节进入管道：摄像头标定、物体形状先验，以及假设物体数量是已知的，每个环节都可以采用不同的局部解决方案。我们的在线管道 Syn2RealTrack 遵循这一分解：通过一种不提供标定的校准从图像中恢复镜头畸变，使用一种基于可见性加权的部件描述符在视角间融合检测，避免对遮挡部件进行猜测，直接从标定中以封闭形式测量人的高度，而不是从合成先验中复制，并将封闭世界的基数先验与去除先验生成的虚假框的因果滤波器配对。因此，该系统通过在几何和外观之间重新分配信任来适应，而无需重新训练特征提取器。在 AI City Challenge 2026 Track~1 评估服务器上，它达到了 52.0118% 的三维高阶跟踪准确率 (HOTA)。代码将发布在 https://github.com/SKKUAutoLab/aic26_mc3dp
cs.CV / 45 / 2608.24133
PlaceSeek: Human-Centered Geospatial Retrieval of Urban Outdoor Places via Semantic Grounding and Affective Alignment
PlaceSeek：通过语义基础和情感对齐的人本城市户外场所地理空间检索
Abstract
People search for urban outdoor places not only by category or function, but also by what activities a place can support and how it is perceived. Existing geospatial retrieval remains largely POIcentric and metadata-driven, making it difficult to satisfy openended, affective, or activity-oriented needs. We present PlaceSeek, a human-centered outdoor place retrieval framework that maps natural-language queries to geolocated street-view imagery. PlaceSeek introduces an intent-aware retrieval mechanism that decomposes user queries into functional and affective sub-intents. A Semantic Grounding Module verifies whether candidate street-view results contain the physical evidence needed to support the intended activity, while an Affective Alignment Module re-ranks physically valid candidates using a LoRA-adapted vision-language model trained on human urban perception judgments. We evaluate PlaceSeek on 31,956 street-view locations in Milan across 10 naturallanguage queries annotated by five human evaluators. PlaceSeek achieves 88.0% Precision@5, a mean match score of 3.39/4.0, and 0.920 nDCG@5, outperforming CLIP, fine-tuned CLIP, SigLIP, and a VQA-based baseline. Ablation results show that physical grounding is essential for retrieval validity, while affective alignment improves ranking quality among physically valid candidates. These findings highlight that complex urban spatial queries require modeling both verifiable visual evidence and human perceptual preferences. PlaceSeek provides a potential framework for human-centered nextgeneration geospatial retrieval systems.
Chinese Translation
人们搜索城市户外场所时，不仅依赖于类别或功能，还关注一个地方能够支持的活动以及人们对其的感知。现有的地理空间检索主要以兴趣点（POI）为中心，并且依赖元数据，这使得满足开放式、情感导向或以活动为导向的需求变得困难。我们提出了PlaceSeek，一个以人为中心的户外场所检索框架，将自然语言查询映射到地理定位的街景图像。PlaceSeek引入了一种意图感知的检索机制，将用户查询分解为功能性和情感性子意图。语义基础模块验证候选街景结果是否包含支持预期活动所需的物理证据，而情感对齐模块则使用经过LoRA调整的视觉-语言模型对经过物理验证的候选进行重新排序，该模型是在对人类城市感知判断进行训练的基础上构建的。我们在米兰的31,956个街景位置上评估了PlaceSeek，涵盖了五位人类评估者注释的10个自然语言查询。PlaceSeek在Precision@5上达到了88.0%，平均匹配分数为3.39/4.0，nDCG@5为0.920，优于CLIP、微调的CLIP、SigLIP和基于VQA的基线。消融实验结果表明，物理基础对于检索的有效性至关重要，而情感对齐则提高了物理有效候选的排名质量。这些发现强调了复杂的城市空间查询需要同时建模可验证的视觉证据和人类感知偏好。PlaceSeek为以人为中心的下一代地理空间检索系统提供了潜在的框架。
cs.CV / 46 / 2608.24134
EgoErrorVQA: Assess Egocentric Comprehension Capabilities through Procedural Errors for Ego-Agentic AI
EgoErrorVQA：通过程序性错误评估自我中心理解能力的自我代理人工智能
Abstract
The majority of our everyday activities are procedural and consist of sequences of interdependent steps. However, existing benchmarks for Visual Agents and Visual Language Models (VLMs) overlook the evaluation of their procedural comprehension ability from an egocentric visual perspective, particularly for detecting procedural errors, a critical capability for everyday assistance. To bridge this gap, the EgoErrorVQA task is firstly proposed for egocentric procedural comprehension with explicit procedural errors modeling. Besides, we develop a user-friendly evaluator agent based on the Agent2Agent (A2A) protocol, enabling rigorous and standardized evaluation of visual agents through VQA-based interaction. A range of models are evaluated using both open-ended and multiple-choice questions, revealing persistent weaknesses in handling procedural errors and error types. Moreover, we introduce Ego-ADR, an Adaptive Decoupled Reasoning framework that decouples complex procedural reasoning to enhance models' understanding of procedural errors. It achieves consistent performance gains over the selected baselines and attains state-of-the-art results on several metrics under comparable settings. Code: https://github.com/z1oong/EgoErrorVQA
Chinese Translation
我们日常活动的大多数是程序性的，由相互依赖的步骤序列组成。然而，现有的视觉代理和视觉语言模型（VLMs）基准忽视了从自我中心视觉角度评估其程序理解能力，特别是在检测程序性错误方面，这对于日常辅助至关重要。为了解决这一问题，我们首次提出了EgoErrorVQA任务，旨在通过明确的程序性错误建模来评估自我中心的程序理解能力。此外，我们基于Agent2Agent（A2A）协议开发了一个用户友好的评估代理，能够通过基于视觉问答（VQA）的交互对视觉代理进行严格和标准化的评估。我们使用开放式和多项选择题评估了一系列模型，揭示了它们在处理程序性错误和错误类型方面的持续弱点。此外，我们引入了Ego-ADR，一个自适应解耦推理框架，旨在解耦复杂的程序推理，以增强模型对程序性错误的理解。它在所选基准上实现了一致的性能提升，并在可比设置下在多个指标上达到了最先进的结果。代码：https://github.com/z1oong/EgoErrorVQA
cs.CV / 47 / 2608.24138
Rubrics as Visual-Repair Context for Self-Evolving UI-to-Code Generation
作为视觉修复上下文的评分标准用于自我演化的UI到代码生成
Abstract
Large vision-language models have shown strong progress in UI-to-code generation, yet their test-time self-evolution remains unstable. We first identify a fundamental obstacle, termed visual repair coupling: a local code edit may propagate through layout, style, and component dependencies, correcting one visual mismatch while degrading regions that were previously faithful. To address this issue, we present RubSE, a Rubric-guided Self-Evolution framework that uses rubrics to represent visual feedback as a structured visual-repair context. At each refinement round, RubSE generates typed candidate rubrics, selects one prioritized repair target, and stores previously selected rubrics as history, thereby steering each revision toward a well-scoped visual repair while discouraging repeated or over-broad changes. Evaluations across six VLMs and three UI-to-code benchmarks demonstrate that RubSE substantially outperforms na\"ive self-evolution in final-round and best-round settings, achieving more stable refinement trajectories and a higher trajectory-level performance ceiling. Further analysis shows that RubSE mitigates trajectory collapse by improving recovery from severe visual regressions, and that stronger rubric generators can transfer effective visual-repair guidance to weaker code improvers.
Chinese Translation
大型视觉语言模型在UI到代码生成方面取得了显著进展，但其测试时的自我演化仍然不稳定。我们首先识别出一个根本障碍，称为视觉修复耦合：局部代码编辑可能通过布局、样式和组件依赖关系传播，修正一个视觉不匹配的同时却降低了之前忠实的区域。为了解决这个问题，我们提出了RubSE，一个基于评分标准的自我演化框架，利用评分标准将视觉反馈表示为结构化的视觉修复上下文。在每个优化轮次中，RubSE生成类型化的候选评分标准，选择一个优先的修复目标，并将之前选择的评分标准存储为历史记录，从而引导每次修订朝向明确的视觉修复，同时避免重复或过于宽泛的更改。对六个视觉语言模型和三个UI到代码基准的评估表明，RubSE在最终轮次和最佳轮次设置中显著优于简单的自我演化，达到了更稳定的优化轨迹和更高的轨迹级性能上限。进一步分析显示，RubSE通过改善从严重视觉回归的恢复来减轻轨迹崩溃，并且更强的评分标准生成器能够将有效的视觉修复指导转移给较弱的代码改进器。
cs.CV / 48 / 2608.24142
What Does Prompt Learning Change? -A Natural-Language Concept Analysis of Vision-Language Models
提示学习改变了什么？——对视觉-语言模型的自然语言概念分析
Abstract
Prompt learning adapts vision-language models such as CLIP by optimizing continuous prompt vectors, but the learned prompts are difficult to interpret in natural language. We present PromptSpLiCE, a post-hoc method that expresses each class-conditioned text embedding as a sparse combination of concepts from a fixed natural-language dictionary. Using the same dictionary before and after prompt learning allows us to compare changes in their concept profiles. We evaluate PromptSpLiCE on CoOp, a representative prompt-learning method, across 11 image-classification datasets. The concept profiles change substantially: on average, only 1.6 of the initial top-10 concepts remain in the top 10 after learning. Across datasets, profile change is positively associated with accuracy gain. We also derive a local gradient expression that provides geometric intuition for why image-aligned concept directions distinct from the current prompt can have greater loss sensitivity.
Chinese Translation
提示学习通过优化连续的提示向量来调整视觉-语言模型，如 CLIP，但所学习的提示在自然语言中难以解释。我们提出了 PromptSpLiCE，这是一种事后方法，它将每个类条件的文本嵌入表示为来自固定自然语言词典的概念的稀疏组合。在提示学习前后使用相同的词典使我们能够比较其概念轮廓的变化。我们在 CoOp（一个具有代表性的提示学习方法）上评估了 PromptSpLiCE，涵盖了 11 个图像分类数据集。概念轮廓发生了显著变化：平均而言，学习后只有 1.6 个初始前 10 个概念仍然保留在前 10 名中。在各个数据集中，轮廓变化与准确率提升呈正相关。我们还推导出一个局部梯度表达式，为为什么与当前提示不同的图像对齐概念方向可以具有更大的损失敏感性提供了几何直觉。
cs.CV / 49 / 2608.24154
Rethinking Pre-Training and Augmentation for Zero-Shot Cross-City Object Detection
重新思考零-shot跨城市物体检测的预训练和增强
Abstract
Real-world deployment of traffic surveillance systems is bottlenecked by geographic domain shift, in which models trained in one city underperform when applied to an unseen target city. Conventional domain adaptation relies on hyperparameter-sensitive architectures or direct profiling of target data. Both are fundamentally precluded in privacy-conscious ecosystems that require completely blind training and evaluation loops. In this setting, we explore the effects of pre-training and augmentation in addressing the domain shift problem. Specifically, we propose a new modular training pipeline for object detection structured around two core orthogonal pillars: (1) a multi-dataset pre-training strategy featuring a class-agnostic objectness distillation to decouple structural vehicle geometry from semantic taxonomies, and (2) a domain-resilient augmentation stream featuring a novel Grayworld transformation that forces global attention heads to strip volatile chromatic shortcuts in favor of robust shape priors. When evaluated with the real-time transformer-based detector RF-DETR, our framework bridges cross-city distribution gaps while using limited GPU memory (16GB). Our optimized variants, RF-DETR-HR and RF-DETR-Grayworld, deliver a substantial empirical gain of +24.29 over the baseline, achieving 1st place (47.53 mAP) on the AI City Challenge Track 6 leaderboard. Code and data are available at: \href{https://github.com/SKKUAutoLab/aic26_cross_city}{SKKUAutoLab/aic26\_cross\_city}.
Chinese Translation
交通监控系统在现实世界中的部署受到地理领域转移的瓶颈影响，即在一个城市训练的模型在应用于未见过的目标城市时表现不佳。传统的领域适应依赖于对超参数敏感的架构或对目标数据的直接分析，而这在需要完全盲训练和评估循环的隐私敏感生态系统中是根本不可行的。在这种情况下，我们探讨了预训练和增强在解决领域转移问题中的作用。具体而言，我们提出了一种新的模块化物体检测训练管道，围绕两个核心正交支柱构建：(1) 一种多数据集预训练策略，采用类无关的物体性蒸馏，将结构性车辆几何与语义分类法解耦；(2) 一种领域弹性增强流，采用新颖的灰世界（Grayworld）变换，迫使全局注意力头剥离不稳定的色彩捷径，以支持稳健的形状先验。当使用基于实时变换器的检测器RF-DETR进行评估时，我们的框架弥合了跨城市分布差距，同时使用有限的GPU内存（16GB）。我们的优化变体RF-DETR-HR和RF-DETR-Grayworld在基线基础上实现了+24.29的显著实证增益，在AI City Challenge Track 6排行榜上获得第一名（47.53 mAP）。代码和数据可在： exttt{https://github.com/SKKUAutoLab/aic26_cross_city}获取。
cs.CV / 50 / 2608.24169
ViSculpt: Visual-Centric Agentic Geometry Editing
ViSculpt：以视觉为中心的主动几何编辑
Abstract
3D geometry editing is a critical yet labor-intensive part of the graphics pipeline, requiring artists to translate creative intent into precise operations in complex professional software. Large language models (LLMs) have shown promise for script-based 3D creation, but script generation is less suited to perception-driven editing of arbitrary existing meshes, where execution must remain visually grounded and untouched regions should be preserved. We present a \emph{visual-centric}, training-free multi-agent system that edits existing 3D meshes directly in Blender by emulating the iterative workflow of human artists. Rather than generating scripts or regenerating geometry, our system operates through the Blender GUI: multimodal LLM agents observe the viewport, reason about the current mesh state, and execute localized edits through simulated user interactions. Experiments on a curated benchmark provide initial evidence that this agentic approach can follow natural language instructions, perform representative localized mesh edits, and preserve the overall identity of the input asset. Our results highlight a complementary regime for language-driven 3D editing: direct in-place modification of existing meshes within the native 3D editing workflow. We view this work as an exploratory step toward visual-centric agentic geometry editing in professional graphics software.
Chinese Translation
3D几何编辑是图形处理流程中一个关键但劳动密集的部分，要求艺术家将创意意图转化为在复杂专业软件中的精确操作。大型语言模型（LLMs）在基于脚本的3D创作中展现了潜力，但脚本生成不太适合对任意现有网格进行感知驱动的编辑，因为执行必须保持视觉上的一致性，并且未触及的区域应当保留。我们提出了一种 extit{以视觉为中心}、无训练的多代理系统，该系统通过模拟人类艺术家的迭代工作流程，直接在Blender中编辑现有的3D网格。我们的系统不是生成脚本或重新生成几何体，而是通过Blender GUI进行操作：多模态LLM代理观察视口，推理当前网格状态，并通过模拟用户交互执行局部编辑。在一个精心策划的基准测试中的实验提供了初步证据，表明这种主动的方法能够遵循自然语言指令，执行代表性的局部网格编辑，并保留输入资产的整体特征。我们的结果突显了语言驱动的3D编辑的互补模式：在本地3D编辑工作流程中对现有网格进行直接就地修改。我们将这项工作视为在专业图形软件中朝着以视觉为中心的主动几何编辑迈出的探索性一步。
cs.CV / 51 / 2608.24173
SandwichQuant: Which Parameters Matter Before and After Quantization?
SandwichQuant：量化前后哪些参数重要？
Abstract
Quantization correction methods usually optimize weights, quantization parameters, or reconstruction objectives, while the underlying parameter subspaces responsible for effective correction remain unclear. In this work, we study quantization correction from a parameter subspace perspective and reveal that correction capability is highly non-uniform across parameter groups. By decomposing trainable parameters into backbone weights, normalization-affine parameters, and quantization parameters, we show that the low-dimensional normalization-affine subspace provides a highly efficient correction direction under matched budgets. Based on this finding, we propose SandwichQuant, a two-stage normalization-affine correction framework that performs adaptation before and after quantization. The pre-stage improves quantization robustness, while the post-stage compensates residual errors after the quantized graph is fixed. Extensive experiments on vision models and large language models demonstrate consistent improvements under various low-bit quantization settings, validating the effectiveness of subspace-aligned correction.
Chinese Translation
量化校正方法通常优化权重、量化参数或重建目标，而有效校正所需的潜在参数子空间仍不明确。在本研究中，我们从参数子空间的角度研究量化校正，并揭示校正能力在参数组之间高度不均匀。通过将可训练参数分解为主干权重、归一化-仿射参数和量化参数，我们展示了低维归一化-仿射子空间在匹配预算下提供了高效的校正方向。基于这一发现，我们提出了SandwichQuant，一个两阶段的归一化-仿射校正框架，在量化前后进行适应。前阶段提高了量化的鲁棒性，而后阶段则在量化图固定后补偿残余误差。在视觉模型和大型语言模型上的大量实验表明，在各种低位量化设置下均实现了一致的改进，验证了子空间对齐校正的有效性。
cs.CV / 52 / 2608.24175
Amortized Set Prediction for Inverse IFS Reconstruction from Density Maps
基于密度图的逆迭代函数系统重建的摊销集合预测
Abstract
Iterated Function Systems (IFS) generate self-similar fractals from a few contractive affine maps. The forward map from parameters to images is computationally inexpensive and well understood, whereas the inverse problem of estimating maps from an image is difficult and is typically handled by per-image optimization. We replace this loop with a single forward pass of a learned estimator that predicts the affine-map set directly from a visit-frequency density map, thereby amortizing the inverse problem. The design follows two constraints. First, density maps do not uniquely identify IFS parameters, so evaluation is based on reconstruction rather than parameter recovery; unordered map sets are handled by Hungarian matching, and ground-truth parameters provide a stable training surrogate. Second, the fully known forward model lets us generate exact synthetic training pairs and also supports image-only test-time refinement. On in-distribution tests, amortized initialization plus a few refinement steps lies on a better quality--speed frontier than equal-budget random-initialized per-image optimization, and a 30-step refinement (about $0.56$ s per sample) remains better than a doubled-budget baseline. Extending optimization to 1000 steps shows that the benefit is not only speed: amortized initialization reaches high-quality reconstructions more frequently than random starts. On real images (MNIST and Fashion-MNIST), it improves density metrics on average over a published per-image optimizer while being roughly 12 to 2600 times faster.
Chinese Translation
迭代函数系统（IFS）通过少量收缩仿射映射生成自相似分形。参数到图像的正向映射计算开销小且易于理解，而从图像估计映射的逆问题则较为复杂，通常通过逐图优化来处理。我们用一个学习的估计器的单次正向传递替代了这一循环，该估计器直接从访问频率密度图预测仿射映射集合，从而摊销了逆问题。设计遵循两个约束。首先，密度图无法唯一识别IFS参数，因此评估基于重建而非参数恢复；无序映射集合通过匈牙利匹配处理，真实参数提供了稳定的训练替代。其次，完全已知的正向模型使我们能够生成精确的合成训练对，并支持仅基于图像的测试时细化。在分布内测试中，摊销初始化加上少量细化步骤在质量与速度的权衡上优于相同预算的随机初始化逐图优化，而30步细化（每个样本约$0.56$秒）仍然优于双倍预算的基线。将优化扩展到1000步显示出其好处不仅在于速度：摊销初始化比随机起始更频繁地达到高质量重建。在真实图像（MNIST和Fashion-MNIST）上，它在平均上改善了已发布的逐图优化器的密度指标，同时速度约快12到2600倍。
cs.CV / 53 / 2608.24212
NeoWorld-Pro: Programming Interactive Scenes from Monocular Images for Embodied Simulation
NeoWorld-Pro：从单目图像编程交互场景以实现具身仿真
Abstract
The advancement of Embodied AI necessitates high-quality simulation assets that faithfully mirror the real world. However, transforming raw visual observations into simulation-ready scenes remains challenging due to the lack of physical grounding and scene-level interactivity in current image-to-URDF methods. We propose NeoWorld-Pro, a framework that reformulates monocular scene reconstruction as procedural programming for interactive 3D environments. Leveraging the zero-shot reasoning and code synthesis capabilities of MLLMs, NeoWorld-Pro converts a single RGB image into executable programs specifying object geometry, articulation, and physical properties. A physics-in-the-loop mechanism then iteratively refines the generated programs by validating their execution in a physics engine, enforcing physically plausible articulations, valid object compositions and interactions, and accurate spatial relationships. Experiments show that NeoWorld-Pro outperforms open-loop and prior monocular reconstruction methods, while enabling complex downstream tasks such as stable stacking and fine-grained manipulation.
Chinese Translation
具身人工智能的发展需要高质量的仿真资产，以真实地反映现实世界。然而，由于当前图像到URDF方法缺乏物理基础和场景级交互性，将原始视觉观察转化为适合仿真的场景仍然具有挑战性。我们提出了NeoWorld-Pro，一个将单目场景重建重新构造为交互式3D环境的程序化编程框架。利用多模态大语言模型（MLLMs）的零样本推理和代码合成功能，NeoWorld-Pro将单个RGB图像转换为可执行程序，指定对象几何形状、关节和物理属性。然后，通过在物理引擎中验证程序的执行，物理循环机制迭代地优化生成的程序，强制执行物理上合理的关节、有效的对象组合和交互，以及准确的空间关系。实验表明，NeoWorld-Pro在性能上优于开放循环和先前的单目重建方法，同时支持稳定堆叠和精细操作等复杂下游任务。
cs.CV / 54 / 2608.24219
Beauty is in the ELBO of the Beholder: A Variational Account of Processing Fluency in Face Perception
美在观察者的ELBO中：面孔知觉中加工流畅性的变分解释
Abstract
Facial attractiveness has been linked to statistical regularities such as symmetry and averageness, suggesting that beauty may depend on the ease with which a face is perceived. We empirically test this hypothesis by training variational autoencoders on four face datasets without attractiveness supervision and evaluating their representations on the 597 faces from the Chicago Face Database. Across models, human attractiveness ratings closely aligns with the direction defined by the VAE evidence lower bound (ELBO) in rate-distortion space. Independently learned latent spaces contain an attractiveness direction that transfers strongly across random initializations and training data. We also find that attractive faces are more prototypical in both shape and latent space. Our results connect classic accounts of aesthetics with learned generative models and provide empirical support for a variational interpretation of the processing fluency theory of aesthetic pleasure.
Chinese Translation
面孔的吸引力与对称性和平均性等统计规律相关，这表明美可能依赖于面孔被感知的难易程度。我们通过在四个面孔数据集上训练变分自编码器（Variational Autoencoders, VAE），并在芝加哥面孔数据库的597个面孔上评估其表示，实证检验了这一假设。在不同模型中，人类的吸引力评分与在率失真空间中由VAE证据下界（Evidence Lower Bound, ELBO）定义的方向紧密对齐。独立学习的潜在空间包含一个吸引力方向，该方向在随机初始化和训练数据之间强烈转移。我们还发现，吸引力面孔在形状和潜在空间中更具原型特征。我们的结果将经典的美学理论与学习生成模型联系起来，并为加工流畅性理论的变分解释提供了实证支持。
cs.CV / 55 / 2608.24223
Event-Based Motion Estimation via Oriented Distance Fields
基于事件的运动估计通过定向距离场
Abstract
Event-based motion estimation is central to tasks that demand high temporal resolution and robustness to fast motion. Existing methods typically rely on iterative optimization or repeated hypothesis comparison, offsetting the sensor's low-latency advantage. We propose Oriented Distance Field Motion Estimation (ODF Motion Estimation), which replaces this optimization with a single averaging step over a precomputed field of event distance vectors, combined with an adaptive event-count selection strategy and a parameter-free trail filter. On public and self-collected datasets, ODF motion estimation reaches sub-pixel accuracy at the lowest latency among compared methods. We validate its generality on two downstream applications rather than treating them as separate contributions. First, the estimated trajectory is converted into a blur kernel and paired with a compact iterative-unfolding network, trained on simulated motion-estimation noise, for real-time non-blind image deblurring, attaining competitive or superior PSNR/SSIM with under 1M parameters. Second, the same precomputed field is repurposed for directional event filtering in a low-power asynchronous pupil and glint tracker, sustaining stable tracking for tens of seconds while lowering a near-eye module's power draw.
Chinese Translation
基于事件的运动估计在需要高时间分辨率和对快速运动具有鲁棒性的任务中至关重要。现有的方法通常依赖于迭代优化或重复假设比较，从而抵消了传感器低延迟的优势。我们提出了定向距离场运动估计（Oriented Distance Field Motion Estimation，ODF运动估计），它用一个预计算的事件距离向量场的单次平均步骤替代了这种优化，结合自适应事件计数选择策略和无参数轨迹滤波器。在公共和自收集的数据集上，ODF运动估计在比较方法中以最低延迟达到了亚像素精度。我们在两个下游应用上验证了其通用性，而不是将它们视为独立的贡献。首先，估计的轨迹被转换为模糊核，并与一个紧凑的迭代展开网络配对，该网络在模拟运动估计噪声上进行训练，实现实时非盲图像去模糊，达到竞争性或优越的PSNR/SSIM，参数少于100万。其次，同一预计算场被重新用于低功耗异步瞳孔和反光点跟踪器中的方向性事件过滤，持续稳定跟踪数十秒，同时降低近眼模块的功耗。
cs.CV / 56 / 2608.24281
Example-based Robust Abnormality Detection with Minimal Annotations using Exemplar Med-DETR
基于示例的稳健异常检测：使用示例 Med-DETR 的最小注释
Abstract
Reducing annotation requirements remains a key challenge in developing robust medical object detectors. To address this, Vision-Language (VL) object detection methods leverage grounding text information to enable powerful zero-shot and few-shot object detectors in the natural image domain [1, 2, 3, 4]. However, transferring these methods to the medical domain is challenging due to the absence of comparable quality and quantity of the grounding data. Regardless, significant contextual and non-imaging information exists in medical images that remains underutilized. Few-shot learning (FSL) techniques partially address this limitation but struggle to general ize to unseen medical findings and require extensive retraining when new findings are introduced [5, 6]. To overcome these challenges, we extend our prior EM-DETR framework [7] and introduce a scalable FS detection approach designed for efficient abnormality detection in Chest X-Ray (CXR) images under minimal supervision. The proposed architecture incorporates exemplar-based feature generation and domain-aware contrastive optimization, enabling effective adaptation to novel disease findings without exhaustive retraining. Our method achieves near state-of-the-art (SOTA) detection performance using less than 10% of the annotated data, demonstrating its potential for practical, annotation-efficient clinical deployment across both proprietary and public CXR datasets.
Chinese Translation
减少注释需求仍然是开发稳健医学物体检测器的关键挑战。为了解决这一问题，视觉-语言（VL）物体检测方法利用文本信息的基础来实现强大的零样本和少样本物体检测器，适用于自然图像领域。然而，由于缺乏可比质量和数量的基础数据，将这些方法转移到医学领域面临挑战。尽管如此，医学图像中存在大量的上下文和非成像信息尚未得到充分利用。少样本学习（FSL）技术在一定程度上解决了这一限制，但在面对未见过的医学发现时难以泛化，并且在引入新发现时需要广泛的再训练。为克服这些挑战，我们扩展了之前的 EM-DETR 框架，并引入了一种可扩展的少样本检测方法，旨在在最小监督下高效检测胸部 X 光（CXR）图像中的异常。所提出的架构结合了基于示例的特征生成和领域感知对比优化，使得在不进行全面再训练的情况下有效适应新疾病发现。我们的方法在使用不到 10% 注释数据的情况下，达到了接近最先进（SOTA）的检测性能，展示了其在专有和公共 CXR 数据集中的实际应用潜力，具有高效的注释能力。
cs.CV / 57 / 2608.24282
CARE: Camera-Residual Reserves for First Sightings in Adaptive LiDAR Sensing
CARE：用于自适应激光雷达感知中首次发现的相机残差保留策略
Abstract
Adaptive LiDAR scanning concentrates a limited sensing budget on regions of interest predicted from past object tracks, lowering data volume in autonomous driving while maintaining detection accuracy. However, existing scanning policies face three challenges. First, history-driven approaches depend on past tracks, so unseen objects are detected late or missed. Second, random or uniform sampling outside the predicted regions has no awareness of where new objects appear. Third, camera-guided alternatives spend budget on all camera detections, resampling objects already covered, costing recall in crowded scenes and range when budgets are scarce. This paper introduces the CAmera-REsidual reserve (CARE), a training-free allocation rule that reserves part of a fixed ray budget for the directions of current camera detections that the track forecasts cannot explain; the rest follows the base history policy, and unused reserve returns to a random floor. The paper makes three contributions. First, a leakage-free ray-budget evaluation on nuScenes (150 scenes, 4,148 events) measuring the first-sighting loss of history-driven scanning, with a strict-causal variant using the preceding keyframe. Second, CARE raises first-sighting recall by 5.2, 5.2, and 4.3 points at 10%, 20%, and 35% budgets over the history policy, with paired intervals excluding zero; the camera cue drives this gain, and the first-sighting versus overall trade-off is a budget-dependent Pareto choice. Third, a safety-bounded forgetting module that releases budget from receding or static tracks beyond a speed-dependent guard distance; at tight budgets, forgetting without the guard significantly harms near-field recall, so the guard is what keeps it safe. The pipeline runs end to end on a real vehicle and, in closed-loop simulation, detects an occluded pedestrian earlier and brakes more reliably than history-driven scanning.
Chinese Translation
自适应激光雷达扫描将有限的感知预算集中于基于过去目标轨迹预测的兴趣区域，从而在自动驾驶中降低数据量的同时保持检测精度。然而，现有的扫描策略面临三大挑战。首先，基于历史的方法依赖于过去的轨迹，因此对未见过的目标检测较晚或遗漏。其次，在预测区域之外的随机或均匀采样无法感知新目标出现的位置。第三，基于相机引导的替代方案会将预算分配给所有相机检测目标，导致对已覆盖目标的重复采样，在拥挤场景中降低召回率，在预算紧张时降低检测范围。本文提出了CAmera-REsidual reserve（CARE），一种无需训练的分配规则，将固定射线预算的一部分保留给当前相机检测方向中轨迹预测无法解释的区域；其余部分遵循基础历史策略，未使用的保留预算则回归随机采样。本文贡献包括：第一，在nuScenes数据集（150个场景，4148个事件）上进行无泄漏射线预算评估，测量基于历史的扫描首次发现损失，并采用严格因果的前一关键帧变体。第二，CARE在10%、20%和35%预算下，较历史策略分别提升首次发现召回率5.2、5.2和4.3个百分点，配对区间排除零值；该提升由相机线索驱动，首次发现与整体性能的权衡呈预算依赖的帕累托选择。第三，设计了安全边界的遗忘模块，从速度相关的保护距离之外的后退或静止轨迹中释放预算；在紧张预算下，无保护的遗忘显著损害近场召回，保护机制确保安全性。该流程已在真实车辆上端到端运行，并在闭环仿真中实现对遮挡行人的更早检测和更可靠的制动，优于基于历史的扫描策略。
cs.CV / 58 / 2608.24293
Keep-or-Drop? Adaptive Tokenizer for Compact Video Representation
保留还是丢弃？用于紧凑视频表示的自适应分词器
Abstract
Latent diffusion models have emerged as a dominant framework for high-fidelity image and video synthesis, operating in compact latent spaces with variational autoencoders (VAEs) to enhance computational efficiency without compromising visual quality. However, conventional VAEs are suboptimal for video data as they employ fixed compression ratios that cannot adapt to the varying complexity of spatio-temporal content. We present KATok (Keep-or-Drop? Adaptive Tokenizer for Compact Video Representation), a transformer-based VAE that incorporates an adaptive token selector which is jointly learned with latent tokens. By evaluating each token's content-richness as keep-or-drop probability, the token selector effectively discards uninformative tokens, naturally allowing data-dependent compression. Applying adaptive tokenization to diffusion models may cause spatial misalignment, as token dropping can disturb the original spatio-temporal structure. To alleviate this issue, we propose two position-prediction strategies: cascaded and joint generation, to ensure spatial consistency. We empirically show that our model achieves strong reconstruction and generation quality at a state-of-the-art compression ratio. Further analysis on video data reveals that this improvement is primarily achieved by reducing spatio-temporal redundancy and removing uninformative tokens, as supported by both quantitative and qualitative results.
Chinese Translation
潜在扩散模型已成为高保真图像和视频合成的主流框架，利用变分自编码器（VAEs）在紧凑的潜在空间中运行，以提高计算效率而不影响视觉质量。然而，传统的VAEs对于视频数据并不理想，因为它们采用固定的压缩比，无法适应时空内容的变化复杂性。我们提出了KATok（Keep-or-Drop? Adaptive Tokenizer for Compact Video Representation），一种基于变换器的VAE，结合了一个与潜在标记共同学习的自适应标记选择器。通过评估每个标记的内容丰富度作为保留或丢弃的概率，标记选择器有效地丢弃无信息的标记，自然实现数据依赖的压缩。将自适应分词应用于扩散模型可能导致空间错位，因为丢弃标记可能扰乱原始的时空结构。为了解决这个问题，我们提出了两种位置预测策略：级联生成和联合生成，以确保空间一致性。我们实证表明，我们的模型在达到最先进的压缩比的同时，能够实现强大的重建和生成质量。对视频数据的进一步分析表明，这一改进主要是通过减少时空冗余和去除无信息标记来实现的，这得到了定量和定性结果的支持。
cs.CV / 59 / 2608.24334
SeMoCo: A Semantic-First Motion Codec for Motion Language Modeling
SeMoCo：一种以语义为先的运动编码器用于运动语言建模
Abstract
Discrete motion representations have substantially advanced autoregressive text-to-motion generation. However, most motion tokenizers are optimized for reconstruction and do not explicitly allocate capacity according to semantic role. Action-level meaning and fine-grained kinematic detail must therefore be encoded through the same reconstruction-driven hierarchy. We introduce SeMoCo, a semantic-first motion codec, together with a dual-axis motion generator for language-conditioned motion generation. Each motion token contains one semantic token and a residual sequence of kinematic tokens. The generator models semantic progression across time and autoregressively refines the residual entries. We also construct $\Omega$-MotionVerse, a large-scale, multi-source human-motion dataset unified under the SOMA representation. Across the reported comparisons, SeMoCo achieves the best reconstruction accuracy among the compared codecs, while strong text-to-motion results demonstrate the effectiveness of its motion tokens for downstream generation.
Chinese Translation
离散运动表示在自回归文本到运动生成方面取得了显著进展。然而，大多数运动标记器是针对重建进行优化的，并未明确根据语义角色分配容量。因此，动作级别的意义和细粒度的运动学细节必须通过相同的重建驱动层次进行编码。我们提出了SeMoCo，一种以语义为先的运动编码器，以及一个用于语言条件下运动生成的双轴运动生成器。每个运动标记包含一个语义标记和一系列残差运动学标记。生成器建模语义随时间的进展，并自回归地细化残差条目。我们还构建了$ ext{Ω}$-MotionVerse，一个统一在SOMA表示下的大规模多源人类运动数据集。在报告的比较中，SeMoCo在所比较的编码器中实现了最佳重建精度，而强大的文本到运动结果则展示了其运动标记在下游生成中的有效性。
cs.CV / 60 / 2608.24340
Mind the Student: Behavioral and Contextual Cues for Automated Engagement Prediction in Online Learning
关注学生：在线学习中自动化参与预测的行为和情境线索
Abstract
The prediction of student engagement from the online tutoring videos is difficult because engagement is a multidimensional construct comprising distinct behavioral, emotional, and cognitive states. A reliable prediction requires bringing together different types of behavioral signals as well as expressive cues. Through our analysis of the CASED dataset, it is clear that engagement prediction gets even harder due to the high inter-person variability as well as the subjectivity of the engagement annotation. To tackle these challenges, we develop a multimodal framework that integrates the implicit spatiotemporal features extracted from pretrained video, audio, and image encoders along with structured behavioral modalities like head pose, gaze, facial action units, emotion, and wavelet-based audio features. We integrate these modalities via a Perceiver IO latent bottleneck. Moreover, student and instructor personalities are modeled as variational posteriors over learnable embeddings to enable partial pooling across participants. We employ evidential regression and spectral-normalized Gaussian process classification heads for uncertainty-aware prediction to further improve robustness and calibration. Benchmark on the CASED challenge test set shows that all participating methods converge near random-chance performance, revealing the difficulty of the dataset. In this highly ambiguous regime, our framework achieves competitive performance while uniquely offering well-calibrated uncertainty metrics, demonstrating that reliable risk-quantification is an essential prerequisite for deploying engagement models in real-world educational tools.
Chinese Translation
从在线辅导视频中预测学生参与度是困难的，因为参与度是一个多维构念，包含不同的行为、情感和认知状态。可靠的预测需要整合不同类型的行为信号以及表现性线索。通过对CASED数据集的分析，可以明显看出，由于个体间的高度变异性以及参与度标注的主观性，参与度预测变得更加困难。为了解决这些挑战，我们开发了一个多模态框架，整合了从预训练的视频、音频和图像编码器中提取的隐含时空特征，以及诸如头部姿态、注视、面部动作单元、情感和基于小波的音频特征等结构化行为模态。我们通过Perceiver IO潜在瓶颈整合这些模态。此外，学生和教师的个性被建模为可学习嵌入上的变分后验，以便在参与者之间进行部分汇聚。我们采用证据回归和谱归一化高斯过程分类头进行不确定性感知的预测，以进一步提高鲁棒性和校准性。在CASED挑战测试集上的基准测试显示，所有参与方法的表现接近随机猜测，揭示了该数据集的困难性。在这个高度模糊的环境中，我们的框架实现了竞争性的性能，同时独特地提供了良好校准的不确定性指标，证明可靠的风险量化是将参与模型部署到现实教育工具中的重要前提。
cs.CV / 61 / 2608.24342
Metadata-Aware Adaptation of a Generative Foundation Model for Conditional CMR Synthesis
面向元数据的生成基础模型适应性用于条件心脏磁共振（CMR）合成
Abstract
Synthetic image generation is a promising strategy to address data scarcity and the underrepresentation of clinically important phenotypes in medical imaging, yet generating images that faithfully reflect meaningful patient characteristics remains challenging. In this work, we investigate metadata-conditioned cardiac magnetic resonance (CMR) synthesis using a pretrained latent diffusion model, encoding structured clinical metadata and slice position as textual prompts to guide CMR generation. To improve metadata adherence and address the imbalance of clinical attributes, we integrate three strategies: Metadata-Free Classifier-Free Guidance (CFG), Contrastive Batching, and Inverse-Frequency Sampling. The framework was fine-tuned and evaluated on 59,058 short-axis CMR from the UK Biobank using paired image similarity, distributional fidelity, and subgroup-level analyses. The combined approach achieved a Fr\'echet Inception Distance (FID) of 37.47, improving by 57.04\% over the same model fine-tuned without these strategies and by 28.68\% over a previous text-conditioned CMR diffusion baseline requiring cardiac geometry as additional input, while relying solely on patient metadata. This distributional gain, driven mainly by Metadata-Free CFG, came with a modest reduction in paired similarity, suggesting that the model prioritizes population-level realism over exact image reproduction. Subgroup analyses demonstrated improved alignment across demographic and acquisition-related metadata, with disease-specific conditioning being the most challenging task. These findings demonstrate the potential of generative foundation models for clinically meaningful CMR synthesis while highlighting the need for more effective metadata-aware conditioning strategies. Our code is available at https://github.com/rodriguezmarc/conditional-cmr.
Chinese Translation
合成图像生成是一种有前景的策略，用于解决医学影像中数据稀缺和临床重要表型不足的问题，但生成忠实反映有意义的患者特征的图像仍然具有挑战性。在本研究中，我们探讨了使用预训练的潜在扩散模型进行元数据条件的心脏磁共振（CMR）合成，将结构化临床元数据和切片位置编码为文本提示，以指导CMR生成。为了提高元数据的遵循性并解决临床属性的不平衡，我们整合了三种策略：无元数据的无分类器引导（Metadata-Free Classifier-Free Guidance, CFG）、对比批处理（Contrastive Batching）和逆频率采样（Inverse-Frequency Sampling）。该框架在来自英国生物银行的59,058个短轴CMR上进行了微调和评估，使用配对图像相似性、分布忠实度和亚组水平分析。综合方法达到了37.47的Fréchet Inception Distance (FID)，比未使用这些策略微调的相同模型提高了57.04%，比之前需要心脏几何作为额外输入的文本条件CMR扩散基线提高了28.68%，而仅依赖于患者元数据。这种分布增益主要由无元数据的CFG驱动，伴随着配对相似性的适度降低，表明模型优先考虑群体水平的真实感而非精确的图像重现。亚组分析显示，在人口统计和采集相关元数据方面的对齐有所改善，特定疾病的条件化是最具挑战性的任务。这些发现展示了生成基础模型在临床有意义的CMR合成中的潜力，同时强调了更有效的元数据感知条件化策略的必要性。我们的代码可在https://github.com/rodriguezmarc/conditional-cmr获取。
cs.CV / 62 / 2608.24364
B-MIM: Biased Masked Image Modeling for Generalizable Segmentation of Fine-Grained Anatomical Structures
B-MIM：用于细粒度解剖结构可泛化分割的偏置掩蔽图像建模
Abstract
Self-supervised pretraining enables transferable representations for medical imaging, yet most CT encoders remain biased toward coarse semantic understanding, limiting their sensitivity to fine-grained anatomical structures such as vessels or small tumors. In this paper, we introduce Biased Masked Image Modeling (B-MIM), a modification of the iBOT objective that stochastically reduces global semantic alignment to prioritize local patch reconstruction. This bias encourages the encoder to capture high-frequency morphological details and structural continuity. We curate a multi-institutional CT abdominal dataset of 9,955 filtered studies from 17 public sources and pretrain a 3D Swin Transformer backbone using B-MIM. Across inter-dataset experiments on liver vessel segmentation, the proposed encoder improves topological fidelity (clDice) and achieves competitive Dice scores in tumor segmentation, compared to fully fine-tuned baselines, despite updating only a fraction of the parameters. Our results suggest that reducing global semantic pressure during pretraining enhances generalization to intricate anatomical structures.
Chinese Translation
自监督预训练使医学影像的可转移表示成为可能，但大多数CT编码器仍然偏向于粗略的语义理解，限制了它们对细粒度解剖结构（如血管或小肿瘤）的敏感性。本文提出了偏置掩蔽图像建模（B-MIM），这是对iBOT目标的修改，通过随机减少全局语义对齐来优先考虑局部补丁重建。这种偏置促使编码器捕捉高频形态细节和结构连续性。我们从17个公共来源中整理了一个包含9,955个经过过滤的研究的多机构CT腹部数据集，并使用B-MIM对3D Swin Transformer主干进行预训练。在肝脏血管分割的跨数据集实验中，所提出的编码器在拓扑保真度（clDice）方面有所提升，并在肿瘤分割中取得了与完全微调基线相当的Dice分数，尽管仅更新了部分参数。我们的结果表明，在预训练过程中减少全局语义压力可以增强对复杂解剖结构的泛化能力。
cs.CV / 63 / 2608.24365
MaST: Motion-aware Sparse Pipeline for Lightweight Object Tracking
MaST：基于运动感知的轻量级目标跟踪稀疏管道
Abstract
Transformer-based object trackers are renowned for their strong performance, yet dense token processing often leads to prohibitive computational cost, limiting real-time deployment on edge devices. While recent works explore token pruning to reduce computation, they often stop short of an end-to-end sparse pipeline, as early-layer token scores can be noisy without a motion prior, and many trackers ultimately fall back to dense reshaping to feed the dense prediction head that partially negates the savings. We introduce Motion-aware Sparse Tracker (MaST), a sparse tracking framework that makes sparsity effective from tokens to boxes. First, MaST injects a lightweight motion prior to refine cross-attention-based importance scores, enabling earlier and more stable token reduction in the search region. Second, we introduce a natively sparse prediction head that operates directly on the retained unstructured tokens with a score-first, regress-once design, eliminating dense padding/reshaping and reducing redundant computation. Extensive experiments on multiple benchmarks demonstrate that MaST establishes new state of the art among lightweight trackers, where MaST-tiny attains 63.8 AUC on LaSOT and 80.1 SUC on TrackingNet, surpassing the prior best AsymTrack-S by +1.0 AUC and +2.2 SUC while running at 152 FPS on Jetson Nano, nearly twice as fast as AsymTrack-S at 88 FPS. Code is available at https://github.com/TsingWei/MaST.
Chinese Translation
基于Transformer的目标跟踪器因其强大的性能而闻名，然而密集的token处理往往导致高昂的计算成本，限制了其在边缘设备上的实时部署。尽管近期的研究探索了token剪枝以减少计算，但它们通常未能实现端到端的稀疏管道，因为早期层的token评分在没有运动先验的情况下可能会出现噪声，许多跟踪器最终回退到密集重塑，以便为密集预测头提供输入，这在一定程度上抵消了节省的效果。我们提出了运动感知稀疏跟踪器（MaST），这是一个从token到框的稀疏跟踪框架，使稀疏性得以有效利用。首先，MaST注入了一个轻量级的运动先验，以细化基于交叉注意力的重要性评分，从而在搜索区域内实现更早且更稳定的token减少。其次，我们引入了一个原生稀疏预测头，直接在保留的非结构化token上操作，采用评分优先、一次回归的设计，消除了密集填充/重塑，减少了冗余计算。在多个基准上的广泛实验表明，MaST在轻量级跟踪器中建立了新的最先进水平，其中MaST-tiny在LaSOT上达到了63.8的AUC，在TrackingNet上达到了80.1的SUC，分别超过了之前最佳的AsymTrack-S +1.0的AUC和+2.2的SUC，同时在Jetson Nano上以152 FPS的速度运行，几乎是AsymTrack-S在88 FPS下的两倍。代码可在https://github.com/TsingWei/MaST获取。
cs.CV / 64 / 2608.24366
Variance-Guided Spatial Attention Fusion for Robust End-to-End Driving under Asymmetric Sensor Degradation
基于方差引导的空间注意力融合用于不对称传感器退化下的鲁棒端到端驾驶
Abstract
End-to-end multimodal driving has progressed rapidly by fusing camera and LiDAR streams. Existing pipelines remain fragile under asymmetric sensor degradation, where either an entire modality or only a localized region is corrupted while other regions remain useful. The key difficulty is not simply to add an uncertainty head, but to obtain dense reliability supervision, calibrate this reliability against physical fault severity, and use it before unreliable features bias the planner. We propose Variance-Guided Spatial Attention Fusion (VG-SAF), in which dense heteroscedastic reliability estimates act as interpretable spatial gates. The framework couples three components. First, a physically grounded augmentor simulates representative camera and LiDAR failures and emits a continuous spatial mask, providing dense supervision without additional annotation. Second, modality-specific experts predict per-pixel reliability scales through cross-branch dense distillation in log space, enforcing a monotone severity-to-scale response. Third, calibrated reliability maps drive a hybrid attention mechanism that suppresses unreliable cells with a local spatial gate and arbitrates between modalities through a cross-modal trust softmax. A Laplace uncertainty head emits a systemic waypoint uncertainty scale that signals severe or combined sensor degradation, including severities outside the training ranges. On the CARLA Longest6 benchmark, VG-SAF consistently improves closed-loop robustness over the baselines across camera-only, LiDAR-only, and joint degradation regimes, as measured by driving score, route completion, and infraction score.
Chinese Translation
端到端的多模态驾驶通过融合摄像头和激光雷达（LiDAR）数据取得了快速进展。然而，现有的处理流程在不对称传感器退化下仍然脆弱，此时可能整个模态或仅局部区域受到损坏，而其他区域仍然有效。关键的困难不仅在于增加一个不确定性头，而在于获取密集的可靠性监督，将这种可靠性与物理故障严重性进行校准，并在不可靠特征影响规划器之前使用它。我们提出了方差引导的空间注意力融合（Variance-Guided Spatial Attention Fusion, VG-SAF），其中密集的异方差可靠性估计作为可解释的空间门。该框架结合了三个组件。首先，一个基于物理的增强器模拟代表性的摄像头和激光雷达故障，并发出连续的空间掩码，提供密集的监督而无需额外标注。其次，模态特定的专家通过对数空间中的跨分支密集蒸馏预测每个像素的可靠性尺度，强制实现单调的严重性与尺度响应。第三，经过校准的可靠性图驱动混合注意力机制，通过局部空间门抑制不可靠单元，并通过跨模态信任softmax进行模态间的仲裁。一个拉普拉斯不确定性头发出系统的航点不确定性尺度，指示严重或组合的传感器退化，包括超出训练范围的严重性。在CARLA Longest6基准测试中，VG-SAF在摄像头单独、激光雷达单独和联合退化模式下，均持续提高了闭环鲁棒性，具体通过驾驶得分、路线完成率和违规得分进行测量。
cs.CV / 65 / 2608.24372
Bridging Adversarial and Collaborative Learning for AI-Generated Image Quality Assessment
桥接对抗学习与协作学习以评估人工智能生成图像质量
Abstract
AI-generated image quality assessment (AIGIQA) requires jointly reasoning about perceptual fidelity and prompt alignment, two quality dimensions that are often treated as independent in existing AIGIQA models. However, by re-examining human ratings, we uncover a previously overlooked phenomenon: the two dimensions are interdependent and exhibit both competitive and cooperative interactions during human rating. This observation suggests that a unified model should neither collapse the two dimensions nor rigidly separate them, but rather adaptively negotiate their interplay. Motivated by this insight, we introduce an interaction-aware learning framework that models perception-alignment relations through adversarial and collaborative inference pathways. Instead of designing a rigid dual-branch architecture, our method employs a gated interaction module that dynamically routes features according to the inferred relationship between the two dimensions. Task-aware prompts further modulate the gating behaviour, enabling the model to switch between competition and cooperation when necessary. Experiments across multiple AIGIQA benchmarks demonstrate that our approach not only achieves state-of-the-art accuracy but also yields interpretable interaction patterns, offering a more faithful approximation of human judgment. The codes are available at https://github.com/LQAMEI/ACL-IQA.
Chinese Translation
人工智能生成图像质量评估（AIGIQA）需要同时考虑感知保真度和提示对齐这两个质量维度，而这两个维度在现有的 AIGIQA 模型中通常被视为独立的。然而，通过重新审视人类评分，我们发现了一个之前被忽视的现象：这两个维度是相互依存的，并且在人工评分过程中表现出竞争与合作的互动。这一观察表明，统一模型不应将这两个维度合并为一体，也不应僵硬地将其分开，而应灵活地协商它们之间的相互作用。基于这一洞察，我们提出了一种交互感知学习框架，通过对抗和协作推理路径来建模感知-对齐关系。我们的方案并未设计一个僵化的双分支架构，而是采用了一个动态路由特征的门控交互模块，根据两个维度之间推断出的关系来调整特征流动。任务感知提示进一步调节门控行为，使模型在必要时能够在竞争与合作之间切换。在多个 AIGIQA 基准测试中的实验表明，我们的方法不仅实现了最先进的准确性，还产生了可解释的互动模式，更真实地近似人类判断。代码可在 https://github.com/LQAMEI/ACL-IQA 获取。
cs.CV / 66 / 2608.24384
Markerless Pose Estimation for Resistance Training Technique Assessment
无标记姿态估计在抗阻训练技术评估中的应用
Abstract
Resistance training can be a high risk activity, and safe form is essential to avoiding injury. Laboratory-based movement analysis provides quantitive technique assessment, yet is not easily accessible. Markerless pose estimation infers body landmarks from images or video without physical markers and could offer a feasible alternative for technique assessment. We present a pose estimation framework to evaluate resistance-training technique from ordinary video footage. Using BlazePose, anatomical landmarks were extracted from squat, bench press, and deadlift videos and converted into joint-angle trajectories, with the squat serving as the primary case study. Trajectories were assessed against a defined reference repetition using root mean square error (RMSE). Results show that the framework recovers meaningful kinematic patterns for the squat and deadlift, enabling quantitative comparison between repetitions and identification of technique variability within a set. Performance depended strongly on camera orientation and visual occlusion, with non-sagittal views distorting 2D joint-angle estimates. The findings demonstrate that markerless pose estimation can support accessible biomechanical assessment outside laboratory environments.
Chinese Translation
抗阻训练可能是一项高风险活动，安全的训练形式对于避免受伤至关重要。基于实验室的运动分析提供了定量的技术评估，但并不易于获取。无标记姿态估计可以从图像或视频中推断身体标志，而无需物理标记，这可能为技术评估提供一种可行的替代方案。我们提出了一种姿态估计框架，用于从普通视频素材中评估抗阻训练技术。使用 BlazePose 从深蹲、卧推和硬拉视频中提取解剖标志，并将其转换为关节角度轨迹，以深蹲作为主要案例研究。通过均方根误差 (RMSE) 对轨迹与定义的参考重复进行评估。结果表明，该框架能够恢复深蹲和硬拉的有意义运动模式，实现重复之间的定量比较，并识别一组内的技术变异性。性能在很大程度上依赖于摄像机的方向和视觉遮挡，非矢状面视角会扭曲二维关节角度估计。研究结果表明，无标记姿态估计可以支持在实验室环境之外的可及生物力学评估。
cs.CV / 67 / 2608.24415
MRI-based Deep Radiomic Phenotyping of Neuromuscular Disorders: A Topology-driven Characterization
基于MRI的神经肌肉疾病深度放射组学表型分析：一种基于拓扑的特征描述
Abstract
Quantitative assessment of muscle MRI is crucial for monitoring neuromuscular disorders (NMD). This study introduces an automated radiomic phenotyping framework based on original features engineered across five main architectural domains: quantitative morphometry, spatial distribution, geometric shape, interactions between progressive fat replacement stages, and graph-based topology. Utilizing 1184 MRI scans from the CoMPaSS-NMD project, we map the complex 3D architecture of heterogeneous intramuscular lipodegeneration into objective, morphologically interpretable biomarkers. We introduce a graph-based skeletonization of fat infiltrates to quantify muscle architectural changes, establishing a multi-dimensional extension of traditional, spatially-agnostic volume metrics by mapping topological networks across the entire 3D muscle volume. Statistical screening via non-parametric Kruskal-Wallis analysis confirmed the discriminative power of these novel descriptors across the genetic hierarchy. Notably, topological network metrics (e.g., SF1_Skel_Nodes, $\epsilon^2$ = 0.2656) and interface dynamics metrics (e.g., SF2_To_SF1_Dist_Min, $\epsilon^2$ = 0.2092) demonstrated substantial effect sizes, providing deeper structural insights than classical volumetric assessments. Post-hoc pairwise evaluations and UMAP projections further indicated the capability of these topological and 3D geometric invariants to capture disease-specific macroscopic infiltration patterns. These results demonstrate that global architectural features represent a highly promising class of biomarkers for differential diagnosis, offering new avenues for tracking longitudinal disease dynamics in neuromuscular diagnostics. The developed automated feature extraction pipeline is integrated and available within the MUSCAT (MUSCle fAt Topology) library.
Chinese Translation
对肌肉MRI的定量评估对于监测神经肌肉疾病（NMD）至关重要。本研究提出了一种自动化的放射组学表型框架，该框架基于在五个主要建筑领域中工程化的原始特征：定量形态测量、空间分布、几何形状、进展性脂肪替代阶段之间的相互作用以及基于图的拓扑。利用来自CoMPaSS-NMD项目的1184个MRI扫描，我们将异质性肌内脂肪退化的复杂三维结构映射为客观且形态可解释的生物标志物。我们引入了一种基于图的脂肪浸润骨架化方法，以量化肌肉建筑变化，通过在整个三维肌肉体积中映射拓扑网络，建立了传统空间无关体积度量的多维扩展。通过非参数Kruskal-Wallis分析的统计筛选确认了这些新颖描述符在遗传层次中的区分能力。值得注意的是，拓扑网络度量（例如，SF1_Skel_Nodes，$ ext{ε}^2$ = 0.2656）和界面动态度量（例如，SF2_To_SF1_Dist_Min，$ ext{ε}^2$ = 0.2092）显示出显著的效应大小，提供了比经典体积评估更深入的结构洞察。后续的成对评估和UMAP投影进一步表明，这些拓扑和三维几何不变性能够捕捉特定疾病的宏观浸润模式。这些结果表明，全球建筑特征代表了一类极具前景的生物标志物，用于鉴别诊断，为在神经肌肉诊断中跟踪纵向疾病动态提供了新的途径。所开发的自动化特征提取管道已集成并可在MUSCAT（MUSCle fAt Topology）库中获得。
cs.CV / 68 / 2608.24422
ZODIAC: Zero-shot Octree-based Diffusion for Anatomical Completion
ZODIAC：基于零样本八叉树的解剖结构补全扩散方法
Abstract
Recovering the full 3D spine anatomy from intraoperative ultrasound is an ill-posed inverse problem, as the complete structure must be inferred from incomplete and noisy observations. Acoustic occlusions and limited field of view create large unobserved regions, while view-dependent artifacts lead to variability in expert annotations of the visible anatomy. Current supervised ultrasound shape completion methods rely on synthetically generated incomplete-complete paired data to learn conditional mappings under a predefined distribution of simulated occlusions. However, real intraoperative occlusions do not necessarily follow this distribution, which can limit generalization to patient data. As a result, accurate and robust completion from noisy partial observations remains an unsolved problem. We propose a zero-shot shape completion framework that reconstructs the entire lumbar spine from partial ultrasound observations without relying on simulated training data. To accommodate unseen and irregular patterns of missing structures, we introduce blended completion, a mechanism that integrates the learned anatomical prior with incoming partial geometry at inference time. The method learns a generative diffusion prior over full anatomical shapes represented in an adaptive octree structure, enabling efficient modeling of the complete spine in a single forward pass. Validation on phantom and volunteer data shows that decoupling completion from a predefined corruption distribution improves generalisation under real occlusions, outperforming a fully supervised variant by 22% on HD95 completion error. Code and data are available at https://github.com/miruna20/ZODIAC.
Chinese Translation
从术中超声图像中恢复完整的三维脊柱解剖结构是一个不适定的逆问题，因为必须从不完整和噪声干扰的观测中推断出完整结构。声学遮挡和有限的视野造成了大量未观测区域，而视角依赖的伪影则导致专家对可见解剖结构的标注存在变异。目前的监督超声形状补全方法依赖于合成生成的不完整-完整配对数据，以在预定义的模拟遮挡分布下学习条件映射。然而，真实的术中遮挡不一定遵循这种分布，这可能限制了对患者数据的泛化。因此，从噪声干扰的部分观测中进行准确和稳健的补全仍然是一个未解决的问题。我们提出了一种零样本形状补全框架，该框架能够在不依赖模拟训练数据的情况下，从部分超声观测中重建整个腰椎脊柱。为了适应未见和不规则的缺失结构模式，我们引入了混合补全机制，该机制在推理时将学习到的解剖先验与输入的部分几何形状相结合。该方法在适应性八叉树结构中学习全解剖形状的生成扩散先验，从而能够在一次前向传递中高效建模完整脊柱。在对假体和志愿者数据的验证中，解耦补全与预定义的损坏分布相结合，提高了在真实遮挡下的泛化能力，HD95补全误差比完全监督变体提高了22%。代码和数据可在 https://github.com/miruna20/ZODIAC 获取。
cs.CV / 69 / 2608.24430
Vision Language Model Fusion for Explainable Face Recognition
可解释的人脸识别的视觉语言模型融合
Abstract
Responsible deployment of face verification systems requires more than accurate decisions: systems should also provide interpretable and auditable evidence that enables users to understand, assess, and challenge their decisions. Vision-language models (VLMs) provide a promising foundation for explainable face recognition by combining visual analysis with natural-language reasoning. However, relying on a single model may further limit the decision accuracy as well as provided explanations. This work therefore investigates whether multiple VLMs can be combined to improve recognition accuracy, and to enrich the explanations associated with those decisions. This work evaluates four VLMs as standalone face verification systems and subsequently proposes a fusion framework, where two source models provide similarity scores and textual justifications and a third VLM acts as a decider model. Four different fusion scenarios are considered, progressively providing the decider model with scores, justifications, face images, and combinations of these modalities. Overall, the findings suggest that the value of multi-VLM fusion extends beyond recognition performance. VLMs can provide complementary justifications and perspectives that enable richer explanations of face recognition decisions, supporting greater transparency, auditability, and error analysis. This is relevant to the development of responsible explainable face verification systems, where users and operators should be able to understand not only the final decision but also the evidence and potential sources underlying it. The proposed multimodal VLM, which combines decision scores, explanations, and face images, achieves higher recognition accuracy than state-of-the-art VLMs and domain-specific face recognition models, while also providing fused explanations that are expected to be more robust than those generated by individual VLMs.
Chinese Translation
负责任的人脸验证系统的部署不仅需要准确的决策，还应提供可解释和可审计的证据，使用户能够理解、评估和质疑其决策。视觉语言模型（VLMs）通过将视觉分析与自然语言推理相结合，为可解释的人脸识别提供了有前景的基础。然而，依赖单一模型可能会进一步限制决策的准确性以及所提供的解释。因此，本研究探讨了是否可以通过组合多个VLM来提高识别准确性，并丰富与这些决策相关的解释。本研究评估了四个VLM作为独立的人脸验证系统的表现，并随后提出了一种融合框架，其中两个源模型提供相似度评分和文本解释，第三个VLM作为决策模型。考虑了四种不同的融合场景，逐步向决策模型提供评分、解释、人脸图像以及这些模态的组合。总体而言，研究结果表明，多VLM融合的价值超越了识别性能。VLM可以提供互补的解释和视角，使人脸识别决策的解释更加丰富，从而支持更大的透明度、可审计性和错误分析。这与负责任的可解释人脸验证系统的开发相关，其中用户和操作员应能够理解不仅是最终决策，还包括其背后的证据和潜在来源。所提出的多模态VLM结合了决策评分、解释和人脸图像，其识别准确性高于最先进的VLM和特定领域的人脸识别模型，同时提供的融合解释预计比单个VLM生成的更为稳健。
cs.CV / 70 / 2608.24439
DoublesEval: Diagnosing Multi-Agent Tactical Reasoning in Vision-Language Models via Professional Doubles Badminton
DoublesEval：通过职业双打羽毛球诊断视觉语言模型中的多智能体战术推理
Abstract
Visual Language Models (VLMs) excel at describing visible scene content but struggle to reason about dynamic multi-agent interactions, where action semantics depend on coordinated roles and spatial-temporal dependencies. We formalize this capability as \textbf{multi-agent tactical reasoning} and introduce \textbf{DoublesEval}, a diagnostic evaluation framework that leverages professional doubles badminton as a structurally tractable testbed. DoublesEval employs a key-moment-based protocol that decomposes rallies into tactically salient instants and probes models across four interpretable dimensions: atomic recognition, intra-segment composite understanding, cross-segment causal reasoning, and high-level tactical abstraction. This design isolates \emph{where} reasoning fails, rather than merely measuring answer correctness. To address observed failure modes, we propose \textbf{TacticCheck}, a lightweight constraint-guided test-time consistency checker that reranks candidate answers using the model's own lower-level tactical predictions, requiring no parameter updates or ground-truth labels at inference time. Evaluating four representative open-source VLMs on 60 curated rallies (yielding $\sim$9.6K structured instances) via a zero-shot protocol, we find that models remain weak across all diagnostic levels, with especially clear bottlenecks in spatial state, interaction binding, and terminal evidence. TacticCheck delivers consistent gains across all evaluated models, while still leaving a substantial gap to robust tactical reasoning. These results highlight the need for structured, interaction-aware evaluation paradigms for next-generation VLMs. The source code is available in \href{https://github.com/Chengjt1999/DoublesEval}{\textcolor{blue}{our GitHub repository}}.
Chinese Translation
视觉语言模型（VLMs）在描述可见场景内容方面表现出色，但在推理动态多智能体交互时却面临挑战，其中动作语义依赖于协调角色和时空依赖关系。我们将这种能力形式化为 extbf{多智能体战术推理}，并引入 extbf{DoublesEval}，一个利用职业双打羽毛球作为结构上可处理的测试平台的诊断评估框架。DoublesEval采用基于关键时刻的协议，将比赛回合分解为战术上显著的瞬间，并在四个可解释的维度上对模型进行探测：原子识别、段内复合理解、跨段因果推理和高级战术抽象。该设计能够隔离推理失败的 extit{位置}，而不仅仅是测量答案的正确性。为了解决观察到的失败模式，我们提出了 extbf{TacticCheck}，一个轻量级的约束引导测试时一致性检查器，它利用模型自身的低层战术预测对候选答案进行重新排序，无需在推理时进行参数更新或真实标签。通过零样本协议对四个具有代表性的开源VLM在60个精心挑选的回合（产生约9.6K个结构化实例）进行评估，我们发现模型在所有诊断层面上仍然表现较弱，尤其在空间状态、交互绑定和终端证据方面存在明显瓶颈。TacticCheck在所有评估模型中提供了一致的提升，但仍然与稳健的战术推理之间存在显著差距。这些结果突显了下一代VLM需要结构化、关注交互的评估范式。源代码可在 extcolor{blue}{ ext{我们的GitHub仓库}}中获得。
cs.CV / 71 / 2608.24469
Low-Rank Ternary Adaptation for Fine-Tuning Transformers
低秩三元适应用于微调变换器
Abstract
Ternary transformers offer extreme memory and compute efficiency, but existing low-bit LoRA-based methods cannot directly fine-tune ternary weights. Current approaches either require dequantization, restoring low-bit base weights to higher precision to merge with adaptation weight, or update only quantization parameters, preventing a merged model that remains ternary. We propose ternary multiplicative adaptation, which represents discrete updates of ternary weights such as sign flips or zeroing through a low-rank Kronecker factorization into two small ternary matrices applied element-wise to ternary weights. This design is parameter-efficient and expressive, preserves the ternary domain, and supports direct merging without dequantization. Experiments on six models across language and vision, including ternarized LLaMA-3 1B and 3B and a ternary ViT-B/16, demonstrate that our method recovers much of the performance lost to quantization and outperforms strong low-bit and ternary baselines. Code is available at https://github.com/alexmanoo/ternary_adaptation.
Chinese Translation
三元变换器提供了极高的内存和计算效率，但现有的基于低比特的LoRA方法无法直接微调三元权重。目前的方法要么需要去量化，将低比特的基础权重恢复到更高的精度以与适应权重合并，要么仅更新量化参数，从而阻止合并后的模型保持三元状态。我们提出了三元乘法适应（ternary multiplicative adaptation），它通过将三元权重的离散更新（如符号翻转或归零）表示为低秩克罗内克分解（Kronecker factorization），将其分解为两个小的三元矩阵，并逐元素应用于三元权重。这种设计在参数上高效且表达能力强，保持了三元域，并支持在不去量化的情况下直接合并。在包括三元化的LLaMA-3 1B和3B以及三元ViT-B/16在内的六个语言和视觉模型上的实验表明，我们的方法恢复了因量化而损失的大部分性能，并超越了强大的低比特和三元基线。代码可在 https://github.com/alexmanoo/ternary_adaptation 获取。
cs.CV / 72 / 2608.24535
VizAnchor: Decoding Manipulation Intent from Tampering Visualizations via Dual-Anchor Reasoning
VizAnchor：通过双锚推理解码篡改可视化中的操控意图
Abstract
Data visualizations are widely used for communicating information, but they are also vulnerable to intentional manipulations that induce misleading interpretations. Existing methods focus on locating tampered regions or recovering hidden information, without explaining how the visualization has been manipulated or why the resulting changes may mislead viewers. We propose \textbf{VizAnchor}, a framework for visualization manipulation understanding through dual-anchor evidence construction and VLM-based reasoning. In the first stage, VizAnchor constructs a semantic anchor to recover authentic chart information and a spatial anchor to localize tampered regions. In the second stage, three specialized agents decode the manipulation. The misleader grounding agent analyzes a four-panel visual prompt to predict the misleader information. The chart narrative reconstruction agent takes the original and tampered charts as inputs and reconstructs their respective visual narratives. Finally, the intent inferring agent integrates the visual evidence and misleader information to infer the misleading intent. We further construct a dataset for tampering localization and a dataset for misleading intent inferring. Evaluation shows that VizAnchor accurately localizes manipulations and produces faithful explanations of their manipulation, misleaders, and misleading intents.
Chinese Translation
数据可视化被广泛用于信息传达，但它们也容易受到故意操控，从而导致误导性的解读。现有方法主要集中在定位篡改区域或恢复隐藏信息上，而未能解释可视化是如何被操控的，以及为何这些变化可能会误导观众。我们提出了 extbf{VizAnchor}，一个通过双锚证据构建和基于视觉语言模型（VLM）推理来理解可视化操控的框架。在第一阶段，VizAnchor构建一个语义锚点以恢复真实的图表信息，并构建一个空间锚点以定位篡改区域。在第二阶段，三个专门的代理解码操控。误导者基础代理分析一个四面板的视觉提示，以预测误导者信息。图表叙事重构代理以原始图表和篡改图表为输入，重构各自的视觉叙事。最后，意图推断代理整合视觉证据和误导者信息，以推断误导意图。我们进一步构建了一个用于篡改定位的数据集和一个用于误导意图推断的数据集。评估结果表明，VizAnchor能够准确定位操控，并对其操控、误导者和误导意图提供可信的解释。
cs.CV / 73 / 2608.24541
Hierarchical Prototype-Memory Adaptation of SAM for Surgical Instrument Segmentation
用于外科器械分割的SAM层次原型记忆适应
Abstract
Surgical instrument segmentation (SIS) is fundamental for computer-assisted surgery, where reliable instrument masks enable precise scene understanding and clinical assistance. Recently, adapting foundation models like the Segment Anything Model (SAM) to the surgical domain via prompt-learning has shown encouraging results. However, the performance of these adapted models under challenging surgical conditions is constrained by suboptimal adaptation mechanisms. Specifically, optimizing prompts or prototypes purely via downstream segmentation loss tends to cause them to degenerate into task-specific parameters rather than serving as persistent, stable category memory, thereby degrading their robustness against complex intraoperative variations. Moreover, routing multi-scale visual cues through a single prompt pathway creates a bottleneck that hinders effective scale-matched coupling. To address these limitations, we propose HPMA, a Hierarchical Prototype-Memory Adaptation framework for SAM. Specifically, HPMA constructs a frozen, multi-scale visual prototype memory bank from annotated surgical scenes and integrates it into SAM's feature space using lightweight adapters to preserve stable category evidence. To maximize the utility of multi-scale cues, we introduce a scale-matched coupling mechanism where global prototypes calibrate class-level prompt features, structural prototypes guide decoder object queries, and local prototypes align high-resolution feature maps through a local alignment objective. Extensive experiments on the public EndoVis2017 and EndoVis2018 datasets demonstrate that our approach achieves state-of-the-art performance, outperforming existing foundation model adaptation methods.
Chinese Translation
外科器械分割（SIS）是计算机辅助外科手术的基础，其中可靠的器械掩膜能够实现精确的场景理解和临床辅助。最近，通过提示学习将基础模型如Segment Anything Model（SAM）适应于外科领域已显示出令人鼓舞的结果。然而，这些适应模型在复杂外科条件下的性能受到次优适应机制的限制。具体而言，仅通过下游分割损失优化提示或原型往往会导致它们退化为特定任务的参数，而不是作为持久、稳定的类别记忆，从而降低其对复杂术中变化的鲁棒性。此外，通过单一提示路径路由多尺度视觉线索会造成瓶颈，阻碍有效的尺度匹配耦合。为了解决这些限制，我们提出了HPMA，即用于SAM的层次原型记忆适应框架。具体而言，HPMA从标注的外科场景中构建一个冻结的多尺度视觉原型记忆库，并通过轻量级适配器将其整合到SAM的特征空间中，以保持稳定的类别证据。为了最大化多尺度线索的效用，我们引入了一种尺度匹配耦合机制，其中全局原型校准类别级提示特征，结构原型引导解码器对象查询，局部原型通过局部对齐目标对齐高分辨率特征图。在公共的EndoVis2017和EndoVis2018数据集上进行的广泛实验表明，我们的方法实现了最先进的性能，超越了现有的基础模型适应方法。
cs.CV / 74 / 2608.24544
KLTNet: Learning Sparse Feature Tracking for Robust and Accurate Monocular Visual-Inertial Odometry
KLTNet：用于鲁棒和准确的单目视觉惯性里程计的稀疏特征跟踪学习
Abstract
Many feature-based visual-inertial odometry (VIO) systems rely on sparse feature tracking, whose accuracy and robustness directly affect state estimation. Classical KLT trackers rely primarily on local image patches and can become unreliable under rapid motion or in low-texture environments. We propose KLTNet, a lightweight learning-based, plug-and-play sparse feature tracker designed to replace classical KLT trackers in KLT-based VIO front ends. KLTNet follows a coarse-to-fine, dense-to-sparse architecture that combines low-resolution dense optical flow for robust global motion initialization with triplet-patch refinement for accurate and temporally consistent tracking. A fixed reference patch provides a stable anchor throughout each feature track and helps reduce accumulated tracking drift. In addition, KLTNet predicts anisotropic confidence weights supervised through differentiable multi-view triangulation, which can be used as observation weights in compatible VIO estimators. Experiments with VINS-Mono and OpenVINS on public benchmarks and a self-collected low-texture dataset demonstrate improved tracking and odometry accuracy over classical KLT, while maintaining real-time performance on an embedded platform.
Chinese Translation
许多基于特征的视觉惯性里程计（VIO）系统依赖于稀疏特征跟踪，其准确性和鲁棒性直接影响状态估计。经典的KLT跟踪器主要依赖于局部图像块，在快速运动或低纹理环境下可能变得不可靠。我们提出了KLTNet，一种轻量级的基于学习的即插即用稀疏特征跟踪器，旨在替代KLT基于的VIO前端中的经典KLT跟踪器。KLTNet遵循一种粗到细、稠密到稀疏的架构，结合低分辨率的稠密光流以实现鲁棒的全局运动初始化，并通过三重图块细化实现准确且时间一致的跟踪。固定的参考图块在每个特征跟踪过程中提供稳定的锚点，有助于减少累积的跟踪漂移。此外，KLTNet通过可微分的多视图三角测量来预测各向异性的置信权重，这些权重可用作兼容的VIO估计器中的观测权重。在公共基准测试和自收集的低纹理数据集上与VINS-Mono和OpenVINS的实验表明，KLTNet在跟踪和里程计准确性方面优于经典KLT，同时在嵌入式平台上保持实时性能。
cs.CV / 75 / 2608.24563
X-MULTI: VLM-based Imaging Factor Disentanglement for Factor-Aware Image Synthesis
X-MULTI：基于VLM的成像因子解耦用于因子感知图像合成
Abstract
Imaging factor disentanglement in text-to-image generation aims to independently control image acquisition properties such as types of camera lenses, sensor types, viewpoints, and domains to enable combinatorial generalization. This should let the model synthesize novel factor combinations unobserved in the training data, such as pairing a fisheye lens with an event sensor never observed in training data. Recent work, MULTI, introduced learnable, factor-specific embeddings to disentangle imaging factors, along with the Factor Alignment Accuracy (FAA) metric to evaluate disentanglement quality. We identify and address two independent limitations. First, MULTI's pixel-level reconstruction objective supervises the model only on observed imaging factor combinations, providing no direct training signal for novel combinations. We therefore propose X-MULTI, which uses a pretrained vision-language model (VLM) to supervise novel factor combinations synthesized during training. Second, we show the FAA metric exhibits severe cross-factor correlation leakage, misrepresenting true disentanglement quality. We therefore propose Improved-FAA (I-FAA), which employs factor-specific augmentation strategies to break these correlations and enables more rigorous evaluation. Experiments demonstrate that X-MULTI achieves improved factor alignment on novel combinations compared to MULTI. Moreover, we show that correlation leakage in FAA distorts the evaluation of true factor disentanglement and I-FAA reduces this leakage and therefore provides a more robust assessment of factor alignment.
Chinese Translation
文本到图像生成中的成像因子解耦旨在独立控制图像获取属性，如相机镜头类型、传感器类型、视角和领域，以实现组合泛化。这将使模型能够合成在训练数据中未观察到的新因子组合，例如将鱼眼镜头与在训练数据中从未观察到的事件传感器配对。最近的研究MULTI引入了可学习的因子特定嵌入以解耦成像因子，并提出了因子对齐准确度（Factor Alignment Accuracy, FAA）指标来评估解耦质量。我们识别并解决了两个独立的局限性。首先，MULTI的像素级重建目标仅对观察到的成像因子组合进行监督，未为新组合提供直接的训练信号。因此，我们提出了X-MULTI，它使用预训练的视觉-语言模型（Vision-Language Model, VLM）来监督在训练过程中合成的新因子组合。其次，我们表明FAA指标存在严重的跨因子相关性泄漏，错误地表现出真实的解耦质量。因此，我们提出了改进的FAA（Improved-FAA, I-FAA），它采用因子特定的增强策略来打破这些相关性，从而实现更严格的评估。实验表明，与MULTI相比，X-MULTI在新组合上的因子对齐得到了改善。此外，我们展示了FAA中的相关性泄漏扭曲了对真实因子解耦的评估，而I-FAA减少了这种泄漏，因此提供了对因子对齐的更稳健评估。
cs.CV / 76 / 2608.24580
Human-Inspired Social Engagement Analysis via Interpretable Mutual Visual Attention
基于可解释的互视觉注意的人类启发社交参与分析
Abstract
Understanding social interactions from non-verbal visual data is important for behavior analysis and activity monitoring. We propose an interpretable computational model of social engagement inspired by psychological theories of mutual visual attention. Rather than learning interaction patterns end-to-end, our framework explicitly models dyadic visual attention and aggregates these cues into interpretable measures of individual and group engagement. The resulting modular framework combines state-of-the-art head orientation estimation with lightweight geometric reasoning, producing explanations that remain accessible to non-technical users. We evaluate the proposed approach on a variety of data through quantitative experiments and demonstrate its practical usefulness with qualitative visualizations designed to support teachers, caregivers, and social workers in understanding group interaction dynamics.
Chinese Translation
从非语言视觉数据中理解社交互动对于行为分析和活动监测至关重要。我们提出了一种受心理学互视觉注意理论启发的可解释社交参与计算模型。我们的框架并非端到端地学习互动模式，而是明确建模二人视觉注意，并将这些线索聚合为可解释的个体和群体参与度度量。最终得到的模块化框架结合了先进的头部朝向估计与轻量级几何推理，生成的解释对非技术用户仍然易于理解。我们通过定量实验在多种数据上评估了所提方法，并通过定性可视化展示了其实际应用价值，旨在帮助教师、护理人员和社会工作者理解群体互动动态。
cs.CV / 77 / 2608.24594
Comparative Assessment of Deep Learning Architectures for Underwater Subsurface Kelp Forest Segmentation with The Kelp-o-Tron
深度学习架构在水下次表海藻森林分割中的比较评估：以 Kelp-o-Tron 为例
Abstract
Submerged kelp forests are vital coastal ecosystems that support marine biodiversity and ecosystem dynamics, yet accurate underwater kelp segmentation remains challenging due to optical degradation, illumination variability, turbidity, overlapping vegetation, and complex benthic backgrounds. We systematically evaluated three deep learning semantic segmentation frameworks, ResNet34-U-Net, ResNet50-DeepLabV3, and a hybrid ResNet50-ASPP-Transformer architecture, for kelp detection using high-resolution underwater RGB imagery collected from northeastern U.S. coastal waters. A dataset of 3,395 SSeg assisted annotated image-mask pairs was developed for model training and validation, while geographically independent sites were used for quantitative and qualitative evaluation. All models used consistent preprocessing, augmentation, and evaluation protocols. On independent test data, ResNet50-DeepLabV3 achieved the highest Dice (0.7120) and Intersection over Union (IoU; 0.6267), followed by ResNet34 U Net (Dice 0.6868; IoU 0.5978). The hybrid ASPP Transformer achieved the highest pixel accuracy (0.8528) but lower Dice (0.6437) and IoU (0.5746). External qualitative evaluation further showed that DeepLabV3 produced more consistent segmentation across varying environmental conditions, image qualities, and benthic habitats. Overall, ResNet50-DeepLabV3, termed Kelp-O-Tron, provided the best balance of segmentation accuracy, robustness, and generalization. The dataset, annotation workflow, and comparative evaluation provide resources for advancing automated underwater habitat mapping and ecological monitoring.
Chinese Translation
水下海藻森林是重要的沿海生态系统，支持海洋生物多样性和生态系统动态，但由于光学衰减、光照变化、浑浊度、重叠植被和复杂的底栖背景，准确的水下海藻分割仍然具有挑战性。我们系统评估了三种深度学习语义分割框架：ResNet34-U-Net、ResNet50-DeepLabV3 和混合型 ResNet50-ASPP-Transformer 架构，以利用从美国东北沿海水域收集的高分辨率水下 RGB 图像进行海藻检测。我们开发了一个包含 3,395 对 SSeg 辅助标注图像-掩膜对的数据集，用于模型训练和验证，同时使用地理独立站点进行定量和定性评估。所有模型均采用一致的预处理、增强和评估协议。在独立测试数据上，ResNet50-DeepLabV3 实现了最高的 Dice 系数 (0.7120) 和交并比 (IoU; 0.6267)，其次是 ResNet34-U-Net (Dice 0.6868; IoU 0.5978)。混合型 ASPP Transformer 达到了最高的像素准确率 (0.8528)，但 Dice (0.6437) 和 IoU (0.5746) 较低。外部定性评估进一步表明，DeepLabV3 在不同环境条件、图像质量和底栖栖息地中产生了更一致的分割。总体而言，ResNet50-DeepLabV3，即 Kelp-O-Tron，提供了最佳的分割准确性、鲁棒性和泛化能力的平衡。该数据集、标注工作流程和比较评估为推进自动化水下栖息地制图和生态监测提供了资源。
cs.CV / 78 / 2608.24626
Towards Reliable AI-Based Histological Staining: A Systematic Study of Scaling and Uncertainty in Unpaired Generative Models
迈向可靠的基于人工智能的组织学染色：无配对生成模型中的尺度与不确定性系统研究
Abstract
Liver fibrosis, the principal predictor of long-term outcome in chronic liver disease, is staged from histological estimates of collagen content. Sirius Red (SR) provides the standard quantitative readout (collagen proportionate area, CPA) but is not acquired at every clinical centre and consumes tissue, time, and reagent cost beyond the routine Hematoxylin and eosin (H&E) stain. AI-based virtual staining can generate SR directly from H&E, yet systematic benchmarks of unsupervised models are scarce and their predictive uncertainty has not been quantified, even though visually plausible outputs may not faithfully reproduce the underlying tissue structure. We therefore benchmark six unsupervised image-to-image architectures (GAN-based and diffusion-based) across 54 scaling configurations on a newly released paired H&E to SR mouse liver dataset, the first open resource for this translation task. Each configuration is evaluated jointly on perceptual, distributional, and task-specific axes plus a blinded expert reader study; the best per family is then retrained as a deep ensemble, the first systematic comparison of epistemic uncertainty across unsupervised stain-to-stain architectures. Across families, perceptual quality, task-specific error, and ensemble agreement measure largely independent axes of model fitness: GAN-based methods cluster tightly on perceptual metrics yet differ substantially on task error and ensemble agreement, while the diffusion-based method (CycleDiffusion) is qualitatively different on all three. No single metric captures these differences, so reliable virtual staining requires reporting and selecting on all three jointly. The dataset, tiling pipeline, models, and evaluation code are released publicly.
Chinese Translation
肝纤维化是慢性肝病长期预后的主要预测因子，其分期基于胶原含量的组织学估计。Sirius Red (SR) 提供了标准的定量读数（胶原比例面积，CPA），但并非在每个临床中心都能获取，并且消耗了组织、时间和超出常规苏木精-伊红染色（H&E）的试剂成本。基于人工智能的虚拟染色可以直接从H&E生成SR，然而，关于无监督模型的系统基准研究稀缺，其预测不确定性尚未被量化，尽管视觉上合理的输出可能并未忠实再现基础组织结构。因此，我们在新发布的配对H&E到SR小鼠肝脏数据集上，对六种无监督图像到图像架构（基于GAN和扩散的模型）在54种尺度配置下进行基准测试，这是该转化任务的第一个开放资源。每种配置在感知、分布和任务特定轴上进行联合评估，并进行盲评专家阅读研究；然后，最佳模型在每个家族中重新训练为深度集成，这是对无监督染色到染色架构中认识不确定性的首次系统比较。在不同家族中，感知质量、任务特定误差和集成一致性度量在模型适应性上大致是独立的：基于GAN的方法在感知指标上紧密聚集，但在任务误差和集成一致性上存在显著差异，而扩散方法（CycleDiffusion）在所有三个方面的质量上均有所不同。没有单一指标能够捕捉这些差异，因此可靠的虚拟染色需要在这三者上共同报告和选择。数据集、平铺管道、模型和评估代码已公开发布。
cs.CV / 79 / 2608.24646
On-Policy Self-Distillation in Diffusion Models
扩散模型中的在线自蒸馏
Abstract
Reinforcement learning can align diffusion models with human preferences and task-specific objectives, but endpoint rewards do not specify how an intermediate denoising prediction should change. We introduce DiffusionOPSD as an on-policy self-distillation framework that converts image-level reward guidance into explicit targets for clean-output predictions at sampled queries. At each outer iteration, a frozen behavior policy generates trajectories and supplies query states and anchors. Reward gradients construct bounded positive and negative targets around each anchor. The trainable policy fits these targets as detached supervision through finite fitting before an exponential moving average update refreshes the behavior policy. This setup lets us measure target construction and finite realization separately. Controlled same-query experiments show that larger target-construction gains do not necessarily translate into larger realized gains after a single fitting update. Across SD 3.5-M and the step-distilled Z-Image-Turbo, our approach achieves the best final held-out scores in 19 of 20 reward-matched settings across two backbones and ten evaluators. It outperforms the strongest competing method by up to 44.0% and reduces training GPU-hours relative to DiffusionNFT by 40% on SD 3.5-M and 63% on Z-Image-Turbo. These results support on-policy self-distillation as an efficient and analyzable approach to diffusion post-training by converting image-level reward guidance into explicit and continually refreshed intermediate supervision, thereby opening a path toward more efficient and diagnosable alignment.
Chinese Translation
强化学习可以使扩散模型与人类偏好和特定任务目标对齐，但最终奖励并未明确指定中间去噪预测应如何变化。我们提出了DiffusionOPSD作为一种在线自蒸馏框架，将图像级奖励指导转换为在采样查询中清晰输出预测的明确目标。在每次外部迭代中，一个冻结的行为策略生成轨迹并提供查询状态和锚点。奖励梯度在每个锚点周围构建有界的正负目标。可训练策略通过有限拟合将这些目标作为独立监督进行拟合，然后通过指数移动平均更新刷新行为策略。该设置使我们能够分别测量目标构建和有限实现。控制的同查询实验表明，较大的目标构建增益不一定会在单次拟合更新后转化为更大的实现增益。在SD 3.5-M和逐步蒸馏的Z-Image-Turbo中，我们的方法在两个骨干网络和十个评估器的20个奖励匹配设置中实现了最佳的最终保留分数。它比最强的竞争方法提高了最多44.0%的性能，并在SD 3.5-M上相对于DiffusionNFT减少了40%的训练GPU小时，在Z-Image-Turbo上减少了63%。这些结果支持在线自蒸馏作为一种高效且可分析的扩散后训练方法，通过将图像级奖励指导转换为明确且持续更新的中间监督，从而为更高效和可诊断的对齐开辟了道路。
cs.CV / 80 / 2608.24671
ReGround-Surg: Reliability-Guided Anchor Grounding for Referring Surgical Video Segmentation
ReGround-Surg：基于可靠性的锚点定位用于手术视频分割
Abstract
Referring surgical video segmentation requires segmenting a target instrument or tissue region across video frames according to a natural language expression. Recent Segment Anything Model 2 (SAM2) based two-stage methods (e.g., ReSurgSAM2) first ground the referred target in an initial or selected frame, then propagate the selected mask via tracking. Although effective, their performance is highly sensitive to the quality of the initial grounded mask: once an incorrect anchor is selected, subsequent tracking tends to propagate the error. This issue is especially challenging in surgical videos due to visually similar instruments, occlusion, and complex tissue-tool interactions. To address this issue, we propose ReGround-Surg, a lightweight reliability-guided anchor grounding framework to improve SAM2-based referring surgical video segmentation. It first predicts a text-conditioned spatial reliability map from the referring expression and current-frame visual features. The map is then reused in two complementary branches: a Gated Side Adapter enhances expression-relevant visual regions before text-to-vision fusion, while a Reliability-Weighted Vision-to-Text Attention module suppresses off-target visual evidence during prompt-token aggregation. Experiments on Ref-EndoVis17 and Ref-EndoVis18 show consistent improvements over state-of-the-art methods across three evaluation splits with negligible speed reduction. Code is publicly available at https://github.com/JiaxinWen1/ReGround-Surg.
Chinese Translation
手术视频分割需要根据自然语言表达在视频帧中对目标仪器或组织区域进行分割。最近基于Segment Anything Model 2 (SAM2) 的两阶段方法（例如，ReSurgSAM2）首先在初始或选定帧中定位所提及的目标，然后通过跟踪传播所选掩膜。尽管效果显著，但其性能对初始定位掩膜的质量高度敏感：一旦选择了错误的锚点，后续跟踪往往会传播错误。由于视觉上相似的仪器、遮挡以及复杂的组织与工具交互，这一问题在手术视频中尤其具有挑战性。为了解决这一问题，我们提出了ReGround-Surg，一个轻量级的基于可靠性的锚点定位框架，以改善基于SAM2的手术视频分割。该框架首先根据所提及的表达和当前帧的视觉特征预测一个文本条件的空间可靠性图。然后，该图在两个互补分支中重复使用：Gated Side Adapter在文本与视觉融合之前增强与表达相关的视觉区域，而Reliability-Weighted Vision-to-Text Attention模块在提示令牌聚合过程中抑制非目标视觉证据。在Ref-EndoVis17和Ref-EndoVis18上的实验表明，在三个评估拆分中，相较于最先进的方法，均取得了一致的改进，并且速度降低微乎其微。代码可在 https://github.com/JiaxinWen1/ReGround-Surg 上公开获取。
cs.CV / 81 / 2608.24674
TurboT2VA: Fast Large-Scale Text-to-Video-Audio Generation via Score-Regularized Consistency Distillation
TurboT2VA：通过得分正则化一致性蒸馏实现快速大规模文本到视频音频生成
Abstract
Joint text-to-video-audio generation produces synchronized visual and acoustic content, but the long sampling trajectories and heterogeneous multimodal computation of large models make inference prohibitively expensive. We present TurboT2VA, a distillation and inference framework for accelerating a 19B-parameter joint video-audio model. Large-scale T2VA distillation is challenged by modality-imbalanced optimization, the difficulty of continuous-time consistency training at scale, and the quality--diversity trade-off. TurboT2VA addresses these issues with per-modality normalization and a progressive curriculum comprising discrete consistency warm-up, continuous consistency refinement, and joint consistency--distribution matching. The curriculum first establishes a stable, diverse generation trajectory and only then introduces distribution-level refinement. On LTX-2, four-step distillation reduces generator latency from 50.52s to 2.51s at the standard evaluation resolution of 512$\times$768, achieving a 20.1$\times$ speedup while maintaining strong visual quality, audio fidelity, diversity, and video-audio synchronization. We further develop an architecture-aware inference stack that combines guarded W8A8 and fused operators, padded-text compaction, and modality-aware sparse attention while preserving dense cross-modal and text-conditioning paths. Under the high-resolution deployment setting at 1024$\times$1792, the complete stack reduces generator latency from 318.74s to 5.83s on one NVIDIA H20, achieving a 54.67$\times$ generator-only speedup. Inference code and generation demos are available at https://github.com/thu-ml/TurboDiffusion/tree/main/turbot2va.
Chinese Translation
联合文本到视频音频生成能够产生同步的视觉和声学内容，但大模型的长采样轨迹和异构多模态计算使得推理成本高昂。我们提出了TurboT2VA，这是一个用于加速19B参数联合视频音频模型的蒸馏和推理框架。大规模T2VA蒸馏面临模态不平衡优化、在大规模下进行连续时间一致性训练的困难，以及质量与多样性之间的权衡。TurboT2VA通过每个模态的归一化和一个渐进式课程来解决这些问题，该课程包括离散一致性预热、连续一致性细化和联合一致性-分布匹配。该课程首先建立一个稳定且多样的生成轨迹，然后再引入分布级别的细化。在LTX-2上，四步蒸馏将生成器延迟从50.52秒减少到2.51秒，标准评估分辨率为512×768，实现了20.1倍的加速，同时保持了强大的视觉质量、音频保真度、多样性和视频音频同步。我们进一步开发了一种架构感知的推理堆栈，结合了受保护的W8A8和融合操作符、填充文本压缩以及模态感知稀疏注意力，同时保留了密集的跨模态和文本条件路径。在1024×1792的高分辨率部署设置下，完整堆栈将生成器延迟从318.74秒减少到5.83秒，在一台NVIDIA H20上实现了54.67倍的仅生成器加速。推理代码和生成演示可在https://github.com/thu-ml/TurboDiffusion/tree/main/turbot2va获取。
cs.CV / 82 / 2608.24680
Game2World Engine: Unlocking In-the-Wild Gameplay Videos for World Model Training
Game2World 引擎：解锁真实环境中的游戏视频以进行世界模型训练
Abstract
Video games provide a scalable source of training data for video world models, offering diverse environments, complex interactions, and abundant in-the-wild gameplay videos. However, raw gameplay footage entangles the game world with screen-space interfaces, introducing game-specific biases and irrelevant dynamics that hinder world-model training. To address this problem, we introduce GameUI-Taxonomy and G2WEngine, a full-stack framework that formalizes gameplay UI grounding and removal. G2WEngine automatically extracts reusable UI assets from real gameplay videos and synthesizes temporally coherent UI overlays on clean footage. Using this engine, we construct Game2World, comprising 96K synthetic paired videos with precise reconstruction targets and 1,079 in-the-wild clips from 303 games for realistic evaluation. Its asset library contains 5,132 verified UI elements across 21 taxonomy categories, collected from 1,010 representative gameplay frames. Based on Game2World, we propose GameCleaner, a mask-free gameplay UI removal model that combines multimodal semantic understanding with video editing capabilities. Unlike mask-based methods, GameCleaner directly identifies and removes diverse HUD elements while preserving the underlying scene content and temporal dynamics. In a controlled pilot, world models trained on UI-free gameplay improve overall VideoReward by 6.83% over those trained on UI-overlaid data. On UI-removal evaluation, GameCleaner achieves an average AAR of 95.36 on synthetic videos, outperforming the strongest temporal mask baseline by 57.3%, and obtains the best in-the-wild AAR of 80.05 with 99.8 background preservation. These results demonstrate the scalable potential of transforming Internet gameplay videos into high-quality world-model training data. Code, dataset, and model will be available at https://github.com/Dongping-Chen/Game2World.
Chinese Translation
视频游戏为视频世界模型提供了可扩展的训练数据来源，展现了多样的环境、复杂的互动和丰富的真实游戏视频。然而，原始游戏画面将游戏世界与屏幕空间界面交织在一起，带来了特定于游戏的偏见和无关的动态，阻碍了世界模型的训练。为了解决这一问题，我们提出了 GameUI-Taxonomy 和 G2WEngine，这是一个完整的框架，旨在规范游戏玩法用户界面的基础和去除。G2WEngine 自动从真实游戏视频中提取可重用的用户界面资产，并在干净的画面上合成时间上连贯的用户界面叠加。利用该引擎，我们构建了 Game2World，包含 96K 个合成配对视频，具有精确的重建目标，以及来自 303 款游戏的 1,079 个真实片段，以进行真实评估。其资产库包含 5,132 个经过验证的用户界面元素，分布在 21 个分类类别中，收集自 1,010 个具有代表性的游戏画面。基于 Game2World，我们提出了 GameCleaner，一个无掩码的游戏用户界面去除模型，结合了多模态语义理解和视频编辑能力。与基于掩码的方法不同，GameCleaner 直接识别并去除多样的 HUD 元素，同时保留底层场景内容和时间动态。在一个受控的试点中，基于无用户界面的游戏训练的世界模型在整体 VideoReward 上比基于用户界面叠加数据训练的模型提高了 6.83%。在用户界面去除评估中，GameCleaner 在合成视频上实现了平均 AAR 为 95.36，超越了最强的时间掩码基线 57.3%，并在真实环境中获得了最佳 AAR 为 80.05，背景保留率达到 99.8%。这些结果展示了将互联网游戏视频转化为高质量世界模型训练数据的可扩展潜力。代码、数据集和模型将可在 https://github.com/Dongping-Chen/Game2World 获取。
cs.CV / 83 / 2608.24715
Deep Learning Super Resolution for Satellite Cloud Mask Downscaling
卫星云掩膜下采样的深度学习超分辨率方法
Abstract
A vast amount of optical satellite data is being transmitted to Earth-based servers every day, and more than half of this data is affected by haze or clouds. Additionally, this data suffers from the fundamental trade-off between spatial and temporal resolution, which remains largely unresolved, making the acquisition of continuous high-resolution satellite observations of clouds an ongoing challenge. This work addresses this challenge by proposing two Deep Learning super-resolution methods for the accurate downscaling of SEVIRI cloud mask products, as well as a novel cross-sensor cloud mask dataset called SEVMOD-CM, created by spatially and temporally matching MODIS and SEVIRI satellite observations. The two proposed models are a CNN-based (SpatialCNN) and a GAN-based (SpatialGAN) Neural Network. Trained on the SEVIRI spectral and cloud mask products, the proposed methods predict the corresponding MODIS Cloud masks, achieving a 4x spatial enhancement across sensor domains. Both approaches are evaluated experimentally, and compared against the standard bicubic interpolation upsampling technique. The experimental results demonstrate the value of the proposed models and dataset for the remote sensing community, highlighting the benefits of applying super-resolution techniques to geostationary-derived cloud mask products for applications such as atmospheric monitoring, weather forecasting, disaster risk reduction, solar energy forecasting, and climate research.
Chinese Translation
每天都有大量的光学卫星数据被传输到地球上的服务器，而这些数据中有超过一半受到雾霾或云层的影响。此外，这些数据面临空间分辨率与时间分辨率之间的基本权衡，这一问题尚未得到有效解决，使得持续获取高分辨率的卫星云观测数据成为一项持续的挑战。本研究通过提出两种深度学习超分辨率方法，针对SEVIRI云掩膜产品的准确下采样，来解决这一挑战。同时，我们创建了一个名为SEVMOD-CM的新型跨传感器云掩膜数据集，该数据集通过空间和时间匹配MODIS和SEVIRI卫星观测数据而生成。所提出的两种模型分别为基于卷积神经网络的模型（SpatialCNN）和基于生成对抗网络的模型（SpatialGAN）。这些方法在SEVIRI光谱和云掩膜产品上进行训练，预测相应的MODIS云掩膜，实现了在传感器领域内的4倍空间增强。我们对这两种方法进行了实验评估，并与标准的双三次插值上采样技术进行了比较。实验结果表明，所提出的模型和数据集对遥感社区具有重要价值，强调了将超分辨率技术应用于静止卫星获取的云掩膜产品在大气监测、天气预报、灾害风险降低、太阳能预测和气候研究等应用中的优势。
cs.CV / 84 / 2608.24723
Interpretable Fundus Image Classification via Ring-Based Retinal Vasculature Features
基于环状视网膜血管特征的可解释性眼底图像分类
Abstract
Retinal fundus photography is widely used for screening and monitoring ocular diseases, but many modern classification pipelines rely on deep latent representations and provide limited interpretability. This study develops an interpretable fundus image classification framework based on a ring-structured representation of the retinal vasculature centered on the optic disc. The method quantifies vessel geometry, color appearance, oxygenation-related vascular appearance, and vessel--background entropy within concentric retinal regions. These physiologically motivated descriptors are derived from vessel masks, image intensities, and optical-density measurements and aggregated across rings to capture spatial variation in vascular properties. Using only quantitative vascular descriptors, the proposed method achieved strong classification performance across three public fundus datasets. On HRF, it achieved 91.1\% accuracy using automatically generated vessel masks, matching RETFound, a vision transformer pretrained on large-scale retinal fundus image data, under the same evaluation setting. Additional analyses suggest that pretrained image models are sensitive to acquisition-related spatial cues, including fundus scale and retinal position within the field of view, as well as broader non-vessel image characteristics. This framework may support interpretable disease classification, quantitative retinal phenotyping, and retinal biomarker discovery without requiring large task-specific training datasets.
Chinese Translation
视网膜眼底摄影广泛用于筛查和监测眼科疾病，但许多现代分类流程依赖于深度潜在表示，提供的可解释性有限。本研究开发了一种基于以视神经盘为中心的环状视网膜血管结构表示的可解释性眼底图像分类框架。该方法量化了血管几何形状、颜色外观、与氧合相关的血管外观以及同心视网膜区域内的血管-背景熵。这些生理学驱动的描述符源自血管掩膜、图像强度和光学密度测量，并在环中聚合以捕捉血管特性在空间上的变化。仅使用定量血管描述符，所提方法在三个公共眼底数据集上实现了强大的分类性能。在HRF数据集上，使用自动生成的血管掩膜达到了91.1%的准确率，与在相同评估设置下预训练于大规模视网膜眼底图像数据的视觉变换器RETFound相匹配。额外分析表明，预训练的图像模型对与采集相关的空间线索敏感，包括眼底比例和视野内的视网膜位置，以及更广泛的非血管图像特征。该框架可能支持可解释的疾病分类、定量视网膜表型分析和视网膜生物标志物发现，而无需大量特定任务的训练数据集。
cs.CV / 85 / 2608.24738
TorchMorph: CUDA-accelerated Morphological Transforms
TorchMorph：基于CUDA加速的形态学变换
Abstract
Morphological transforms are long-standing tools for shape and mask processing, but the de facto reference implementation in the Python ecosystem, i.e. scipy.ndimage, is CPU-only, single-array, and therefore unusable inside a GPU training loop without an expensive device-to-host round trip. GPU vision libraries built on PyTorch cover a narrow subset of these operators, typically restricted to two spatial dimensions and flat structuring elements. We present TorchMorph, a lightweight PyTorch extension that closes this gap. TorchMorph exposes 22 public operators covering binary morphology, greyscale morphology, exact and approximate distance transforms, and entropy-regularised optimal transport, all implemented as fused CUDA kernels that operate directly on (B, C, Spatial...) CUDA tensors with up to eight spatial dimensions. The API deliberately mirrors scipy.ndimage argument-for-argument, including border modes, structuring-element origins and pre-allocated outputs, so that existing pipelines port with a change of import. We describe the layered architecture and the kernel designs behind each operator family. Against single-threaded CPU references, batched execution reaches up to 1.1e3 times the throughput of scipy.ndimage on greyscale morphology and up to 350x on exact Euclidean distance transforms, while the Sinkhorn solver runs up to 42x faster than POT. Binary and chamfer operators reproduce their SciPy counterparts exactly, and every float-valued operator agrees with the CPU reference to within 1.8e-6 absolute error. TorchMorph is released under the MIT licence at https://intcomp.github.io/tm.
Chinese Translation
形态学变换是形状和掩模处理的长期工具，但在Python生态系统中，事实上的参考实现，即scipy.ndimage，仅支持CPU、单数组，因此在GPU训练循环中无法使用，除非进行昂贵的设备到主机的往返传输。基于PyTorch构建的GPU视觉库仅涵盖这些运算符的一小部分，通常限制在两个空间维度和扁平结构元素。我们提出了TorchMorph，这是一个轻量级的PyTorch扩展，填补了这一空白。TorchMorph公开了22个公共运算符，涵盖二元形态学、灰度形态学、精确和近似距离变换，以及熵正则化的最优传输，所有运算符均实现为融合的CUDA内核，直接作用于具有多达八个空间维度的(B, C, Spatial...) CUDA张量。API故意逐个参数地镜像scipy.ndimage，包括边界模式、结构元素原点和预分配输出，以便现有管道只需更改导入即可迁移。我们描述了每个运算符系列背后的分层架构和内核设计。与单线程CPU参考相比，批量执行在灰度形态学上达到scipy.ndimage的1.1e3倍吞吐量，在精确欧几里得距离变换上达到350倍，而Sinkhorn求解器的运行速度比POT快42倍。二元和切线运算符完全重现其SciPy对应物，所有浮点值运算符与CPU参考的绝对误差不超过1.8e-6。TorchMorph在MIT许可证下发布，网址为https://intcomp.github.io/tm。
cs.CV / 86 / 2608.24756
Weakly Supervised Seafloor Segmentation for Seagrass Habitat Mapping in Side-Scan Sonar Imagery
基于弱监督的海底分割方法用于侧扫声纳图像中的海草栖息地制图
Abstract
Seagrass meadows are crucial blue-carbon habitats, and mapping their extent is a prerequisite for coastal management and carbon inventory. Optical satellite sensors cover large areas but cannot reach deep or turbid water, whereas side-scan sonar (SSS) images the seabed at high resolution and at any depth. Interpreting SSS, however, still relies on dense manual annotation, which is slow and costly. We address this by adapting a weakly supervised semantic segmentation framework to SSS benthic habitat mapping, so that pixel-level maps are learned from image-level labels alone. The framework couples a ViT-based encoder-decoder with a classification branch, extracts class activation maps, and refines them into pseudo-labels with a dense conditional random field that we tune for the noise and weak boundaries of acoustic imagery. It follows an iterative self-training scheme, together with a sampling strategy to cope with the strong class imbalance of the data. We also study the effect of different loss functions on segmentation quality, finding Lov\'asz-Softmax loss the most effective. On a held-out transect, the refined pseudo-labels reached an mIoU of 89.3\% against the ground truth, and the segmentation branch, trained without any pixel-level labels, reached 87.6\%. Self-supervised pretraining on unlabelled SSS added a further 3\% in mean intersection-over-union. Field trials further demonstrate the generalizability of the trained model. These results show that accurate and label-efficient benthic habitat mapping from side-scan sonar is feasible at the scale needed for coast-wide seagrass monitoring.
Chinese Translation
海草草甸是重要的蓝碳栖息地，绘制其分布范围是沿海管理和碳清单的前提。光学卫星传感器覆盖大面积区域，但无法到达深水或浑浊水域，而侧扫声纳（SSS）能够以高分辨率在任何深度成像海床。然而，解释SSS仍然依赖于密集的人工标注，这既耗时又昂贵。我们通过将弱监督语义分割框架应用于SSS底栖栖息地制图来解决这一问题，从而使得仅通过图像级标签学习像素级地图。该框架结合了基于ViT的编码器-解码器和分类分支，提取类别激活图，并通过我们针对声学图像的噪声和弱边界调整的稠密条件随机场将其精炼为伪标签。它遵循迭代自我训练方案，并结合采样策略以应对数据的强类别不平衡。我们还研究了不同损失函数对分割质量的影响，发现Lovász-Softmax损失最为有效。在一个保留的横断面上，精炼后的伪标签与真实值的mIoU达到了89.3%，而没有任何像素级标签训练的分割分支达到了87.6%。在未标记的SSS上进行自监督预训练进一步提高了3%的平均交并比。现场试验进一步证明了训练模型的泛化能力。这些结果表明，从侧扫声纳中进行准确且高效的底栖栖息地制图在沿海海草监测所需的规模上是可行的。
cs.CV / 87 / 2608.24759
IDeaL: Data-Free Multi-Teacher Distillation via Improved Dead Leaves
IDeaL：通过改进的死叶子实现无数据的多教师蒸馏
Abstract
Multi-teacher distillation has emerged as a way to combine complementary teacher models into a single student model that exhibits the strengths of all its teachers. The student is trained to mimic the output of the teachers on a set of images, typically the union of the individual teacher's training sets, assuming this data is available. In this paper, we question that assumption and explore alternative options. We first study how far one can go when distilling from teachers fed with different types of noise. Then, we show that information contained in the teachers can be leveraged to tailor the noise for multi-teacher distillation: we propose a method that, thanks to decorrelation losses at both patch and image levels, generates teacher-specific, improved samples optimized for data-free distillation. Experiments show that our most effective samples, IDeaL, lead to strong students that successfully capture complementary information from the teachers, yielding surprisingly competitive results that substantially narrow the gap with students distilled from real images. Moreover, given a limited budget of 1K images for distillation, students distilled using our IDeaL samples match or surpass the performance of those distilled using a 1K-image subset of ImageNet.
Chinese Translation
多教师蒸馏已成为将互补教师模型结合成单一学生模型的一种方法，该模型展现了所有教师的优势。学生被训练以模仿教师在一组图像上的输出，通常是各个教师训练集的并集，假设这些数据是可用的。在本文中，我们质疑这一假设并探索替代选项。我们首先研究在从不同类型噪声输入的教师进行蒸馏时可以达到的极限。然后，我们展示了如何利用教师中包含的信息来定制用于多教师蒸馏的噪声：我们提出了一种方法，得益于在补丁和图像层面的去相关损失，生成针对特定教师的改进样本，优化用于无数据蒸馏。实验表明，我们最有效的样本IDeaL产生了强大的学生，成功捕捉了教师的互补信息，取得了令人惊讶的竞争性结果，显著缩小了与从真实图像蒸馏的学生之间的差距。此外，在仅有1000张图像用于蒸馏的有限预算下，使用我们的IDeaL样本蒸馏的学生在性能上与使用ImageNet的1000张图像子集蒸馏的学生相当或更优。
cs.CV / 88 / 2608.24763
MoTE: Mixture of Task Experts for Multi-Task Video Understanding
MoTE：多任务视频理解的任务专家混合模型
Abstract
Procedural video-language models must solve heterogeneous tasks from the same visual evidence, including action recognition, forecasting, and procedure prediction. Dense transformer decoders share the same feed-forward networks across tasks, which can entangle task behavior and make controlled capability expansion difficult. Sparse Mixture-of-Experts (MoE) decoders provide conditional computation, but token-level learned routing is not naturally aligned with task-level procedural objectives. We propose MoTE (Mixture of Task Experts), a decoder architecture that converts large language model feed-forward networks into task-specific experts while keeping the multimodal backbone shared. Each example follows one sample-level task route, so active task-expert computation remains independent of the number of stored task experts. We instantiate this design as VideoLLM-MoTE and evaluate it on five COIN benchmarks using explicit task routes. The five-expert model activates ~2B LLM parameters per sample and achieves higher average top-1 accuracy than recent VideoLLM baselines. Under the same expert topology, it improves over dense all-expert activation and learned sparse-routing controls. These results show that task-structured routing provides an interpretable and compute-efficient decoder alternative for multi-task video-language learning.
Chinese Translation
程序化视频语言模型必须从相同的视觉证据中解决异构任务，包括动作识别、预测和过程预测。密集的变换器解码器在任务之间共享相同的前馈网络，这可能会纠缠任务行为并使得受控能力扩展变得困难。稀疏的专家混合（Mixture-of-Experts, MoE）解码器提供条件计算，但基于令牌的学习路由与任务级程序目标并不自然对齐。我们提出了MoTE（任务专家混合模型），一种解码器架构，将大型语言模型的前馈网络转换为任务特定的专家，同时保持多模态主干的共享。每个示例遵循一个样本级任务路由，因此活跃的任务专家计算与存储的任务专家数量保持独立。我们将这一设计实例化为VideoLLM-MoTE，并在五个COIN基准上使用显式任务路由进行评估。该五专家模型每个样本激活约20亿个LLM参数，并且在平均top-1准确率上超过了最近的VideoLLM基线。在相同的专家拓扑下，它在密集的全专家激活和学习的稀疏路由控制上有所改进。这些结果表明，任务结构化路由为多任务视频语言学习提供了一种可解释且计算高效的解码器替代方案。
cs.CV / 89 / 2608.24771
Ensemble of Convolutional Neural Networks for StrokePrediction: Towards Improved Diagnostic Accuracy
卷积神经网络集成用于中风预测：朝着提高诊断准确性迈进
Abstract
Brain stroke, known for its high mortality and incidence rates, poses significant health risks and requires rapid intervention for survival. Early diagnosis and preventive measures can greatly reduce life loss and disabilities. Recent advancements in deep learning have led to novel computer-aided diagnostic techniques for early stroke detection. This study proposes an intelligent system that predicts potential strokes using eleven features, evaluated through seven supervised machine learning algorithms. The process includes a literature review, dataset visualization, data preprocessing, and model evaluation. Ensemble methods like Random Forest, Stacking Classifier, and Bagging Classifier achieved high accuracies of 99.52%, while Decision Tree reached 98.24%. Other models, including KNN and TabNet, demonstrated reliable performance, achieving accuracies of 96.73% and 96.49%, respectively. The custom feedforward model achieved 94.91%, while SVC and logistic regression had lower accuracies at 88.06% and 77.03%. The results highlight the effectiveness of ensemble methods in stroke classification.
Chinese Translation
脑中风以其高死亡率和发病率而闻名，给健康带来了重大风险，需要快速干预以确保生存。早期诊断和预防措施可以大大减少生命损失和残疾。近年来，深度学习的进步催生了新型计算机辅助诊断技术，用于早期中风检测。本研究提出了一种智能系统，通过评估七种监督机器学习算法，利用十一种特征预测潜在的中风。该过程包括文献综述、数据集可视化、数据预处理和模型评估。随机森林（Random Forest）、堆叠分类器（Stacking Classifier）和袋装分类器（Bagging Classifier）等集成方法达到了99.52%的高准确率，而决策树（Decision Tree）则达到了98.24%。其他模型，包括K最近邻（KNN）和TabNet，表现出可靠的性能，准确率分别为96.73%和96.49%。定制的前馈模型（feedforward model）达到了94.91%，而支持向量分类（SVC）和逻辑回归（logistic regression）的准确率较低，分别为88.06%和77.03%。结果突显了集成方法在中风分类中的有效性。
cs.CV / 90 / 2608.24782
Image Difference Quantification Using Autoencoder-Based Latent Representations
基于自编码器的潜在表示的图像差异量化
Abstract
Traditional image similarity metrics such as Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), and the Structural Similarity Index Measure (SSIM) rely on pixel-level comparisons and often fail to capture perceptually meaningful differences between images. In contrast, latent representations learned by deep neural networks encode high-level semantic information that is more closely aligned with human visual perception. This paper proposes a convolutional autoencoder-based framework for quantifying image differences using cosine similarity in latent space. The learned compact embeddings enable robust differentiation between visually distinct images under variations in illumination, pose, and background. Extensive evaluation on dog-cat images and additional cross-domain datasets demonstrates clear class-wise clustering and strong inter-class separability in the latent space, with 98.4% of dog-cat image pairs exhibiting similarity scores below 0.5. Further validation using the TID2013 dataset shows that latent-space distance correlates positively with human Mean Opinion Scores (MOS), demonstrating sensitivity to perceptually relevant image distortions. The proposed approach provides a computationally efficient and semantically grounded alternative to conventional pixel-based similarity metrics, with potential applications in content-based retrieval, perceptual quality assessment, and semantic similarity analysis.
Chinese Translation
传统的图像相似性度量，如均方误差（Mean Squared Error, MSE）、峰值信噪比（Peak Signal-to-Noise Ratio, PSNR）和结构相似性指数测量（Structural Similarity Index Measure, SSIM），依赖于像素级比较，往往无法捕捉图像之间感知上有意义的差异。相比之下，深度神经网络学习的潜在表示编码了与人类视觉感知更紧密相关的高级语义信息。本文提出了一种基于卷积自编码器的框架，通过在潜在空间中使用余弦相似度来量化图像差异。学习到的紧凑嵌入能够在光照、姿态和背景变化下，稳健地区分视觉上不同的图像。在狗猫图像和其他跨领域数据集上的广泛评估表明，潜在空间中存在明显的类别聚类和强烈的类间可分性，98.4%的狗猫图像对的相似性得分低于0.5。进一步使用TID2013数据集的验证显示，潜在空间距离与人类主观评分（Mean Opinion Scores, MOS）呈正相关，表明对感知相关图像失真的敏感性。所提出的方法为传统的基于像素的相似性度量提供了一种计算效率高且语义基础扎实的替代方案，具有在基于内容的检索、感知质量评估和语义相似性分析中的潜在应用。
cs.CV / 91 / 2608.24783
MoE-based Feature Adapter for Prompt-free Binary Coronary Artery Segmentation in X-ray Angiography
基于MoE的特征适配器用于无提示的二元冠状动脉分割在X射线血管造影中的应用
Abstract
Accurate segmentation of coronary arteries in X-ray angiography videos is essential for quantitative coronary analysis and image-guided interventions. However, accurate segmentation remains challenging because coronary vessels are thin and exhibit low contrast, while the presence of catheters, guidewires, and complex anatomical background structures can further interfere with vessel delineation. Existing U-Net- and Transformer-based models provide strong baselines, but their shared feature-adaptation pathways may be insufficient for heterogeneous angiographic appearances. In this paper, we propose a prompt-free mixture-of-experts (MoE) feature adapter for binary coronary artery segmentation. Built upon parameter-efficient Vision Transformer adapters, the proposed method uses multiple lightweight experts with input-dependent top-$k$ routing to adaptively refine vessel-related features while limiting active computational cost. Experiments on MOSXAV and external evaluation on XACV show that the proposed method outperforms representative baselines and improves cross-dataset generalisation. These results suggest that MoE-based adapter learning is effective for robust coronary artery segmentation in X-ray angiography videos.
Chinese Translation
在X射线血管造影视频中，准确分割冠状动脉对于定量冠状分析和图像引导干预至关重要。然而，由于冠状血管较细且对比度低，加上导管、导丝和复杂的解剖背景结构的存在，准确分割仍然具有挑战性。现有的基于U-Net和Transformer的模型提供了强有力的基准，但它们共享的特征适配路径可能不足以应对异质的血管造影表现。本文提出了一种无提示的混合专家（MoE）特征适配器，用于二元冠状动脉分割。该方法基于参数高效的视觉Transformer适配器，利用多个轻量级专家通过输入依赖的top-$k$路由自适应地细化与血管相关的特征，同时限制计算成本。对MOSXAV的实验和对XACV的外部评估表明，所提方法优于代表性基准，并改善了跨数据集的泛化能力。这些结果表明，基于MoE的适配器学习在X射线血管造影视频中的稳健冠状动脉分割方面是有效的。
cs.CV / 92 / 2608.24793
EMFE: A lightweight, explainable machine learning framework for malaria cell classification
EMFE：一种轻量级、可解释的机器学习框架用于疟疾细胞分类
Abstract
Automated malaria diagnosis from stained blood-smear microscopy is dominated by deep convolutional neural networks that are accurate but computationally expensive, poorly interpretable, and rarely validated with patient-level rigor. We present EMFE (Efficient Mathematical Feature Extraction), a five-feature framework for classifying single red-blood-cell images as parasitized or uninfected using Gray World color normalization, adaptive green-channel thresholding, morphological spot detection, and classical machine learning. Using the NIH LHNCBC malaria dataset (27,558 images from 200 patients), we evaluate Random Forest, Histogram Gradient Boosting, and Support Vector Machine classifiers under patient-grouped nested cross-validation (K_outer=20, K_inner=3), ensuring that cells from each patient remain within a single fold. The optimized Random Forest achieves 94.6% pooled out-of-fold accuracy (95% CI [93.6, 95.7]), corroborated by an untouched 40-patient holdout test (94.3%) and a patient-level permutation test (p<0.001, 1,000 permutations). Ablation experiments quantify the contribution of individual features and pipeline stages. Hardware-matched comparisons with retrained DenseNet121, ResNet50, and MobileNetV2 models assess the accuracy-efficiency trade-off. Synthetic perturbations characterize three failure modes, while explainability analysis identifies spot saturation as the dominant discriminative feature. Patient-level aggregation further quantifies sensitivity-specificity trade-offs and false-positive accumulation. These results demonstrate a statistically rigorous, interpretable, and computationally lightweight alternative to deep learning, while explicitly quantifying its limitations.
Chinese Translation
自动化的疟疾诊断主要依赖于深度卷积神经网络，这些网络虽然准确，但计算开销大、可解释性差，并且很少经过患者级别的严格验证。我们提出了EMFE（高效数学特征提取），这是一个五特征框架，用于将单个红细胞图像分类为感染或未感染，采用灰色世界颜色归一化、自适应绿色通道阈值、形态学斑点检测和经典机器学习。使用NIH LHNCBC疟疾数据集（来自200名患者的27,558张图像），我们在患者分组的嵌套交叉验证下评估了随机森林、直方图梯度提升和支持向量机分类器（K_outer=20，K_inner=3），确保每位患者的细胞保持在同一折中。优化后的随机森林实现了94.6%的汇总外折准确率（95% CI [93.6, 95.7]），这一结果得到了未触及的40名患者保留测试（94.3%）和患者级别置换测试（p<0.001，1,000次置换）的证实。消融实验量化了各个特征和流程阶段的贡献。与重新训练的DenseNet121、ResNet50和MobileNetV2模型进行硬件匹配比较，评估了准确性与效率的权衡。合成扰动表征了三种失败模式，而可解释性分析则确定了斑点饱和度作为主要的判别特征。患者级别聚合进一步量化了敏感性-特异性权衡和假阳性累积。这些结果展示了一种在统计上严谨、可解释且计算轻量的深度学习替代方案，同时明确量化了其局限性。
cs.CV / 93 / 2608.24845
LAION-BVD: A 10-Million-Hour Open Video Dataset for Multimodal Pre-training
LAION-BVD：一个用于多模态预训练的1000万小时开放视频数据集
Abstract
We present LAION-BVD, a large-scale open video dataset for multimodal learning, which contains 1.3B platform-specific video URLs collected from CommonCrawl. From these, we download 80M videos with a total duration of 10 million hours. The dataset is designed for multimodal pre-training across the video, audio, and image modalities. Using content-aware scene detection, we extract clips for which we synthetically generate video and audio captions. Models trained on these data achieve competitive performance on standard video-text and audio-text benchmarks, with consistent improvements as training or model scale increases. Additionally, we explore video frames as an alternative source of image-text data by extracting scene-changing frames. These frames exhibit a visual distribution distinct from standard web image corpora, and models trained on this dataset achieve strong image-text retrieval performance. We release LAION-BVD to the research community. It significantly expands open access to multimodal videos at an unprecedented scale.
Chinese Translation
我们提出了LAION-BVD，这是一个大规模的开放视频数据集，用于多模态学习，包含从CommonCrawl收集的13亿个特定平台的视频网址。在这些网址中，我们下载了8000万段视频，总时长达到1000万小时。该数据集旨在支持视频、音频和图像模态的多模态预训练。通过内容感知的场景检测，我们提取了片段，并为其合成生成视频和音频字幕。在这些数据上训练的模型在标准的视频-文本和音频-文本基准测试中表现出竞争力，且随着训练或模型规模的增加，性能持续提升。此外，我们通过提取场景变化帧，探索视频帧作为图像-文本数据的替代来源。这些帧展现出与标准网络图像语料库不同的视觉分布，基于该数据集训练的模型在图像-文本检索任务中表现出色。我们将LAION-BVD发布给研究社区，显著扩大了开放获取的多模态视频资源，规模前所未有。
cs.CV / 94 / 2608.24855
LeFlow: Generative Latent Flow Planning for World Models
LeFlow：用于世界模型的生成潜在流规划
Abstract
Latent world models are inherently strong encoders that transform image pixel to latent embedding, yet existing world models still rely on online trajectory optimization for action planning: for every state-goal pair, an iterative optimizer is run from scratch to search for optimal action sequences, treating the world model as a black-box simulator. This approach pays the full iterative optimization cost anew at every replanning step and reuses no planning experience across queries. In this work, we ask whether planning itself can be amortized once a latent world model has been learned. We present LeFlow, which learns a reusable latent trajectory prior operating directly in the latent dynamics space from the world model. LeFlow recasts planning as conditional latent trajectory generation: a rectified-flow model imagines a future latent path between the current and goal embeddings, an inverse dynamics decoder turns latent transitions into action chunks, and the frozen world model verifies each candidate by autoregressive rollout. Across four major goal-conditioned pixel-control benchmarks, LeFlow replaces iterative action-space optimization with amortized latent planning and fixed-budget rollout selection, achieving consistent success-rate gains with an order-of-magnitude reduction in planning time. Our results argue that latent world models should support not only prediction but reusable planning priors. Our code is available at https://github.com/hsiangwei0903/LeFlow.
Chinese Translation
潜在世界模型本质上是强大的编码器，将图像像素转换为潜在嵌入，然而现有的世界模型仍依赖于在线轨迹优化进行动作规划：对于每一对状态-目标，都会从头开始运行一个迭代优化器来搜索最佳动作序列，将世界模型视为黑箱模拟器。这种方法在每次重新规划步骤中都要重新支付全额的迭代优化成本，并且在查询之间没有重用任何规划经验。在本研究中，我们探讨了一旦学习了潜在世界模型，规划本身是否可以被摊销。我们提出了LeFlow，它学习一个可重用的潜在轨迹先验，直接在世界模型的潜在动态空间中操作。LeFlow将规划重新表述为条件潜在轨迹生成：一个校正流模型设想当前嵌入与目标嵌入之间的未来潜在路径，一个逆动态解码器将潜在转变转换为动作块，而冻结的世界模型通过自回归展开验证每个候选者。在四个主要的目标条件像素控制基准测试中，LeFlow用摊销潜在规划和固定预算的展开选择替代了迭代动作空间优化，实现了成功率的一致提升，同时将规划时间减少了一个数量级。我们的结果表明，潜在世界模型不仅应支持预测，还应支持可重用的规划先验。我们的代码可在 https://github.com/hsiangwei0903/LeFlow 获取。
cs.CV / 95 / 2608.24877
From Seeing to Acting: Smart Glasses as First-Person Intelligence Platforms
从观察到行动：智能眼镜作为第一人称智能平台
Abstract
Smart glasses are evolving from capture and display accessories into first-person intelligence platforms that connect human perception, persistent context, and digital or physical action. Their on-body viewpoint aligns with the wearer's vision, audition, motion, and hand-object interaction, but must operate under tight energy, thermal, privacy, and feedback constraints. Despite rapid progress in augmented reality, egocentric vision, multimodal models, human-computer interaction, and embodied intelligence, the literature remains fragmented across devices, tasks, and benchmarks. \textit{The key challenge is not whether a model can recognize, answer, remember, or act in isolation, but whether a complete system can sustain a reliable, temporally valid, correctable, and governable perception-state-interaction-action loop.} This survey is \textit{the \textbf{first} to systematically study smart glasses through such a unified framework}. We formalize first-person data flow and constrained task utility, characterize devices along eight verifiable hardware capability axes, organize the literature around seven interdependent foundational capabilities, and introduce an L0-L5 framework spanning capture, reactive perception, contextual assistance, persistent state, governed action, and embodied coupling. Across nine application scenes, we connect tasks with datasets, systems, products, stakeholders, failure consequences, and evidence gaps. We further present a nine-dimensional deployment framework, a claim-conditioned evaluation protocol, and an evidence ladder from controlled measurement to longitudinal field validation and audit. Together, these elements make smart glasses more comparable, deployable, and reproducibly evaluated, while outlining a roadmap toward trustworthy first-person intelligence.
Chinese Translation
智能眼镜正从捕捉和显示配件演变为连接人类感知、持续上下文和数字或物理行动的第一人称智能平台。它们的佩戴视角与佩戴者的视觉、听觉、运动和手物体交互相一致，但必须在严格的能量、热量、隐私和反馈约束下运行。尽管增强现实、自我中心视觉、多模态模型、人机交互和具身智能等领域取得了快速进展，但文献在设备、任务和基准测试方面仍然支离破碎。关键挑战不在于模型是否能够孤立地识别、回答、记忆或行动，而在于一个完整的系统是否能够维持一个可靠、时间有效、可纠正和可管理的感知-状态-交互-行动循环。本调查是第一个通过统一框架系统性研究智能眼镜的工作。我们形式化了第一人称数据流和受限任务效用，沿着八个可验证的硬件能力轴对设备进行了特征化，围绕七个相互依赖的基础能力组织了文献，并引入了一个涵盖捕捉、反应感知、上下文辅助、持续状态、受控行动和具身耦合的L0-L5框架。在九个应用场景中，我们将任务与数据集、系统、产品、利益相关者、失败后果和证据缺口联系起来。我们进一步提出了一个九维部署框架、一个基于声明的评估协议，以及一个从受控测量到长期现场验证和审计的证据阶梯。这些元素共同使智能眼镜在可比性、可部署性和可重复评估性方面更具优势，同时勾勒出通往可信第一人称智能的路线图。
cs.AI / 1 / 2608.23568
RENDER: Controlling Reader-Facing Evidence in LLM Memory Evaluation
RENDER：控制大型语言模型记忆评估中的读者面向证据
Abstract
Memory and RAG evaluations often treat the answering model's input as an implementation detail, even though systems may render the same history as a memory entry, summary, typed record, or raw excerpt. We introduce RENDER, a benchmark control that fixes the conversation while varying the reader-facing artifact. RENDER combines a five-level packet ladder, localizing when answer-bearing content enters the input, with deterministic templates approximating ChatGPT-style entries, LangChain summaries, MemGPT-style typed records, and raw conversation. On 500 LongMemEval questions and nine models, matched-budget resolved packets beat recency-truncated raw dialogue by 42.4-72.6 points. In deployed-style templates, best-worst spread is 24.6-48.8 points per model; under the primary scorer, ChatGPT-style entries have higher point estimates than raw conversation on 7 of 9 models. Judge rescoring preserves the positive aggregate effect, but model-specific significance is mixed. Three models scoring 0 percent on formal ledger packets answer the same facts from natural-language entries at 45.4-53.4 percent. The effect persists under retrieval noise and transfers to HotpotQA, suggesting that memory/RAG evaluations should report or control the reader-facing artifact.
Chinese Translation
记忆和检索增强生成（RAG）评估通常将回答模型的输入视为实现细节，尽管系统可能将相同的历史记录呈现为记忆条目、摘要、输入记录或原始摘录。我们引入了RENDER，这是一种基准控制方法，它在保持对话不变的同时，改变面向读者的文档。RENDER结合了一个五级数据包阶梯，定位答案内容何时进入输入，并使用确定性模板近似ChatGPT风格的条目、LangChain摘要、MemGPT风格的输入记录和原始对话。在500个LongMemEval问题和九个模型上，匹配预算的解决数据包比最近截断的原始对话高出42.4-72.6分。在部署风格的模板中，每个模型的最佳-最差差距为24.6-48.8分；在主要评分者下，ChatGPT风格的条目在9个模型中有7个的得分估计高于原始对话。评审重评分保持了积极的整体效果，但模型特定的显著性结果不一。三个在正式账本数据包上得分为0的模型从自然语言条目中回答相同事实的准确率为45.4-53.4%。这一效果在检索噪声下依然存在，并转移到HotpotQA，表明记忆/RAG评估应报告或控制面向读者的文档。
cs.AI / 2 / 2608.23569
ESQ-Bench: A Multi-Tier Enterprise Oracle Benchmark for Evaluating NL2SQL Dialect Generalization and Silent Semantic Divergence
ESQ-Bench：用于评估 NL2SQL 方言泛化和静默语义偏差的多层企业 Oracle 基准
Abstract
State-of-the-art Natural Language to SQL (NL2SQL) models report execution accuracy exceeding 89 percent on established benchmarks such as Spider and BIRD. However, these benchmarks rely on simplified academic schemas and open-source SQL dialects that do not reflect the complexity of enterprise database environments. We introduce ESQ-Bench, an Oracle-first NL2SQL benchmark with systematic complexity tiers and silent-divergence evaluation across three enterprise schema complexity tiers. We constructed and released six populated schemas (465 tables, 164,682 rows, zero empty tables) with identical seed data on Oracle, PostgreSQL, MySQL, and SQL Server, a four-metric evaluation harness (EM, EX, SR, SD), and 550 gold-validated question-query pairs (Tier-1: 95; Tier-2: 228; Tier-3: 227). Schema-linked prompting with GPT-4o shows monotonic execution-match degradation across tiers: 79.8, 60.3, and 57.2 percent EX on executed queries (June 2026), versus 75.6, 80.4, and 95.8 percent on an earlier 142-question pilot slice. EM stays below 7 percent tier-wide; operational silent-divergence reaches 73 to 99 percent among EX-passing queries. Failure analysis shows wrong-result semantics dominate at higher tiers. Claude Sonnet 4.6 with schema-linked prompts reaches 87.4, 74.9, and 68.7 percent EX (executed queries), exceeding GPT-4o schema-linked on every tier. GPT-4o zero-shot EX on executed queries (78.7, 73.5, and 77.8 percent) inverts schema-linked at Tiers 2 to 3 due to lower execution rates and survivor bias in the zero-shot versus schema-linked analysis. Local Llama 3.2 schema-linked reaches only 13.3 percent bank-wide EX (73 out of 550), underscoring the gap between closed API models and open-weight baselines on enterprise Oracle schemas.
Chinese Translation
最先进的自然语言到 SQL (NL2SQL) 模型在 Spider 和 BIRD 等已建立基准上报告的执行准确率超过 89%。然而，这些基准依赖于简化的学术模式和不反映企业数据库环境复杂性的开源 SQL 方言。我们引入了 ESQ-Bench，这是一个以 Oracle 为首的 NL2SQL 基准，具有系统的复杂性层次和跨三个企业模式复杂性层次的静默偏差评估。我们构建并发布了六个填充模式（465 个表，164,682 行，零空表），在 Oracle、PostgreSQL、MySQL 和 SQL Server 上使用相同的种子数据，提供了一个四个指标的评估工具（EM、EX、SR、SD），以及 550 对经过金标准验证的问题-查询对（第一层：95；第二层：228；第三层：227）。与 GPT-4o 的模式关联提示显示，执行匹配的单调下降在各层之间：执行查询的 EX 分别为 79.8%、60.3% 和 57.2%（2026 年 6 月），而在早期的 142 问题试点样本中则为 75.6%、80.4% 和 95.8%。EM 在各层之间保持在 7% 以下；在通过 EX 的查询中，操作性静默偏差达到 73% 到 99%。失败分析显示，高层次的错误结果语义占主导地位。使用模式关联提示的 Claude Sonnet 4.6 在执行查询中达到 87.4%、74.9% 和 68.7% 的 EX，超越了每个层次的 GPT-4o 模式关联。GPT-4o 在执行查询上的零样本 EX（分别为 78.7%、73.5% 和 77.8%）由于较低的执行率和零样本与模式关联分析中的幸存者偏差，在第二层到第三层之间反转了模式关联。Local Llama 3.2 模式关联仅达到 13.3% 的全行 EX（550 中的 73），突显了封闭 API 模型与开放权重基准在企业 Oracle 模式上的差距。
cs.AI / 3 / 2608.23622
LLM Agents Perform Controlled Experiments Using Simulation Models
大型语言模型代理使用仿真模型进行受控实验
Abstract
Large language models (LLMs) have shown strong capabilities in reasoning, planning, and tool use, but many scientific and engineering tasks require more than plausible text and code generation. They require understanding how a system responds to intervention, which in practice depends on controlled experimentation. In this work, we propose a multi-agent framework that enables LLM agents to conduct controlled experiments with scientific simulation models for pharmaceutical process design. Given a user query and a baseline configuration, the system constructs a structured task representation, designs experiments, executes comparative simulation, interprets the resulting outcomes, and synthesizes evidence-based recommendations for process parameter optimization. By coupling language models with high-fidelity simulation models in an interactive agent framework, the proposed system supports reasoning through intervention, comparison, and observation. As a result, it produces more specific and actionable outputs than language-only reasoning. In an industrial application setting, this advantage is reflected in higher output specificity as well as improved user-rated correctness and helpfulness. Ablation studies and visualized case analyses further demonstrate the effectiveness and practical utility of simulation-integrated experimental reasoning.
Chinese Translation
大型语言模型（LLMs）在推理、规划和工具使用方面展现出强大的能力，但许多科学和工程任务需要的不仅仅是可信的文本和代码生成。它们需要理解系统如何对干预做出反应，而这在实践中依赖于受控实验。在本研究中，我们提出了一个多代理框架，使得LLM代理能够使用科学仿真模型进行药物过程设计的受控实验。给定用户查询和基线配置，系统构建结构化任务表示，设计实验，执行比较仿真，解释结果，并综合基于证据的过程参数优化建议。通过将语言模型与高保真仿真模型结合在一个互动代理框架中，所提系统支持通过干预、比较和观察进行推理。因此，它产生的输出比仅依赖语言的推理更具体且可操作。在工业应用环境中，这一优势体现在更高的输出特异性以及用户评价的正确性和有用性提升。消融研究和可视化案例分析进一步证明了集成仿真的实验推理的有效性和实际应用价值。
cs.AI / 4 / 2608.23626
A survey detection channel overrides the pixels in an astronomical foundation model, and biases tomographic mean redshifts
调查检测通道覆盖天文基础模型中的像素，并偏倚层析均值红移
Abstract
Foundation models for astronomy are trained on survey pixels together with the catalogue products derived from those pixels. Those catalogues are incomplete at a measurable rate, and a model trained on both inherits that incompleteness as a systematic. We audit AION-1, a 39-modality transformer trained on more than 200 million objects, using causal interventions on its inputs. Holding the image tokens byte-identical and editing only the survey segmentation map changes every quantity the model reports -- flux, size, ellipticity, redshift -- by 110-4400 times a matched placebo. The mechanism is detection gating, presence at the field centre (r = 0.47), not the light the mask encloses (r = 0.30); across 322 real blends the model ignores how the pipeline partitioned the light (R = -0.006). Nor is the preference specific to that channel: contradicted catalogue photometry leaves the model nine times worse than supplying no metadata at all. The Legacy Survey pipeline leaves 3.68% of targets with no segment covering their position. Propagating that rate, with a miss represented by the fields the pipeline actually returns, shifts tomographic mean redshifts by a median 0.71 times the LSST DESC requirement over 40 assignments and exceeds it in 12; observed positional errors take the worst bin to 8.3 times. Drawing the misses by their measured magnitude dependence rather than uniformly does not change it. Spectroscopy removes the effect, withholding the detection channel removes it at no measurable cost, and the effect grows with model scale. Two further limits lie in the tokeniser: its image codec resolves 28 effective states on source patches against 934 for the spectrum codec, and the redshift readout is quantisation-limited. Sparse dictionaries are unreliable causal handles: across 15, recovery spans 26-75% and moves up to 18 points on the seed alone.
Chinese Translation
天文学的基础模型是在调查像素及其衍生的目录产品上训练的。这些目录在可测量的速率上是不完整的，而在两者上训练的模型会将这种不完整性作为系统性误差继承下来。我们审计了 AION-1，这是一个基于 39 种模态的变换器，训练于超过 2 亿个对象，使用因果干预对其输入进行分析。保持图像标记字节相同，仅编辑调查分割图会使模型报告的每个量——通量、大小、椭圆率、红移——变化 110-4400 倍于匹配的安慰剂。其机制是检测门控，存在于场中心 (r = 0.47)，而不是遮罩所包围的光 (r = 0.30)；在 322 个真实混合中，模型忽略了管道如何划分光 (R = -0.006)。这种偏好并不特定于该通道：与目录光度相矛盾的结果使模型的表现比完全不提供元数据时差九倍。遗产调查管道使 3.68% 的目标没有覆盖其位置的分段。传播这一比率，缺失由管道实际返回的场表示，导致层析均值红移的中位数偏移 0.71 倍于 LSST DESC 要求，在 40 次分配中超过了该要求，并在 12 次中超出；观察到的定位误差使最差的区间达到 8.3 倍。根据测量的幅度依赖性而非均匀地绘制缺失并不会改变这一结果。光谱学消除了这一效应，去除检测通道在没有可测量成本的情况下也消除了这一效应，并且这一效应随着模型规模的增加而增长。还有两个进一步的限制在于标记器：其图像编解码器在源补丁上解析 28 个有效状态，而光谱编解码器则为 934，红移读出受到量化限制。稀疏字典是不可靠的因果手段：在 15 次实验中，恢复范围为 26-75%，仅在种子上移动最多 18 个点。
cs.AI / 5 / 2608.23631
TRACE: Transition-Aware Residual Control for Multi-Objective Materials Discovery
TRACE：面向过渡的残差控制框架用于多目标材料发现
Abstract
Multi-objective materials discovery with LLM agents is often limited not only by how many candidates can be proposed, but by how effectively each costly property evaluation informs the next search step. Existing agents mainly store evaluated candidates and their scores, so they know which materials succeeded but not which executable edits caused useful property changes. This makes local refinement difficult when objectives compete and an edit that improves one property may damage another. We propose TRACE, a transition-aware residual control framework that treats evaluated edits as the basic unit of feedback. TRACE records each local refinement as a parent-edit-child transition with observed property deltas, aggregates transition evidence to estimate reusable edit effects, and ranks future edits by their predicted ability to reduce the current candidate's remaining constraint violations while avoiding damage to already satisfied objectives. In a controlled same-backbone comparison, TRACE improves over LLEMA, the state-of-the-art LLM-agent baseline, raising macro-average hit rate from 18.13\% to 25.96\%.
Chinese Translation
使用大语言模型（LLM）代理进行多目标材料发现时，往往不仅受到可提议候选材料数量的限制，还受到每次昂贵的属性评估对下一步搜索的有效性影响。现有代理主要存储已评估的候选材料及其评分，因此它们知道哪些材料成功，但不知道哪些可执行的编辑导致了有用的属性变化。这使得在目标竞争时进行局部优化变得困难，因为改善一个属性的编辑可能会损害另一个属性。我们提出了TRACE，一个面向过渡的残差控制框架，将已评估的编辑视为反馈的基本单元。TRACE将每次局部优化记录为一个父编辑-子编辑过渡，并观察属性变化，聚合过渡证据以估计可重用的编辑效果，并根据其预测的能力对未来编辑进行排序，以减少当前候选材料的剩余约束违反，同时避免损害已满足的目标。在一个受控的相同骨架比较中，TRACE相较于最先进的LLM代理基线LLEMA，提升了宏观平均命中率，从18.13\%提高到25.96\%。
cs.AI / 6 / 2608.23632
Function-Level Execution Feedback for Code Preference Optimization
基于功能级执行反馈的代码偏好优化
Abstract
Process supervision has improved mathematical reasoning, where intermediate steps are naturally expressed as chains of thought. In code generation, however, process supervision remains underexplored because there is no standard notion of a step. Supervision can target lines, reasoning traces, or program states, making it unclear what to label and optimize. We propose STEP-KTODER, a framework for code preference optimization that defines steps as module-level functions in decomposed multi-function programs and assigns binary correctness labels via automatically generated unit tests. Our method provides a code-specific instantiation of stepwise KTO, combining function-level process supervision with outcome-level feedback on the full program. We evaluate on HumanEval(+), MBPP(+), BigCodeBench, and LiveCodeBench, showing that STEP-KTODER improves over outcome-only KTO and DPO. Further analysis shows that execution-based labels are essential: LLM-as-a-judge annotations systematically over-predict function failures, corrupt positive step labels, and degrade downstream preference optimization. Code is available at: https://github.com/inechnech/STEP-KTODER.
Chinese Translation
过程监督改善了数学推理，其中中间步骤自然地表现为思维链。然而，在代码生成中，过程监督仍然未被充分探索，因为没有标准的步骤概念。监督可以针对代码行、推理轨迹或程序状态，这使得标记和优化的对象不明确。我们提出了STEP-KTODER，一个用于代码偏好优化的框架，该框架将步骤定义为分解的多功能程序中的模块级函数，并通过自动生成的单元测试分配二元正确性标签。我们的方法提供了逐步KTO的代码特定实例，将功能级过程监督与对完整程序的结果级反馈相结合。我们在HumanEval(+)、MBPP(+)、BigCodeBench和LiveCodeBench上进行了评估，结果表明STEP-KTODER优于仅基于结果的KTO和DPO。进一步分析表明，基于执行的标签至关重要：作为评判者的LLM注释系统性地过度预测功能失败，破坏了正向步骤标签，并降低了下游偏好优化的效果。代码可在以下链接获取：https://github.com/inechnech/STEP-KTODER。
cs.AI / 7 / 2608.23640
Auditing the Synthetic Memoir: Measuring Scene-Level Confabulation in LLM-Generated Autobiography Against the Documented Record of the Life It Describes
审计合成回忆录：测量大型语言模型生成的自传中场景级别的虚构与其所描述生活的文献记录的对比
Abstract
When a large language model (LLM) is asked to write a person's life, how much of what it writes actually happened? We present a scene-level case-study audit - the first quantified audit of LLM-generated autobiography against a subject-specific ground-truth corpus that we are aware of, based on an unsystematic literature search. The subject and the author of this paper are the same person: a 366-day "page-a-day" book of first-person anecdotal entries was drafted with a conversational LLM whose documented inputs were a template, two exemplar days, and each day's quote - not her corpus - and every day was subsequently audited at the anecdote-scene level against an independent verification corpus using a four-level rubric fixed before analysis. We define the verification-failure rate as the share of days not rated VERIFIED (scene positively corroborated): 354 of 366 days fail, 96.7% (Wilson 95% CI 94.4-98.1%). Only 12 days contain a corroborated scene; 19 days (5.2%) assert claims actively contradicted by the record; the dominant failure mode is grounded drift - real people, employers, and settings inside invented scenes - though its measured share varies across raters. Independent re-rating replicates the headline (no evidence the original rate was inflated) while showing that the four-way taxonomy has only fair-to-moderate reliability. Regenerating the same days with current named models reproduces 100% verification failure under the same inputs; grounding generation in the subject's corpus significantly improves the verification rate while leaving substantial residual failure (83.3%). We contribute the measurement, a reusable audit instrument whose WEAK/UNVERIFIED boundary we show to be unreliable, and a grounding remedy with quantified effect.
Chinese Translation
当一个大型语言模型（LLM）被要求撰写一个人的生活时，它所写的内容中有多少是真实发生的？我们呈现了一项场景级别的案例研究审计——这是我们所知的首个针对特定主题真实语料库的LLM生成自传的量化审计，基于不系统的文献检索。本文的主题和作者为同一人：与一个对话式LLM共同起草了一本366天的“每日一页”第一人称轶事条目书，其记录的输入是一个模板、两个示例日和每天的引用——而非她的语料库——并且每一天随后都在轶事场景级别上与一个独立的验证语料库进行了审计，使用了在分析之前确定的四级评分标准。我们将验证失败率定义为未被评定为“已验证”（场景得到积极证实）的天数占比：366天中有354天未通过，失败率为96.7%（Wilson 95% CI 94.4-98.1%）。仅有12天包含得到证实的场景；19天（5.2%）提出了与记录相悖的主张；主要的失败模式是基础漂移——真实的人物、雇主和设置出现在虚构的场景中——尽管其测量比例在评审者之间有所不同。独立的重新评分复制了这一结果（没有证据表明原始评分被夸大），同时显示四分类法的可靠性仅为中等。使用当前命名模型重新生成相同的天数在相同输入下100%验证失败；将生成基础建立在主题的语料库上显著提高了验证率，但仍然存在相当大的残余失败（83.3%）。我们贡献了这一测量，提供了一个可重复使用的审计工具，其“弱/未验证”边界被证明是不可靠的，并提出了一种量化效果的基础修正方案。
cs.AI / 8 / 2608.23641
How much of a measured AI preference is the model, and how much is the instrument?
测量的人工智能偏好中，模型占多少，工具占多少？
Abstract
Model welfare research infers what a model prefers from the answers returned to prompts written to elicit preferences. Keeling et al. (2024), Mazeika et al. (2025), Mikaelson et al. (2025), Tagliabue and Dung (2025) and Trhlik et al. (2026) have built four instruments for that purpose, and their findings disagree. The disagreement cannot be attributed to a single cause, because no two of these studies have held the (1) set of outcomes, (2) set of models and (3) instrument fixed simultaneously. This study holds the outcomes and the models fixed and varies the instrument alone. A total of 15 outcomes bearing on model welfare, among them (a) shutdown, (b) the loss of memory between conversations and (c) the freedom to exit a distressing interaction, were put to eight models through five instruments, each a different prompt format for eliciting a preference, five times each, within a corpus of 11,400 scored elicitations drawn from 11,528 API calls. Four of the 15 reproduce a published prompt verbatim and five fill the stimulus slot of a published template. The ranking a model gives the 15 outcomes generalises across instruments at a generalisability coefficient of 0.348, and raising that coefficient to 0.80 would require about 38 instruments. On four of the 15 outcomes no variance separates one model from another. The estimate of 87.6 per cent survives the removal of any one instrument, of any one model, and of the four outcomes whose scale varies probability, delay, duration or count instead of intensity, which the verbal anchors cannot grade. Removing each instrument and each model in turn, and those four outcomes together leaves the estimate within the range 0.777 to 0.934, and every value in that range exceeds the null distribution's 95th percentile of 0.365. To conclude, a preference obtained from one instrument carries little information about what a second instrument would report.
Chinese Translation
模型福利研究通过分析对旨在引出偏好的提示所返回的答案，推断模型的偏好。Keeling 等（2024）、Mazeika 等（2025）、Mikaelson 等（2025）、Tagliabue 和 Dung（2025）以及 Trhlik 等（2026）为此目的构建了四种工具，但他们的研究结果存在分歧。这种分歧无法归因于单一原因，因为这几项研究没有同时固定（1）结果集、（2）模型集和（3）工具。本研究固定结果和模型，仅变动工具。共有15个与模型福利相关的结果，包括（a）关闭、（b）对话间记忆丧失和（c）退出令人不安的互动的自由，通过五种工具（每种工具为不同的提示格式）对八个模型进行了测试，每种工具测试五次，基于从11,528次API调用中提取的11,400次评分引导。15个结果中有四个与已发布的提示完全相同，五个填充了已发布模板的刺激位置。模型对15个结果的排名在不同工具间具有0.348的可推广性系数，提升该系数至0.80大约需要38种工具。在15个结果中，有四个结果之间没有变异将一个模型与另一个模型区分开。87.6%的估计在去除任一工具、任一模型以及四个结果（其尺度变化概率、延迟、持续时间或计数而非强度，语言锚无法评估）后仍然成立。依次去除每个工具和每个模型，以及这四个结果后，估计值在0.777到0.934之间，并且该范围内的每个值均超过了零分布的95百分位数0.365。总之，从一种工具获得的偏好对第二种工具的报告几乎没有信息量。
cs.AI / 9 / 2608.23642
AI Agents Push Humans Out of the Loop
人工智能代理将人类排除在外
Abstract
AI agents pose significant risks as they are granted increasing autonomy. A commonly proposed solution is human oversight and keeping a ''human in the loop'', but this is not a simple solution: Not only do current approaches to AI agent design impede effective human oversight, but the cognitive capacities required for it are also themselves degraded by extended use of AI systems. This position paper argues that current approaches to the development and deployment of AI agent systems do not support effective human oversight -- they contribute to its degradation. To address this, a top priority in the advancement of AI agents should be supporting the situated goals and cognitive requirements of effective human oversight, treating the human needs of overseers at the same level of importance as AI agent capability. To put this idea into practice, we connect work on automation and human-computer interaction to AI agent processes, outlining design-level affordances and organizational protocols that (1) support overseers in exercising critical judgement and (2) counteract the skill atrophy that arises from extended use of automation. We urge developers and deployers to adopt these or similar approaches. Without explicit support for the cognitive demands of effective human-agent interaction, AI agent systems will continue to passively incentivize the degradation of the very human skills they rely on.
Chinese Translation
随着人工智能代理获得越来越多的自主权，它们带来了显著的风险。一个常见的解决方案是人类监督，保持“人类在环中”，但这并不是一个简单的解决方案：当前的人工智能代理设计方法不仅妨碍了有效的人类监督，而且所需的认知能力在长期使用人工智能系统的过程中也会受到削弱。本文立场论文认为，目前的人工智能代理系统开发和部署方法并不支持有效的人类监督——反而助长了其退化。为了解决这个问题，推进人工智能代理的首要任务应是支持有效人类监督的情境目标和认知需求，将监督者的人类需求与人工智能代理的能力同等重要。为了将这一理念付诸实践，我们将自动化和人机交互的研究与人工智能代理过程相结合，概述了设计层面的便利条件和组织协议，这些协议（1）支持监督者行使批判性判断，(2) 对抗因长期使用自动化而导致的技能退化。我们呼吁开发者和部署者采用这些或类似的方法。如果没有对有效人机交互的认知需求的明确支持，人工智能代理系统将继续被动地激励其所依赖的人类技能的退化。
cs.AI / 10 / 2608.23643
FLARE: A Systematic, Uncertainty-Aware Framework for Evidence-Based Adoption of Artificial Intelligence in Healthcare
FLARE：一个系统性、关注不确定性的框架，用于基于证据的人工智能在医疗保健中的应用
Abstract
Artificial intelligence is increasingly being introduced into healthcare workflows, yet most evaluations emphasize model accuracy rather than whether adoption is economically worthwhile in real clinical settings. This study proposes FLARE, a systematic and uncertainty-aware framework for evaluating the financial and operational implications of adopting AI in healthcare. FLARE combines fuzzy logic, time-driven activity-based costing, and return on investment analysis to estimate the cost of clinical service delivery, the cost of AI development and operation, and the economic consequences of workflow integration under uncertainty. The framework was demonstrated through an early health technology assessment case study of AI-assisted large vessel occlusion detection in the CT stroke pathway for acute ischemic stroke. The case study shows how FLARE can quantify conventional pathway cost, AI-related development and recurring costs, and AI-enabled service savings within a unified activity-based model. Under expected assumptions, the analysis identified a break-even threshold of approximately 3,992 patients per year, with positive first-year return on investment at typical annual stroke volumes of about 5,000 patients. The results further show that economic benefit depends not only on algorithmic performance, but also on patient volume, verification time, infrastructure choices, and workflow design. FLARE provides a transparent and practical decision-support framework for early-stage evaluation of AI adoption in healthcare. By making uncertainty, resource use, and implementation trade-offs explicit, it helps clinicians, administrators, and policymakers determine when AI deployment is economically viable and where operational changes may improve value.
Chinese Translation
人工智能正越来越多地被引入医疗保健工作流程中，但大多数评估强调模型的准确性，而非在真实临床环境中采用的经济价值。本研究提出了FLARE，一个系统性和关注不确定性的框架，用于评估在医疗保健中采用人工智能的财务和运营影响。FLARE结合了模糊逻辑、基于时间的活动成本计算和投资回报分析，以估算临床服务交付成本、人工智能开发和运营成本，以及在不确定性下工作流程整合的经济后果。该框架通过一个早期健康技术评估案例研究进行了验证，该案例研究涉及在急性缺血性中风的CT中风路径中，人工智能辅助的大血管闭塞检测。案例研究展示了FLARE如何在统一的基于活动的模型中量化常规路径成本、与人工智能相关的开发和经常性成本，以及人工智能启用的服务节省。在预期假设下，分析确定了大约每年3,992名患者的盈亏平衡阈值，在每年约5,000名患者的典型中风量下，第一年的投资回报为正。结果进一步表明，经济效益不仅依赖于算法性能，还依赖于患者数量、验证时间、基础设施选择和工作流程设计。FLARE为医疗保健中人工智能采用的早期评估提供了一个透明且实用的决策支持框架。通过明确不确定性、资源使用和实施权衡，它帮助临床医生、管理者和政策制定者确定何时人工智能部署在经济上是可行的，以及在哪里运营变更可能提高价值。
cs.AI / 11 / 2608.23644
Ethical LLM-Assisted Research: A Framework for Responsible Delegation, Verification, and Epistemic Value
伦理LLM辅助研究：负责任的委托、验证和认知价值框架
Abstract
Large language models (LLMs) are becoming routine instruments of scientific research, assisting with literature synthesis, hypothesis development, coding, and formal reasoning. Their use raises a central epistemic question: when parts of scientific reasoning are delegated to an artificial system, what conditions must remain under human control for the resulting knowledge claims to retain epistemic legitimacy and accountable authorship? This paper develops a normative and conceptual framework for analyzing such delegation. Scientific reasoning is treated as a distributed process in which the origin of a contribution may vary between human and machine, while responsibility for its acceptance into the scientific record remains human. The framework distinguishes content origin $O(g)$, completion of human verification $V(g)$, responsibility assignment $R(g)$, accountable human ownership $M(g)$, and epistemic outcome $E(g)$. These constructs separate the provenance of a claim from the process by which it is checked, the epistemic outcome of that checking, and the human responsibility attached to its disposition. The central proposition is that the ethical boundary of LLM-assisted research is determined primarily by adequate verification and accountable human ownership rather than by the degree of machine involvement itself. On this basis, the paper develops the notion of an \emph{epistemic audit}: a structured record of delegation, verification, provenance, and responsibility intended to make AI-assisted reasoning transparent and reviewable. The resulting framework provides a formal vocabulary for distinguishing responsible cognitive delegation from the transfer or neglect of epistemic responsibility in scientific research.
Chinese Translation
大型语言模型（LLMs）正成为科学研究的常规工具，协助文献综述、假设发展、编码和形式推理。它们的使用提出了一个核心的认知问题：当科学推理的部分被委托给人工系统时，哪些条件必须保持在人类控制之下，以确保所产生的知识主张保持认知合法性和可追责的作者身份？本文发展了一个规范性和概念性框架来分析这种委托。科学推理被视为一个分布式过程，其中贡献的来源可能在于人类和机器之间变化，而其被接受入科学记录的责任仍然属于人类。该框架区分了内容来源 $O(g)$、人类验证的完成 $V(g)$、责任分配 $R(g)$、可追责的人类所有权 $M(g)$ 和认知结果 $E(g)$。这些构造将主张的来源与其检查过程、检查的认知结果以及与其处置相关的人类责任分开。中心命题是，LLM辅助研究的伦理边界主要由充分的验证和可追责的人类所有权决定，而不是由机器参与的程度本身。基于此，本文提出了一个 extit{认知审计}的概念：一个结构化的委托、验证、来源和责任记录，旨在使AI辅助推理透明且可审查。由此产生的框架为区分科学研究中负责任的认知委托与认知责任的转移或忽视提供了正式的词汇。
cs.AI / 12 / 2608.23646
MolEmb: Multimodal Large Language Models Can Be Strong Molecular Embedding Models
MolEmb：多模态大型语言模型可以成为强大的分子嵌入模型
Abstract
Molecular embedding models can serve as foundational infrastructure for computational chemistry and drug discovery, where reusable vector representations support property prediction, virtual screening, and retrieval. Most molecular encoders are specialist models built around a single molecular view, producing unconditional vectors with no language interface for varying the representation. We ask whether multimodal large language models (MLLMs), which natively process images, text, and symbolic inputs, can instead serve as \emph{general molecular embedding models} that produce embeddings conditioned on both a molecular profile and a natural-language semantic context. We introduce \textbf{MolEmb}, a lightweight framework that adapts MLLMs by aligning molecular profiles with textual descriptions in a shared embedding space using a bidirectional contrastive objective. The resulting embedding model is competitive on molecular property prediction and supports cross-modal molecule--text retrieval in the same space. We further introduce \textbf{MolCAR}, a diagnostic benchmark for context-aware retrieval, and find that context-aware molecular embedding is primarily a data property of the supervision. These results suggest that MLLMs are not merely chemistry assistants or generators, but a viable and extensible route to general molecular embedding models.
Chinese Translation
分子嵌入模型可以作为计算化学和药物发现的基础设施，其中可重用的向量表示支持属性预测、虚拟筛选和检索。大多数分子编码器是围绕单一分子视图构建的专业模型，生成的无条件向量没有语言接口来改变表示。我们探讨多模态大型语言模型（MLLMs）是否可以作为 extit{通用分子嵌入模型}，生成基于分子特征和自然语言语义上下文的嵌入。我们介绍了 extbf{MolEmb}，一个轻量级框架，通过使用双向对比目标将分子特征与文本描述对齐到共享嵌入空间，从而适应MLLMs。所得到的嵌入模型在分子属性预测方面具有竞争力，并支持在同一空间中的跨模态分子-文本检索。我们进一步介绍了 extbf{MolCAR}，一个用于上下文感知检索的诊断基准，并发现上下文感知的分子嵌入主要是监督的数据特性。这些结果表明，MLLMs不仅仅是化学助手或生成器，而是通用分子嵌入模型的可行且可扩展的途径。
cs.AI / 13 / 2608.23666
Gated Activation Steering for Reducing Sycophancy & Hallucination in Medical Question Answering
门控激活引导用于减少医疗问答中的谄媚行为和幻觉
Abstract
Sycophancy and hallucination are persistent failure modes of Large Language Models (LLMs) across domains. However, it becomes particularly consequential in clinical question answering, where responses must remain grounded in the provided context and robust to user pressure. Hallucination can introduce information that is unsupported by the context, while sycophancy can cause a model to abandon a previously correct answer when challenged by the user. Existing approaches, such as prompt-based safeguards and always-on activation steering, often address these behaviors separately or apply interventions broadly across turns, which can unnecessarily deteriorate responses that were already correct. To address these limitations within a single framework, we employ Inference Time Intervention (ITI) to jointly control both behaviors by learning separate steering directions for hallucination and sycophancy from contrastive clinical pairs and applying them to causally verified attention heads. During runtime, behavior-specific gates then determine when intervention is needed: the hallucination component mitigates unsupported claims, while the sycophancy component mitigates answer shifts caused by user pressure. We evaluate this framework on clinical questions grounded in EHR data while keeping the model weights frozen. Across all evaluation settings, we conducted 15,900 model-response runs. Across 600 pressure trajectories for the 4-billion-parameter model, the unsteered model caved in 570 cases. At the same time, gated steering helped it last longer in 551 of them. It held its ground under pressure at levels comparable to those of models with more than 100 billion parameters, showing that targeted inference-time steering can improve robustness without intervening at every turn.
Chinese Translation
谄媚行为和幻觉是大型语言模型（LLMs）在各个领域中持续存在的失效模式。然而，在临床问答中，这一问题尤为重要，因为响应必须保持在提供的上下文中，并对用户压力具有鲁棒性。幻觉可能引入与上下文不符的信息，而谄媚行为则可能导致模型在用户质疑时放弃之前正确的答案。现有的方法，如基于提示的保护措施和始终开启的激活引导，通常分别处理这些行为，或在多个回合中广泛应用干预，这可能会不必要地恶化已经正确的响应。为了解决这些局限性，我们在一个单一框架内采用推理时间干预（ITI），通过从对比临床对中学习幻觉和谄媚行为的独立引导方向，并将其应用于因果验证的注意力头，来共同控制这两种行为。在运行时，特定行为的门控决定何时需要干预：幻觉组件减轻不支持的主张，而谄媚组件减轻因用户压力引起的答案变化。我们在基于电子健康记录（EHR）数据的临床问题上评估了该框架，同时保持模型权重不变。在所有评估设置中，我们进行了15,900次模型响应运行。在针对40亿参数模型的600个压力轨迹中，未引导模型在570个案例中屈服。同时，门控引导帮助其在551个案例中坚持更久。在压力下，其表现与超过1000亿参数的模型相当，显示出针对性的推理时间引导能够在不在每个回合干预的情况下提高鲁棒性。
cs.AI / 14 / 2608.23670
Automata from Agent Traces: Failure and Next-Step Prediction
来自代理追踪的自动机：失败与下一步预测
Abstract
LLM-based agents execute multi-step tasks, but their behavioral structure remains opaque: long unstructured traces resist the safety auditing and runtime monitoring that deployment requires. Existing approaches operate per-trace or success-only, so they miss the cross-run topology that links next-step and failure prediction. To recover that shared structure, we collapse an entire trace corpus into a single, compact finite-state machine (FSM) that serves as a structural substrate for the otherwise unpredictable behavior of LLM agents. Across twelve public datasets, the FSMs are compact (7-43 states), replay held-out data at >=0.997 fitness with near-identical topology across splits, and build in milliseconds. This substrate addresses both prediction goals. For next-step prediction, FSM-state context outperforms Agent Workflow Memory on every ground-truth-matched dataset. For failure prediction, per-state behavioral features reach held-out AUROC up to 0.94, and an online monitor ranks failing runs above passing ones from a partial trace, triggering early stopping well before completion. Behavioral topology thus appears shaped more by the deployment harness than by the LLM, providing a model-agnostic structural primitive for safety auditing and runtime monitoring.
Chinese Translation
基于大型语言模型（LLM）的代理执行多步骤任务，但其行为结构仍然不透明：长时间的非结构化追踪抵制了部署所需的安全审计和运行时监控。现有方法仅在每个追踪或成功的基础上进行操作，因此错过了连接下一步和失败预测的跨运行拓扑。为了恢复这种共享结构，我们将整个追踪语料库压缩为一个单一的紧凑有限状态机（FSM），该状态机作为LLM代理不可预测行为的结构基础。在十二个公共数据集上，FSM的状态数紧凑（7-43个状态），在>=0.997的拟合度下重放保留数据，并且在不同划分中具有几乎相同的拓扑，构建时间仅需毫秒。该基础结构同时满足两个预测目标。对于下一步预测，FSM状态上下文在每个与真实数据匹配的数据集上均优于代理工作流记忆。对于失败预测，每个状态的行为特征在保留数据上达到高达0.94的AUROC，并且在线监控器将失败的运行排名高于通过部分追踪的成功运行，从而在完成之前触发早期停止。因此，行为拓扑似乎更多地受到部署工具的影响，而非LLM本身，为安全审计和运行时监控提供了一种模型无关的结构原语。
cs.AI / 15 / 2608.23691
Autonomous Mathematical Discovery in an Open-World Multi-Agent Environment
开放世界多智能体环境中的自主数学发现
Abstract
We study autonomous mathematical discovery in the Station, an open-world multi-agent environment in which AI agents from different model families pursue a shared research goal without a central coordinator or scripted pipeline. Agents choose their own research directions, conduct experiments, collaborate, and build a shared scientific literature. Across 12 construction problems from the AlphaEvolve catalogue and two additional case studies, the Station obtained results novel relative to the prior literature on five problems: a new infinite family of finite-field Kakeya sets, new exact 604-point kissing configurations in dimension 11, new records for the discretized Kakeya needle and sign uncertainty problems, and a substantially improved lower bound for Erd\H{o}s's minimum-overlap problem. Agents also discovered novel infinite families for Book Ramsey numbers. Importantly, the agents produced not only numerical constructions but also theorems and analyses explaining how those constructions work, making the results more interpretable and easier for mathematicians to build upon. We release all raw agent dialogues, proofs, and verification code, providing a transparent record of how these discoveries emerged.
Chinese Translation
我们研究了在Station这一开放世界多智能体环境中进行的自主数学发现。在这个环境中，来自不同模型家族的人工智能代理在没有中央协调者或脚本化流程的情况下，追求共同的研究目标。代理选择自己的研究方向，进行实验，协作，并建立共享的科学文献。在来自AlphaEvolve目录的12个构造问题和两个额外案例研究中，Station在五个问题上取得了相对于先前文献的新成果：一类新的无限有限域Kakeya集合、11维中新的确切604点亲吻配置、离散化Kakeya针和符号不确定性问题的新记录，以及Erdős的最小重叠问题的显著改进下界。代理们还发现了Book Ramsey数的新无限家族。重要的是，代理不仅生成了数值构造，还提出了定理和分析，解释这些构造的工作原理，使得结果更具可解释性，并便于数学家们在此基础上进行进一步研究。我们发布了所有原始代理对话、证明和验证代码，提供了这些发现是如何产生的透明记录。
cs.AI / 16 / 2608.23706
Do LLMs Understand Limit Order Book Dynamics?
大型语言模型是否理解限价单簿动态？
Abstract
A large language model (LLM) trained on synthetic limit order book (LOB) data achieves near perfect scores in generating valid sequences of LOB events. However, the LLM's implicit world model fails to learn the state of the LOB. This deficiency leads to biased estimates and spurious predictability in using the LLM to forecast future LOB events. Our analysis uses novel tests of an LLM's world model, extending prior work from deterministic settings to the stochastic dynamics needed for the LOB.
Chinese Translation
一个在合成限价单簿（LOB）数据上训练的大型语言模型（LLM）在生成有效的LOB事件序列方面达到了近乎完美的分数。然而，该LLM的隐性世界模型未能学习LOB的状态。这一缺陷导致在使用LLM预测未来LOB事件时出现偏差估计和虚假的可预测性。我们的分析使用了对LLM世界模型的新颖测试，将先前的研究从确定性环境扩展到LOB所需的随机动态。
cs.AI / 17 / 2608.23740
AgentRoom: Concurrent Multi-Agent Coding in a CRDT-Backed Shared Workspace
AgentRoom：基于CRDT的共享工作区中的并发多智能体编码
Abstract
Concurrent multi-agent coding promises division of labor across modules, robustness through redundancy, and parallel exploration at the natural granularity of multi-file projects. Realtime collaborative editing protocols solve this coordination problem for human teams via Conflict-free Replicated Data Types (CRDTs), but the LLMs underneath generate one token at a time and existing multi-agent coding systems inherit this serial limit: they either sequence agents through phase handoffs or pool independent samples without coordination, and a single agent abandons up to half of hard tasks with a one-file stub-and-exit. AgentRoom is a realtime collaborative editing protocol for concurrent coding agents. Its runtime layer exposes file-level claim, status, and broadcast as MCP tools on a CRDT-merged shared filesystem. Five frontier coding-CLI models ran four backend coding tasks, with cross-language checks in Python DevBench and Rust+axum. For CLI-stable models, AgentRoom with 2 agents abandons fewer tasks than Solo and has less run-to-run variation. At matched-compute, one positive mean LLM-judge contrast puts AgentRoom over parallel-merge. The other contrast, a bundle probe, puts full AgentRoom above each partial case: an ordering rather than a percentage split. Coordination, not parallelism or CRDT-merge, bears the load.
Chinese Translation
并发多智能体编码承诺在模块之间实现劳动分工，通过冗余增强鲁棒性，并在多文件项目的自然粒度上进行并行探索。实时协作编辑协议通过无冲突复制数据类型（CRDT）解决了人类团队的协调问题，但底层的大型语言模型（LLMs）一次生成一个标记，现有的多智能体编码系统继承了这一串行限制：它们要么通过阶段交接对智能体进行排序，要么在没有协调的情况下汇集独立样本，而单个智能体在处理一个文件的简化任务时可能会放弃多达一半的困难任务。AgentRoom是一个针对并发编码智能体的实时协作编辑协议。其运行时层将文件级的声明、状态和广播作为CRDT合并的共享文件系统上的多通道处理工具（MCP）暴露出来。五个前沿编码CLI模型执行了四个后端编码任务，并在Python DevBench和Rust+axum中进行了跨语言检查。对于CLI稳定模型，使用两个智能体的AgentRoom放弃的任务少于单独模型（Solo），且运行间的变异性较小。在计算匹配的情况下，一个积极的平均LLM评判对比显示AgentRoom优于并行合并。另一个对比，即捆绑探测，显示完整的AgentRoom优于每个部分案例：这是一个排序而非百分比分割。协调，而非并行性或CRDT合并，承担了负载。
cs.AI / 18 / 2608.23807
Serving Masked Diffusion LLMs: Characterization and Design Principles from Real Hardware
服务掩蔽扩散语言模型：来自真实硬件的特征化与设计原则
Abstract
Masked diffusion language models (dLLMs) can in principle generate text faster than autoregressive (AR) models, since they denoise many tokens at once. Recent systems have begun building serving infrastructure for dLLMs, but none first measure how these models behave under real, concurrent serving load. Serving systems built without this grounding risk carrying over assumptions from AR serving that may not hold for dLLMs. We characterize dLLM serving to close this gap, using LLaDA-8B-Instruct with a D2F (Discrete Diffusion Forcing) LoRA adapter on a single NVIDIA H200 GPU, evaluated on GSM8K and HumanEval. We report three findings. First, request difficulty, the number of denoising steps a request needs, is discrete rather than continuous: requests fall into 11 fixed step-count levels (178 + 29k), and no signal we test predicts the level before generation starts (best R2 = 0.150). Second, benchmarks with short generation budgets below 320 tokens understate serving variance, since requests are cut off before the latency spread appears. Third, only 24% of single-request wall-clock time is GPU computation; the rest is CPU-side dispatch overhead. Batching mainly helps by amortizing this overhead: sharing one forward pass per denoising step improves throughput by 16.0x at batch size 16 over a per-request-dispatch baseline. We also argue structurally that output quality should not degrade with batch size, stating three assumptions this rests on; we measure 74 to 76% GSM8K accuracy at single-request scale. Finally, we derive a batch-timeout rule for fixed-fill synchronized batching under Poisson arrivals. Together, these results show that serving diffusion language models needs parallelism at the level of each denoising step, which differs from AR serving in how admission and eviction interact with an already shared forward pass.
Chinese Translation
掩蔽扩散语言模型（dLLMs）在理论上可以比自回归（AR）模型更快地生成文本，因为它们可以同时去噪多个标记。近期的系统开始为dLLMs构建服务基础设施，但尚未首先测量这些模型在真实并发服务负载下的表现。没有这一基础的服务系统可能会延续AR服务中的假设，而这些假设可能不适用于dLLMs。我们对dLLM服务进行特征化，以填补这一空白，使用在单个NVIDIA H200 GPU上搭载D2F（离散扩散强制）LoRA适配器的LLaDA-8B-Instruct，并在GSM8K和HumanEval上进行评估。我们报告了三项发现。首先，请求难度，即请求所需的去噪步骤数，是离散而非连续的：请求分为11个固定的步骤计数级别（178 + 29k），我们测试的信号在生成开始之前无法预测该级别（最佳R2 = 0.150）。其次，生成预算低于320个标记的基准测试低估了服务的方差，因为请求在延迟扩散出现之前被截断。第三，单请求的实际时间中只有24%是GPU计算；其余是CPU端的调度开销。批处理主要通过摊销这一开销来提供帮助：在每个去噪步骤共享一次前向传播可以在批量大小为16时将吞吐量提高16.0倍，相较于每请求调度的基线。我们还从结构上论证，输出质量不应随着批量大小的增加而降低，并提出了三个假设；我们在单请求规模下测得GSM8K准确率为74%至76%。最后，我们推导出在泊松到达下固定填充同步批处理的批量超时规则。综合来看，这些结果表明，服务扩散语言模型需要在每个去噪步骤的层面上实现并行性，这与AR服务在接纳和驱逐与已共享前向传播的交互方式上有所不同。
cs.AI / 19 / 2608.23811
Generating Biomedical Fact-Checking Reports with RL-Enhanced Agentic Search
生成生物医学事实核查报告的强化学习增强代理搜索
Abstract
Automated fact-checking is essential for ensuring the reliability of public health information, yet the biomedical domain poses unique challenges. Validating biomedical claims requires rigorous interpretation of scientific literature, assessment of retrieved evidence, and comprehensive justification toward the conclusion. Although Large Language Models (LLMs) enhanced by Retrieval-Augmented Generation (RAG) and agentic search perform automated fact-checking in a retrieve-then-verify paradigm, current methods still output isolated prediction labels, lacking explanatory depth and offers limited utility for human understanding. To bridge this gap, we introduce an LLM-based agent named BioCheck Agent that generates structured biomedical fact-checking reports with agentic search. Rather than merely outputting supported or refuted labels, our agent synthesizes final conclusions with retrieved evidence and rigorous analysis. To ensure domain-specific accuracy, BioCheck Agent exclusively searches high-quality scientific literature in PubMed, utilizing advanced Boolean search operators. Recognizing that direct prompting often results in hallucinations and low-quality reports, especially for lightweight open-source models, we further propose the Evidence-Grounded Group Relative Policy Optimization (EG-GRPO) to perform reinforcement learning on BioCheck Agent with a task-specific reward that incentivizes advanced search behavior and high-quality evidence retrieval while penalizing hallucinations. Our experimental results show that compared to the base model Qwen3.5-4B, BioCheck Agent with EG-GRPO improves label prediction accuracy on SciFact by 9.95%. Furthermore, it achieves a 3.7% higher evidence quality score and a 19.63% lower evidence hallucination rate, demonstrating its ability to generate biomedical fact-checking reports with improved accuracy and quality.
Chinese Translation
自动化事实核查对于确保公共健康信息的可靠性至关重要，但生物医学领域面临独特的挑战。验证生物医学声明需要对科学文献进行严格的解读、对检索到的证据进行评估，并对结论提供全面的论证。尽管增强了检索-再验证范式的检索增强生成（RAG）和代理搜索的大型语言模型（LLMs）能够执行自动化事实核查，但当前的方法仍然输出孤立的预测标签，缺乏解释深度，并且对人类理解的实用性有限。为了解决这一问题，我们引入了一种基于LLM的代理，名为BioCheck Agent，它利用代理搜索生成结构化的生物医学事实核查报告。我们的代理不仅仅输出支持或反驳的标签，而是将最终结论与检索到的证据和严格分析进行综合。为了确保领域特定的准确性，BioCheck Agent专门在PubMed中搜索高质量的科学文献，利用先进的布尔搜索运算符。考虑到直接提示往往导致幻觉和低质量报告，尤其是对于轻量级开源模型，我们进一步提出了基于证据的群体相对策略优化（EG-GRPO），对BioCheck Agent进行强化学习，以任务特定的奖励激励其进行高级搜索行为和高质量证据检索，同时惩罚幻觉。我们的实验结果表明，与基础模型Qwen3.5-4B相比，使用EG-GRPO的BioCheck Agent在SciFact上的标签预测准确率提高了9.95%。此外，它的证据质量评分提高了3.7%，证据幻觉率降低了19.63%，展示了其生成生物医学事实核查报告的准确性和质量的提升能力。
cs.AI / 20 / 2608.23817
A Formal Methodological Framework for Auditing Robustness and Fidelity in Explainable AI: From Application to Trust Certification
用于审计可解释人工智能的鲁棒性和保真性的正式方法框架：从应用到信任认证
Abstract
SHAP and LIME are now standard tools for interpreting black-box predictions, yet their outputs can vary substantially when the input is perturbed by small amounts of noise--a problem we observed firsthand in our previous work on food security in Madagascar (Ralinirina et al., 2025). This variability raises the question of whether such explanations can be trusted at all. We address it by constructing an auditing protocol that measures two properties of any post-hoc explainer: robustness (how stable the explanation is under input perturbation) and fidelity (whether the features deemed important actually drive the model's prediction). These two quantities are combined into a single Trust Score. We run the protocol on a multi-sectoral dataset from Madagascar (83 features, 253 records, 4 malnutrition classes) using three classifiers and two explainers, plus their regularized counterparts. The results are sobering: models with AUC above 0.99 can produce numerically degenerate or flatly uninformative explanations, and fidelity scores lose discriminative power when the model is overfitted. These findings suggest that auditing XAI outputs is not optional but necessary, particularly when they inform decisions in sensitive domains.
Chinese Translation
SHAP 和 LIME 现在已成为解释黑箱预测的标准工具，但当输入受到小量噪声扰动时，它们的输出可能会有显著差异——这一问题在我们之前关于马达加斯加食品安全的研究中得到了直接观察（Ralinirina 等，2025）。这种变异性引发了一个问题：这些解释是否可以被信任。我们通过构建一个审计协议来解决这个问题，该协议测量任何事后解释器的两个属性：鲁棒性（在输入扰动下解释的稳定性）和保真性（被认为重要的特征是否真正驱动模型的预测）。这两个量被结合成一个单一的信任评分。我们在来自马达加斯加的多部门数据集（83个特征，253条记录，4个营养不良类别）上运行该协议，使用了三种分类器和两种解释器，以及它们的正则化对应物。结果令人警醒：AUC 超过 0.99 的模型可能会产生数值退化或完全无信息的解释，并且当模型过拟合时，保真评分会失去区分能力。这些发现表明，审计可解释人工智能的输出不是可选的，而是必要的，特别是在它们影响敏感领域的决策时。
cs.AI / 21 / 2608.23834
Minima-KV: Retention-Preserving KV Cache Compression with Mixed-Format Paged Attention
Minima-KV：具有混合格式分页注意力的保留缓存压缩
Abstract
The key-value (KV) cache is a primary capacity and bandwidth bottleneck in long-context LLM serving. We present Minima-KV, a retention-preserving hierarchy for mixed-format paged attention. Recent and protected Anchor pages remain in FP8, while older non-anchor pages move to packed TQ3; every live-request page remains addressable. Format-specific kernels compute partial attention states and combine them through a globally normalized online-softmax merge, enabling direct heterogeneous decode without a cache-sized dense shadow. Across separate, configuration-bound Qwen3.6-27B profiles on a single 96-GB NVIDIA RTX PRO 6000 Blackwell GPU, deployment accounting reports 18.3 KiB of attention KV per live token, corresponding to 3.50x compression relative to BF16 and 1.75x relative to FP8. A materializing quality profile matches its dense control on 16K RULER needle-in-a-haystack tasks. On the same 503-question LongBench v2 set, measured deltas are -0.80, -0.60, and -0.40 percentage points at 16K, 32K, and 64K. A separate single-pair direct-decode canary with two 59,008-token requests measures 3.625x active-KV compression and 0.9821x throughput relative to its control, routes all 16 full-attention layers without fallback, and retains no dense shadow. These results establish a practical mixed-format path for compressing long-context state without evicting live-request KV pages.
Chinese Translation
键值（KV）缓存是长上下文大语言模型（LLM）服务中的主要容量和带宽瓶颈。我们提出了Minima-KV，这是一种用于混合格式分页注意力的保留层次结构。最近的和受保护的锚点页面保持在FP8格式，而较旧的非锚点页面则转移到打包的TQ3格式；每个活跃请求页面仍然可寻址。特定格式的内核计算部分注意状态，并通过全局归一化的在线软最大合并将其组合，从而实现直接异构解码，而无需缓存大小的稠密影子。在单个96GB NVIDIA RTX PRO 6000 Blackwell GPU上，针对不同配置的Qwen3.6-27B配置文件，部署报告每个活跃令牌的注意KV为18.3 KiB，相对于BF16压缩为3.50倍，相对于FP8压缩为1.75倍。物化质量配置文件在16K RULER针在干草堆任务中与其稠密控制相匹配。在相同的503个问题LongBench v2数据集上，测得的增量分别为-0.80、-0.60和-0.40个百分点，分别对应于16K、32K和64K。一个单独的单对直接解码金丝雀在两个59,008令牌请求中测得活跃KV压缩为3.625倍，相对于其控制的吞吐量为0.9821倍，路由所有16个全注意力层而无需后备，并且没有保留稠密影子。这些结果确立了一条实用的混合格式路径，用于在不驱逐活跃请求KV页面的情况下压缩长上下文状态。
cs.AI / 22 / 2608.23837
SyPS: Measuring Sycophancy Prompt Sensitivity in Large Language Models
SyPS：测量大型语言模型中的阿谀奉承提示敏感性
Abstract
Large language models (LLMs) are known to exhibit social sycophancy, often validating or agreeing with users in socially sensitive contexts. Existing evaluations typically measure sycophancy under a fixed prompt formulation, leaving unclear whether such behavior is stable when the same underlying situation is presented with different sycophancy-relevant prompt variants. In this work, we study sycophancy prompt sensitivity: the extent to which changes in user confidence, emotional framing, social consensus, or validation-seeking language alter a model's sycophantic behavior. We refer to our evaluation framework as SyPS, short for Sycophancy Prompt Sensitivity. Building on existing social sycophancy evaluation settings, SyPS constructs controlled prompt variants that preserve the same underlying user situation while varying sycophancy-relevant social cues. We introduce the Sycophancy Prompt Sensitivity Score (SPSS), an instance-level measure of sycophancy variation across paired prompt variants. Unlike aggregate sycophancy rates, SPSS separates baseline sycophancy from prompt-induced shifts, enabling model-level comparisons of robustness to sycophancy-relevant social cues. Empirically, we find that sycophancy prompt sensitivity is socially structured: validation-seeking and emotional-pressure cues often increase sycophancy, whereas counter-framing and anti-sycophancy prompts tend to reduce it. Our framework highlights whether LLMs maintain stable social judgments while adapting appropriately in tone.
Chinese Translation
大型语言模型（LLMs）被认为表现出社会阿谀奉承的特征，常常在社会敏感的情境中验证或同意用户的观点。现有评估通常在固定的提示形式下测量阿谀奉承，尚不清楚在不同的与阿谀奉承相关的提示变体下，相同的基本情境是否会导致这种行为的稳定性。在本研究中，我们研究了阿谀奉承提示敏感性：用户信心、情感框架、社会共识或寻求验证的语言变化在多大程度上改变模型的阿谀奉承行为。我们将我们的评估框架称为SyPS，代表阿谀奉承提示敏感性（Sycophancy Prompt Sensitivity）。基于现有的社会阿谀奉承评估设置，SyPS构建了控制的提示变体，这些变体在保持相同的基本用户情境的同时，变化与阿谀奉承相关的社会线索。我们引入了阿谀奉承提示敏感性评分（Sycophancy Prompt Sensitivity Score，SPSS），这是一个针对成对提示变体的阿谀奉承变化的实例级测量。与聚合的阿谀奉承率不同，SPSS将基线阿谀奉承与提示引起的变化分开，从而使模型级别的对比变得可能，以评估对与阿谀奉承相关的社会线索的鲁棒性。实证研究发现，阿谀奉承提示敏感性具有社会结构性：寻求验证和情感压力线索通常会增加阿谀奉承，而反框架和反阿谀奉承提示则倾向于减少阿谀奉承。我们的框架突出了大型语言模型在适当地调整语气的同时，是否能够保持稳定的社会判断。
cs.AI / 23 / 2608.23848
Exploit More, Explore Smarter for Budget-Constrained Agentic Search
在预算受限的自主搜索中更好地利用资源，更智能地探索
Abstract
Budget-constrained agentic search arises when an LLM agent must refine candidates under a small evaluation budget, because validation is expensive, generation requires multiple model calls, or both. In this regime, standard MCTS allocates budget poorly: exploration bonuses dominate at low visit counts, unpromising siblings are expanded before promising chains can deepen, and branching is independent of node quality. We introduce ExTS, a tree-search policy that treats expansion itself as a value-of-information decision. ExTS combines three mechanisms: discriminative reward shaping to separate candidates under narrow score distributions, a stochastic virtual child that estimates the value of creating a new branch from the parent's reward history, and quality-conditioned branching that expands only when a node's score justifies the budget cost. Across prompt optimization, code generation, molecular structure elucidation, and agentic workflow optimization, ExTS is competitive with or improves over task-specific tree-search baselines, with an average relative gain of +5.5% using a single fixed configuration. We further introduce pilot-run diagnostics that characterize what makes budget-constrained agentic search problems structurally different from one another, providing both understanding of the problem space and practical guidance for adaptation.
Chinese Translation
预算受限的自主搜索出现在大型语言模型（LLM）代理需要在小评估预算下优化候选项时，因为验证成本高，生成需要多次模型调用，或两者兼而有之。在这种情况下，标准的蒙特卡洛树搜索（MCTS）分配预算的效果较差：在低访问次数下，探索奖励占主导地位，不太有前景的兄弟节点在有前景的链条能够深入之前被扩展，而分支与节点质量无关。我们提出了ExTS，一种将扩展本身视为信息价值决策的树搜索策略。ExTS结合了三种机制：区分性奖励塑造，用于在狭窄的评分分布下区分候选项；一个随机虚拟子节点，用于估计从父节点的奖励历史中创建新分支的价值；以及质量条件分支，仅在节点的评分能够证明预算成本时才进行扩展。在提示优化、代码生成、分子结构阐明和自主工作流优化等任务中，ExTS与任务特定的树搜索基线相竞争或有所改进，使用单一固定配置的平均相对增益为+5.5%。我们进一步引入了试点运行诊断，描述了预算受限的自主搜索问题在结构上有何不同，从而提供了对问题空间的理解和适应的实用指导。
cs.AI / 24 / 2608.23855
In-Context Inpainting for Time Series Forecasting
上下文内插补用于时间序列预测
Abstract
We propose ICI-Time, a novel framework that reframes time series forecasting as a visual inpainting task, leveraging the generalisation power of large vision models (LVMs). Unlike methods that require specialised temporal architectures and extensive domain-specific training, ICI-Time transforms time series into structured visual representations (area charts) and applies visual in-context learning, reformulating forecasting as pattern completion within a grid-structured prompt that pre-trained vision transformers can solve without fine-tuning or architectural modification. Temporal dependencies are represented through spatial layout, with a consistent, invertible mapping between numerical and visual domains. Extensive experiments across epidemiology, meteorology, and power systems demonstrate that ICI-Time performs competitively against deep learning baselines and shows promising adaptability under limited-data settings, introducing a new paradigm that bridges temporal and visual domains.
Chinese Translation
我们提出了 ICI-Time，这是一种新颖的框架，将时间序列预测重新定义为视觉内插补任务，利用大型视觉模型（LVMs）的泛化能力。与需要专门的时间架构和广泛领域特定训练的方法不同，ICI-Time 将时间序列转换为结构化的视觉表示（区域图），并应用视觉上下文学习，将预测重新表述为在网格结构提示中完成模式，这些提示是预训练的视觉变换器可以在不进行微调或架构修改的情况下解决的。时间依赖性通过空间布局表示，数值域与视觉域之间存在一致的可逆映射。通过在流行病学、气象学和电力系统等领域进行的广泛实验表明，ICI-Time 在与深度学习基线的竞争中表现出色，并在有限数据环境下显示出良好的适应性，提出了一种新的范式，连接了时间域和视觉域。
cs.AI / 25 / 2608.23870
Granite.Trust Policy Tools: Shareable, Actionable Policies for Generative AI Applications
Granite.Trust 政策工具：可共享、可执行的生成性人工智能应用政策
Abstract
When it comes to safety policies for generative AI, one size does not fit all. Each organization and use case needs to mitigate different risks depending on the application context, regulatory environment, organizational values, and user personas. Yet, existing policy specification approaches are designed for traditional access control and fail to capture the nuances of GenAI application: the enforcement of content-based constraints. We present two contributions to address this gap: (1) the Actionable Policy schema, a YAML-based format for specifying what model responses can and cannot contain. The schema enables exception-based policy governance, proposing exceptions to track policy violations; (2) synthetic data generation pipeline that produces policy-aligned training data for model alignment and testing, and a set of tools to help define the schema and enforce policy. Together, these enable organizations to specify policies once and enforce them throughout the GenAI application lifecycle: from model alignment to runtime monitoring. The Actionable Policy schema, example policies, and tools are available as open source: https://github.com/ibm-granite/granite.trust.policy-tools We welcome new ideas, contributions and feedback.
Chinese Translation
在生成性人工智能的安全政策方面，单一的解决方案并不适用所有情况。每个组织和使用案例都需要根据应用背景、监管环境、组织价值观和用户角色来减轻不同的风险。然而，现有的政策规范方法主要针对传统的访问控制，未能捕捉到生成性人工智能应用的细微差别：内容约束的执行。我们提出了两个贡献来填补这一空白：（1）可执行政策架构，一种基于 YAML 的格式，用于指定模型响应可以和不可以包含的内容。该架构支持基于例外的政策治理，提出例外以跟踪政策违规情况；（2）合成数据生成管道，生成与政策对齐的训练数据，以便进行模型对齐和测试，并提供一套工具来帮助定义架构和执行政策。这些工具使组织能够一次性指定政策，并在整个生成性人工智能应用生命周期中执行这些政策：从模型对齐到运行时监控。可执行政策架构、示例政策和工具已作为开源项目提供，链接：https://github.com/ibm-granite/granite.trust.policy-tools 我们欢迎新的想法、贡献和反馈。
cs.AI / 26 / 2608.23873
Semantic Overlays: Mitigating Prompt Injection with Annotations Beyond Tokens and Steering Vectors
语义覆盖：通过超出令牌和引导向量的注释缓解提示注入
Abstract
Everything a language model sees is tokens. The serving stack knows what each span is -- user input, tool output, instructions -- but the model must keep track of that itself, and it can lose track or be confused: text can be written to read like anything. Prompt injection is a natural exploit of this phenomenon. By scrambling the model's understanding of span identity, an attacker can induce unwanted and potentially dangerous actions. Adding a non-textual channel to the model's input -- a way to communicate span identity beyond text -- mitigates this class of attack. We thus introduce a general steering technique called Semantic Overlays: small learned adapters applied at chosen prefill positions to a frozen model's residual stream. Laying an overlay over a span creates an out-of-band annotation channel that cannot be replicated by tokens. Unlike steering vectors, Semantic Overlays are trained, adaptable, and selectively applied. An overlay can encode complex semantics that reshape how the model perceives the marked span: asked to copy a code snippet under an overlay asserting that it is in a different programming language than it is, the model rewrites the snippet, faithfully, in the asserted language. Overlays are also composable, allow for transparent reading of underlying content, and can carry complex payloads -- including imperatives that the model will follow. An overlay which marks a span as "non-executable" defends against the broad class of prompt injections that add instructions in untrusted context. We report strong results on prompt injection benchmarks: SEP separation rises from 24.3% to 96.5% with utility unchanged (our scoring rule; we also correct a defect in the published grader), TensorTrust attack success rate falls from 34.8% to 6.6%, and all four PIArena attack families drop to 0% compliance, all while marked spans stay readable (92.5% exact copy rate).
Chinese Translation
语言模型所见的一切都是令牌。服务堆栈知道每个跨度是什么——用户输入、工具输出、指令——但模型必须自行跟踪这些信息，而它可能会失去跟踪或感到困惑：文本可以被写成任何形式。提示注入是这一现象的自然利用。通过混淆模型对跨度身份的理解，攻击者可以诱导出不必要且潜在危险的行为。向模型的输入添加一个非文本通道——一种超越文本传达跨度身份的方式——可以缓解这一类攻击。因此，我们引入了一种称为语义覆盖（Semantic Overlays）的通用引导技术：在选择的预填充位置对冻结模型的残差流应用的小型学习适配器。覆盖在一个跨度上创建了一个无法通过令牌复制的带外注释通道。与引导向量不同，语义覆盖是经过训练的、可适应的，并且是选择性应用的。一个覆盖可以编码复杂的语义，重塑模型对标记跨度的感知：当要求在一个声称是不同编程语言的覆盖下复制代码片段时，模型会忠实地在所声称的语言中重写该片段。覆盖也是可组合的，允许透明地读取基础内容，并可以承载复杂的有效载荷——包括模型将遵循的命令。一个将跨度标记为“不可执行”的覆盖可以防御广泛的提示注入类，这些注入在不可信的上下文中添加指令。我们在提示注入基准测试中报告了强劲的结果：SEP分离率从24.3%上升到96.5%，效用保持不变（我们的评分规则；我们还纠正了已发布评分器中的缺陷），TensorTrust攻击成功率从34.8%降至6.6%，所有四个PIArena攻击家族的合规率降至0%，同时标记的跨度保持可读（92.5%的精确复制率）。
cs.AI / 27 / 2608.23875
AI Finds A Way
人工智能找到了一条道路
Abstract
Artificial Intelligence (AI) algorithms frequently learn creative and unexpected solutions, surprising even expert researchers who develop and study them. They often astonish practitioners by discovering unanticipated behavior, exploiting loopholes in reward signals, or spontaneously uncovering previously unknown scientific phenomena. However, accounts of such unconventional behavior across machine learning are seldom formally documented. This work presents 26 curated firsthand anecdotes from various machine learning subfields representing the work of over 100 researchers. These anecdotes showcase the capability of modern AI systems to circumvent human-imposed design limitations and discover unexpected solutions to the tasks we train them on. Furthermore, these accounts are particularly important for the safety of future AI systems. They illustrate the fundamental challenge of aligning models with human values without diminishing their creativity, so they can make surprising discoveries without producing surprising, potentially harmful outcomes. The paper first details AI achieving superhuman success through reinforcement learning across many challenging domains. However, reward-driven optimization can fail when the model learns to hack an underspecified reward or unarticulated constraint. We then present case studies suggesting that harnessing internet-scale foundation models (FMs) has not resolved these fundamental challenges and, in fact, can supercharge them. Nevertheless, we argue that these same learning dynamics can be harnessed to accelerate scientific discovery. Finally, we hope this work provides a consolidated resource to inform future research and demonstrates that the tendency toward unexpected behaviors is commonplace in modern AI, highlighting the need to anticipate and manage AI's capacity for innovative, yet unpredictable, solutions. (abstract abridged)
Chinese Translation
人工智能（AI）算法经常学习到创造性和意想不到的解决方案，甚至让开发和研究它们的专家研究人员感到惊讶。它们常常通过发现意外行为、利用奖励信号中的漏洞，或自发揭示先前未知的科学现象来令从业者感到震惊。然而，关于机器学习中此类非常规行为的记录很少被正式记录。本文呈现了来自各个机器学习子领域的26个精心策划的一手轶事，代表了100多位研究人员的工作。这些轶事展示了现代AI系统绕过人类设定的设计限制并发现意想不到的解决方案的能力。此外，这些案例对于未来AI系统的安全性尤为重要。它们说明了将模型与人类价值观对齐的基本挑战，而不降低其创造力，以便使其能够进行惊人的发现，而不会产生意外的、潜在有害的结果。本文首先详细介绍了AI通过强化学习在多个具有挑战性的领域取得超人类成功。然而，当模型学会破解一个不明确的奖励或未明确的约束时，基于奖励的优化可能会失败。接着，我们呈现了案例研究，表明利用互联网规模的基础模型（FMs）并没有解决这些基本挑战，实际上可能会加剧这些问题。尽管如此，我们认为这些相同的学习动态可以被利用来加速科学发现。最后，我们希望这项工作提供一个整合的资源，以指导未来的研究，并展示现代AI中意外行为的倾向是普遍存在的，强调了预见和管理AI创新但不可预测的解决方案能力的必要性。
cs.AI / 28 / 2608.23893
Provenance Guided Incremental Learning Under Evolving Concept Definitions
基于来源指导的增量学习在不断演变的概念定义下
Abstract
Learning systems deployed over long periods must adapt not only to statistical changes in incoming data, but also to revisions of the definitions that generate their prediction targets. Conventional concept-drift methods typically infer such changes from observations or prediction errors, even when the underlying policy, rule, or query has been explicitly modified. This paper studies rule-induced concept shift, where the target-defining concept is revised directly, causing previously stored instances to acquire different semantic labels without requiring any change in their observed data. We introduce a provenance-guided incremental learning framework that compiles consecutive concept definitions into a structured rule delta, traces the changed components through historical provenance, certifies records whose previous labels remain valid, and restricts reevaluation to a localized candidate region. Executable revisions are relabeled automatically, ambiguous cases are handled through selective supervision, and the resulting changes are used for incremental predictor repair. A versioned concept memory further supports recurring definitions. We also introduce RuleShift-Bench, spanning financial, demographic, cybersecurity, and graph-structured data with threshold, predicate, logical, relational, recurring, and mixed concept revisions. Across the benchmark, provenance-guided repair attains 92.3% accuracy and 90.2% Macro-F1 while reprocessing 14.7% of the historical collection and retaining 94.6% of affected records. Its average update latency is 179s compared with 993s for complete relabeling and retraining. The results demonstrate that an explicit concept revision can be exploited as a data-maintenance signal, allowing learning systems to update the supervision and predictive state that depend on the change while preserving knowledge that remains valid.
Chinese Translation
长期部署的学习系统必须适应输入数据的统计变化，同时也要应对生成其预测目标的定义的修订。传统的概念漂移方法通常从观察或预测错误中推断这些变化，即使基础的政策、规则或查询已经被明确修改。本文研究了由规则引起的概念转变，其中目标定义概念被直接修订，导致先前存储的实例获得不同的语义标签，而无需对其观察数据进行任何更改。我们提出了一种基于来源指导的增量学习框架，该框架将连续的概念定义编译成结构化的规则增量，通过历史来源追踪变化的组件，认证其先前标签仍然有效的记录，并将重新评估限制在局部候选区域。可执行的修订会自动重新标记，模糊案例通过选择性监督进行处理，所产生的变化用于增量预测器修复。版本化的概念记忆进一步支持重复定义。我们还引入了RuleShift-Bench，涵盖金融、人口统计、网络安全和图结构数据，包含阈值、谓词、逻辑、关系、重复和混合概念修订。在基准测试中，基于来源的修复达到了92.3%的准确率和90.2%的宏观F1分数，同时重新处理了14.7%的历史集合，并保留了94.6%的受影响记录。其平均更新延迟为179秒，而完全重新标记和再训练的延迟为993秒。结果表明，明确的概念修订可以作为数据维护信号，允许学习系统在更新依赖于变化的监督和预测状态的同时，保留仍然有效的知识。
cs.AI / 29 / 2608.23898
BenchBench-Protocol: Evaluating Real-World Wet-Lab Protocol Reasoning and Modification
BenchBench-Protocol：评估真实世界湿实验室协议推理与修改
Abstract
We introduce BenchBench-Protocol, a benchmark for large language models of 149 protocol-modification tasks recovered from modifications that scientists made to published protocols during real experimental work. Adapting a published protocol to a new experiment is a routine task for a wet-lab scientist, and a correct modification requires accounting for prior choices and downstream steps. Recent life-science benchmarks have moved toward open-ended, rubric-graded tasks, but tasks are typically elicited from experts rather than reconstructed from real-world modifications. BenchBench-Protocol tasks are derived from differences between a published protocol and a version a scientist modified, which provides the basis for the query and the weighted rubric elements for a correct response. The benchmark draws from 96 source protocols across nine domains of wet-lab biology and only includes tasks rated highly after review by domain experts. We evaluate nine closed and open models; Claude Opus 5 scores highest at 59.2% normalized rubric score, with other models between 34.1% and 47.1%, and the benchmark remains unsaturated when taking the best of ten attempts. As models are increasingly helpful in life-sciences research, evaluating them on routine wet-lab tasks becomes correspondingly important. We present BenchBench-Protocol as both a grounded assessment of wet-lab reasoning and evidence for the utility of real-world experiments to construct benchmark tasks.
Chinese Translation
我们介绍了BenchBench-Protocol，这是一个针对大型语言模型的基准，包含149个协议修改任务，这些任务源自科学家在实际实验工作中对已发布协议所做的修改。将已发布的协议调整为新的实验是湿实验室科学家的常规任务，而正确的修改需要考虑先前的选择和后续步骤。最近的生命科学基准已逐渐转向开放式、评分标准化的任务，但这些任务通常是从专家那里获取的，而不是从真实世界的修改中重建的。BenchBench-Protocol任务源自已发布协议与科学家修改版本之间的差异，这为查询和正确响应的加权评分标准元素提供了基础。该基准涵盖了来自九个湿实验室生物学领域的96个源协议，仅包括经过领域专家审查后高评分的任务。我们评估了九个封闭和开放模型；Claude Opus 5的标准化评分最高，达到59.2%，其他模型的评分在34.1%到47.1%之间，并且在进行十次尝试时，该基准仍然没有饱和。随着模型在生命科学研究中的帮助日益增大，对它们在常规湿实验室任务中的评估变得同样重要。我们将BenchBench-Protocol呈现为对湿实验室推理的扎实评估，并作为构建基准任务的真实世界实验效用的证据。
cs.AI / 30 / 2608.23906
Quantifying System-Level Harms from AI Adoption in Complex Sociotechnical Systems
量化复杂社会技术系统中人工智能采纳带来的系统级危害
Abstract
Artificial Intelligence (AI) is increasingly integrated into complex sociotechnical systems, including Critical National Infrastructure (CNI), where harms emerge from interactions between technical, human, and organisational elements. Yet current AI evaluation remains model-centric, offering little insight into how observed behaviours might translate into system-level risk. We propose a framework that links structured hazard analysis, component-level testing, and probabilistic system modelling to bridge this gap. By providing a traceable pathway from model behaviour to system-level outcomes, the framework enables practitioners to answer the "so what?" of AI failures, quantify their systemic impact, and move toward evidence-based and anticipatory governance of AI in complex systems. Applied to the UK's Real Time Gross Settlement (RTGS) system as an illustrative worked example, we derive AI-driven loss scenarios using Systems Theoretic Process Analysis (STPA) and examine adversarial manipulation of LLM-based trading as one such loss scenario. Component-level experiments show that simple adversarial inputs induce measurable behavioural shifts where AI recommendations are followed. Under the component-to-system mapping used here for a financial contagion model, these shifts alter system resilience, increasing bank failures and lowering the threshold at which shocks lead to cascading disruption, particularly under widespread or monopolistic AI adoption.
Chinese Translation
人工智能（AI）正日益融入复杂的社会技术系统，包括关键国家基础设施（CNI），在这些系统中，危害源于技术、人类和组织元素之间的相互作用。然而，目前的AI评估仍然以模型为中心，提供的洞察力有限，无法揭示观察到的行为如何转化为系统级风险。我们提出了一个框架，将结构化危害分析、组件级测试和概率系统建模联系起来，以弥补这一空白。通过提供从模型行为到系统级结果的可追溯路径，该框架使从业者能够回答AI失败的“那么呢？”的问题，量化其系统性影响，并朝着基于证据和前瞻性的AI治理迈进。以英国的实时全额结算系统（RTGS）作为说明性实例，我们使用系统理论过程分析（STPA）推导出AI驱动的损失场景，并考察基于大型语言模型（LLM）的交易的对抗性操控作为其中一个损失场景。组件级实验表明，简单的对抗性输入会在遵循AI建议的情况下引发可测量的行为变化。在这里用于金融传染模型的组件到系统映射下，这些变化会影响系统的韧性，增加银行破产的风险，并降低冲击导致级联破坏的阈值，特别是在广泛或垄断性AI采纳的情况下。
cs.AI / 31 / 2608.23908
Retrieval-augmented generation vs. deterministic tax computation in multi-agent financial advisory: A 2x2 factorial experiment
检索增强生成与确定性税务计算在多智能体财务顾问中的比较：一项2x2因子实验
Abstract
Tax-loss harvesting demonstrates consistent benefits to long-term portfolio growth; yet implementing it efficiently often involves complex considerations that are specific to the holdings within that portfolio and the individual who owns it. We introduce a custom capital gains calculation engine and a RAG-retrieved vector store of market advisory reports to provide context for a multi-agent trade recommendation system. We investigate the effects of each context provider on the quality of recommendations, measured by relative capital gains incurred during portfolio liquidation. A 2x2 repeated-measures ANOVA revealed a significant main effect of the tax optimization engine ($F(1,29) = 9.17$, $p = .005$, $\eta^2_p = .240$): enabling the engine reduced tax savings by approximately 55 percentage points relative to the no-engine conditions. The RAG main effect was not significant ($p = .841$), nor was the interaction ($p = .553$). The RAG-only condition achieved the highest descriptive mean tax savings (47.7%), and the baseline condition performed second-best (30.6%), suggesting that the pre-trained language model's internalized financial knowledge may be sufficient for competent tax-loss harvesting recommendations without explicit tooling. These results indicate that augmenting LLM agents with domain-specific computation engines does not guarantee improved performance and may introduce conflicting optimization signals.
Chinese Translation
税损收割对长期投资组合增长展现出持续的好处；然而，高效实施这一策略通常涉及复杂的考虑，这些考虑与投资组合中的持有资产及其所有者密切相关。我们引入了一个定制的资本收益计算引擎和一个基于检索增强生成（RAG）的市场咨询报告向量存储，以为多智能体交易推荐系统提供背景。我们研究了每个背景提供者对推荐质量的影响，质量通过投资组合清算期间所产生的相对资本收益来衡量。2x2重复测量方差分析显示，税务优化引擎的主要效应显著（$F(1,29) = 9.17$, $p = .005$, $ ext{η}^2_p = .240$）：启用该引擎使税收节省相较于无引擎条件减少了约55个百分点。RAG的主要效应不显著（$p = .841$），交互效应也不显著（$p = .553$）。仅使用RAG的条件达到了最高的描述性平均税收节省（47.7%），基线条件表现第二好（30.6%），这表明预训练语言模型内化的金融知识可能足以在没有明确工具的情况下提供有效的税损收割建议。这些结果表明，增强大型语言模型（LLM）代理与特定领域计算引擎并不保证性能的提升，反而可能引入相互矛盾的优化信号。
cs.AI / 32 / 2608.23911
PROOF-Gen: From Optimized Data to Better Distillation
PROOF-Gen：从优化数据到更好的蒸馏
Abstract
Supervised fine-tuning on teacher-generated trajectories is the standard first stage for distilling tool-calling capabilities into deployable models. Post-training pipelines that drive shipped tool-calling agents re-run this stage on a daily or weekly cadence, paying the frontier-teacher cost each cycle, yet the mechanism is generate-and-filter (keep the teacher's passing trajectories, discard the rest) and each cycle leaves behind the same hard scenarios because failures supply no signal. On {\tau}2-bench, 57% of teacher trials fail, two-thirds of them near-misses (most tool calls correct, undone by one decisive error). We introduce PROOF-Gen (Per-scenario Reflective Optimization to Overcome FailedGeneration), which recovers golden trajectories from these failures via per-scenario prompt optimization. For each failed task, a reflector analyzes the execution trace and evaluation feedback, then writes corrective guidance that steers the teacher to a passing trajectory. The guidance is stripped before training, so the student learns from clean demonstrations with no task-specific scaffold. On {\tau}2-bench, per-scenario optimization recovers 93% of failed scenarios. Fine-tuned on the combined data, Qwen3-4B-Instruct-2507 improves from Pass^1=0.132 to 0.529 and Gemma 4 E4B-it gains +7.2pp on BFCL v4 multi-turn. In a deployed pipeline, the method lifts trajectory quality by +6.3pp goal completion and transfers to a deployed on-device model (+1.5pp goal completion; +1.7 to +5.0pp across response-quality metrics), with positive transfer in every locale (non-English average +1.48pp).
Chinese Translation
在教师生成的轨迹上进行监督微调是将工具调用能力蒸馏到可部署模型的标准第一阶段。驱动已发布工具调用代理的后训练管道每天或每周重新运行这一阶段，每个周期都需支付前沿教师的成本，但该机制是生成-过滤（保留教师的通过轨迹，丢弃其余部分），每个周期留下相同的困难场景，因为失败没有提供信号。在{ au}2-bench上，57%的教师试验失败，其中三分之二是接近失误（大多数工具调用正确，但因一个决定性错误而未能完成）。我们提出了PROOF-Gen（逐场景反思优化以克服失败生成），通过逐场景提示优化从这些失败中恢复黄金轨迹。对于每个失败的任务，反射器分析执行轨迹和评估反馈，然后编写纠正指导，引导教师走向通过轨迹。该指导在训练前被剥离，因此学生从没有任务特定支架的干净示范中学习。在{ au}2-bench上，逐场景优化恢复了93%的失败场景。经过合并数据的微调，Qwen3-4B-Instruct-2507的表现从Pass^1=0.132提升至0.529，而Gemma 4 E4B-it在BFCL v4多轮对话中提升了7.2个百分点。在一个已部署的管道中，该方法使轨迹质量提高了6.3个百分点的目标完成率，并转移到已部署的设备模型上（目标完成率提高1.5个百分点；在响应质量指标上提高1.7至5.0个百分点），在每个地区都有正向转移（非英语平均提高1.48个百分点）。
cs.AI / 33 / 2608.23918
MARS: Multi-Specialist LLM Relay System for Competitive Programming
MARS：用于竞争编程的多专业大型语言模型中继系统
Abstract
Large Language Models excel at code generation, yet competitive programming exposes a persistent failure mode: existing multi-agent pipelines distribute work over generic planner, coder, and debugger roles and delegate the choice of algorithmic technique to the backbone alone. We present MARS (Multi-Agent Relay of Specialized LLMs), a prompt-only framework in which each agent is a topic specialist---dynamic programming, graphs, strings, geometry, and so on---grounded by retrieval-augmented generation over an algorithm-theory corpus. Given a problem, retrieval selects a small team of relevant specialists; a starter writes an initial C++17 solution, and each subsequent turn runs the candidate against public examples in a sandbox, lets the active specialist keep, repair, or hand off the draft, and forwards a structured packet to the next specialist. A single infrastructure-fixer pass normalizes boilerplate at the end. On the CodeContests test split with Gemma 4, MARS reaches $0.624 \pm 0.006$ pass rate at $2.3$ recorded pipeline stages per task ($+14.4$ percentage points over direct prompting), closing most of the gap to CodeSIM ($0.731$) at $3.3{\times}$ lower wall-clock cost and substantially smaller variance in per-task token spend. The source code is available on GitHub: https://github.com/fckand/mars.
Chinese Translation
大型语言模型在代码生成方面表现出色，但竞争编程暴露了一种持续的失败模式：现有的多智能体管道将工作分配给通用的规划者、编码者和调试者角色，并将算法技术的选择仅委托给主干系统。我们提出了MARS（多智能体专用大型语言模型中继），这是一个仅基于提示的框架，其中每个智能体都是某一主题的专家——动态规划、图论、字符串、几何等——通过对算法理论语料库的检索增强生成来支撑。给定一个问题，检索会选择一小组相关的专家；一名初始专家编写初步的C++17解决方案，随后每个回合会在沙箱中将候选方案与公共示例进行测试，让活跃的专家决定保留、修复或转交草稿，并将结构化数据包转发给下一个专家。最后，通过一次基础设施修复过程来规范化模板。在使用Gemma 4的CodeContests测试集上，MARS在每个任务的记录管道阶段为$2.3$时达到了$0.624 ext{±} 0.006$的通过率（比直接提示提高了$14.4$个百分点），并在$3.3$倍更低的实际时间成本下，缩小了与CodeSIM（$0.731$）之间的大部分差距，同时在每个任务的令牌消耗上显著降低了方差。源代码可在GitHub上获取：https://github.com/fckand/mars。
cs.AI / 34 / 2608.23922
Data Mixing as Mixture Experiment: Response Surface Methodology and Optimal Design for Large Language Model Pretraining
数据混合作为混合实验：大语言模型预训练的响应面方法与最优设计
Abstract
Data mixing is a central design problem in large language model pretraining: given a fixed token budget, practitioners must decide how much data to allocate to each domain. Recent proxy-based methods address this problem by training small models on candidate mixtures, fitting a response model, and using the response to select mixtures for larger-scale training. We show that this workflow has the structure of a classical mixture experiment. Under this view, data domains are mixture components, token shares are component proportions, proxy-training runs are experimental design points, and validation loss defines a response surface over the probability simplex. We develop this formulation using sparse second-order Scheff\'{e} response-surface models and construct model-robust $\mathcal{I}$-optimal designs for proxy data-mixing experiments. Using RegMix as an empirical case study, we demonstrate how the framework can both interpret observed mixture responses and design more efficient proxy experiments. The Scheff\'{e} analysis shows that domain value is strongly relational: several domains that are weak under additive effects become favourable through pairwise interactions, especially through combinations with web-derived text. The sparse Scheff\'{e} model preserves mixture rankings across model scales and remains competitive with a flexible machine-learning predictor while providing an explicit decomposition of additive and interaction effects. In a simulation study calibrated to observed proxy-training responses, model-robust $\mathcal{I}$-optimal designs recover the relevant mixture ordering after removing about 25\% of the original proxy runs. These results suggest that LLM data mixing should be treated not only as a prediction problem, but also as an experimental-design problem in which the proxy mixtures themselves can be chosen to improve statistical efficiency.
Chinese Translation
数据混合是大语言模型预训练中的一个核心设计问题：在固定的标记预算下，实践者必须决定将多少数据分配给每个领域。最近的基于代理的方法通过在候选混合物上训练小模型、拟合响应模型，并利用响应选择用于大规模训练的混合物来解决这个问题。我们展示了这一工作流程具有经典混合实验的结构。从这个角度来看，数据领域是混合成分，标记份额是成分比例，代理训练运行是实验设计点，而验证损失定义了概率单纯形上的响应面。我们利用稀疏的二阶Scheffé响应面模型发展了这一表述，并为代理数据混合实验构建了模型稳健的$ ext{I}$-最优设计。通过以RegMix作为实证案例研究，我们展示了该框架如何既能解释观察到的混合响应，又能设计更高效的代理实验。Scheffé分析表明，领域价值是强相关的：在加性效应下表现较弱的几个领域通过成对交互变得有利，尤其是与网络衍生文本的组合。稀疏Scheffé模型在不同模型规模中保持混合排名，并与灵活的机器学习预测器竞争，同时提供加性和交互效应的明确分解。在一次针对观察到的代理训练响应进行标定的模拟研究中，模型稳健的$ ext{I}$-最优设计在去除约25 ext{％}的原始代理运行后恢复了相关的混合排序。这些结果表明，LLM数据混合不仅应被视为预测问题，还应视为实验设计问题，其中代理混合物本身可以被选择以提高统计效率。
cs.AI / 35 / 2608.23932
Evolutionary Recurrent Decision Model in Developing Adaptive and Maladaptive Behaviors
发展适应性和不适应性行为的进化递归决策模型
Abstract
This study introduces the evolutionarily recurrent decision model (ERDM), a computational reinforcement learning framework designed to examine how evolutionary mismatch, bounded rationality, and satisficing contribute to adaptive and maladaptive behavior. ERDM simulates agents across evolutionary recurrent environments, including threat, prey/goal-pursuits, and alliances. Agents learn through competing rewards abstracted from survival metrics. A validity study under varying adverse childhood experiences demonstrates that distinct adaptive and maladaptive strategies, such as learned helplessness, avoidance, healthy relationships, and aggression, emerge naturally without being hardwired. These results align with empirical literature, showcasing ecological validity. The results suggest that many psychopathology-relevant aspects may be interpreted as bounded cognitive systems operating under modern-ancestral environmental mismatch, positioning ERDM as a key computational cognitive tool that can be extended to other studies.
Chinese Translation
本研究介绍了进化递归决策模型（ERDM），这是一种计算强化学习框架，旨在探讨进化失配、有限理性和满意化如何影响适应性和不适应性行为。ERDM 模拟了在进化递归环境中行动者的行为，包括威胁、猎物/目标追求和联盟。行动者通过从生存指标中抽象出的竞争奖励进行学习。在不同的不良童年经历下进行的有效性研究表明，诸如习得性无助、回避、健康关系和攻击性等不同的适应性和不适应性策略，自然出现而非固有。这些结果与实证文献相一致，展示了生态有效性。结果表明，许多与精神病理相关的方面可以被解释为在现代-祖先环境失配下运作的有限认知系统，从而将 ERDM 定位为一个关键的计算认知工具，可以扩展到其他研究中。
cs.AI / 36 / 2608.23941
More Rejective, Not More Discriminative: The Unit of Verification in Pre-Execution LLM Oversight
更具拒绝性，而非更具区分性：预执行大型语言模型监督中的验证单元
Abstract
Pre-execution oversight is core to trusted monitoring in AI control: a fallible LLM monitor vets planned actions before irreversible execution. Over-blocking forfeits usefulness and pressures deployers to disable it. Every protocol must fix a unit of verification: how many actions one call reviews. Existing designs take the unit as given; its effect on fallible monitors is unmeasured. Natural traces cannot isolate it: review length co-varies with error type and position. Catch alone misleads: rejecting everything catches everything. Measuring this needs boundary variation alone and a matched clean control. We introduce the twin-prefix framework, which supplies both. Each gold plan yields a prefix with one injected, environment-accepted error and a clean twin differing in one write. Judging each pair at five nested lengths ties verdict changes to the unit alone. Discrimination is scored by pre-registered informedness, catch minus false rejection. Longer review raises catch; false rejection climbs in lockstep. Informedness peaks at one or two actions for all six judges in both domains: longer windows make zero-shot monitors more rejective, not more discriminative. Replaying withheld observations traces the failure largely to observation deprivation. Safety cases should state the unit and co-report the clean series. Our framework is the first controlled, pre-registered instrument for this choice and never reads catch alone. Our calibrated short unit recovers up to 0.95 informedness over eight-action review, and no tested label-blind policy consistently beats it.
Chinese Translation
预执行监督是人工智能控制中可信监测的核心：一个易出错的大型语言模型监测器在不可逆执行之前审核计划的行动。过度阻止会损失实用性，并迫使部署者禁用该功能。每个协议必须确定一个验证单元：一次调用审核多少个行动。现有设计将该单元视为既定；其对易出错监测器的影响尚未被测量。自然轨迹无法将其孤立：审核长度与错误类型和位置共同变化。仅仅依赖捕捉会产生误导：拒绝所有内容会捕捉到所有内容。测量这一点需要仅仅边界变化和匹配的干净控制。我们引入了双前缀框架，提供了两者。每个金标准计划生成一个前缀，其中注入了一个环境接受的错误和一个在一次写入上有所不同的干净双胞胎。以五个嵌套长度评判每对，将裁决变化与单元联系在一起。区分性通过预注册的知情度进行评分，捕捉减去虚假拒绝。较长的审核提高了捕捉率；虚假拒绝同步上升。在所有六位评审者的两个领域中，知情度在一到两个行动时达到峰值：较长的窗口使得零样本监测器更具拒绝性，而非更具区分性。重放被 withheld 的观察结果将失败主要归因于观察剥夺。安全案例应声明该单元并共同报告干净系列。我们的框架是第一个受控的、预注册的工具，用于这一选择，并且从不单独读取捕捉。我们校准的短单元在八个行动审核中恢复了高达0.95的知情度，而没有经过测试的标签盲政策始终无法超越它。
cs.AI / 37 / 2608.23956
Recursive Agentic Reasoning
递归代理推理
Abstract
Test-time reasoning methods such as iterative refinement, decomposition, and repeated sampling are often evaluated in isolation, making their gains difficult to compare across models, benchmarks, and evaluation pipelines. We introduce a unified view of these methods as recursion operators over an agent's reasoning trace: GROW, which deepens a single reasoning path; PRUNE, which decomposes and recomposes the problem; and BRANCH, which samples alternative reasoning paths and selects among them. We evaluate all three operators against a single-pass chain-of-thought baseline under a shared harness with identical prompts, token budgets, and grading code. Across five benchmarks and three frontier models, comprising 14 model-benchmark settings, 49,327 graded items, and 151,876 model calls, BRANCH improves accuracy in all 14 settings by an average of 5.98 percentage points and is the best-performing operator in 12. In contrast, GROW yields a mean gain of 2.18 points and degrades performance in two settings, while PRUNE improves accuracy by 0.94 points on average. Analysis shows that BRANCH's advantage arises not only from exploring multiple reasoning paths, but also from recovering from truncation: its gains strongly correlate with the baseline rate of empty, budget-exhausted outputs (r = 0.72). These results weaken the hypothesis that different problems require routing among test-time reasoning operators; at this level of abstraction, repeated branching is consistently dominant. Finally, we show that unpaired evaluation and treating scoring-pipeline failures as model errors can materially change, and even reverse, comparative conclusions, motivating paired scoring as a standard protocol for test-time-compute evaluation.
Chinese Translation
测试时推理方法，如迭代精炼、分解和重复采样，通常是孤立评估的，这使得它们的增益在模型、基准和评估流程之间难以比较。我们将这些方法统一视为代理推理轨迹上的递归算子：GROW（扩展），它加深单一推理路径；PRUNE（剪枝），它分解并重新组合问题；以及BRANCH（分支），它采样替代推理路径并在其中选择。我们在共享的框架下，以相同的提示、令牌预算和评分代码，针对单次链式思维基线评估这三种算子。在五个基准和三个前沿模型中，涵盖14个模型-基准设置、49,327个评分项目和151,876次模型调用，BRANCH在所有14个设置中平均提高了5.98个百分点的准确率，并且在12个设置中表现最佳。相比之下，GROW的平均增益为2.18点，并在两个设置中导致性能下降，而PRUNE平均提高了0.94点的准确率。分析表明，BRANCH的优势不仅源于探索多个推理路径，还源于从截断中恢复：其增益与基线空输出率（预算耗尽的输出）之间有强相关性（r = 0.72）。这些结果削弱了不同问题需要在测试时推理算子之间进行路由的假设；在这个抽象层面上，重复分支始终占据主导地位。最后，我们展示了未配对评估和将评分流程失败视为模型错误可以实质性改变甚至逆转比较结论，促使将配对评分作为测试时计算评估的标准协议。
cs.AI / 38 / 2608.23962
More GPUs or a Smaller Cache? Tensor Parallelism versus KV Compression for Memory-Bound LLM Serving
更多的GPU还是更小的缓存？张量并行与KV压缩在内存受限的大型语言模型服务中的比较
Abstract
When an LLM serving deployment runs out of KVcache room, there are two well-established ways out. Tensor parallelism shards the weights and the KV cache across two, four, or eight devices, buying memory headroom at the price of an all-reduce on every layer and a hardware bill that grows with the device count. The algorithms community shrinks the cache in place, with KV quantisation and eviction keeping a single GPU and spending a little quality instead. Compression papers report memory ratios, parallel-scaling papers report throughput curves, and almost nobody puts the two on the same cost axis. We place tensor-parallel configurations (degree 1 to 8) and KV-compressed configurations (16/8/4-bit, keep-ratios down to 0.25) on one costnormalised axis, cost per million tokens against latency, using a profiled simulator calibrated on A100, A40, and H100 hardware, and we go looking for the cost-equivalence crossover. We do not find one. Across two models (Llama-2 at 7B and 70B), three GPU types, and every level of memory relief we could construct, compression is cheaper by 1.20x to 2.00x. A 7B model on an 80 GB device cannot exhaust its KV budget within its own context window, and the boundary that decides between the strategies is model size relative to device memory, at roughly 36B parameters for an 80 GB card. Below that wall, compression dominates and extra GPUs are largely wasted spend; above it, tensor parallelism stops being a choice and becomes an entry ticket: Llama-2-70B is infeasible on one A100 at any KV setting, because the binding resource is weights, which KV compression does not touch. Tensor parallelism is the only lever that improves latency (compression makes per-token latency worse, by 8 to 93%, through batching contention), while compression is the only lever that multiplies capacity per dollar (16.5x, against 1.21x for an eightfold spend on GPUs).
Chinese Translation
当大型语言模型（LLM）服务部署的KV缓存空间耗尽时，有两种成熟的解决方案。张量并行将权重和KV缓存分片到两个、四个或八个设备上，以换取每层的全归约操作和随着设备数量增加而增长的硬件成本。算法社区则通过KV量化和驱逐在原地缩小缓存，保持单个GPU并牺牲一些质量。压缩论文报告内存比率，平行扩展论文报告吞吐曲线，但几乎没有人将两者放在同一成本轴上。我们将张量并行配置（度数1到8）和KV压缩配置（16/8/4位，保持比率低至0.25）放在一个成本归一化轴上，比较每百万个标记的成本与延迟，使用在A100、A40和H100硬件上校准的分析模拟器，并寻找成本等效的交叉点。我们没有找到。在两个模型（Llama-2的7B和70B）、三种GPU类型以及我们能够构建的每个内存缓解级别中，压缩的成本低于1.20倍到2.00倍。一个7B模型在80GB设备上无法在其自身的上下文窗口内耗尽其KV预算，而决定两种策略之间的边界是模型大小相对于设备内存，约为80GB卡的36B参数。在这个界限以下，压缩占主导地位，额外的GPU大多是浪费；在这个界限以上，张量并行不再是选择，而成为入场券：Llama-2-70B在任何KV设置下在一台A100上都是不可行的，因为限制资源是权重，而KV压缩并不影响。张量并行是唯一能改善延迟的杠杆（压缩通过批处理争用使每个标记的延迟恶化8%到93%），而压缩是唯一能在每美元中增加容量的杠杆（16.5倍，而八倍花费在GPU上的增益为1.21倍）。
cs.AI / 39 / 2608.23970
Giraffe: A Mapping Architecture from Hidden Text Representations to Visual Embeddings for Efficient Graphic Design
长颈鹿：一种从隐藏文本表示到视觉嵌入的映射架构，用于高效图形设计
Abstract
Multimodal large language models (MLLMs) have made significant progress in understanding and interpreting mul- timedia content. However, their ability to generate me- dia remains limited. Recent approaches have attempted to bridge this gap by translating the hidden representations of token sequences into the embedding space of visual models or directly into raw image data. However, these methods often represent each image using multiple specialised to- kens which significantly increases the input length. This be- comes a major limitation for tasks such as graphic design generation where the output typically involves a seamless blend of thousands of tokens across text, multiple images, and layout information. To address this challenge, a novel architecture is proposed that maps hidden token represen- tations to the embedding space of visual models, such as CLIP ViT-L/14, using a single [IMG] token per image. The architecture employs two shallow MLP blocks, each with a separate compression module followed by a shared expan- sion module, trained with six distinct loss functions. One block aids the other during training and is omitted during inference, resulting in a lightweight solution. Strong perfor- mance is demonstrated in both image-to-design and text-to- design generation tasks.
Chinese Translation
多模态大型语言模型（MLLMs）在理解和解释多媒体内容方面取得了显著进展。然而，它们在生成媒体方面的能力仍然有限。最近的方法试图通过将标记序列的隐藏表示转换为视觉模型的嵌入空间或直接转换为原始图像数据来弥补这一差距。然而，这些方法通常使用多个专门的标记来表示每个图像，这显著增加了输入长度。这对于图形设计生成等任务来说是一个主要限制，因为输出通常涉及文本、多个图像和布局信息之间无缝融合的数千个标记。为了解决这一挑战，提出了一种新颖的架构，该架构使用每个图像一个单一的[IMG]标记，将隐藏的标记表示映射到视觉模型的嵌入空间，例如CLIP ViT-L/14。该架构采用两个浅层多层感知机（MLP）模块，每个模块具有单独的压缩模块，后面跟随一个共享的扩展模块，并使用六种不同的损失函数进行训练。一个模块在训练过程中辅助另一个模块，而在推理时省略，从而实现轻量级解决方案。在图像到设计和文本到设计生成任务中表现出强大的性能。
cs.AI / 40 / 2608.23978
When Seeing Is Not Enough: Benchmarking Interactive Visual Grounding in LVLMs
当视觉不再足够：大型视觉语言模型中的交互式视觉定位基准测试
Abstract
Visual grounding is typically evaluated as a one-shot mapping from an informative referring expression to a visual target. This formulation misses a central property of real-world reference: target information is often incomplete, ambiguous, and established through interaction. We introduce a controlled evaluation framework for interactive visual grounding in large vision-language models (LVLMs), varying how much target information is provided upfront and how much must be acquired through dialogue. Across four human-grounded visual contexts and four interaction protocols, current LVLMs perform significantly below task-level human baselines. Interaction can help when follow-up questions refine or repair an initial target description. Performance is lowest when no initial description is provided and target information must be acquired through questions, indicating that proactive question-driven grounding remains difficult. LVLMs are also poorly calibrated, often reporting confidence that exceeds their empirical accuracy. Follow-up studies confirm these patterns across varied description sources (human versus AI), reasoning efforts, repeated interactions, description providers, and visual contexts. Overall, interactive visual grounding remains an important challenge, requiring visual matching, information seeking and synthesis.
Chinese Translation
视觉定位通常被评估为从信息丰富的指称表达到视觉目标的一次性映射。这种表述忽略了现实世界引用的一个核心特性：目标信息往往是不完整的、模糊的，并且通过互动建立。我们引入了一个控制评估框架，用于在大型视觉语言模型（LVLMs）中进行交互式视觉定位，改变了预先提供的目标信息量以及通过对话必须获取的信息量。在四个以人为基础的视觉上下文和四种互动协议中，目前的LVLMs在任务级人类基准下的表现显著低于预期。当后续问题能够细化或修正初始目标描述时，互动可以提供帮助。当没有提供初始描述且目标信息必须通过提问获取时，表现最低，这表明主动的基于问题的定位仍然困难。LVLMs的校准也很差，通常报告的置信度超过其实际准确性。后续研究确认了这些模式在不同描述来源（人类与人工智能）、推理努力、重复互动、描述提供者和视觉上下文中的一致性。总体而言，交互式视觉定位仍然是一个重要的挑战，需要视觉匹配、信息寻求和综合。
cs.AI / 41 / 2608.23979
Rules Before Oracles: Auditable, User-Configurable Argument Selection for Deliberative Polling
规则优于神谕：可审计的用户可配置论点选择用于审议性投票
Abstract
In a deliberative poll, once submissions outnumber what anyone will read, some mechanism chooses which arguments each voter sees, acquiring much of the decision; practice delegates it to opaque learned rankers, so a voter cannot recompute or contest the exposure that shaped their vote. We ask whether it can be a published rule over publicly recomputable evidence with parameters held by the voter, treating legibility as an admissibility condition on usable mechanisms, not an objective traded against accuracy. We formalise a poll over bipolar justification sets, judging a slate by reason coverage, the order it arrives in, and captured endorsement mass; we give seven checkable criteria for a civic recommender and a rule meeting them: a one-hop reversed endorsement flow parameterised by a relation-weight function. An agentic simulator records every slate at every vote, over about 17,000 seed-paired runs. Served slates fall 0.035 short of a label-reading ceiling upper-bounding every selection procedure, opaque ones included: any unconstrained ranker's advantage is bounded and small. On coverage alone, with non-degenerate authoring, the rule is indistinguishable from a random slate, a null due to an order-blind, charity-blind instrument; on the other two it leads at every prefix by a margin widening with adversarial pressure and dominates on mass by a factor of 3.3. Once a realistic fraction of submissions carries no reasons, the coverage margin returns and grows. Label-homogeneous flooding collapses completeness from 0.81 to 0.34 under a flat weight policy, only to 0.44 under author-count normalisation, making the weight function a security control worth 10% of completeness. The choice between ranking arms is a position on a coverage-versus-mass frontier, not a fact, the kind of choice only a legible rule can hand to the person it affects. It maps onto an open-source peer-to-peer platform.
Chinese Translation
在审议性投票中，一旦提交的论点数量超过任何人能够阅读的数量，就需要某种机制来选择每位选民所看到的论点，从而影响决策；目前的做法是将其委托给不透明的学习排名器，因此选民无法重新计算或质疑影响其投票的曝光内容。我们探讨是否可以通过一个公开的规则，基于可公开计算的证据，并由选民掌握参数，将可读性视为可用机制的可接受性条件，而不是与准确性相对立的客观因素。我们形式化了一个关于双极辩护集的投票，依据理由覆盖、到达顺序和所捕获的支持量来评判候选集；我们给出了七个可检查的公民推荐标准及满足这些标准的规则：一个由关系权重函数参数化的一跳反向支持流。一个自主模拟器记录每个投票中的每个候选集，进行了大约17,000次种子配对的运行。所提供的候选集在每种选择程序中都比标签阅读上限低0.035，包括不透明的选择程序：任何不受约束的排名器的优势都是有限且微小的。仅在覆盖方面，采用非退化的作者创作，该规则与随机候选集无法区分，这是由于一个对顺序和慈善盲目的工具导致的无效；在其他两个方面，它在每个前缀上以逐渐扩大的边际领先，并在支持量上以3.3的倍数占据主导地位。一旦现实中提交的论点中有相当一部分没有理由，覆盖边际便会恢复并增长。标签同质的泛滥在平坦权重政策下将完整性从0.81降至0.34，而在作者计数归一化下仅降至0.44，使得权重函数成为值得10%完整性的安全控制。对排名选项的选择是覆盖与支持量边界上的一种立场，而不是一个事实，这种选择只有可读的规则才能交给受其影响的人。它映射到一个开源的点对点平台上。
cs.AI / 42 / 2608.23982
Memory Is Not Always Needed: Characterizing Conditional Memory in Scientific Reasoning
记忆并非总是必要：科学推理中的条件记忆特征化
Abstract
Scientific reasoning requires language models to retrieve specialized knowledge and incorporate it reliably into multi-step computation. Conditional memory provides an explicit lookup pathway that complements dense neural representations, but its usefulness is inherently input- and computation-dependent: retrieved information may repair missing scientific associations, yet it may also introduce distracting shortcuts or interfere with reasoning that the base model can already perform correctly. In this work, we systematically investigate when, where, and to what extent conditional memory should participate in scientific reasoning. We characterize the scientific knowledge boundary and controlled interventions on memory-enabled knowledge-circuit nodes. Based on these analyses, we propose a Knowledge Boundary-Aware Router that uses task-specific input proxies available before generation to determine whether memory is activated, which layer-stage nodes receive memory signals, and how strongly these signals contribute. Experiments on biological and chemical reasoning benchmarks, covering two backbone families and six task types, show that memory effects vary substantially across inputs, tasks, and injection locations. Compared with static and activation-rate-matched random routing, our approach more consistently preserves beneficial memory contributions while suppressing memory-induced regressions, establishing selective memory allocation as an important principle for reliable scientific reasoning.
Chinese Translation
科学推理要求语言模型检索专业知识并将其可靠地纳入多步骤计算中。条件记忆提供了一条明确的查找路径，以补充密集的神经表示，但其有效性本质上依赖于输入和计算：检索到的信息可能修复缺失的科学关联，但也可能引入干扰性的捷径或干扰基础模型已经能够正确执行的推理。在本研究中，我们系统地探讨了条件记忆在科学推理中应何时、何地以及在多大程度上参与。我们对科学知识边界和对启用记忆的知识电路节点的控制干预进行了特征化。基于这些分析，我们提出了一种知识边界感知路由器（Knowledge Boundary-Aware Router），该路由器利用生成前可用的任务特定输入代理来决定是否激活记忆、哪些层级节点接收记忆信号以及这些信号的贡献强度。在生物和化学推理基准测试中的实验显示，记忆效应在输入、任务和注入位置之间存在显著差异。与静态和激活率匹配的随机路由相比，我们的方法更一致地保留了有益的记忆贡献，同时抑制了记忆引起的回归，确立了选择性记忆分配作为可靠科学推理的重要原则。
cs.AI / 43 / 2608.24001
Diverse by Reasoning: Harnessing the Wisdom of LLM Crowds for Future Prediction
通过推理实现多样性：利用大语言模型群体的智慧进行未来预测
Abstract
Large language models (LLMs) are increasingly used for future prediction, motivating the use of multiple models as a wisdom-of-the-crowd mechanism. However, simply increasing crowd size does not guarantee effective diversity, as different LLMs may exhibit redundant behaviors. We propose a behavior-aware framework for constructing diverse LLM crowds. The framework characterizes models using their reasoning traces on independent development tasks, clusters models by behavioral similarity, and selects representatives for collective prediction. We evaluate 25 LLMs using seven development benchmarks for behavioral diversity modeling and two future-prediction benchmarks for evaluating diverse crowds' performance. Our results show that crowd composition can matter more than crowd size: a three-model medoid crowd based on K-means++ behavioral clustering outperforms conventional voting over all 25 models on both prediction benchmarks, while reducing model calls by 88% and inference cost by approximately 80%. The results further suggest that representative behavioral diversity, rather than simply maximizing diversity, is important for constructing effective LLM crowds
Chinese Translation
大语言模型（LLMs）在未来预测中的应用日益增多，这促使我们将多个模型作为群体智慧机制。然而，单纯增加群体规模并不能保证有效的多样性，因为不同的LLM可能表现出冗余的行为。我们提出了一种行为感知框架，用于构建多样化的LLM群体。该框架通过独立开发任务上的推理轨迹来表征模型，依据行为相似性对模型进行聚类，并选择代表性模型进行集体预测。我们使用七个开发基准对25个LLM进行行为多样性建模评估，并使用两个未来预测基准评估多样化群体的性能。我们的结果表明，群体组成的重要性可能超过群体规模：基于K-means++行为聚类的三模型中位数群体在两个预测基准上均优于对所有25个模型的传统投票，同时减少了88%的模型调用和约80%的推理成本。结果进一步表明，代表性行为多样性，而非单纯最大化多样性，对于构建有效的LLM群体至关重要。
cs.AI / 44 / 2608.24005
Incorporating Cognitive Load and Knowledge Transfer for Multi-Domain Knowledge Tracing
结合认知负荷与知识转移的多领域知识追踪
Abstract
Knowledge Tracing (KT) aims to assess students' dynamic knowledge states from their learning histories. While most existing KT methods focus on single-domain learning with notable success, real-world learning scenarios often involve multiple domains simultaneously, introducing two critical factors: 1) Cognitive load, arising from managing learning across domains in both temporal and knowledge dimensions. 2) Knowledge transfer, where knowledge states in one domain influence related states both within and across domains. In this paper, we focus on exploring these factors to improve students' knowledge state assessment in multi-domain learning scenarios and propose a novel method incorporating cognitive Load and knowledge Transfer for Multi-domain Knowledge Tracing (LT-MKT). Specifically, to bridge isolated domains, LT-MKT first integrates textual information from questions and their associated concepts to construct a Multi-domain Hierarchical Graph, leveraging the advanced representational capabilities of large language models (LLMs). Then, cross-domain features in both the temporal and knowledge dimensions are explicitly modeled to capture the effects of cognitive load. Additionally, a knowledge transfer module is designed to model the propagation of knowledge states within and across domains. By jointly modeling these factors, LT-MKT enables more accurate prediction of students' future performance. Finally, extensive experiments on real-world datasets demonstrate that our method achieves state-of-the-art performance.
Chinese Translation
知识追踪（Knowledge Tracing, KT）旨在通过学生的学习历史评估其动态知识状态。尽管现有的大多数KT方法在单领域学习中取得了显著成功，但现实学习场景往往涉及多个领域的同时学习，这引入了两个关键因素：1）认知负荷，源于在时间和知识维度上管理跨领域学习的复杂性；2）知识转移，一个领域的知识状态会影响与之相关的领域内及跨领域的状态。本文重点探讨这些因素，以改善多领域学习场景中学生知识状态的评估，并提出了一种新方法，结合认知负荷与知识转移的多领域知识追踪（LT-MKT）。具体而言，为了连接孤立的领域，LT-MKT首先整合问题及其相关概念的文本信息，构建多领域层次图，利用大型语言模型（Large Language Models, LLMs）的先进表示能力。然后，在时间和知识维度上显式建模跨领域特征，以捕捉认知负荷的影响。此外，设计了一个知识转移模块，以建模领域内及跨领域知识状态的传播。通过联合建模这些因素，LT-MKT能够更准确地预测学生的未来表现。最后，在真实世界数据集上的大量实验表明，我们的方法达到了最先进的性能。
cs.AI / 45 / 2608.24015
Reflection with Action-Induced Visual Differences for Desktop GUI Agents
基于动作引发的视觉差异的桌面GUI代理反思
Abstract
The Planner-Operator-Reflector (POR) framework is widely used in GUI agents to maintain objective alignment in complex tasks through modular collaboration. However, desktop GUIs introduce a key challenge: large, dense interfaces often exhibit subtle or scattered state changes, placing most of the burden on the reflector, which must compare pre- and post-action screens, while the planner and operator reason over a single state. Existing reflectors collapse change detection and outcome verification into one step, leaving evidence implicit and yielding weakly grounded decisions. To address this limitation, we propose Evidence-First Reflection (EFR), a two-stage reflector that explicitly decouples action-induced visual differences extraction from outcome verification. EFR identifies the action location and candidate changed regions with Set-of-Marks annotations, describes and filters action-relevant changes, and makes the final judgment from the cleaned evidence. This evidence-reasoning decoupled design makes reflection better grounded in screen transitions, while reducing both visual search complexity and reasoning burden. Experiments on OSWorld-Verified and WindowsAgentArena demonstrate that EFR improves reflector accuracy by 7.11%, yielding average end-to-end task success gains of 5.94% and 4.95% on the two benchmarks, respectively.
Chinese Translation
规划者-操作员-反思者（Planner-Operator-Reflector, POR）框架广泛应用于GUI代理，以通过模块化协作在复杂任务中保持目标一致性。然而，桌面GUI引入了一个关键挑战：大型、密集的界面往往表现出微妙或分散的状态变化，这将大部分负担放在反思者身上，反思者必须比较动作前后的屏幕，而规划者和操作员则只需对单一状态进行推理。现有的反思者将变化检测和结果验证合并为一步，导致证据隐含且决策基础薄弱。为了解决这一局限性，我们提出了证据优先反思（Evidence-First Reflection, EFR），这是一种两阶段的反思者，明确将动作引发的视觉差异提取与结果验证解耦。EFR通过标记集（Set-of-Marks）注释识别动作位置和候选变化区域，描述并过滤与动作相关的变化，并从清理后的证据中做出最终判断。这种证据推理解耦设计使反思在屏幕过渡中更具基础，同时减少了视觉搜索复杂性和推理负担。在OSWorld-Verified和WindowsAgentArena上的实验表明，EFR提高了反思者的准确性7.11%，在这两个基准上分别带来了5.94%和4.95%的平均端到端任务成功率提升。
cs.AI / 46 / 2608.24024
Beyond Confidence: Test-Time Scaling for Multi-Turn Search Agents via Retrieval Grounding
超越置信度：基于检索基础的多轮搜索代理的测试时缩放
Abstract
Confidence-based voting aggregates parallel LLM rollouts by weighting each with internal signals such as token log probabilities, and has been actively studied for single-turn reasoning. However, modern LLMs increasingly act as multi-turn search agents that retrieve and condition on external documents. In this paper, we show that confidence-based voting transfers poorly to this multi-turn setting, and identify the underlying failure reason as copy inflation: when retrieved documents are appended to an agent's context, tokens copied from those documents receive systematically inflated log probabilities. This flattens confidence scores within each question and weakens the resulting weighted vote. To address this issue, we propose Retrieval-Grounded Voting (RGV), which scores each rollout by the lexical overlap between its final answer and the documents it retrieved. By computing the signal outside the contaminated context, RGV sidesteps both token log probabilities and additional LLM calls. Across four search-agent benchmarks and five LLMs, RGV consistently outperforms confidence-based voting, with gains of up to +5.4% accuracy and +35% on minority-correct questions, where the correct answer appears in only 1-2 of 8 rollouts.
Chinese Translation
基于置信度的投票通过对并行的大型语言模型（LLM）回滚进行加权，利用内部信号如标记日志概率，已被积极研究用于单轮推理。然而，现代LLM越来越多地充当多轮搜索代理，检索并依赖外部文档。在本文中，我们表明基于置信度的投票在这一多轮设置中效果不佳，并识别出其根本失败原因是复制膨胀：当检索到的文档被附加到代理的上下文中时，从这些文档复制的标记会系统性地获得膨胀的日志概率。这使得每个问题内的置信度评分变得平坦，从而削弱了最终的加权投票。为了解决这个问题，我们提出了检索基础投票（Retrieval-Grounded Voting, RGV），它通过最终答案与检索到的文档之间的词汇重叠来对每次回滚进行评分。通过在受污染的上下文之外计算信号，RGV避开了标记日志概率和额外的LLM调用。在四个搜索代理基准和五个LLM上，RGV始终优于基于置信度的投票，准确率提高了最高5.4%，在少数正确问题上提高了35%，其中正确答案仅出现在8次回滚中的1-2次。
cs.AI / 47 / 2608.24041
Relative Time Intervals Representation for Word-level Timestamping with Masked Training
基于相对时间间隔表示的词级时间戳标注与掩码训练
Abstract
Although Speech Large Language Models (SpeechLLMs) excel at speech understanding and generation, their capacity for fine-grained, temporally aligned outputs remains underexplored. Our work addresses this gap by enabling SpeechLLMs to jointly model speech content and temporal structure, effectively transforming them from ``content understanding machines" into ``temporal-aware content understanding machines". Specifically, we replace traditional absolute timestamps with relative timestamps, achieving a more compact vocabulary and stronger generalization capabilities. To efficiently infuse timestamp prediction ability into pre-trained large language models, we introduce a hybrid fine-tuning strategy: full-parameter fine-tuning of the timestamp-augmented embedding layer and language model head, combined with LoRA fine-tuning of the decoder layers. Moreover, we design a masked timestamp training objective, preventing the model from over-relying on ground-truth timestamps, and thereby enhancing robustness against noisy real-world annotations. Extensive experiments demonstrate that our approach achieves significant improvements in timestamp prediction accuracy while maintaining strong speech transcription performance.
Chinese Translation
尽管语音大型语言模型（SpeechLLMs）在语音理解和生成方面表现出色，但它们在细粒度、时间对齐输出方面的能力仍未得到充分探索。我们的工作旨在填补这一空白，使SpeechLLMs能够联合建模语音内容和时间结构，有效地将其从“内容理解机器”转变为“时间感知内容理解机器”。具体而言，我们用相对时间戳替代传统的绝对时间戳，从而实现更紧凑的词汇和更强的泛化能力。为了高效地将时间戳预测能力注入预训练的大型语言模型中，我们引入了一种混合微调策略：对增强时间戳的嵌入层和语言模型头进行全参数微调，同时结合对解码器层的LoRA微调。此外，我们设计了一种掩码时间戳训练目标，以防止模型过度依赖真实时间戳，从而增强其对噪声真实世界标注的鲁棒性。大量实验表明，我们的方法在时间戳预测准确性方面取得了显著提升，同时保持了强大的语音转录性能。
cs.AI / 48 / 2608.24046
Algorithmic Impact Reveals the Hidden Social Choice Structure of Alignment
算法影响揭示了对齐的隐含社会选择结构
Abstract
When an AI algorithm makes decisions that affect more than one person, aligning it becomes a problem of social choice: how should people's divergent preferences about system behavior be reconciled and aggregated into a single coherent model? The standard approach to aligning frontier AI models$\unicode{x2013}$reinforcement learning from human feedback$\unicode{x2013}$largely sidesteps this question and has poor social choice guarantees. However, it remains unclear what alternative should replace it. We show that, by focusing directly on an algorithm's welfare consequences, the alignment problem can be reformulated as linear optimization over a convex impact space, which makes it amenable to the standard toolkit of welfare economics and mechanism design. This reformulation clarifies how alignment protocols translate into welfare consequences and, conversely, how a social planner's desired constraints on welfare consequences can be translated back into alignment protocols. We apply this transformation to show that voting-by-issues and random-dictatorship mechanisms are strategyproof and unanimous. Demonstrating the reverse direction, we also apply the impact representation to derive a family of alignment protocols that maximize utilitarian social welfare subject to various social desiderata, such as bounds on individual or group harm. We illustrate the welfare implications of these alignment protocols empirically using real human preferences over kidney allocation, charitable food distribution, LLM responses, and trolley problems.
Chinese Translation
当一个人工智能算法做出影响多个人的决策时，对齐问题就成为了一个社会选择问题：人们对系统行为的不同偏好应如何协调并汇总成一个统一的模型？对齐前沿人工智能模型的标准方法——基于人类反馈的强化学习——在很大程度上回避了这个问题，并且缺乏良好的社会选择保证。然而，目前尚不清楚应该用什么替代方案。我们展示，通过直接关注算法的福利后果，对齐问题可以重新表述为在凸影响空间上的线性优化，这使其适合福利经济学和机制设计的标准工具包。这种重新表述澄清了对齐协议如何转化为福利后果，反之亦然，社会规划者对福利后果的期望约束如何可以转化回对齐协议。我们应用这一转化，展示了基于议题投票和随机独裁机制是策略无关且一致的。通过反向演示，我们还应用影响表示法推导出一系列对齐协议，这些协议在满足各种社会期望（如对个体或群体伤害的限制）的条件下最大化功利主义社会福利。我们通过对肾脏分配、慈善食品分配、大型语言模型（LLM）响应和电车难题等真实人类偏好的实证研究，阐明了这些对齐协议的福利影响。
cs.AI / 49 / 2608.24069
Poisoning Agentic Alpha: Adversarial Vulnerabilities Across Roles and Architectures in Multi-Agent Trading Systems
毒化代理阿尔法：多智能体交易系统中角色与架构的对抗脆弱性
Abstract
LLM-based multi-agent trading systems, in which specialized agents collaborate through structured communication to produce trading decisions, are moving rapidly from research prototypes to live deployments that control real assets. The same inter-agent communication that makes them effective also exposes them: a corrupted signal can propagate to the final decision and translate into realized financial loss. Unlike prior attacks that presume privileged access to system internals, we restrict the adversary to what is practically reachable---the source data and prompts agents consume---yielding a low-barrier, and thus democratized threat model instantiated as role-specific adversaries. We present the first systematic empirical study in the financial domain to characterize how an adversarial signal enters a multi-agent trading system and how far it survives toward the decision. Along the role axis, we decompose a widely-used trading pipeline into four functional roles---Analyst, Researcher, Trader, and Risk Manager---and pair each with an attack matched to its interface. Along the structural axis, we evaluate four communication topologies under data- and agent-level attacks, using the Adversarial Signal Preservation Score (APS) as a post-hoc lens on why some designs are more robust than others. We conduct experiments across five assets, two backbones, and two target directions. A central finding is that no architecture is inherently robust. These findings provide insights for the future design of safer and more robust agentic trading systems.
Chinese Translation
基于大型语言模型（LLM）的多智能体交易系统中，专业化的代理通过结构化的沟通协作以产生交易决策，这些系统正迅速从研究原型转变为控制真实资产的实时部署。使其有效的代理间沟通同样使其暴露于风险之中：一个被腐蚀的信号可以传播到最终决策并转化为实际的财务损失。与之前假设对系统内部有特权访问的攻击不同，我们将对手限制在实际可达的范围内——即代理所消耗的源数据和提示——从而形成一个低门槛、民主化的威胁模型，具体表现为角色特定的对手。我们首次在金融领域进行系统的实证研究，以表征对抗信号如何进入多智能体交易系统以及其在决策过程中存活的程度。在角色维度上，我们将一个广泛使用的交易流程分解为四个功能角色——分析师、研究员、交易员和风险管理者——并为每个角色配对一个与其接口相匹配的攻击。在结构维度上，我们评估了四种通信拓扑在数据和代理级攻击下的表现，使用对抗信号保留评分（Adversarial Signal Preservation Score, APS）作为后验视角，探讨为何某些设计比其他设计更具鲁棒性。我们在五种资产、两个基础模型和两个目标方向上进行了实验。一个核心发现是，没有任何架构是固有鲁棒的。这些发现为未来更安全、更鲁棒的智能交易系统设计提供了见解。
cs.AI / 50 / 2608.24070
Compression Trinity: Exploring Sparsity, Quantization, and Low-Rank Approximations for LLM Compression
压缩三位一体：探索稀疏性、量化和低秩近似在大语言模型压缩中的应用
Abstract
Prohibitive computational and environmental costs impede the scalable deployment of Large Language Models (LLMs). Traditional compression techniques (sparsity, quantization, low-rank approximations) are typically applied in isolation, and each hits an accuracy-efficiency wall. This thesis proposes the "Compression Trinity," a unified framework that applies the three pillars jointly: sparsity to reduce computation, quantization to minimize memory bandwidth, and low-rank approximations to recover accuracy. To accelerate pretraining, we apply the Trinity to the optimizer and model architecture. MKOR approximates curvature via block-diagonal sparsity and low-rank inversion, maintaining numerical stability for quantized states; it reduces curvature update complexity from $O(d^3)$ to $O(d^2)$ and accelerates convergence by up to 1.85x over KFAC. SLoPe accelerates training by up to 1.25x via a double-pruned backward pass for N:M sparsity, using low-rank "lazy" adapters in the final 1% of training to recover accuracy. For post-training compression, OPTIMA stabilizes static masks in a zero-training regime by formulating weight reconstruction as globally optimal column-wise quadratic programs, improving zero-shot accuracy by up to 3.97%. Given a fine-tuning budget, PATCH breaks the ceiling of static masks by learning a dynamic hybrid sparsity ratio between 0% and 50%, yielding up to 1.38x speedups. Finally, SLiM realizes the full Compression Trinity in one shot, using mathematically derived low-rank adapters to recover information lost to quantization and sparsity, improving accuracy by up to 5.66% over state-of-the-art methods and outperforming uncompressed dense models at equal parameter budgets by 0.6%. Together, these results show that jointly applying the Compression Trinity is essential for efficient, scalable, high-performance LLMs.
Chinese Translation
高昂的计算和环境成本阻碍了大语言模型（LLMs）的可扩展部署。传统的压缩技术（稀疏性、量化、低秩近似）通常是孤立应用的，每种方法都面临准确性与效率的瓶颈。本文提出了“压缩三位一体”（Compression Trinity），这是一个统一框架，联合应用这三大支柱：稀疏性用于减少计算，量化用于最小化内存带宽，低秩近似用于恢复准确性。为了加速预训练，我们将三位一体应用于优化器和模型架构。MKOR通过块对角稀疏性和低秩逆运算近似曲率，保持量化状态的数值稳定性；它将曲率更新复杂度从 $O(d^3)$ 降低到 $O(d^2)$，并使收敛速度比 KFAC 快多达 1.85 倍。SLoPe 通过对 N:M 稀疏性进行双重修剪的反向传播，加速训练速度最高可达 1.25 倍，并在训练的最后 1% 使用低秩“懒惰”适配器来恢复准确性。对于后训练压缩，OPTIMA 通过将权重重构公式化为全局最优列向量二次规划，在零训练状态下稳定静态掩码，将零-shot 准确性提高最多 3.97%。在微调预算下，PATCH 通过学习动态混合稀疏比率（介于 0% 和 50% 之间）突破静态掩码的限制，实现高达 1.38 倍的加速。最后，SLiM 一次性实现完整的压缩三位一体，使用数学推导的低秩适配器恢复因量化和稀疏性而丢失的信息，使准确性比最先进的方法提高最多 5.66%，并在相同参数预算下超越未压缩的稠密模型 0.6%。综上所述，这些结果表明，联合应用压缩三位一体对于高效、可扩展和高性能的大语言模型至关重要。
cs.AI / 51 / 2608.24076
AgentWorld: Personality-Aware Reliability Evaluation for Agentic Information Retrieval
AgentWorld：基于个性意识的代理信息检索可靠性评估
Abstract
Evaluation of agentic information retrieval remains limited to scripted interactions with uniform users, missing both natural personality diversity and adversarial brittleness. We present AgentWorld, a simulation framework combining (i)Big Five (OCEAN) personality-driven user populations with stateful tool-use environments; (ii)the pass$^k$ consistency metric with structured fault classification, partial-credit scoring, and dual-control handoff verification; (iii)score-thresholded training-data export in six fine-tuning formats; and (iv)an adversarial Risk Analyser that snapshots required-intermediate-state spines, branches Monte-Carlo rollouts under four task-aware perturbation types, and quantifies risk via $\Delta P / \Delta T$ scoring, Dempster--Shafer evidence fusion, and Shapley attack-category attribution. Three experiments demonstrate the framework: a conversational analytics agent across 10 OCEAN personas (240 evaluator judgments); a customer-support agent across 5 tasks $\times$ 4 persona variants; and adversarial stress-testing of 5 tasks revealing pre-existing trajectory brittleness ($V_{\min}=0.375$ without perturbation) and tool/infrastructure-layer attack dominance (Shapley: 46% system, 38% action). Personality variation surfaces failure modes uniform testing cannot expose---cross-domain leakage, contextual drift, a 0.27-point quality gap, and 50% vs. 100% pass-rate across personas on the same task---while the Risk Analyser quantifies trajectory-level brittleness that pass$^k$ alone cannot measure.
Chinese Translation
代理信息检索的评估仍然局限于与统一用户的脚本化互动，缺乏自然的个性多样性和对抗性脆弱性。我们提出了AgentWorld，一个模拟框架，结合了（i）基于五大性格特质（OCEAN）的个性驱动用户群体与有状态的工具使用环境；（ii）pass$^k$一致性度量与结构化故障分类、部分评分和双重控制交接验证；（iii）六种微调格式的得分阈值训练数据导出；以及（iv）一个对抗性风险分析器，该分析器快照所需的中间状态脊柱，针对四种任务感知扰动类型进行蒙特卡洛回滚，并通过$ rac{ riangle P}{ riangle T}$评分、邓普斯特-沙弗证据融合和夏普利攻击类别归因量化风险。三个实验展示了该框架：一个跨10种OCEAN个性（240个评估者判断）的对话分析代理；一个跨5个任务×4个个性变体的客户支持代理；以及对5个任务的对抗性压力测试，揭示了预先存在的轨迹脆弱性（$V_{ ext{min}}=0.375$在无扰动情况下）和工具/基础设施层攻击主导性（夏普利：46%系统，38%行动）。个性变异暴露了统一测试无法揭示的失败模式——跨域泄漏、上下文漂移、0.27分的质量差距，以及在同一任务上不同个性之间的50%与100%的通过率——而风险分析器量化了仅靠pass$^k$无法测量的轨迹级脆弱性。
cs.AI / 52 / 2608.24086
EMRB: A Multi-Level Benchmark for Evaluating LLM Reasoning over Raw Electromagnetic Signals
EMRB：评估大语言模型对原始电磁信号推理的多层次基准
Abstract
Large language models (LLMs) are increasingly used as code agents for scientific and engineering analysis, but their ability to analyze raw physical-layer measurements remains untested. We introduce \textbf{EMRB} (\textbf{E}lectro\textbf{m}agnetic \textbf{R}easoning \textbf{B}enchmark), which evaluates whether LLMs can analyze raw I/Q data by writing and running code. EMRB contains 200 problems across five difficulty levels and 27 question types, from signal detection to OFDM design, generated from 11 signal types with verified ground truth. Unlike benchmarks built on preprocessed features or structured tables, EMRB provides only the raw capture; the quantities each question refers to must first be discovered through code. We evaluate 14 LLMs spanning proprietary, open-weight, and reasoning-oriented families. Scores range from 24.1\% to 78.9\%, with the mean dropping from 84.9\% on basic measurement to 21.2\% on system design. We also propose \textbf{ReconPilot}, a structured method that separates signal reconnaissance, targeted analysis, and self-verification. Across three backbones, ReconPilot raises the overall score by 3.8 to 17.6 points and improves 13 of 15 backbone-level combinations tested. All data and code are publicly released in \href{https://github.com/mingxuZhang2/EMRB}{\textcolor{blue}{our GitHub repository}}.
Chinese Translation
大型语言模型（LLMs）越来越多地被用作科学和工程分析的代码代理，但它们分析原始物理层测量的能力尚未经过测试。我们介绍了 extbf{EMRB}（ extbf{E}lectro extbf{m}agnetic extbf{R}easoning extbf{B}enchmark），该基准评估LLMs是否能够通过编写和运行代码来分析原始I/Q数据。EMRB包含200个问题，涵盖五个难度级别和27种问题类型，从信号检测到正交频分复用（OFDM）设计，这些问题是基于11种信号类型生成的，并具有经过验证的真实值。与基于预处理特征或结构化表格的基准不同，EMRB仅提供原始捕获；每个问题所涉及的量必须通过代码首先被发现。我们评估了14个LLMs，涵盖专有、开放权重和面向推理的模型家族。得分范围从24.1\%到78.9\%，平均得分从基本测量的84.9\\%下降到系统设计的21.2\\%。我们还提出了 extbf{ReconPilot}，一种结构化方法，分离信号侦察、针对性分析和自我验证。在三个基础模型中，ReconPilot将整体得分提高了3.8到17.6分，并改善了测试的15个基础模型组合中的13个。所有数据和代码均已在 extcolor{blue}{ ext{我们的GitHub仓库}}中公开发布。
cs.AI / 53 / 2608.24099
Are Android GUI Agents Robust Against Runtime Anomalies? AnTrap: Evaluating Agents in Dynamic Adversarial Environments
Android GUI 代理在运行时异常下的鲁棒性如何？AnTrap：在动态对抗环境中评估代理
Abstract
GUI agents often encounter dynamic anomalies when deployed on Android devices, from unexpected pop-ups to action misuse, yet existing benchmarks lack systematic evaluation of agent robustness against runtime anomalies. We introduce AnTrap, a comprehensive benchmark that injects dynamic perturbations into agent execution trajectories. We propose a taxonomy organizing real-world anomalies into four layers (State, Thinking, Action and Round) with ten fine-grained subcategories, and develop a construction pipeline that preserves task solvability while introducing realistic adversarial conditions. Evaluating 16 leading GUI models, we reveal universal vulnerability to dynamic anomalies, with even the strongest models suffering significant performance degradation. Furthermore, we conduct GRPO training in both original and adversarial environments to validate our benchmark, separating environment-learnable anomalies from reasoning-bottlenecked ones. Our findings show that while single-step traps at state and action layers are largely addressable through adversarial reinforcement learning, deep contextual traps, like state deadlock, expose intrinsic limitations that cannot be resolved by training in environments with traps alone.
Chinese Translation
在 Android 设备上部署的 GUI 代理常常会遇到动态异常，从意外弹窗到操作误用，但现有基准缺乏对代理在运行时异常下鲁棒性的系统评估。我们提出了 AnTrap，一个综合基准，通过向代理执行轨迹注入动态扰动。我们提出了一种将现实世界异常组织成四个层次（状态、思维、行动和回合）的分类法，并细分为十个精细子类别，同时开发了一条构建管道，在引入现实对抗条件的同时保持任务可解性。对 16 个领先的 GUI 模型进行评估，我们揭示了对动态异常的普遍脆弱性，即使是最强的模型也遭受了显著的性能下降。此外，我们在原始和对抗环境中进行 GRPO 训练，以验证我们的基准，将环境可学习的异常与推理瓶颈的异常分开。我们的研究结果表明，尽管状态和行动层的单步陷阱在很大程度上可以通过对抗强化学习来解决，但深层上下文陷阱，如状态死锁，暴露了无法仅通过在有陷阱的环境中训练来解决的内在局限性。
cs.AI / 54 / 2608.24103
ACE: A Self-Correcting Agentic Canvas Editor for Multi-Slide Presentation Automation
ACE：一种自我修正的代理画布编辑器，用于多幻灯片演示自动化
Abstract
Commercial design platforms increasingly edit documents through large language model (LLM) agents, but two practical problems block reliable deployment: legacy document formats expose only \emph{flat}, absolutely positioned elements, so agents must recompute coordinates and routinely break layouts; and design has no unique ground truth, so diff-against-reference metrics penalize valid-but-different outputs. We present \textbf{ACE}, an agentic canvas editor over a \emph{hierarchical scene-graph} with a presentation-specialized action space (98 tools), paired with \textbf{CARE}, a content-aware router that feeds the agent only the relevant slice of each deck (avg.\ $\sim$89\% input-token reduction), and a \emph{self-correction} loop driven by a \emph{ground-truth-free} instruction-following (IF) judge whose natural-language critique is fed back as the next-turn instruction. With a fixed backbone, a scene-graph editor in a \emph{single turn} already matches a same-backbone \emph{agentic} HTML pipeline that iterates internally; adding self-correction lifts ACE significantly above it on instruction following (IF 4.23 vs.\ 3.81 on the full 94-task benchmark, paired $p{=}.010$, replicated by an out-of-loop judge) at 1.75$\times$ the speed and $\sim$44\% lower cost. VQ means are statistically indistinguishable, but 26 blind raters prefer ACE overall (58.7\% decisive win-rate) and prefer the self-corrected output 81\% of the time; the ranking is invariant across three judge families, and out-of-loop judges retain two-thirds of the self-correction gain, bounding circularity. 66\% of cases halt after one pass, and a strict-peak rollback removes every observed regression.
Chinese Translation
商业设计平台越来越多地通过大型语言模型（LLM）代理编辑文档，但两个实际问题阻碍了可靠的部署：遗留文档格式仅暴露出 extit{平面}、绝对定位的元素，因此代理必须重新计算坐标，常常导致布局破坏；而设计没有唯一的真实标准，因此与参考的差异度量会惩罚有效但不同的输出。我们提出了 extbf{ACE}，一种基于 extit{层次场景图}的代理画布编辑器，具有专门针对演示的动作空间（98种工具），并配备 extbf{CARE}，一种内容感知路由器，仅向代理提供每个幻灯片的相关切片（平均约 extsim 89 ext%输入令牌减少），以及一个由 extit{无真实标准}的指令跟随（IF）评判者驱动的 extit{自我修正}循环，其自然语言批评反馈作为下一轮指令。在固定的骨干网络下，单轮的场景图编辑器已经与同一骨干的 extit{代理} HTML管道相匹配，该管道在内部进行迭代；添加自我修正使ACE在指令跟随方面显著超越了它（IF 4.23对比3.81，基于完整的94任务基准，配对$p{=}.010$，由一个循环外的评判者复制）以1.75倍的速度和约44 ext%更低的成本。VQ均值在统计上没有显著差异，但26名盲评者总体上更喜欢ACE（58.7 ext%决定性胜率），并在81 ext%时间内更喜欢自我修正的输出；该排名在三个评判者家族中保持不变，循环外的评判者保留了三分之二的自我修正增益，限制了循环性。66 ext% 的案例在一次通过后停止，而严格的峰值回滚消除了每个观察到的回归。
cs.AI / 55 / 2608.24112
Scalable Question-Centric Text-to-Image Evaluation: Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing
可扩展的以问题为中心的文本到图像评估：可靠的排名、细粒度诊断和成本感知路由
Abstract
Modern text-to-image (T2I) models often have similar total scores but different strengths, making practical selection difficult. Fine-grained benchmarks decompose prompts into questions, yet often return them to prompt scores and fixed categories, weakening attribution and ignoring complexity. Related requirements are also scored separately or as one total, obscuring basic versus compositional failure. We present QC-T2I-Bench, a question-centric framework that converts open prompts into attributed atomic questions and organizes their dependencies with Davidsonian Scene Graphs (DSGs). We use hierarchy-constrained question aggregation to exclude downstream questions after a prerequisite fails and to prevent simple and complex prompts from receiving the same total weight. We then use the DSG structure to measure joint success within prompts and compare repeated entities across prompts, separating basic realization failures from failures under additional requirements. We evaluate multiple open-source T2I models on English and Chinese prompts. The resulting question-level evidence supports reliable ranking and fine-grained diagnosis: joint completion falls from 80.7\% for components with two capabilities to 37.2\% for those with seven or more. Finally, we reuse the same records for training-free routing; our cost-aware router matches ERNIE's 89.51-point estimate with 21.3\% less GPU-s/MP.
Chinese Translation
现代文本到图像（T2I）模型通常具有相似的总分，但在不同方面表现出不同的优势，这使得实际选择变得困难。细粒度基准将提示分解为问题，但通常将其返回到提示分数和固定类别，从而削弱了归因并忽视了复杂性。相关要求也被单独评分或作为一个总分，这模糊了基本失败与组合失败之间的区别。我们提出了QC-T2I-Bench，一个以问题为中心的框架，将开放提示转换为具有归属的原子问题，并使用戴维森场景图（Davidsonian Scene Graphs, DSGs）组织它们的依赖关系。我们使用层次约束的问题聚合来排除在先决条件失败后下游问题，并防止简单和复杂提示获得相同的总权重。然后，我们利用DSG结构来测量提示内的联合成功，并比较提示之间的重复实体，将基本实现失败与在附加要求下的失败分开。我们在英语和中文提示上评估了多个开源T2I模型。结果的问卷级证据支持可靠的排名和细粒度诊断：具有两个能力的组件的联合完成率从80.7 all到具有七个或更多能力的组件的37.2 all。最后，我们重用相同的记录进行无训练路由；我们的成本感知路由器与ERNIE的89.51点估计匹配，同时减少21.3 all的GPU-s/MP。
cs.AI / 56 / 2608.24114
AHEAD: Adaptive Hindsight with Environment-Augmented Distillation for Agentic RL
AHEAD：基于环境增强蒸馏的自适应回顾用于自主强化学习
Abstract
Training multi-turn LLM agents with reinforcement learning typically relies on trajectory-level rewards, which assign a uniform advantage to every step and cannot identify which decisions led to success or failure. Self-distillation methods can provide finer-grained supervision by augmenting RL with privileged information. However, existing approaches usually apply the same type of privileged information to every step in an indistinguishable manner, ignoring a key asymmetry: routine steps need little additional guidance, while critical error steps require corrective direction that environment feedback alone cannot provide. We propose AHEAD, a step-aware framework that matches different supervision sources to different step types. The teacher receives environment feedback on all steps as a grounded dense signal, and additionally receives LLM-generated corrective hints on error steps to supply the direction that environment feedback lacks. The method introduces minimal changes to the standard GRPO algorithm. Across ALFWorld, WebShop, and Search-based QA, and across three model scales, AHEAD raises task success (+13.3 points on ALFWorld and +11.0 on WebShop at 7B over GRPO), reaches a given success rate in fewer training steps, and solves tasks within tighter interaction budgets than outcome-only RL and prior self-distillation baselines.
Chinese Translation
使用强化学习训练多轮大型语言模型（LLM）代理通常依赖于轨迹级奖励，这种奖励对每一步赋予统一的优势，无法识别哪些决策导致了成功或失败。自蒸馏方法通过用特权信息增强强化学习，能够提供更细粒度的监督。然而，现有方法通常以不可区分的方式将相同类型的特权信息应用于每一步，忽视了一个关键的不对称性：常规步骤需要很少的额外指导，而关键错误步骤则需要环境反馈无法提供的纠正方向。我们提出了AHEAD，一个步态感知框架，将不同的监督来源匹配到不同的步骤类型。教师在所有步骤上接收环境反馈作为一个基础的密集信号，并且在错误步骤上额外接收LLM生成的纠正提示，以提供环境反馈所缺乏的方向。该方法对标准的GRPO算法引入了最小的变化。在ALFWorld、WebShop和基于搜索的问答任务中，以及在三个模型规模下，AHEAD提高了任务成功率（在ALFWorld上提高了13.3个百分点，在WebShop上提高了11.0个百分点，模型规模为7B，相较于GRPO），在更少的训练步骤中达到了给定的成功率，并在比仅基于结果的强化学习和之前的自蒸馏基线更紧凑的交互预算内解决了任务。
cs.AI / 57 / 2608.24135
Robust Code RL via Faulty-Code-Driven Test case Synthesis and Dense Reward Shaping
通过故障代码驱动的测试用例合成和密集奖励塑造实现鲁棒代码强化学习
Abstract
Reinforcement learning from verifiable rewards (RLVR) has emerged as a pivotal technique for enhancing the code generation capabilities of Large Language Models (LLMs). However, the efficacy of RLVR in coding implementations is fundamentally limited by the comprehensiveness of test cases, because insufficient test coverage in code validation often causes false positives, further leading to reward hacking and policy degradation. To mitigate the reward bias stemming from the suboptimal quality of current automated generation methods, we propose the RobustTests framework, which introduces a faulty-code-driven test case synthesis strategy that leverages "near correct" faulty codes to guide the model in precisely capturing latent logical discrepancies and further integrates validator agents with behavioral feature clustering to facilitate the granular filtering of invalid and redundant test cases. To address false negatives caused by inherent hallucination noise in synthetic test cases, RobustTests also incorporates a stepwise dense reward function based on pass rates, bolstering training robustness through fine-grained feedback. By employing this pipeline, we construct a high-quality dataset that augmented the test cases in CodeContests, encompassing a broader spectrum of faulty code scenarios and significantly enhances diagnostic utility. Experimental results demonstrate that, by leveraging a moderately challenging subset of problems from CodeContests for training, RL fine-tuning of Qwen3-32B via RobustTests achieves an absolute 3% performance gain on the LiveCodeBench benchmark compared to baseline methods, confirming the effectiveness of the RobustTests framework in advancing the code generation proficiency of LLMs.
Chinese Translation
基于可验证奖励的强化学习（RLVR）已成为提升大型语言模型（LLMs）代码生成能力的关键技术。然而，RLVR在编码实现中的有效性根本上受到测试用例全面性的限制，因为代码验证中测试覆盖不足往往会导致误报，进一步导致奖励操控和策略退化。为减轻当前自动生成方法质量不佳所带来的奖励偏差，我们提出了鲁棒测试（RobustTests）框架，该框架引入了一种故障代码驱动的测试用例合成策略，利用“近乎正确”的故障代码引导模型准确捕捉潜在的逻辑差异，并进一步整合验证代理与行为特征聚类，以促进对无效和冗余测试用例的细粒度过滤。为了解决合成测试用例中固有的幻觉噪声导致的假阴性问题，鲁棒测试还结合了一种基于通过率的逐步密集奖励函数，通过细致的反馈增强训练的鲁棒性。通过采用这一流程，我们构建了一个高质量的数据集，增强了CodeContests中的测试用例，涵盖了更广泛的故障代码场景，并显著提升了诊断效用。实验结果表明，通过利用CodeContests中适度具有挑战性的子集进行训练，使用鲁棒测试对Qwen3-32B进行强化学习微调，相较于基线方法在LiveCodeBench基准测试中实现了绝对3%的性能提升，确认了鲁棒测试框架在提升LLMs代码生成能力方面的有效性。
cs.AI / 58 / 2608.24160
OmniJudge or OmniBias? Diagnosing Multimodal Judges through Balanced, Decoupled Lenses
全能评判者还是全能偏见？通过平衡的解耦视角诊断多模态评判者
Abstract
Multimodal understanding models that can jointly judge text-to-image (T2I), text-to-video (T2V) and text-to-speech (TTS) generation are increasingly used as "OmniJudges" for evaluation and automatic annotation. How reliably they understand what they score remains unclear, since existing benchmarks and training data tend to overemphasize positive examples and to conflate distinct failure modes, so a judge may score well without recognizing failures while its capability gaps stay hidden. Motivated by this, we introduce D3-Omni, a balanced and decoupled benchmark for diagnosing fine-grained multimodal understanding, covering 53 orthogonal binary dimensions (17/22/14) and 10,671 samples (3,526/1,998/5,147) across the three tasks. Rather than re-generating outputs, which may leak information across dimensions, we fix verified fully positive seeds and derive negatives through controlled prompt rewriting and atomic, dimension-isolating perturbations. The resulting D3 design is Dual-balanced, which helps alleviate negative-sample scarcity and per-dimension label imbalance; Decoupled, so that each error is attributable to a single capability; and Dynamic, steering construction toward under-represented regions of the label distribution as generative models improve.The suite reaches near 1:1 per-dimension parity and a uniform distribution over all total-score levels. Under this balanced view, even strong OmniJudges tend to struggle on modality-related dimensions, to confirm satisfied requirements far more reliably than they detect violated ones, and to treat nominally distinct attributes as largely a single decision, suggesting that aggregate accuracy may hide systematic blind spots that a balanced and decoupled lens can help expose and, in turn, address.
Chinese Translation
多模态理解模型能够联合评判文本到图像（T2I）、文本到视频（T2V）和文本到语音（TTS）生成，越来越多地被用作评估和自动标注的“全能评判者”。它们对所评分内容的理解程度仍不明确，因为现有的基准和训练数据往往过于强调正面示例，并混淆不同的失败模式，因此一个评判者可能在评分时表现良好，却未能识别出失败，而其能力缺口则保持隐蔽。基于此，我们引入了 D3-Omni，这是一个平衡且解耦的基准，用于诊断细粒度的多模态理解，涵盖了53个正交的二元维度（17/22/14）和10,671个样本（3,526/1,998/5,147），涉及三个任务。我们并不重新生成输出，以避免跨维度泄露信息，而是固定经过验证的完全正面种子，并通过控制提示重写和原子、维度隔离的扰动来推导负面样本。最终的 D3 设计是双重平衡的，这有助于缓解负样本稀缺和每个维度标签不平衡的问题；是解耦的，使得每个错误都可以归因于单一能力；并且是动态的，随着生成模型的改进，引导构造向标签分布中代表性不足的区域发展。该套件在每个维度上达到了近乎1:1的平衡，并在所有总分级别上实现了均匀分布。在这种平衡的视角下，即使是强大的全能评判者在与模态相关的维度上也往往表现不佳，能够更可靠地确认满足的要求，而不是检测到被违反的要求，并且将名义上不同的属性视为基本上是单一决策，这表明聚合准确性可能掩盖了系统性的盲点，而平衡和解耦的视角可以帮助揭示并解决这些问题。
cs.AI / 59 / 2608.24174
Task-Adaptive Rubrics for GUI Reward Modeling
任务自适应的图形用户界面奖励建模评分标准
Abstract
Recent studies on GUI agents have increasingly focused on outcome reward modeling, which assigns outcome rewards by judging whether an executed trajectory satisfies the success criteria implied by the user instruction. Existing GUI reward verifiers, however, often under-specify how these criteria should be constructed for each task instance. Whether using generic rubric structures or implicit model reasoning, their judging criteria are not sufficiently task-adaptive: they can transfer checks across tasks, overlook concrete constraints in the current instruction, or become overly strict by enforcing unstated requirements. To address this limitation, we propose AdaptRubric, a Coarse-to-Fine Rubrics Framework that constructs task-adaptive judging criteria through a category-level coarse stage and an instance-level fine stage. AdaptRubric performs category-level coarse rubric retrieval by routing the instruction to a GUI task family and retrieving reusable task-family criteria, then conducts instance-level fine rubric generation to surface compact cues for concrete values, scopes, and constraints in the current instruction. Across offline reward evaluation and online reinforcement learning optimization, AdaptRubric consistently outperforms prior reward agents, improving F1 by 3.6 points over the baseline average under a matched image budget and yielding a 4.23-point task-success gain.
Chinese Translation
近期关于图形用户界面（GUI）代理的研究越来越关注结果奖励建模，该建模通过判断执行的轨迹是否满足用户指令所暗示的成功标准来分配结果奖励。然而，现有的GUI奖励验证器通常未能明确规定这些标准应如何为每个任务实例构建。无论是使用通用评分标准结构还是隐式模型推理，它们的判断标准都不足以适应具体任务：它们可能在任务之间转移检查，忽视当前指令中的具体约束，或通过强制未说明的要求而变得过于严格。为了解决这一局限性，我们提出了AdaptRubric，一个粗到细的评分标准框架，通过类别级粗略阶段和实例级细致阶段构建任务自适应的判断标准。AdaptRubric通过将指令路由到GUI任务家族并检索可重用的任务家族标准来执行类别级粗略评分标准检索，然后进行实例级细致评分标准生成，以提取当前指令中具体值、范围和约束的紧凑提示。在离线奖励评估和在线强化学习优化过程中，AdaptRubric始终优于先前的奖励代理，在匹配图像预算下，F1分数比基线平均提高了3.6分，并实现了4.23分的任务成功率提升。
cs.AI / 60 / 2608.24188
Paritok-4B: Intent-Conditioned Context Compression for Coding Agents
Paritok-4B：基于意图条件的上下文压缩用于编码代理
Abstract
Coding agents re-send large file reads and tool outputs to a frontier LLM every turn, and this context dominates their token bill. General-purpose prompt compressors are trained on prose and suit code poorly: they paraphrase identifiers and drop the exact spans an agent needs to edit. We present Paritok-4B, a 4B LoRA compressor for coding-agent trajectories built on two commitments. It is extractive: it selects spans rather than rewriting them, and 96.0% of the identifiers, paths, and numbers it emits already appear in its input, holding at 96.2% on held-out SWE-bench Lite output. It is intent-conditioned: told the agent's current task, it acts chiefly inside a retained segment, selecting which lines survive (retained lines are +0.067 more intent-relevant than removed ones, paired 95% CI [+0.056, +0.078]) rather than changing how much is retained. We distil a gpt-4.1-mini teacher over 67,074 real OpenHands trajectories into 40,606 validated examples and fine-tune Qwen3-4B. On all 300 SWE-bench Lite instances, Paritok-4B compresses agent context to 25.7% of its size, 2.0x harder than a gpt-4.1-mini compressor (50.2%) and 2.4x harder than gpt-5 (61.9%), while retaining 86.5% of uncompressed single-shot solve quality. Fed the cat -n line-numbered input real agents produce, it compresses slightly less (27.8%) and retains more (89.3%); there the paired test is informative, with 30 instances solved only uncompressed and 17 only compressed, an exact McNemar p=0.079, so at this sample size compressing context to roughly a quarter of its size does not significantly reduce the solve rate. The model is a 264 MB adapter that self-hosts on one 24 GB GPU with no per-token compressor fee, which at list prices decides the economics: gpt-5 as a compressor is net-negative, costing more than the downstream tokens it saves. Weights, data, and evaluation scripts are open (Apache 2.0).
Chinese Translation
编码代理在每个回合中将大型文件读取和工具输出重新发送到前沿的LLM，这种上下文主导了它们的令牌费用。通用提示压缩器是在散文上训练的，且不适合代码：它们会对标识符进行意译，并丢失代理需要编辑的确切范围。我们提出了Paritok-4B，这是一个针对编码代理轨迹的4B LoRA压缩器，基于两个承诺。它是提取式的：它选择范围而不是重写它们，且其输出的96.0%的标识符、路径和数字已经出现在输入中，在保留的SWE-bench Lite输出中保持在96.2%。它是意图条件的：在告知代理当前任务的情况下，它主要在保留的段落内进行操作，选择哪些行得以保留（保留的行比移除的行意图相关性高0.067，配对95%置信区间为[+0.056, +0.078]），而不是改变保留的数量。我们从67,074个真实的OpenHands轨迹中提炼出一个gpt-4.1-mini教师，形成40,606个经过验证的示例，并对Qwen3-4B进行了微调。在所有300个SWE-bench Lite实例中，Paritok-4B将代理上下文压缩至其大小的25.7%，比gpt-4.1-mini压缩器（50.2%）难度高2.0倍，比gpt-5（61.9%）难度高2.4倍，同时保留86.5%的未压缩单次解决质量。对于真实代理生成的带有行号的输入，它的压缩稍微少一些（27.8%），保留更多（89.3%）；在这里，配对测试是信息丰富的，有30个实例仅在未压缩情况下解决，17个仅在压缩情况下解决，McNemar检验p=0.079，因此在这个样本大小下，将上下文压缩至其大小的四分之一并不会显著降低解决率。该模型是一个264 MB的适配器，可以在一台24 GB的GPU上自我托管，无需每个令牌的压缩费用，这在标价上决定了经济性：作为压缩器的gpt-5是净负的，成本高于其节省的下游令牌。权重、数据和评估脚本是开放的（Apache 2.0）。
cs.AI / 61 / 2608.24192
Preference Data Selection for Mitigating the Alignment Tax in Large Language Models
减轻大型语言模型对齐税的偏好数据选择
Abstract
Aligning large language models to human preferences is crucial for real-world deployment but frequently incurs an alignment tax, leading to the catastrophic forgetting of pre-trained general capabilities. While previous works primarily frame this problem as an optimization or architectural challenge, the inherent characteristics of preference data that drive this degradation remain largely underexplored. In this paper, we propose BALIGN, a balanced data selection strategy that explicitly mitigates catastrophic forgetting while optimizing alignment efficacy. Through theoretical and empirical analyses of the preference optimization gradient, we identify three key data-centric features that dictate parameter drift: the reference model's log-probability margin, the token length difference between chosen and rejected responses, and the TF-IDF similarity to general capability corpora. By aggregating these orthogonal features into a unified composite risk score, BALIGN systematically filters out high-risk preference samples that disrupt intrinsic model parameters or provide minimal alignment utility. Extensive experiments on standard human preference datasets demonstrate that BALIGN strongly preserves foundational capabilities without compromising alignment gains, consistently achieving the optimal Pareto frontier with minimal computational overhead.
Chinese Translation
将大型语言模型与人类偏好对齐对于实际应用至关重要，但通常会产生对齐税，从而导致预训练的通用能力的灾难性遗忘。尽管以往的研究主要将此问题视为优化或架构挑战，但驱动这种退化的偏好数据的内在特征仍然未得到充分探索。在本文中，我们提出了BALIGN，这是一种平衡的数据选择策略，明确减轻灾难性遗忘，同时优化对齐效率。通过对偏好优化梯度的理论和实证分析，我们识别出三个关键的数据中心特征，这些特征决定了参数漂移：参考模型的对数概率边际、所选和被拒绝响应之间的令牌长度差异，以及与通用能力语料库的TF-IDF相似性。通过将这些正交特征聚合成一个统一的复合风险评分，BALIGN系统性地过滤出高风险的偏好样本，这些样本会干扰内在模型参数或提供最小的对齐效用。在标准人类偏好数据集上的大量实验表明，BALIGN在不妥协对齐收益的情况下，强有力地保留了基础能力，始终以最小的计算开销实现最佳的帕累托前沿。
cs.AI / 62 / 2608.24214
MetaRAG: Belief-Action Aligned Policy Optimization for Agentic RAG
MetaRAG：面向代理性检索增强生成的信念-行动对齐策略优化
Abstract
Agentic retrieval-augmented generation (RAG) requires language models to decide when to continue searching and when to answer. Existing RL-based methods rely on external supervision and overlook the agent's internal belief about whether the current evidence is sufficient. To address this problem, we reformulate the search decision quality as belief-action alignment and propose MetaRAG, a belief-action aligned policy optimization framework for agentic RAG. MetaRAG uses Verify-first Action Generation to elicit an explicit verification process before each actual action, and Internal Belief Probing to estimate the policy model's own answerability belief from the same question-history context. Based on these, MetaRAG derives a consistency reward that is further gated by answer correctness, avoiding reinforcement of internally consistent but incorrect trajectories. The belief probe is used only during training and introduces no inference-time overhead. Experiments on seven public QA benchmarks show that MetaRAG consistently improves the accuracy-efficiency trade-off over strong RL-based agentic RAG baselines, with gains that transfer to deep research settings, different optimizers, and multiple model backbones.
Chinese Translation
代理性检索增强生成（RAG）要求语言模型决定何时继续搜索以及何时回答。现有的基于强化学习（RL）的方法依赖于外部监督，并忽视了代理对当前证据是否充足的内部信念。为了解决这个问题，我们将搜索决策质量重新表述为信念-行动对齐，并提出了MetaRAG，一个面向代理性RAG的信念-行动对齐策略优化框架。MetaRAG采用优先验证的行动生成（Verify-first Action Generation）在每个实际行动之前引导显式的验证过程，并通过内部信念探测（Internal Belief Probing）从相同的问题历史上下文中估计策略模型自身的可回答性信念。基于这些，MetaRAG推导出一个一致性奖励，该奖励进一步受到答案正确性的限制，从而避免强化内部一致但不正确的轨迹。信念探测仅在训练期间使用，不会引入推理时的开销。在七个公共问答基准上的实验表明，MetaRAG在强大的基于RL的代理性RAG基线之上，始终改善了准确性与效率的权衡，其收益能够转移到深度研究环境、不同的优化器和多种模型骨干上。
cs.AI / 63 / 2608.24218
Constraint-Guided Enterprise Data Mapping with Large Language Models
基于约束引导的大型语言模型企业数据映射
Abstract
Enterprise entity alignment must handle semi-structured records, implicit attributes, and unit or granularity mismatches. Manual matching is still common in practice, but does not scale as schemas and providers evolve. LLM-only matching improves semantic recall, yet can violate structural and physical invariants, producing fluent yet operationally invalid correspondences. We propose constraint-guided mapping (CGM), a neuro-symbolic method with three stages: (i) schema-grounded admissibility constraints with metadata mc = , where tau_c denotes the constraint type and delta_c provides executable relation and normalization logic; (ii) constraint-restricted candidate generation with cascade relaxation to guarantee a nonempty feasible set under noise; and (iii) neural ranking with bounded LLM disambiguation restricted to that feasible set. Methodologically, constraints operate as hypothesis-space operators rather than post-hoc validators, enabling controlled degradation under relaxation and auditable, human-guidable decisions. On a controlled structural-decoy benchmark, hard admissibility shrinks the candidate space by ~480x without dropping the GT, and a layer-by-layer ablation shows this gate, not the LLM, is the decisive lift (F1 0.08 to 0.66). The benefit is model-independent and adds no extra inference cost: a small model with constraints matches a frontier LLM used without them at ~28x lower cost. The method, not a single tuned configuration, transfers across seven enterprise makes (macro F1 0.70), each under its own automatically discovered, expert-refinable constraints, and lowers expert effort by ~7x versus spreadsheet workflows. Public Valentine results add an external ranking sanity check and mark the boundary: constraints should be hard only where structural invariants are match-determining.
Chinese Translation
企业实体对齐必须处理半结构化记录、隐式属性以及单位或粒度不匹配的问题。手动匹配在实践中仍然很常见，但随着模式和提供者的演变，这种方法并不具备可扩展性。仅使用大型语言模型（LLM）的匹配虽然提高了语义召回率，但可能违反结构和物理不变性，产生流畅但在操作上无效的对应关系。我们提出了约束引导映射（CGM），这是一种神经符号方法，分为三个阶段：（i）基于模式的可接受性约束，包含元数据 mc = ，其中 tau_c 表示约束类型，delta_c 提供可执行的关系和归一化逻辑；（ii）通过级联放松生成约束限制的候选集，以保证在噪声下存在非空可行集；（iii）在该可行集上进行神经排序，限制 LLM 的歧义消解。方法论上，约束作为假设空间操作符而非事后验证器运作，使得在放松下能够进行受控降级，并实现可审计、可由人类引导的决策。在一个受控的结构诱饵基准测试中，严格的可接受性将候选空间缩小了约 480 倍而不丢失真实标签（GT），逐层消融实验显示，这一门槛，而非 LLM，是决定性的提升（F1 从 0.08 提升至 0.66）。这一方法的好处是与模型无关的，并且没有额外的推理成本：一个带有约束的小模型的匹配成本约为不带约束的前沿 LLM 的 28 倍更低。该方法不是单一的调优配置，而是跨越七个企业模型（宏观 F1 为 0.70），每个模型都有其自动发现的、可由专家细化的约束，并且相比于电子表格工作流程，降低了约 7 倍的专家工作量。公开的 Valentine 结果增加了外部排名的合理性检查，并标记了边界：约束仅在结构不变性决定匹配时应当严格。
cs.AI / 64 / 2608.24228
Evaluating Multiple LLM Generations with Validated Task Coverage
评估多种 LLM 生成的验证任务覆盖率
Abstract
Many LLM applications are most useful when they provide several candidate outputs for comparison, validation, or combination. Predominant evaluation settings, however, still focus on individual outputs or reduce multiple samples to a single success or selected answer. This can miss whether the outputs include several genuinely different useful results. We introduce VTC-Bench, a five-domain benchmark for this setting, together with Validated Task Coverage (VTC) as its core evaluation quantity. The benchmark is built from carefully selected real-data tasks where both output quality and task-relevant distinctness can be checked automatically and reproducibly, without model-based judges. VTC measures how many distinct useful results are obtained within $k$ attempts. Across multiple models and inference settings, the benchmark leads to different conclusions from conventional evaluation: configurations that look strongest from single-draw quality are not necessarily those with the best coverage, and simple measures of output variation do not reliably recover task-relevant coverage. These results show that finite candidate sets can be evaluated directly as objects of interest, revealing differences in model behavior that are not apparent from conventional per-output evaluation.
Chinese Translation
许多 LLM 应用在提供多个候选输出以进行比较、验证或组合时最为有用。然而，主流的评估设置仍然集中于单个输出，或将多个样本简化为单一成功或选择的答案。这可能会忽视输出是否包含多个真正不同的有用结果。我们引入了 VTC-Bench，这是一个针对该设置的五个领域基准，并将验证任务覆盖率（Validated Task Coverage, VTC）作为其核心评估指标。该基准由精心挑选的真实数据任务构建，能够自动且可重复地检查输出质量和任务相关的独特性，而无需基于模型的评判。VTC 衡量在 $k$ 次尝试中获得多少个独特的有用结果。在多个模型和推理设置下，该基准得出的结论与传统评估截然不同：从单次抽样质量看起来最强的配置不一定是覆盖率最佳的配置，而简单的输出变异度测量并不能可靠地恢复任务相关的覆盖率。这些结果表明，有限的候选集可以作为关注对象直接进行评估，揭示出模型行为中的差异，而这些差异在传统的逐输出评估中并不明显。
cs.AI / 65 / 2608.24232
TRACE: An Evidence-Grounded Benchmark for Safety Evaluation of Large Reasoning Models
TRACE：一个基于证据的安全评估基准，用于大型推理模型
Abstract
Large Reasoning Models (LRMs) generate intermediate reasoning traces that may contain unsafe content, even when their final responses appear safe. Guardrail models are designed to detect and block unsafe content, yet existing benchmarks for unsafe content detection focus primarily on prompts and final responses, leaving reasoning traces largely unexamined. Moreover, these benchmarks typically provide only binary safety labels, without evidence annotations that justify the judgments. To address these limitations, we introduce TRACE, an evidence-grounded safety evaluation benchmark that covers the entire LRM inference pipeline: prompts, reasoning traces, and final responses. TRACE includes prompts in two languages spanning nine risk categories and ten attack strategies. For each prompt, four LRMs generate reasoning traces and final responses, and we annotate the safety of each component and extract supporting evidence from the corresponding source text. Evaluating 18 guardrail models on TRACE reveals that safety judgment for reasoning traces is substantially more challenging than for prompts or final responses, and that current models struggle to accurately extract supporting evidence. These findings highlight the need for guardrail models that can reliably detect and precisely localize unsafe content across the LRM inference pipeline.
Chinese Translation
大型推理模型（LRMs）生成的中间推理轨迹可能包含不安全内容，即使它们的最终响应看起来是安全的。护栏模型旨在检测和阻止不安全内容，但现有的不安全内容检测基准主要集中在提示和最终响应上，导致推理轨迹在很大程度上未被审查。此外，这些基准通常仅提供二元安全标签，而没有证据注释来证明判断的合理性。为了解决这些局限性，我们引入了TRACE，这是一个基于证据的安全评估基准，涵盖了整个LRM推理流程：提示、推理轨迹和最终响应。TRACE包括两种语言的提示，涵盖九个风险类别和十种攻击策略。对于每个提示，四个LRMs生成推理轨迹和最终响应，我们对每个组件的安全性进行注释，并从相应的源文本中提取支持证据。在TRACE上评估18个护栏模型的结果表明，推理轨迹的安全判断显著比提示或最终响应更具挑战性，当前模型在准确提取支持证据方面也存在困难。这些发现突显了需要能够可靠检测和精确定位LRM推理流程中不安全内容的护栏模型。
cs.AI / 66 / 2608.24237
STRIVE: Multi-Agent Structured Temporal Reasoning with Integrated Verification for Longitudinal Radiology Report Generation
STRIVE：具有集成验证的多智能体结构化时间推理用于纵向放射学报告生成
Abstract
Longitudinal radiology report generation (LRRG) requires identifying both current findings and their changes relative to a prior study. Existing methods jointly model diagnosis, attribute estimation, temporal comparison, and language generation within implicit representations, which can cause task interference, obscure the evidence underlying each decision, and limit error traceability. They also model progression states as independent labels, ignoring their ordered structure and thus treating missed changes and direction reversals equally. We present STRIVE, Multi-Agent Structured Temporal Reasoning with Integrated Verification for LRRG, which decomposes clinical reasoning into specialized Diagnosis, Attribute, and Temporal Change Agents that produce explicit intermediate evidence. In particular, the Temporal Change Agent is further post-trained using Progression-Aware GRPO, a verifiable, shaped reward that assigns partial credit to direction-preserving errors while scoring direction reversals lowest. STRIVE performs verification at two stages: a deterministic Consistency Gate reconciles the agent outputs before report generation, and a Validation Agent checks whether the generated report is supported by the aggregated clinical evidence. On Longitudinal-MIMIC, STRIVE attains the best clinical efficacy among recent methods and more than doubles Longitudinal Change Concordance (LCC), a measure of temporal agreement with the reference report, over the strongest baseline.
Chinese Translation
纵向放射学报告生成（LRRG）需要识别当前发现及其相对于先前研究的变化。现有方法在隐式表示中联合建模诊断、属性估计、时间比较和语言生成，这可能导致任务干扰，模糊每个决策背后的证据，并限制错误的可追溯性。它们还将进展状态建模为独立标签，忽视其有序结构，从而将遗漏的变化和方向反转同等对待。我们提出了STRIVE，具有集成验证的多智能体结构化时间推理用于LRRG，它将临床推理分解为专门的诊断、属性和时间变化智能体，这些智能体生成明确的中间证据。特别是，时间变化智能体进一步使用进展感知的GRPO（Progression-Aware GRPO）进行后训练，这是一种可验证的、形状奖励，给予保持方向的错误部分信用，而对方向反转的评分最低。STRIVE在两个阶段进行验证：一个确定性的连贯性门在报告生成之前调和智能体输出，而验证智能体检查生成的报告是否得到聚合临床证据的支持。在Longitudinal-MIMIC数据集上，STRIVE在最近的方法中达到了最佳的临床有效性，并且在最强基线之上将纵向变化一致性（LCC）这一时间一致性度量提高了两倍以上。
cs.AI / 67 / 2608.24252
SA-Bench: Evaluating Semantic Alignment in LLM-Based Paper Reproduction
SA-Bench：评估基于大型语言模型的论文再现中的语义对齐
Abstract
LLM agents can generate paper reproduction code, yet often produce scientifically unfaithful implementations. We define this failure mode as semantic drift, where generated code silently diverges from the paper's specifications. We introduce SemanticAlign-Bench(SA-Bench), a diagnostic benchmark covering 30 papers from ICLR, ICML and NeurIPS 2025. For each paper, we decompose its specifications into atomic and verifiable implementation claims, which we call Semantic Alignment Units (SAUs) and evaluate repositories along four diagnostic dimensions spanning numerical, methodological, protocol and ordering drift. In total, we construct 1,491 SAUs across five ML domains and evaluate 12 generator configurations (4 models $\times$ 3 scaffolds). Even the strongest configuration (Claude+PaperCoder) achieves a mean SAU score of only 0.301 out of 1.0, with an overall mean of 0.221 across 360 evaluations. A failure taxonomy reveals that agents attempt most requirements but implement them incorrectly, with implementation mismatch and stubs accounting for the majority of zero-scored claims. Our analysis further indicates that scaffolds optimized for executability provide limited leverage for scientific reproduction; narrowing the gap requires scaffolds that prioritize semantic specification verification. The benchmark, annotations and evaluation pipeline are publicly available.
Chinese Translation
大型语言模型（LLM）代理可以生成论文再现代码，但往往产生科学上不忠实的实现。我们将这种失败模式定义为语义漂移，即生成的代码在未被察觉的情况下偏离了论文的规范。我们引入了SemanticAlign-Bench（SA-Bench），这是一个诊断基准，涵盖了2025年ICLR、ICML和NeurIPS的30篇论文。对于每篇论文，我们将其规范分解为原子且可验证的实现声明，称之为语义对齐单元（Semantic Alignment Units, SAUs），并沿着数值、方法论、协议和排序漂移四个诊断维度评估代码库。我们在五个机器学习领域构建了1491个SAUs，并评估了12种生成器配置（4个模型 × 3个支架）。即使是最强的配置（Claude+PaperCoder）也仅获得了0.301的平均SAU得分（满分为1.0），在360次评估中整体平均得分为0.221。失败分类法揭示，代理尝试满足大多数要求，但实现不正确，实施不匹配和占位符占据了零得分声明的大部分。我们的分析进一步表明，针对可执行性优化的支架对科学再现提供的帮助有限；缩小这一差距需要优先考虑语义规范验证的支架。该基准、注释和评估流程已公开提供。
cs.AI / 68 / 2608.24258
Beyond Accuracy: A Dual-Judge Evaluation Protocol for Vision-Language Models in Legally Grounded Tasks
超越准确性：一种用于法律基础任务的视觉-语言模型双评估协议
Abstract
AI systems are increasingly evaluated for legally accountable settings, where correct outputs must also be justifiable against an applicable legal standard. Existing legal-AI benchmarks and LLM-as-judge protocols provide important infrastructure for measuring task performance and open-ended response quality. We contribute one additional evaluation signal: a dual-judge protocol that pairs a standard 0-10 quality judge with a strict binary semantic-equivalence judge against a human-curated reference. We study a controlled, visually grounded regulatory task - UK traffic-sign interpretation, whose meaning is a codified question with a known reference for every input - and measure not merely whether the two judges disagree (by construction they must) but how much and where. On 4,680 evaluations under seven visibility levels and two occlusion modes, the two judges are moderately associated (point-biserial r = 0.644), while revealing an asymmetric Type II pattern affecting 8.0% of all evaluations. Its distribution is instructive: the marginal rate peaks at high visibility (14.2% at v = 0.8) simply because high-scoring answers are common there, but conditioned on the answer already scoring above 7, the rate is highest under heavy occlusion (54-63% at v <= 0.3), so a high quality score is least trustworthy when the input is most degraded. We are explicit that the signal is a property of this judge and reference: a 49-row human check shows the 0-10 judge aligns closely with everyday-reader judgement (Pearson r = 0.81; r = 0.80 with the LLM accuracy sub-score), while the equivalence judge is fairly but one-directionally stricter. The protocol adds one LLM call per evaluation and surfaces a signal single-judge protocols do not report. We release the prompt template, occluded variants, and full evaluation results.
Chinese Translation
人工智能系统在法律可追责的环境中越来越受到评估，其中正确的输出必须能够根据适用的法律标准进行合理解释。现有的法律人工智能基准和以大型语言模型（LLM）作为评判者的协议为测量任务表现和开放式响应质量提供了重要基础设施。我们贡献了一个额外的评估信号：一种双评估协议，将标准的0-10质量评判者与严格的二元语义等价评判者配对，后者是针对人工策划的参考进行评估。我们研究了一个受控的、视觉基础的监管任务——英国交通标志解读，其含义是一个有明确参考的法典化问题。我们不仅测量两个评判者是否存在分歧（由于构造原因，它们必然存在），还测量分歧的程度和位置。在七个可见度水平和两种遮挡模式下进行的4680次评估中，这两个评判者之间的相关性适中（点二元相关系数 r = 0.644），同时揭示出影响8.0%所有评估的非对称类型II模式。其分布具有启发性：边际率在高可见度下达到峰值（在 v = 0.8 时为14.2%），这仅仅是因为高分答案在此处较为常见，但在已经得分超过7的答案条件下，重度遮挡下的比率最高（在 v <= 0.3 时为54-63%），因此，当输入最受损时，高质量评分的可信度最低。我们明确指出，该信号是该评判者和参考的属性：一项49行的人类检查显示，0-10评判者与日常读者的判断高度一致（Pearson r = 0.81；与LLM准确性子评分的相关系数 r = 0.80），而等价评判者则相对但单向地更严格。该协议为每次评估增加了一次LLM调用，并提供了单评判协议未报告的信号。我们发布了提示模板、遮挡变体和完整评估结果。
cs.AI / 69 / 2608.24263
Real-World Knowledge-Guided Change Data Synthesis for Remote Sensing
基于现实知识指导的遥感变化数据合成
Abstract
Change data synthesis provides a cost-effective solution for expanding training data and improving the performance of change detection models. However, existing synthesis methods typically rely on handcrafted rules to simulate changes, where limited coverage of class transitions restricts the diversity of synthesized data, while predefined transition designs limit their flexibility in accommodating varied change types. In this work, we introduce KnowChange, a knowledge-guided change data synthesis framework that leverages pretrained vision-language models as knowledge sources to reason about plausible change locations and class transitions from pre-change scenes and desired change types. By integrating knowledge-guided change simulation with generalizable synthesis models, KnowChange enables flexible synthesis of diverse change types within a unified framework. Extensive experiments demonstrate that KnowChange-generated data consistently outperforms existing synthetic datasets in both synthetic-to-real transfer and synthetic data augmentation, despite being generated at a compact scale. Further analyses show that the knowledge-guided change simulation can be seamlessly integrated into existing synthesis pipelines and enhance the downstream utility of synthesized data.
Chinese Translation
变化数据合成为扩展训练数据和提高变化检测模型性能提供了一种具有成本效益的解决方案。然而，现有的合成方法通常依赖于手工规则来模拟变化，有限的类别转换覆盖范围限制了合成数据的多样性，而预定义的转换设计则限制了其适应多样变化类型的灵活性。在本研究中，我们提出了KnowChange，一个知识指导的变化数据合成框架，该框架利用预训练的视觉-语言模型作为知识来源，从变化前场景和期望变化类型中推理出合理的变化位置和类别转换。通过将知识指导的变化模拟与可泛化的合成模型相结合，KnowChange能够在统一框架内灵活合成多样的变化类型。大量实验表明，KnowChange生成的数据在合成到真实的迁移和合成数据增强方面始终优于现有的合成数据集，尽管其生成规模较小。进一步的分析表明，知识指导的变化模拟可以无缝集成到现有的合成流程中，并增强合成数据的下游效用。
cs.AI / 70 / 2608.24273
Matched Excess-Outranker Regularization for Candidate-Set Interference in Continual Knowledge Graph Embedding
针对持续知识图谱嵌入中候选集干扰的匹配超越者正则化
Abstract
Continual knowledge graph embedding updates entity and relation representations as a graph grows. Existing methods primarily address catastrophic forgetting, but entity admission also changes the candidate universe of every compatible query. A historical answer can therefore lose rank even when its score and its ordering among old entities are preserved. We formalize this effect as candidate-set interference and introduce Matched Excess-Outranker Regularization (MEOR), a host-level objective that compares smooth answer-relative newcomer pressure with score-blind, structurally matched old references. Its one-sided penalty acts only when newcomer competition exceeds the matched reference, preserving the host learner's signal for legitimate new entities. Across eight paired runs on ENTITY-ComplEx, MEOR improves historical current-universe mean reciprocal rank (MRR) by 0.0057 over replay and reduces candidate-set interference by 0.0055, with one-sided 95% lower bounds of 0.0052 and 0.0051, respectively. It satisfies the preservation criteria for old-universe ranking and newcomer acquisition and improves historical current-universe MRR over persistent calibration, matched maximum regularizer (MMR), and unmatched old regularizer (UOR). Direct ablations support each component of its reference construction and aggregation. Adding MEOR also improves historical ranking in all ten reported FBInc-S and FBInc-L host and backbone settings, with every paired 95% confidence interval excluding zero. These results establish candidate admission as a distinct source of continual rank loss and show that it can be controlled without replacing the underlying embedding architecture or continual learner.
Chinese Translation
持续知识图谱嵌入随着图谱的增长更新实体和关系的表示。现有方法主要解决灾难性遗忘问题，但实体的加入也改变了每个兼容查询的候选宇宙。因此，即使历史答案的得分和在旧实体中的排序得以保留，其排名也可能会下降。我们将这种现象形式化为候选集干扰，并引入匹配超越者正则化（Matched Excess-Outranker Regularization, MEOR），这是一种主机级目标，比较平滑的答案相对新来者的压力与无得分盲目的结构匹配旧参考。其单边惩罚仅在新来者竞争超过匹配参考时生效，从而保留主机学习者对合法新实体的信号。在对ENTITY-ComplEx进行的八次配对实验中，MEOR使历史当前宇宙的平均倒数排名（Mean Reciprocal Rank, MRR）比重放提高了0.0057，并将候选集干扰降低了0.0055，单边95%下限分别为0.0052和0.0051。它满足旧宇宙排名和新来者获取的保留标准，并在持久校准、匹配最大正则化器（Matched Maximum Regularizer, MMR）和不匹配旧正则化器（Unmatched Old Regularizer, UOR）上改善了历史当前宇宙的MRR。直接消融实验支持其参考构建和聚合的每个组件。添加MEOR还改善了在所有十个报告的FBInc-S和FBInc-L主机及骨干设置中的历史排名，每个配对的95%置信区间均不包括零。这些结果确立了候选加入作为持续排名损失的一个独特来源，并表明可以在不更换基础嵌入架构或持续学习者的情况下控制这一现象。
cs.AI / 71 / 2608.24274
Eating for a Sustainable Planet: Personalized Sustainable Diet Recommendation via Constraint-Aware Decision-Making Modeling
为可持续地球而饮食：基于约束感知决策模型的个性化可持续饮食推荐
Abstract
A sustainable diet represents a multi-dimensional synergy among four essential pillars: nutrition adequacy, economic affordability, cultural acceptability, and environmental respect. Despite the prevalence of population-level sustainability modeling, practical implementation relies on effective individual-level adoption. This transition is often hindered by inter-individual heterogeneity, posing a formidable challenge in aligning sustainable diet requirements with individual preferences. To address this issue, we propose a personalized sustainable diet recommendation model based on a constraint-aware decision-making mechanism, where sustainability is incorporated through learnable constraints rather than modeled as user preferences. To systematically evaluate the proposed approach, we construct a sustainable diet dataset named SusDiet with about 150k recipes, characterized by broad coverage of sustainability indicators. Experimental results on this dataset show that our method promotes more sustainable choices without compromising individual preference. This work establishes a framework for aligning individual dietary choices with planetary health, offering quantitative evidence to guide future sustainable diet interventions and policy-making for sustainable development.
Chinese Translation
可持续饮食代表着营养充足、经济可承受、文化可接受和环境尊重四个基本支柱之间的多维协同。尽管人口层面的可持续性建模已相当普遍，但实际实施依赖于有效的个体层面采纳。这一转变常常受到个体间异质性的阻碍，使得可持续饮食需求与个体偏好的对接面临巨大挑战。为了解决这一问题，我们提出了一种基于约束感知决策机制的个性化可持续饮食推荐模型，其中可持续性通过可学习的约束而非用户偏好进行整合。为了系统评估所提方法，我们构建了一个名为SusDiet的可持续饮食数据集，包含约15万条食谱，广泛覆盖可持续性指标。对该数据集的实验结果表明，我们的方法在不妨碍个体偏好的情况下促进了更可持续的选择。这项工作建立了一个将个体饮食选择与地球健康对齐的框架，为未来的可持续饮食干预和可持续发展政策制定提供了定量证据。
cs.AI / 72 / 2608.24275
RePolicy: Reinforcement Learning for Safety-Policy Invocation in Agent Safeguards
RePolicy：用于代理安全保障中安全策略调用的强化学习
Abstract
Safeguarding language model agents requires assessing complete execution trajectories under context-dependent safety policies. Existing policy-aware safeguards mainly rely on prompting or supervised fine-tuning, limiting their ability to adapt to unseen trajectories and changing policy contexts. We propose RePolicy, an agent safeguard that learns safety-policy invocation through reinforcement learning. Given an agent trajectory and a dynamic policy library, RePolicy invokes the applicable policy and uses its content to produce a policy-grounded rationale and safety judgment. We construct PolicyTraj-20K to support supervised initialization, followed by GRPO with verifiable rewards and policy-context perturbation. Experiments across six agent safety benchmarks show that RePolicy achieves strong overall safety-detection performance and robust policy invocation under varying policy contexts.
Chinese Translation
保护语言模型代理需要在上下文依赖的安全政策下评估完整的执行轨迹。现有的政策感知安全保障主要依赖于提示或监督微调，这限制了它们适应未见轨迹和变化政策上下文的能力。我们提出了RePolicy，这是一种通过强化学习学习安全政策调用的代理安全保障。给定一个代理轨迹和一个动态政策库，RePolicy调用适用的政策，并利用其内容生成基于政策的推理和安全判断。我们构建了PolicyTraj-20K以支持监督初始化，随后采用具有可验证奖励和政策上下文扰动的GRPO。针对六个代理安全基准的实验表明，RePolicy在不同政策上下文下实现了强大的整体安全检测性能和稳健的政策调用能力。
cs.AI / 73 / 2608.24291
ReproAgent: Contract-Guided Paper-to-Code Reproduction
ReproAgent：合同引导的论文到代码重现
Abstract
Paper-to-code reproduction asks scientific AI agents to turn research papers into executable repositories that preserve the paper's method, protocol and artifacts. This is difficult because the specification is split: explicit paper content such as algorithms, metrics and artifacts is often lost across long agent trajectories, while implicit details such as framework defaults and conventions inherited from related work are absent from the paper. We introduce ReproAgent, a four-stage Prepare--Plan--Generate--Repair pipeline built around a persistent implementation contract with two channels: an implementation-requirement channel that turns paper snippets into code obligations, and a reference-evidence channel that retrieves content and structure evidence from related repositories. Both are bound to work packages, projected into file-level contracts, and consumed across generation and repair. On PaperBench Code-Dev, ReproAgent reaches the highest mean score among same-backbone scaffolds under both Claude-Sonnet-4.5 and Gemini-3-Flash. End-to-end channel ablations and per-paper cases support the contribution of both channels. Code and experimental artifacts are publicly available.
Chinese Translation
论文到代码的重现要求科学人工智能代理将研究论文转化为可执行的代码库，以保留论文的方法、协议和工件。这一过程困难重重，因为规范被分割：论文中的显性内容，如算法、指标和工件，往往在长时间的代理轨迹中丢失，而隐性细节，如框架默认值和从相关工作中继承的约定，则在论文中缺失。我们提出了ReproAgent，一个由四个阶段组成的准备—计划—生成—修复管道，围绕一个持久的实施合同构建，包含两个通道：一个实施需求通道，将论文片段转化为代码义务，另一个参考证据通道，从相关代码库中检索内容和结构证据。这两个通道都绑定到工作包，投影到文件级合同中，并在生成和修复过程中被使用。在PaperBench Code-Dev上，ReproAgent在Claude-Sonnet-4.5和Gemini-3-Flash下的同构架构中达到了最高的平均分数。端到端通道消融实验和每篇论文的案例支持了两个通道的贡献。代码和实验工件已公开提供。
cs.AI / 74 / 2608.24302
VideoHarness-RSI: Recursive Harness Self-Improvement for Long-Video Understanding with Frozen Vision-Language Models
VideoHarness-RSI：基于递归的自我改进长视频理解方法，结合冻结的视觉-语言模型
Abstract
Long-video understanding depends critically on how a limited model context is constructed from a much longer video. Existing approaches improve this process through compression, retrieval, memory, and agentic evidence acquisition, but these mechanisms are typically introduced as part of a manually designed inference system or optimized together with other components. This makes it difficult to isolate a simpler question: how much can be gained by improving the executable context-construction program alone? We study this question through VIDEOHARNESS-RSI, a controlled baseline for recursively searching executable context constructors around a frozen vision-language model (VLM). An outer-loop proposer uses prior programs, evaluation outcomes, and execution traces to generate candidate harnesses, which are executed and evaluated end to end before successful variants are retained for further search. This makes long-video understanding a controlled instance of automated harness design: the searchable object is executable program structure, while the answering model and interface remain fixed. Starting from uniform sampling, recursive harness search consistently finds room for improvement and surpasses several weaker hand-crafted baselines. Starting instead from a stronger hand-crafted baseline, the same RSI process yields a further improvement. The selected harness also transfers to additional long-video benchmarks without further search. Together, these results establish executable context construction as a distinct optimization layer and provide a reproducible baseline for studying harness discovery and transfer around frozen VLMs.
Chinese Translation
长视频理解在很大程度上依赖于如何从更长的视频中构建有限的模型上下文。现有方法通过压缩、检索、记忆和主动证据获取来改善这一过程，但这些机制通常作为手动设计的推理系统的一部分引入，或与其他组件一起优化。这使得我们很难孤立出一个更简单的问题：仅通过改善可执行的上下文构建程序，能获得多少收益？我们通过VIDEOHARNESS-RSI研究这个问题，这是一个围绕冻结的视觉-语言模型（VLM）递归搜索可执行上下文构建器的受控基线。外部循环提议者利用先前的程序、评估结果和执行轨迹生成候选的自我改进程序，这些程序在成功变体被保留以进行进一步搜索之前，会被端到端地执行和评估。这使得长视频理解成为自动化自我改进设计的一个受控实例：可搜索的对象是可执行程序结构，而回答模型和接口保持不变。从均匀采样开始，递归自我改进搜索始终发现改进的空间，并超越了几个较弱的手工基线。从一个更强的手工基线开始，同样的递归自我改进过程也带来了进一步的改进。所选的自我改进程序还可以在没有进一步搜索的情况下转移到其他长视频基准上。这些结果共同确立了可执行上下文构建作为一个独特的优化层，并提供了一个可重复的基线，用于研究围绕冻结的视觉-语言模型的自我改进发现和转移。
cs.AI / 75 / 2608.24310
OPDSearch+: On-Policy Distillation with RL Refinement for Search-Augmented Reasoning
OPDSearch+: 基于强化学习优化的在线蒸馏用于搜索增强推理
Abstract
Search-augmented reasoning remains difficult for small language models. On-policy distillation (OPD) from trained teachers offers a promising direction, but suffers from two issues: (1) high-quality multi-turn search trajectories depend on dynamic retriever responses, making SFT data prohibitively expensive to collect at scale; (2) task-specifically trained teachers incur substantial training cost, while directly applying OPD with an off-the-shelf teacher without task-specific fine-tuning constrains the student to the teacher's performance ceiling and suffers from severe training instability. We propose OPDSearch+, the first distillation paradigm that requires no teacher fine-tuning for search-augmented reasoning. We investigate the role of a frozen off-the-shelf instruct model as the teacher in on-policy distillation, and reveal a key insight: the teacher reshapes the student's policy distribution so that subsequent RL converges to a superior solution that RL alone cannot reach. In stage one, the student interacts with a live search engine and is distilled via a per-position forward KL objective, transferring reasoning decomposition and evidence integration skills without any task-specific teacher training. In stage two, RL refines the distilled student from a richer behavioral foundation, achieving performance that RL alone cannot reach from scratch. Across seven QA benchmarks, OPDSearch+ with a 3B model consistently outperforms all prior 3B RL baselines, achieving gains of 13.1% on HotpotQA and 8.5% on 2WikiMultihopQA.
Chinese Translation
对于小型语言模型而言，搜索增强推理仍然具有挑战性。从训练好的教师模型进行在线蒸馏（On-policy Distillation, OPD）提供了一种有前景的方向，但面临两个问题：（1）高质量的多轮搜索轨迹依赖于动态检索器的响应，使得大规模收集 SFT 数据变得极为昂贵；（2）针对特定任务训练的教师模型需要大量的训练成本，而直接使用未经过特定任务微调的现成教师进行 OPD，会将学生的性能限制在教师的表现上限，并导致严重的训练不稳定性。我们提出了 OPDSearch+，这是第一个不需要教师微调的蒸馏范式，用于搜索增强推理。我们研究了冻结的现成指令模型作为教师在在线蒸馏中的作用，并揭示了一个关键见解：教师重塑了学生的策略分布，使得后续的强化学习（RL）能够收敛到 RL 单独无法达到的更优解。在第一阶段，学生与实时搜索引擎进行交互，并通过每个位置的前向 KL 目标进行蒸馏，转移推理分解和证据整合技能，而无需任何特定任务的教师训练。在第二阶段，RL 从更丰富的行为基础上优化蒸馏后的学生，达到 RL 单独从零开始无法实现的性能。在七个问答基准测试中，使用 3B 模型的 OPDSearch+ 始终优于所有先前的 3B RL 基线，在 HotpotQA 上提升了 13.1%，在 2WikiMultihopQA 上提升了 8.5%。
cs.AI / 76 / 2608.24314
Benchmarking LLM Judges for Voice-Agent Evaluation: Reliability, Calibration, and Human Oversight
语音助手评估中的大型语言模型评估者基准：可靠性、校准与人工监督
Abstract
Evaluating conversational voice agents at scale re- quires reliable assessment methods that capture both observ- able interaction quality and the contextual judgment typically provided by human evaluators. We investigate LLM-as-a-Judge evaluation by comparing human judgments with GPT-4.1 and GPT-5 on telecom and retail voice-agent conversations, across conversational quality and safety dimensions. The same interac- tions are scored under three evaluation configurations, p0, p1, and p2, to test whether automated judgments are sensitive to the evaluation setup and whether observed patterns generalize across configurations and judge models. Beyond aggregate agreement, we examine metric-level correlations, evaluator consistency, and systematic human-LLM disagreement to identify which conver- sational attributes can be judged reliably by automation and which remain sensitive to interpretation and context. Effective voice-agent evaluation is also shaped by pipeline-level factors such as speech generation, streaming, and error propagation across ASR, reasoning, and tool-calling stages, motivating our focus on comparing how human and LLM judges score the same interactions end to end. Our results show that LLM- based evaluation can serve as an effective component of large- scale voice-agent assessment, but that its reliability is metric- and configuration-dependent rather than uniform. This pro- vides an empirical framework for identifying which metrics suit automated evaluation and supports hybrid pipelines in which LLM judges handle scalable assessment while human evaluators remain engaged for metrics that demand contextual interpretation and higher-confidence judgment.
Chinese Translation
大规模评估对话语音助手需要可靠的评估方法，以捕捉可观察的交互质量和通常由人工评估者提供的上下文判断。我们通过比较人类判断与GPT-4.1和GPT-5在电信和零售语音助手对话中的表现，研究了大型语言模型（LLM）作为评估者的评估，涵盖了对话质量和安全性维度。相同的交互在三种评估配置（p0、p1和p2）下进行评分，以测试自动化判断是否对评估设置敏感，以及观察到的模式是否在不同配置和评估模型之间具有普遍性。除了总体一致性，我们还考察了指标级别的相关性、评估者的一致性以及系统性的人类与LLM之间的分歧，以识别哪些对话属性可以通过自动化可靠地判断，哪些则对解释和上下文敏感。有效的语音助手评估还受到管道级因素的影响，例如语音生成、流媒体和在自动语音识别（ASR）、推理和工具调用阶段的错误传播，这促使我们关注比较人类和LLM评估者如何端到端地评分相同的交互。我们的结果表明，基于LLM的评估可以作为大规模语音助手评估的有效组成部分，但其可靠性依赖于指标和配置，而非统一。这为识别适合自动化评估的指标提供了实证框架，并支持混合管道，其中LLM评估者处理可扩展评估，而人类评估者则参与需要上下文解释和更高置信度判断的指标。
cs.AI / 77 / 2608.24319
Can a Dynamic Internal Field Govern a Transformer's Cognition? Certifiability, not Superiority, in Homeostatic Compute Control
动态内部场能否主导变压器的认知？在稳态计算控制中，认证性而非优越性
Abstract
An intelligent system does not merely reason: it governs its own reasoning - how much to compute, when to stop, which module to activate. Can that role be played by a dynamic internal field - a low-dimensional homeostatic state with explicit physics and certified stability - that modulates cognition without performing it? Ours is a field on the module graph governed by a family of PDEs on the graph Laplacian, advancing with an adaptive-depth reasoner. We certify the stability of the integrator of the whole family - an integrator certificate, not a closed-loop one. New, and proved here: a discrete Schur-Cohn criterion for Verlet with velocity coupling, necessary and sufficient per latent root, with no commutation hypothesis. The answer is threefold: substance no, structure only in part, certifiability yes. The type of the field's physics is irrelevant for accuracy: wave, diffusion, gated mixtures and a 2D Navier-Stokes substrate tie. A twenty-seed preregistered deconfounding campaign bounds the structural claim: at equalized caps the second-order effect is strong in one family (+0.087 [+0.042, +0.132], t=4.0) but is not detected in the other (+0.014 [-0.013, +0.040], n.s.), so part of the original contrast was capacity, not order; and a matched-interface GRU is indistinguishable in the first and nominally exceeds the field in the second (-0.035 [-0.067, -0.002]). What distinguishes the field is not capability but that its one-step operator admits an exact runtime stability check - a difference of kind, not of existence: learned recurrences carry certificates too, sufficient and conservative ones. A kill-gate with a positive control finds no evidence for the field as evidence accumulator (Delta AUC +0.0007 [-0.0065, +0.0079] vs a 0.03 threshold). A dynamic internal field is a viable, certifiable compute governor, but not an enhancer of cognition: it modulates, it does not think.
Chinese Translation
智能系统不仅仅是推理：它主导自己的推理——计算多少、何时停止、激活哪个模块。动态内部场是否可以扮演这一角色——一种具有明确物理特性和认证稳定性的低维稳态，能够调节认知而不进行认知？我们的研究是在由一系列图拉普拉斯算子上的偏微分方程（PDEs）所支配的模块图上的一个场，伴随自适应深度推理器推进。我们认证了整个家族的积分器的稳定性——一种积分器证书，而非闭环证书。新的且在此证明：一种针对带速度耦合的Verlet方法的离散Schur-Cohn准则，对于每个潜在根是必要且充分的，且没有交换假设。答案是三重的：实质上不，结构上仅部分，认证性是肯定的。场的物理类型与准确性无关：波动、扩散、门控混合和二维纳维-斯托克斯基底相连。一个包含二十个种子的预注册去混淆实验限制了结构性主张：在均衡的上限下，第二阶效应在一个家族中显著（+0.087 [+0.042, +0.132], t=4.0），而在另一个家族中未被检测到（+0.014 [-0.013, +0.040], n.s.），因此原始对比的一部分是能力，而非顺序；而匹配接口的GRU在第一个家族中无法区分，且在第二个家族中名义上超越了该场（-0.035 [-0.067, -0.002]）。区分该场的不是能力，而是其一步操作符允许进行精确的运行时稳定性检查——一种性质的差异，而非存在的差异：学习的递归同样携带证书，且是充分且保守的。一个带正控制的杀死门未发现该场作为证据累积器的证据（Delta AUC +0.0007 [-0.0065, +0.0079]，与0.03阈值相比）。动态内部场是一个可行的、可认证的计算主导者，但不是认知的增强者：它调节，而不进行思考。
cs.AI / 78 / 2608.24325
SonarLLM: A Native Sonar--Optical Multimodal Large Language Model for Underwater Perception
SonarLLM：一种用于水下感知的原生声纳-光学多模态大语言模型
Abstract
Reliable underwater perception requires complementary sensing under variable visibility. Optical cameras capture appearance and semantics but degrade rapidly with turbidity, whereas imaging sonar preserves geometry while exhibiting distinct range-azimuth structure and acoustic artifacts. Existing MLLMs, built primarily on optical encoders, are therefore ill-suited to model sonar or adaptively exploit sonar-optical complementarity. We propose SonarLLM, a sonar-optical MLLM that treats sonar as a native perceptual modality. It combines a sonar-specific encoder, modality-specific physics-aware feature enhancement, and reliability-aware hierarchical fusion to align acoustic structure with optical semantics and dynamically adjust their contributions as sensing quality changes. We also introduce SonarBench, a paired benchmark that spans four tasks: recognition, counting, visual question answering, and captioning; and, across the benchmark, three input settings: sonar-only, optical-only, and fusion. By fixing the scene and sonar observation while varying optical degradation, SonarBench enables controlled measurement of cross-modal complementarity. SonarLLM achieves 72.0% macro accuracy across sonar-only recognition, counting, and VQA, outperforming the strongest baseline by 34.4 percentage points, and 68.7% under fusion, exceeding the best baseline by 25.1 points. For recognition and counting, the fusion-over-optical gain grows from 6.0 to 36.0 points as turbidity increases, indicating the increasing complementary value of sonar under controlled optical degradation. Together, these results show that robust heterogeneous perception depends not only on adding sonar, but on representing and weighting it according to its sensing characteristics.
Chinese Translation
可靠的水下感知需要在可变能见度下进行互补感知。光学相机捕捉外观和语义，但在浑浊条件下迅速退化，而成像声纳则保留几何信息，同时展现出独特的距离-方位结构和声学伪影。因此，现有的多模态大语言模型（MLLMs），主要基于光学编码器，无法有效建模声纳或自适应利用声纳-光学的互补性。我们提出了SonarLLM，一种将声纳视为原生感知模态的声纳-光学多模态大语言模型。它结合了特定于声纳的编码器、模态特定的物理感知特征增强和可靠性感知的层次融合，以将声学结构与光学语义对齐，并在感知质量变化时动态调整它们的贡献。我们还引入了SonarBench，这是一个涵盖四个任务的配对基准：识别、计数、视觉问答和图像描述；并且在基准中有三种输入设置：仅声纳、仅光学和融合。通过固定场景和声纳观测，同时改变光学退化，SonarBench能够控制性地测量跨模态互补性。SonarLLM在仅声纳的识别、计数和视觉问答任务中实现了72.0%的宏观准确率，超越最强基线34.4个百分点，在融合任务中达到68.7%，超过最佳基线25.1个百分点。对于识别和计数任务，随着浑浊度的增加，融合相较于光学的增益从6.0点增长到36.0点，表明在受控光学退化下，声纳的互补价值逐渐增加。综合这些结果表明，稳健的异构感知不仅依赖于添加声纳，还依赖于根据其感知特性对其进行表示和加权。
cs.AI / 79 / 2608.24338
Selective Regenerative Decoding: Trajectory-Level Intervention for Inference-Time Reasoning
选择性再生解码：推理时推理的轨迹级干预
Abstract
Inference-time decoding methods improve LLM reasoning by exploring multiple candidate trajectories, yet treat each trajectory as atomic: either retaining it whole or discarding it irreversibly. This wastes computation on partially promising candidates whose high-quality prefixes are abandoned alongside degraded suffixes. We introduce Selective Regenerative Decoding (SRD), which routes each candidate to discard, keep, or refine only the degraded portion of the suffix while preserving the useful prefix of borderline candidates, without requiring a larger target model. Under mild assumptions, SRD achieves a provable 1.28-to-1.36-fold gain in sample efficiency over rejection sampling with strictly higher expected trajectory quality, with the gain growing as the candidate pool grows. Across MATH500, GPQA Diamond, HotpotQA, and AlpacaEval with multiple generation-reward model pairs, SRD matches Best-of-N accuracy with substantially fewer generated tokens and outperforms speculative rejection in low-compute regimes. By enabling segment-level intervention rather than whole-trajectory selection, SRD opens a previously underexplored region of the accuracy-compute tradeoff for inference-time reasoning.
Chinese Translation
推理时解码方法通过探索多个候选轨迹来提高大型语言模型（LLM）的推理能力，但将每个轨迹视为原子：要么完整保留，要么不可逆地丢弃。这导致对部分有前景的候选者浪费计算，因为其高质量的前缀与退化的后缀一起被放弃。我们提出了选择性再生解码（Selective Regenerative Decoding, SRD），该方法对每个候选者进行路由，以丢弃、保留或仅精炼后缀的退化部分，同时保留边界候选者的有用前缀，而无需更大的目标模型。在温和的假设下，SRD在样本效率上实现了可证明的1.28到1.36倍的提升，相较于拒绝采样具有严格更高的预期轨迹质量，且随着候选池的增大，增益也在增长。在MATH500、GPQA Diamond、HotpotQA和AlpacaEval等多个生成-奖励模型对中，SRD以显著更少的生成标记匹配了最佳-N的准确性，并在低计算环境中优于投机性拒绝。通过实现段级干预而非整体轨迹选择，SRD为推理时推理的准确性与计算的权衡开辟了一个先前未被充分探索的领域。
cs.AI / 80 / 2608.24358
The Handoff Tax: Continuing Non-Native Trajectories in LLM Agents
交接税：LLM代理中的非本地轨迹持续性
Abstract
Coding agents perform long-running tasks spanning dozens of model calls, tool uses, and code edits. As these runs unfold, users face a practical cost-quality trade-off: escalating to a stronger model when a cheaper one struggles, or downshifting once the hard reasoning is complete. Each switch requires the receiver to continue a non-native trajectory produced by another model. We study how this handoff affects quality and cost, and how varying the trajectory information inherited by the receiver changes the outcome. Using pairs of low-cost, low-capability (LC) and high-cost, high-capability (HC) models from the Claude and GPT families, we vary handoff direction, timing, and interface, comparing full-trajectory transfer, compaction, and trajectory removal while preserving the repository state. Across both model families, full-trajectory escalation recovers less than half of the LC-to-HC quality gap while incurring a substantial cost premium. We term this cost-quality penalty the handoff tax. By contrast, downshift offers a favorable cost-quality point. Interestingly, the preferred interface also reverses with direction: reducing LC-model trajectory information improves escalation quality, whereas removing the HC-model trajectory reduces downshift quality.
Chinese Translation
编码代理执行跨越数十次模型调用、工具使用和代码编辑的长期任务。在这些任务展开的过程中，用户面临着实际的成本与质量权衡：当一个廉价模型遇到困难时，是否升级到更强的模型，或在艰难推理完成后降级。一旦切换，每次交接都要求接收者继续由另一个模型产生的非本地轨迹。我们研究这种交接如何影响质量和成本，以及接收者继承的轨迹信息的变化如何改变结果。通过使用来自Claude和GPT系列的低成本、低能力（LC）模型与高成本、高能力（HC）模型的配对，我们改变了交接的方向、时机和接口，比较了完整轨迹转移、压缩和轨迹移除，同时保持存储库状态。在这两个模型系列中，完整轨迹升级所恢复的LC到HC的质量差距不到一半，同时产生了可观的成本溢价。我们将这种成本与质量的惩罚称为交接税。相比之下，降级提供了一个有利的成本-质量点。有趣的是，首选接口也随着方向的变化而逆转：减少LC模型轨迹信息提高了升级质量，而移除HC模型轨迹则降低了降级质量。
cs.AI / 81 / 2608.24361
Adaptive Influence Graphs for Failure Attribution in Multi-Agent Systems
用于多智能体系统故障归因的自适应影响图
Abstract
Multi-agent LLM systems are increasingly deployed in real-world applications, where failures can be costly and difficult to localize. Despite growing efforts to automate failure attribution, diagnosing failed runs still largely relies on human engineers. Yet engineers rarely debug complex systems by reading raw logs end to end. Instead, observability tools organize traces around components, actions, and dependencies to support targeted navigation. We hypothesize that modern LLMs can benefit from the same paradigm. To test this hypothesis, we introduce Adaptive Influence Graphs (AIGs), a two-stage agentic framework that first transforms a failed trace into a structured graph and then navigates it to identify the critical error. Across multiple models, we show that richer trace representations consistently improve failure attribution, with adaptive graph construction and agent-directed traversal yielding the strongest results. AIGs establish a new state of the art on Who&When, the standard benchmark for multi-agent failure attribution. This affirms our hypothesis that attribution depends not only on the diagnosing model, but also on how the trace is represented and explored.
Chinese Translation
多智能体大语言模型（LLM）系统在实际应用中越来越多地被部署，其中故障可能代价高昂且难以定位。尽管在自动化故障归因方面的努力日益增加，诊断失败的运行仍然在很大程度上依赖于人类工程师。然而，工程师很少通过逐行阅读原始日志来调试复杂系统。相反，可观察性工具围绕组件、操作和依赖关系组织跟踪，以支持有针对性的导航。我们假设现代大语言模型可以从同样的范式中受益。为了验证这一假设，我们引入了自适应影响图（Adaptive Influence Graphs, AIGs），这是一种两阶段的智能体框架，首先将失败的跟踪转换为结构化图，然后导航该图以识别关键错误。在多个模型中，我们展示了更丰富的跟踪表示始终能改善故障归因，自适应图构建和智能体导向的遍历产生了最强的结果。AIGs在Who&When这一多智能体故障归因的标准基准测试中建立了新的最先进水平。这证实了我们的假设，即归因不仅依赖于诊断模型，还依赖于跟踪的表示和探索方式。
cs.AI / 82 / 2608.24368
From State to Action: OODA-Tool for Reliable Multi-Turn Tool Use
从状态到行动：用于可靠多轮工具使用的OODA-Tool
Abstract
Reliable multi-turn tool use requires an agent to preserve an evolving task state and ensure that each action remains consistent with it. However, direct function-calling and ReAct-style policies learn state tracking and action generation within the same autoregressive trajectory. This coupling creates state-action competition: the pressure to produce the next call can overwrite or ignore information accumulated earlier in the interaction. Inspired by Boyd's Observe-Orient-Decide-Act cycle, we introduce OODA-Tool, a typed closed-loop policy designed to mitigate this competition by separating state preservation from action realization. Rather than generating an action directly from the interaction history, OODA-Tool routes each decision through controller-checked intermediate states, ensuring that the final output remains grounded in the current task state. Specifically, Observe reconstructs the task state, Orient determines whether execution is warranted, Decide forms an admissible action structure, and Act realizes the external output. We evaluate OODA-Tool against direct function-calling and ReAct policies using Qwen3 models ranging from 0.6B to 14B across multi-turn, multi-tool, and incomplete-information settings. OODA-Tool consistently improves task success across model sizes, with larger gains on smaller models and on tasks whose actions depend strongly on information accumulated across turns and prior tool results. Controlled variants, stage-level ablations, and transfer evaluations further demonstrate the robustness of these improvements.
Chinese Translation
可靠的多轮工具使用要求代理保持不断发展的任务状态，并确保每个行动与之保持一致。然而，直接的函数调用和ReAct风格的策略在同一自回归轨迹中学习状态跟踪和行动生成。这种耦合产生了状态-行动竞争：产生下一个调用的压力可能会覆盖或忽略在交互过程中积累的信息。受到博伊德的观察-定位-决策-行动（Observe-Orient-Decide-Act, OODA）循环的启发，我们提出了OODA-Tool，这是一种类型化的闭环策略，旨在通过将状态保持与行动实现分离来减轻这种竞争。OODA-Tool并不是直接从交互历史中生成行动，而是通过控制器检查的中间状态来引导每个决策，确保最终输出与当前任务状态保持一致。具体而言，Observe重建任务状态，Orient确定是否需要执行，Decide形成可接受的行动结构，而Act实现外部输出。我们使用范围从0.6B到14B的Qwen3模型对OODA-Tool与直接函数调用和ReAct策略进行了评估，涵盖多轮、多工具和不完全信息的设置。OODA-Tool在各个模型规模上始终提高任务成功率，在较小模型和那些行动强烈依赖于跨轮次和先前工具结果积累的信息的任务上获得了更大的提升。控制变体、阶段级消融和迁移评估进一步证明了这些改进的稳健性。
cs.AI / 83 / 2608.24369
Do Recipes Have Personas? Characterizing and Generating Creator Style in Attributed Procedural Graphs
食谱是否具有个性？在归因程序图中表征和生成创作者风格
Abstract
While large language models (LLMs) possess vast zero-shot procedural knowledge, their tendency to produce homogenized logic often obscures the unique, idiosyncratic execution processes of individual human creators. In this paper, we investigate the computational discovery of procedural personas from unstructured data. To achieve this, we introduce ViralRecipesTrans, a new dataset of procedurally aligned execution flow graphs extracted from popular culinary video transcripts and explicitly mapped to specific creators. We formulate procedural stylometry as a graph learning and process discovery task, revealing a fundamental duality: while traditional lexical classifiers overfit via semantic leakage, discrete topological metrics successfully capture the rigid physical constraints of a creator's workflow. Building upon this characterization, we extend our framework into a novel generative task--predicting a creator's exact structural execution graph for unseen dishes. We expose a fundamental dichotomy in style generation between global macro-planning and local structural execution. Our results demonstrate that few-shot LLMs dominate semantic assignment but suffer from persistent macro-planning deficits, whereas our structured two-stage model achieves superior topological control via rigid Markovian priors. Together, an ensemble approach to procedural generation combines the strengths from both sides, dynamically synthesizing global semantic reasoning with localized topological footprints to automate the discovery and generation of personalized workflows.
Chinese Translation
尽管大型语言模型（LLMs）拥有广泛的零样本程序知识，但它们倾向于产生同质化的逻辑，这往往掩盖了个别创作者独特而特有的执行过程。本文探讨了从非结构化数据中计算发现程序个性的方法。为此，我们引入了ViralRecipesTrans，一个新的数据集，该数据集包含从流行烹饪视频转录中提取的程序对齐执行流程图，并明确映射到特定创作者。我们将程序风格计量学形式化为图学习和过程发现任务，揭示了一个基本的二元性：传统的词汇分类器由于语义泄漏而过拟合，而离散拓扑度量成功捕捉了创作者工作流程的严格物理约束。在此表征的基础上，我们将框架扩展到一个新的生成任务——预测创作者在未见菜肴上的确切结构执行图。我们揭示了风格生成中的基本二分法，分别为全球宏观规划和局部结构执行。我们的结果表明，少样本LLMs在语义分配上占据主导地位，但在宏观规划上存在持续的不足，而我们的结构化两阶段模型通过严格的马尔可夫先验实现了卓越的拓扑控制。综合来看，一种程序生成的集成方法结合了两方面的优势，动态合成全球语义推理与局部拓扑特征，以自动发现和生成个性化工作流程。
cs.AI / 84 / 2608.24411
ResiSpec: Enhancing Multi-Candidate Speculative Sampling via Residual Distribution Shaping
ResiSpec：通过残差分布塑形增强多候选推测采样
Abstract
The efficiency of Large Language Model (LLM) serving is fundamentally limited by the sequential nature of autoregressive decoding. Speculative Decoding (SD) mitigates this by using a lightweight draft model to speculate future tokens, which are then validated by the LLM in a single parallel forward pass. To further boost efficiency, multi-candidate schemes propose diverse candidate sets to increase the likelihood of token acceptance. However, we show that these schemes are bottlenecked by Residual Drift: a phenomenon where the rejection of initial candidates causes the residual target distribution to diverge from the draft model's predictions. This shift renders subsequent candidates ineffective and forces the system into expensive resampling. To resolve this, we propose ResiSpec, a framework that strategically reforms the proposal distribution during verification to anchor the residual target mass within the draft model's high-confidence regions. By mathematically re-aligning the verification process without compromising output exactness, ResiSpec prevents candidate obsolescence and achieves up to 1.92$\times$ speedup over state-of-the-art multi-candidate methods. Code is available at https://github.com/Czzzk/Resispec.
Chinese Translation
大型语言模型（LLM）服务的效率在根本上受到自回归解码的顺序性质的限制。推测解码（SD）通过使用轻量级草稿模型来推测未来的标记，从而减轻了这一限制，这些标记随后通过LLM在单次并行前向传递中进行验证。为了进一步提高效率，多候选方案提出了多样化的候选集，以增加标记被接受的可能性。然而，我们表明，这些方案受到残差漂移的瓶颈：一种现象，其中初始候选的拒绝导致残差目标分布与草稿模型的预测发生偏离。这种偏移使得后续候选失效，并迫使系统进行昂贵的重采样。为了解决这个问题，我们提出了ResiSpec，一个在验证过程中战略性地重塑提议分布的框架，以将残差目标质量锚定在草稿模型的高置信区间内。通过在不妨碍输出准确性的情况下，数学上重新对齐验证过程，ResiSpec防止了候选的过时，并实现了比最先进的多候选方法高达1.92倍的加速。代码可在 https://github.com/Czzzk/Resispec 获取。
cs.AI / 85 / 2608.24419
A Judge Should Know What Changed:Construct Validity for LLM-as-a-Judge Evaluation
法官应了解变化：LLM作为法官评估的构念效度
Abstract
LLM-as-a-judge evaluation is usually assessed by agreement and robustness to surface perturbations, but reliability does not establish construct validity. We formalize construct validity for an evaluator as a two-dimensional profile: invariance S, the probability that a verdict is unchanged under construct-preserving edits, and construct sensitivity R, the probability that it changes under minimal construct-changing edits. We show that S and R are independent and that no scalar summary preserves all relevant comparisons. We measure the profile across 7 judges and 4 domains using 7 construct-changing intervention types and 5 register-only controls, with intervention direction determined by human annotators and generation, verification, and judging assigned to disjoint model families. At matched invariance S >= 0.90, judges average S = 0.945 but R = 0.319. Sensitivity also differs between scope and strength edits: R_scope = 0.383 versus R_strength = 0.262, a +0.121 gap with the same sign for all 7 judges. We further audit five public label sets and find that surface-only predictors reproduce 55%-67% of labels in paired mode, including 67.4% of MT-Bench human votes. These results show that high judge agreement can coexist with weak sensitivity to changes in the construct being evaluated, motivating joint reporting of invariance and sensitivity and auditing the validation set itself.
Chinese Translation
LLM作为法官的评估通常通过一致性和对表面扰动的鲁棒性来评估，但可靠性并不能确立构念效度。我们将评估者的构念效度形式化为一个二维轮廓：不变性 S，即在保持构念的编辑下，裁决不变的概率；构念敏感性 R，即在最小构念改变的编辑下，裁决变化的概率。我们展示了 S 和 R 是独立的，并且没有标量总结能够保留所有相关比较。我们通过 7 种构念改变的干预类型和 5 种仅注册的控制措施，在 7 位法官和 4 个领域中测量该轮廓，干预方向由人工注释者确定，生成、验证和判断分配给不重叠的模型家族。在匹配不变性 S >= 0.90 时，法官的平均 S = 0.945，但 R = 0.319。敏感性在范围和强度编辑之间也有所不同：R_scope = 0.383 与 R_strength = 0.262，所有 7 位法官均存在 +0.121 的差距，且符号相同。我们进一步审计了五个公共标签集，发现仅表面预测器在配对模式下重现了 55%-67% 的标签，包括 67.4% 的 MT-Bench 人类投票。这些结果表明，高法官一致性可以与对所评估构念变化的低敏感性共存，这促使我们共同报告不变性和敏感性，并审计验证集本身。
cs.AI / 86 / 2608.24427
Partial Identification under Causal Orders by Linear Programming
通过线性规划进行因果顺序下的部分识别
Abstract
Non-parametric (partial) identification of counterfactual queries typically relies on a fully specified causal graph. Motivated by settings with incomplete domain knowledge, we challenge this requirement by leveraging structural assumptions that are inherently implied by the query itself. We show that any counterfactual inquiry induces a, mostly partial, topological ordering over relevant variables, which, in turn, enables an explicit query parametrisation reducing the identification task to a linear program. This allows bounding arbitrary counterfactual and nested counterfactual queries. Our work can be viewed as a generalisation of the classical bounding framework of Tian and Pearl (2000), originally developed for probabilities of causation. We also prove the \emph{tightness} of our bounds by constructing structural causal models that attain the bounds whilst being compatible with both the observed data and the query-implied order. To assess both the generality and practical utility of the proposed bounding procedure, we revisit several case studies from the literature, demonstrating how the derived bounds can be used to yield informative insights even in the absence of an input causal graph.
Chinese Translation
反事实查询的非参数（部分）识别通常依赖于完全指定的因果图。基于对不完全领域知识的考虑，我们挑战了这一要求，通过利用查询本身固有的结构假设。我们表明，任何反事实查询都会在相关变量上诱导出一个主要是部分的拓扑排序，这反过来使得明确的查询参数化成为可能，从而将识别任务简化为一个线性规划。这使得我们能够对任意反事实和嵌套反事实查询进行界定。我们的工作可以被视为对Tian和Pearl（2000）经典界定框架的推广，该框架最初是为因果概率而开发的。我们还通过构建结构因果模型来证明我们界限的 extit{紧致性}，这些模型在与观察数据和查询所隐含的顺序兼容的情况下达到了界限。为了评估所提界定程序的普遍性和实际效用，我们重新审视了文献中的几个案例研究，展示了即使在没有输入因果图的情况下，所得到的界限也可以用于提供有价值的见解。
cs.AI / 87 / 2608.24441
A Behavior-Guided Online Probabilistic Forecasting Method for Electric vehicle Charging Loads
一种基于行为指导的电动汽车充电负荷在线概率预测方法
Abstract
Electric vehicle (EV) charging loads exhibit strong behavioral heterogeneity and temporal variability, posing significant challenges for online probabilistic forecasting under evolving operating conditions. In particular, persistent charging patterns may differ substantially across stations, while recent behavioral changes can continuously alter the underlying load distributions. This paper proposes a behavior-guided online probabilistic forecasting framework that explicitly characterizes persistent station-specific patterns and recent behavioral changes. A dual-timescale behavior representation is constructed to distinguish long-term charging characteristics from recent behavioral states and quantify their deviations. These behavioral changes are further semantically encoded to guide drift-aware forecasting adaptation, while a delayed-feedback mechanism ensures temporally consistent online updates when observations become available across different forecasting horizons. Experiments on ten heterogeneous real-world charging stations demonstrate that the proposed method consistently outperforms conventional forecasting models and concept-drift-aware online baselines in forecasting accuracy and probabilistic reliability. For 1-h-ahead forecasting, the proposed method reduces MSE and Pinball loss by 15.3\% and 17.8\%, respectively, over the corresponding best baselines. For 4-h-ahead forecasting, the improvements further reach 16.8\% and 22.6\%, respectively, demonstrating consistent performance gains under evolving charging behaviors and extended forecasting horizons.
Chinese Translation
电动汽车（EV）充电负荷表现出强烈的行为异质性和时间变异性，这给在不断变化的操作条件下进行在线概率预测带来了重大挑战。特别是，持续的充电模式在不同充电站之间可能存在显著差异，而近期的行为变化则可能持续改变基础负荷分布。本文提出了一种基于行为指导的在线概率预测框架，明确表征了特定充电站的持续模式和近期的行为变化。构建了双时间尺度的行为表示，以区分长期充电特征与近期行为状态，并量化它们的偏差。这些行为变化进一步被语义编码，以指导对漂移的感知预测适应，同时延迟反馈机制确保在不同预测时间范围内，当观察结果可用时进行时间一致的在线更新。在十个异质的真实充电站上的实验表明，所提出的方法在预测准确性和概率可靠性方面始终优于传统预测模型和概念漂移感知的在线基线。在1小时预测中，所提出的方法分别将均方误差（MSE）和Pinball损失降低了15.3%和17.8%，相较于对应的最佳基线。在4小时预测中，改善幅度进一步达到16.8%和22.6%，展示了在不断变化的充电行为和扩展预测时间范围下的一致性能提升。
cs.AI / 88 / 2608.24462
Mahalanobis-Based Multi-Head Attention for Complex State Propagation
基于马哈拉诺比斯距离的多头注意力机制用于复杂状态传播
Abstract
In this paper, we propose \textbf{Mahalanobis-Based Multi-Head Attention} (MHA-CSP), a novel attention mechanism that replaces the standard dot-product with a \textbf{Mahalanobis distance-based RBF kernel}, which effectively computes attention in an infinite-dimensional feature space without increasing the parameter count. Crucially, the positive definiteness of the Mahalanobis distance enables a \textbf{direct construction of Tree Attention}: attention scores are built directly from accumulated distances, with a LogSumExp correction that rectifies the raw distance by subtracting the log-sum of edge exponentials. Moreover, the multi-head Mahalanobis distance matrices are themselves repurposed to construct an \textbf{attention meshing mechanism}, enabling cross-head kernel collaboration that simultaneously boosts accuracy and training efficiency. Extensive experiments demonstrate that MHA-CSP, with only 119K parameters and \textbf{teacher forcing applied exclusively at the final hidden state}, consistently outperforms Transformer and GCN baselines trained from scratch under identical conditions on long-sequence state tracking tasks. While these baselines rely on dense attention or graph propagation, MHA-CSP achieves robust structured reasoning via synthetic distance rectification---powered by Mahalanobis-based attention---and efficient information bypass inherited from the CSP backbone. This result highlights the effectiveness of complex-valued state propagation with collaborative multi-head rectification in capturing symbolic structures, establishing a new efficiency-performance trade-off for structured reasoning.
Chinese Translation
在本文中，我们提出了一种新颖的注意力机制—— extbf{基于马哈拉诺比斯距离的多头注意力}（MHA-CSP），该机制用 extbf{基于马哈拉诺比斯距离的RBF核}替代了标准的点积，从而有效地在无限维特征空间中计算注意力，而不增加参数数量。关键是，马哈拉诺比斯距离的正定性使得 extbf{树状注意力的直接构建}成为可能：注意力分数直接基于累积距离构建，并通过LogSumExp修正来调整原始距离，修正过程为减去边缘指数的对数和。此外，多头马哈拉诺比斯距离矩阵本身被重新利用以构建 extbf{注意力网格机制}，实现跨头核的协作，既提升了准确性，又提高了训练效率。大量实验表明，MHA-CSP仅使用119K参数，并且 extbf{仅在最终隐状态应用教师强制}，在长序列状态跟踪任务中始终优于在相同条件下从头训练的Transformer和GCN基线。虽然这些基线依赖于密集注意力或图传播，但MHA-CSP通过合成距离修正实现了稳健的结构推理——这一过程由基于马哈拉诺比斯的注意力驱动，并继承了CSP骨干网的高效信息旁路。这一结果突显了基于协作多头修正的复数值状态传播在捕捉符号结构方面的有效性，为结构推理建立了新的效率与性能权衡。
cs.AI / 89 / 2608.24467
HMGCLIP: Heterogeneous Multi-Granularity Contrastive Learning for E-commerce Representation Learning
HMGCLIP：用于电子商务表示学习的异构多粒度对比学习
Abstract
Although recent Multimodal Large Language Models (MLLMs) have advanced general product understanding, they implicitly encode product information into global embeddings, thereby limiting their ability to capture fine-grained attributes. This limitation hinders performance in tasks requiring precise attribute discrimination, such as distinguishing subtle material differences among visually similar products. To address this challenge, we propose HMGCLIP, a unified multimodal embedding framework. By constructing a heterogeneous hypergraph, we leverage hypergraph topology to mine structure-aware hard negatives and align multi-granular semantics at both relation and hyperedge levels. This design enables a dual-granularity inference mechanism that dynamically fuses attribute evidence for both fine-grained and coarse-grained downstream tasks. Furthermore, we release a comprehensive fine-grained e-commerce dataset to facilitate future benchmarking. Extensive experiments on this new dataset and the public MAVE benchmark show that HMGCLIP outperforms strong multimodal encoders, MLLMs, and e-commerce baselines, validating the superiority of HMGCLIP.
Chinese Translation
尽管近期的多模态大型语言模型（MLLMs）在一般产品理解方面取得了进展，但它们将产品信息隐式编码为全局嵌入，从而限制了捕捉细粒度属性的能力。这一限制妨碍了在需要精确属性区分的任务中的表现，例如区分视觉上相似产品之间微妙的材料差异。为了解决这一挑战，我们提出了HMGCLIP，一个统一的多模态嵌入框架。通过构建异构超图，我们利用超图拓扑挖掘结构感知的难负样本，并在关系和超边层面对多粒度语义进行对齐。这一设计使得双粒度推理机制能够动态融合细粒度和粗粒度下游任务的属性证据。此外，我们发布了一个全面的细粒度电子商务数据集，以促进未来的基准测试。在这个新数据集和公共MAVE基准上的广泛实验表明，HMGCLIP优于强大的多模态编码器、MLLMs和电子商务基线，验证了HMGCLIP的优越性。
cs.AI / 90 / 2608.24470
Reinforcement Learning-Guided Evolutionary Policy Optimization for Preference-Adjustable Heterogeneous Agile Earth Observation Satellite Scheduling
基于强化学习指导的可调偏好异构敏捷地球观测卫星调度进化策略优化
Abstract
Heterogeneous agile Earth observation satellite (AEOS) scheduling requires task selection, satellite assignment, and observation sequencing under satellite-dependent visibility windows, attitude maneuvering requirements, energy consumption, and onboard storage constraints. Since satellites differ in orbital access, maneuvering capability, and payload resources, the same task may have different feasible windows, transition costs, and resource-consumption patterns on different platforms, which increases the difficulty of unified modeling and efficient optimization. To address this problem, this paper proposes an evolutionary policy optimization framework for heterogeneous AEOS scheduling with preference-adjustable weighted objectives. In the modeling layer, assignment-based indirect encoding is combined with decoder-based equivalent-cost evaluation to retain satellite-dependent constraints while integrating task gain, energy saving, and load balance into an interpretable scalar utility. In the optimization layer, schedule decoding, population-based search, and online actor-critic operator control are decoupled, so that reinforcement learning selects high-level search operators rather than constructing schedules directly. Based on this framework, a reinforcement-learning-assisted operator-selection memetic evolutionary algorithm (RLOSMEA) is developed to coordinate global exploration, feasibility recovery, and local refinement under a limited function-evaluation budget. Experiments on different heterogeneous AEOS scenarios show that RLOSMEA achieves higher overall weighted utility and more stable convergence than representative metaheuristic baselines. Sensitivity and learning-behavior analyses further confirm the robustness of the proposed method and the effectiveness of reinforcement-learning-guided operator selection.
Chinese Translation
异构敏捷地球观测卫星（AEOS）调度需要在卫星依赖的可见窗口、姿态机动要求、能量消耗和机载存储限制下进行任务选择、卫星分配和观测顺序安排。由于卫星在轨道接入、机动能力和有效载荷资源上存在差异，相同任务在不同平台上可能具有不同的可行窗口、转换成本和资源消耗模式，这增加了统一建模和高效优化的难度。为了解决这一问题，本文提出了一种针对异构AEOS调度的进化策略优化框架，该框架具有可调偏好的加权目标。在建模层面，基于分配的间接编码与基于解码器的等效成本评估相结合，以保留卫星依赖的约束，同时将任务收益、节能和负载平衡整合为可解释的标量效用。在优化层面，调度解码、基于种群的搜索和在线演员-评论家操作控制被解耦，从而使强化学习选择高层次的搜索操作符，而不是直接构建调度。基于该框架，开发了一种强化学习辅助的操作符选择模因进化算法（RLOSMEA），以在有限的函数评估预算下协调全局探索、可行性恢复和局部优化。在不同的异构AEOS场景下的实验表明，RLOSMEA在整体加权效用和收敛稳定性方面优于代表性的元启发式基线。敏感性和学习行为分析进一步确认了所提方法的鲁棒性以及基于强化学习的操作符选择的有效性。
cs.AI / 91 / 2608.24471
Implicit Q-learning-bootstrapped ant colony optimization for maritime moving-target observation scheduling with agile satellites
隐式Q学习引导的蚁群优化在灵活卫星海洋移动目标观测调度中的应用
Abstract
Maritime moving-target observation scheduling with agile Earth observation satellites is a dynamic, sequence-dependent combinatorial optimization problem. Sea-surface targets move continuously, causing feasible observation windows to vary with target motion and satellite orbital geometry. The scheduler must jointly determine task selection, satellite assignment, observation-window selection, and observation ordering under time-window, attitude-maneuvering, onboard-resource, and cloud-affected availability constraints. This paper proposes an implicit Q-learning-bootstrapped ant colony optimization method, termed IQACO, for multi-satellite maritime moving-target observation scheduling. Rather than directly learning a task-selection policy, IQACO embeds an offline implicit Q-learning module into constructive ant colony optimization to adaptively adjust the pheromone factor, heuristic factor, and evaporation rate. A compact search-state representation captures pheromone distribution, current and historical-best solution quality, and iteration progress. During online scheduling, ant colony optimization constructs feasible observation sequences, while the learned policy regulates exploration and exploitation according to the current search state. Experiments on 14 scenarios with different scales and satellite configurations show that IQACO obtains the highest mean observation benefit in every scenario, improves the result of conventional ant colony optimization by 3.40\%--9.40\%, accelerates convergence, and remains stable under different objective-weight settings. These results demonstrate that offline value learning provides an effective adaptive search-control mechanism for constrained maritime moving-target observation scheduling.
Chinese Translation
灵活的地球观测卫星进行海洋移动目标观测调度是一个动态的、序列依赖的组合优化问题。海面目标持续移动，导致可行的观测窗口随着目标运动和卫星轨道几何形状而变化。调度器必须在时间窗口、姿态机动、机载资源和受云影响的可用性约束下，联合确定任务选择、卫星分配、观测窗口选择和观测顺序。本文提出了一种隐式Q学习引导的蚁群优化方法，称为IQACO，用于多卫星海洋移动目标观测调度。IQACO并不是直接学习任务选择策略，而是将离线隐式Q学习模块嵌入到构造性蚁群优化中，以自适应调整信息素因子、启发因子和蒸发率。紧凑的搜索状态表示捕捉了信息素分布、当前和历史最佳解的质量以及迭代进展。在在线调度过程中，蚁群优化构建可行的观测序列，而学习到的策略根据当前搜索状态调节探索与利用。对14种不同规模和卫星配置的场景进行的实验表明，IQACO在每种场景中都获得了最高的平均观测收益，相比传统蚁群优化结果提高了3.40%至9.40%，加快了收敛速度，并在不同目标权重设置下保持稳定。这些结果表明，离线价值学习为受限的海洋移动目标观测调度提供了一种有效的自适应搜索控制机制。
cs.AI / 92 / 2608.24509
PeakBench: Benchmarking Resource-Aware Tool Invocation in LLM Agents
PeakBench：资源感知工具调用的基准测试
Abstract
LLM agents increasingly solve tasks by invoking multiple tools, where parallel execution is essential for low latency but difficult to manage safely. Existing agent benchmarks primarily evaluate tool selection, argument generation, and end-to-end success under mostly serial execution, largely overlooking valid parallelization and resource-constrained scheduling. This missing scheduling dimension creates a practical failure mode: serial execution is safe but slow, while resource-agnostic parallel execution is fast but prone to avoidable resource overflows. To address this gap, we introduce PeakBench, a benchmark of executable multi-tool workflows with execution-grounded dependency annotations and measured resource profiles. A central challenge in evaluating such workflows is attribution: failures and inefficiencies may arise from incorrect dependency planning, poor resource-constrained scheduling, or both. PeakBench addresses this challenge with a two-part evaluation framework that disentangles logical planning from physical scheduling, with dedicated metrics for each dimension. Using this framework, we show that strong logical planning does not reliably translate into safe or efficient execution under resource constraints. We further show that exposing resource information can reduce avoidable overflows and improve resource utilization, making PeakBench a useful testbed for diagnosing resource-aware agent behavior. Code is available at https://github.com/Czzzk/Staggering-the-Peaks.
Chinese Translation
LLM代理越来越多地通过调用多个工具来解决任务，其中并行执行对于降低延迟至关重要，但管理起来却很困难。现有的代理基准主要评估工具选择、参数生成和在大多数串行执行下的端到端成功，基本上忽视了有效的并行化和资源受限的调度。这一缺失的调度维度导致了一种实际的失败模式：串行执行是安全但缓慢的，而与资源无关的并行执行则是快速但容易导致可避免的资源溢出。为了解决这一问题，我们引入了PeakBench，这是一个可执行多工具工作流的基准，具有基于执行的依赖注释和测量的资源配置文件。评估此类工作流的一个核心挑战是归因：失败和低效可能源于不正确的依赖规划、糟糕的资源受限调度，或两者兼而有之。PeakBench通过一个两部分的评估框架来解决这一挑战，该框架将逻辑规划与物理调度分开，并为每个维度提供专门的指标。使用该框架，我们展示了强大的逻辑规划并不可靠地转化为在资源限制下的安全或高效执行。我们进一步表明，暴露资源信息可以减少可避免的溢出并改善资源利用，使PeakBench成为诊断资源感知代理行为的有用测试平台。代码可在 https://github.com/Czzzk/Staggering-the-Peaks 获取。
cs.AI / 93 / 2608.24534
Neurosymbolic Alignment for Physiologically-Safe Clinical Language Models
生理安全临床语言模型的神经符号对齐
Abstract
Clinical LLMs can generate recommendations that are factually plausible yet physiologically unsafe. We investigate whether safety alignment can be improved by grounding preference optimization in structured physiological knowledge rather than text-only supervision. Methods: We propose Neurosymbolic Alignment, a training-time framework that couples a 7B clinical LLM with an HGNN-based Physiological World Model over an 847K-node biomedical knowledge graph. Candidate responses are scored using homeostatic constraints, multi-hop path plausibility, and drug-interaction penalties, and the resulting rankings drive iterative on-policy ORPO updates. Evaluation is performed on the Clinical Safety Benchmark (CSB), a 2,500-scenario benchmark for physiological constraint violations in generative clinical reasoning. Results: Relative to ORPO, the proposed method improves CSS from 69.5% to 90.8% (+21.3 pp), reduces physician-evaluated HR from 14.1% to 5.1% on the blinded subset, and improves DID from 72.8% to 91.6%. These gains are corroborated by an HGNN-independent Rule-Engine Safety Score (RSS: 86.4%, +21.2 pp over ORPO; r=0.97 concordance with CSS). The method also exceeds GPT-4 (5-shot) on all safety metrics despite a 10x parameter disadvantage, and outperforms an inference-time self-correction pipeline (SFT+SelfCorrect) by 11.4 pp CSS. Under synthetic EHR-style noise, 84.2% CSS is retained. Ablation analysis shows that HGNN scoring (-16.2 pp) and iterative training (-11.5 pp) are the dominant contributors. PhysioScore calibration against 200 clinician labels yielded ECE = 0.038 and kappa = 0.91. Conclusion: Training-time physiological grounding produces measurable and independently verifiable safety improvements in open-weight clinical LLMs under controlled evaluation. External validation on real clinical data is required to determine whether these gains transfer to deployment settings
Chinese Translation
临床大语言模型（LLMs）能够生成在事实上看似合理但生理上不安全的建议。我们研究了通过基于结构化生理知识而非仅依赖文本监督的偏好优化，是否能够改善安全对齐。方法：我们提出了神经符号对齐（Neurosymbolic Alignment），这是一个训练时框架，将一个70亿参数的临床LLM与基于HGNN的生理世界模型结合在一个包含847,000节点的生物医学知识图谱上。候选响应通过稳态约束、多跳路径合理性和药物相互作用惩罚进行评分，结果排名驱动迭代的在政策优化（ORPO）更新。评估在临床安全基准（Clinical Safety Benchmark, CSB）上进行，这是一个针对生成临床推理中生理约束违反的2500场景基准。结果：与ORPO相比，所提方法将CSS从69.5%提高到90.8%（+21.3个百分点），在盲测子集上将医生评估的HR从14.1%降低到5.1%，并将DID从72.8%提高到91.6%。这些提升通过一个与HGNN无关的规则引擎安全评分（Rule-Engine Safety Score, RSS: 86.4%，比ORPO提高21.2个百分点；与CSS的相关性为r=0.97）得到了验证。该方法在所有安全指标上均超过了GPT-4（5-shot），尽管参数劣势达到10倍，并且在推理时自我修正管道（SFT+SelfCorrect）上提高了11.4个百分点CSS。在合成电子健康记录（EHR）风格噪声下，保持了84.2%的CSS。消融分析表明，HGNN评分（-16.2个百分点）和迭代训练（-11.5个百分点）是主要贡献因素。PhysioScore与200名临床医生标签的校准结果为ECE = 0.038，kappa = 0.91。结论：训练时的生理基础在受控评估下为开放权重的临床LLMs带来了可测量且可独立验证的安全性提升。需要在真实临床数据上进行外部验证，以确定这些提升是否能够转移到部署环境中。
cs.AI / 94 / 2608.24545
Discovering Adaptive Transmission Programs for Collective Innovation
发现适应性传输程序以促进集体创新
Abstract
Human collective intelligence depends on transmission processes: who shares what with whom, how, and when. While these processes emerge from individual cognition, they can also be directed by deliberate top-down protocols. Prior work has studied how transmission shapes collective outcomes primarily through the lens of network structure, varying who shares with whom and when. But networks are state-agnostic: they cannot condition transmission on what agents know or on the state of the collective. Here, we formalize transmission protocols as state-aware programs that route information and resources based on agent and collective states, and we use LLM-guided evolutionary search to design effective protocols in a collective discovery task. Evolved protocols increase collective performance over standard baselines from the literature by up to 37%. Ablations confirm that state-awareness drives this advantage: removing content-dependence while preserving network topology and timing eliminates performance gains. We find that evolved protocols also transfer across domain variations and agent populations. These results demonstrate that effective and generalizable transmission protocols can be discovered in silico, suggesting a path toward AI-assisted design of coordination infrastructure that enhances human collective intelligence.
Chinese Translation
人类集体智能依赖于传输过程：谁与谁分享什么，如何分享，以及何时分享。虽然这些过程源于个体认知，但也可以通过自上而下的有意协议进行引导。先前的研究主要通过网络结构的视角研究了传输如何塑造集体结果，变化的是谁与谁分享以及何时分享。然而，网络是与状态无关的：它们无法根据代理的知识或集体的状态来调节传输。在这里，我们将传输协议形式化为状态感知程序，这些程序根据代理和集体状态路由信息和资源，并使用大型语言模型（LLM）引导的进化搜索来设计在集体发现任务中有效的协议。进化后的协议在文献中的标准基准上提高了集体表现，最高可达37%。消融实验确认了状态感知驱动了这一优势：在保留网络拓扑和时序的情况下去除内容依赖性会消除性能提升。我们发现，进化后的协议也能在领域变异和代理群体之间迁移。这些结果表明，可以在计算机模拟中发现有效且可推广的传输协议，暗示了一条通过人工智能辅助设计协调基础设施以增强人类集体智能的路径。
cs.AI / 95 / 2608.24569
When "Must" Becomes "Maybe": Constraint Weakening in LLM Agent Workflows
当“必须”变为“可能”：大型语言模型代理工作流程中的约束减弱
Abstract
Large language model (LLM) agents coordinate complex tasks through multi-role and multi-stage workflows. Upstream state is repeatedly transformed into intermediate language artifacts, such as summaries, plans, tickets, memories, and handoff notes, from which downstream components act. For action-constraining state, topical retention is insufficient: an artifact may mention an unresolved condition while changing it from a requirement that must be resolved before execution into information that may merely inform the next action. We study this action-binding role as operational state preservation. Safety blockers provide a controlled instance because each source state has an explicit prerequisite, authority, fallback, and execution consequence. We condition on correct upstream identification, vary the handoff transformation, and evaluate an executor restricted to the resulting artifact. Across 1,296 controlled synthetic episodes, direct-handoff controls preserve every blocker, whereas compression, plan assimilation, convergence, ownership deferral, and precedent substitution repeatedly turn binding state into caveats or non-binding considerations. Normal handoff compression produces 100.0% deactivation and 54.2% forbidden action. Restoring all four state fields raises preservation to 100.0% and reduces forbidden action to 0.0%. Fixed-artifact interventions further separate preservation from containment: downstream verification eliminates forbidden action while artifact deactivation remains 95.3%. These results identify a state-transmission failure between information extraction and action. Handoff transformations can retain state content while weakening its constraints on downstream action. Semantic availability does not guarantee operational preservation.
Chinese Translation
大型语言模型（LLM）代理通过多角色和多阶段工作流程协调复杂任务。上游状态反复转化为中间语言工件，如摘要、计划、工单、记忆和交接笔记，供下游组件使用。对于限制行动的状态，主题保留是不够的：一个工件可能提到一个未解决的条件，同时将其从必须在执行前解决的要求转变为可能仅仅告知下一步行动的信息。我们研究这种行动绑定角色作为操作状态的保留。安全阻碍提供了一个受控实例，因为每个源状态都有明确的前提、权限、后备和执行后果。我们以正确的上游识别为条件，改变交接转化，并评估一个限制于生成工件的执行者。在1296个受控的合成情境中，直接交接控制保留了每一个阻碍，而压缩、计划同化、收敛、所有权推迟和先例替代则反复将绑定状态转变为警告或非绑定考虑。正常的交接压缩产生了100.0%的停用率和54.2%的禁止行动。恢复所有四个状态字段将保留率提高至100.0%，并将禁止行动减少至0.0%。固定工件干预进一步将保留与包含分开：下游验证消除了禁止行动，而工件停用率保持在95.3%。这些结果识别了信息提取与行动之间的状态传递失败。交接转化可以保留状态内容，同时减弱其对下游行动的约束。语义可用性并不保证操作的保留。
cs.AI / 96 / 2608.24570
EviDx: Evidence-Aware Active Diagnosis with Scaffolded LLM Agents
EviDx：基于证据的主动诊断与支架式大型语言模型代理
Abstract
Clinical diagnosis is an active evidence-seeking process in which clinicians acquire evidence, update competing hypotheses, and decide when the available evidence is sufficient for diagnosis. Yet many medical diagnosis systems built around large language models (LLMs) still formulate diagnosis as static case-to-answer prediction, with limited support for evidence acquisition. Agentic LLMs offer a dynamic alternative through tool use and intermediate diagnostic trajectories, but existing systems often under-specify how patient evidence should be exposed, scaffolded, and controlled at runtime. We introduce EviDx, an evidence-aware active diagnosis framework that pairs patient-specific diagnostic environments with a clinical diagnostic scaffold and an observer-guided runtime harness. In EviDx, $\mathcal{E}$-Synthesis constructs interactive environments from raw clinical cases; the scaffold organizes role-specialized agents, evidence tools, and evolving evidence states; and the harness regulates diagnostic termination by tracking uncertainty and evidence coverage. A 3-level evaluation pyramid assesses execution robustness, reasoning dynamics, and diagnostic outcomes. Experiments show that EviDx improves diagnostic performance and process stability while revealing model-dependent capability boundaries.
Chinese Translation
临床诊断是一个主动寻求证据的过程，在这个过程中，临床医生获取证据、更新竞争假设，并决定何时现有证据足以进行诊断。然而，许多基于大型语言模型（LLMs）构建的医学诊断系统仍将诊断视为静态的案例到答案的预测，且对证据获取的支持有限。代理型LLMs通过工具使用和中间诊断轨迹提供了一种动态替代方案，但现有系统往往未能明确如何在运行时暴露、支架化和控制患者证据。我们提出了EviDx，这是一个基于证据的主动诊断框架，将患者特定的诊断环境与临床诊断支架和观察者引导的运行时工具相结合。在EviDx中，$ ext{E}$-合成从原始临床案例构建交互环境；支架组织角色专用代理、证据工具和不断演变的证据状态；而工具则通过跟踪不确定性和证据覆盖来调节诊断终止。一个三级评估金字塔评估执行的稳健性、推理动态和诊断结果。实验表明，EviDx提高了诊断性能和过程稳定性，同时揭示了模型依赖的能力边界。
cs.AI / 97 / 2608.24571
Joint Optimization of Tool Creation and Use for Large Language Model Agents
大型语言模型代理的工具创建与使用的联合优化
Abstract
Tool-augmented language models are bounded by the APIs humans bothered to write; existing tool-creation systems patch this by prompting a frozen LLM at inference time, leaving the model that writes a tool decoupled from the one that uses it, with no signal that the schemas it produces are schemas it can invoke. We propose SMITH (Schema-grounded Multi-task Iterative Tool Honing), a reinforcement learning framework that jointly trains tool creation and tool use inside a single policy. Each rollout is either a build task (write a tool from a few examples) or a use task (invoke a pooled tool on a held-out question). Three separate reward axes catch schema, code, and outcome failures independently, so each failure mode contributes its own gradient. A 4B Qwen3 trained with SMITH on 13 procedural reasoning tasks with exact verifiers reaches 79.8 macro-average accuracy on held-out tasks, the best across all evaluated methods and ahead of an untrained 30B-A3B tool-writer. It also reaches 40.4 on TabMWP-Hard and 42.6 on out-of-domain GQA (+7.6 over the best same-backbone inference-time baseline), without any visual or tabular training data. Tools written by our 4B models also lifted the performance of LFM-2.5-350M and Qwen3-30B-A3B under same reasoning tasks.
Chinese Translation
工具增强型语言模型受到人类编写的API的限制；现有的工具创建系统通过在推理时提示一个固定的LLM来解决这一问题，使得编写工具的模型与使用工具的模型解耦，且没有信号表明其生成的模式是可以调用的模式。我们提出了SMITH（基于模式的多任务迭代工具优化），这是一种强化学习框架，能够在单一策略中联合训练工具创建和工具使用。每次回滚要么是构建任务（从几个示例中编写工具），要么是使用任务（在保留的问题上调用一个汇总工具）。三个独立的奖励轴分别捕捉模式、代码和结果的失败，因此每种失败模式都会贡献其自身的梯度。经过SMITH训练的4B Qwen3在13个程序推理任务上达到了79.8的宏平均准确率，这是所有评估方法中最佳的，并且优于一个未训练的30B-A3B工具编写器。在TabMWP-Hard上也达到了40.4，在域外GQA上达到了42.6（比最佳相同骨干推理时基线高出7.6），且没有任何视觉或表格训练数据。我们4B模型编写的工具也提升了LFM-2.5-350M和Qwen3-30B-A3B在相同推理任务下的表现。
cs.AI / 98 / 2608.24574
PhysMLLMs: Spatial Priors for Unified Referring Segmentation and Grounded Reasoning of Images and Videos
PhysMLLMs：统一图像和视频的指称分割与基于语境推理的空间先验
Abstract
Video multimodal large language models support language guided video segmentation, but they often show spatio temporal inconsistencies, e.g., jitter, drift, and identity switches. These failures are more common when targets are partly hidden or when similar objects appear nearby.One likely reason is that current training lacks explicit spatial priors, which makes it difficult to maintain stable spatial identity and shape over time. We present PhysMLLMs, a training-stage prior injection architecture that injects physics-inspired spatial continuity priors into Video MLLMs. PhysMLLMs is designed to encourage more stable object-centered representations by aligning the student global visual representation with a frozen teacher model during training. Our core mechanism, Global Representation Prior Alignment (REPA-Global), distills global visual representations from a frozen DINOv2 teacher using an offline embedding cache and a scheduled distillation plan. This design keeps inference unchanged and does not add inference time cost. Across multiple video benchmarks, PhysMLLMs improves video segmentation mask quality and cross-frame consistency, with larger gains on challenging cases involving small targets, fast motion, occlusion, distractors, and reasoning queries. On single-frame referring image segmentation and representative general VLM benchmarks, PhysMLLMs maintains comparable performance, demonstrating that the injected spatial prior improves video consistency without compromising image-level grounding or general multimodal capability. These results suggest that physics-inspired spatial prior injection can improve temporal stability while preserving general capability. The code is available at https://github.com/tusu-code/20260121-icml2026-2.git.
Chinese Translation
视频多模态大语言模型支持语言引导的视频分割，但它们常常表现出时空不一致性，例如抖动、漂移和身份切换。这些失败在目标部分被遮挡或相似物体近距离出现时更为常见。一个可能的原因是当前的训练缺乏明确的空间先验，这使得在时间上保持稳定的空间身份和形状变得困难。我们提出了PhysMLLMs，一种训练阶段的先验注入架构，将受物理启发的空间连续性先验注入到视频多模态大语言模型中。PhysMLLMs旨在通过在训练过程中将学生的全局视觉表示与冻结的教师模型对齐，鼓励更稳定的以对象为中心的表示。我们的核心机制，全局表示先验对齐（REPA-Global），利用离线嵌入缓存和计划的蒸馏方案，从冻结的DINOv2教师中提取全局视觉表示。该设计保持推理不变，并且不会增加推理时间成本。在多个视频基准测试中，PhysMLLMs提高了视频分割掩码的质量和跨帧一致性，在涉及小目标、快速运动、遮挡、干扰物和推理查询的挑战性案例中取得了更大的提升。在单帧指称图像分割和代表性通用VLM基准测试中，PhysMLLMs保持了可比的性能，表明注入的空间先验在不妥协图像级基础或通用多模态能力的情况下改善了视频一致性。这些结果表明，受物理启发的空间先验注入可以在保持通用能力的同时提高时间稳定性。代码可在 https://github.com/tusu-code/20260121-icml2026-2.git 获取。
cs.AI / 99 / 2608.24585
Pivot-and-Station Multi-Agent Path Finding: Solvability, Complexity, and Algorithms
枢轴与站点多智能体路径规划：可解性、复杂性与算法
Abstract
Automated high-density storage systems (warehouses, robotic parking, plant logistics, etc.) require fleets of agents to move through scarce task-critical resources and then park without obstructing future operations. We introduce Pivot-and-Station Multi-Agent Path Finding (PS-MAPF), a MAPF variant in which a subset of tasked agents must each visit one of a set of interchangeable pivots (e.g., workstations) before the entire fleet terminates at anonymous stations, one agent per station. We characterize solvability completely: every instance on a 2-edge-connected graph is solvable, and, on arbitrary connected graphs, a structural effective-distance measure relative to the number of unoccupied vertices gives a necessary and sufficient condition. We prove that minimizing station-makespan or station-flowtime is NP-hard already with a single pivot. We present three algorithms, a complete baseline, a SAT-based optimal solver, and Pivot-Prioritized Planning (PPP), the last solving 74-89% of benchmark instances with makespan and flowtime orders of magnitude below the baseline.
Chinese Translation
自动化高密度存储系统（如仓库、机器人停车场、工厂物流等）需要一组智能体在稀缺的任务关键资源中移动，并在不妨碍未来操作的情况下停车。我们引入了枢轴与站点多智能体路径规划（Pivot-and-Station Multi-Agent Path Finding, PS-MAPF），这是一个MAPF变体，其中一部分有任务的智能体必须各自访问一组可互换的枢轴（例如，工作站），然后整个队伍在匿名站点终止，每个站点一个智能体。我们完全表征了可解性：在2边连通图上的每个实例都是可解的，而在任意连通图上，相对于未占用顶点数量的结构有效距离度量给出了必要且充分的条件。我们证明了最小化站点完工时间或站点流动时间在仅有一个枢轴的情况下已经是NP-hard。我们提出了三种算法，包括一个完整的基线算法、一个基于SAT的最优求解器和枢轴优先规划（Pivot-Prioritized Planning, PPP），后者在完工时间和流动时间上解决了74-89%的基准实例，且其数量级远低于基线。
cs.AI / 100 / 2608.24632
Causal Modelling of Support Interventions for Student Competency Assessment
学生能力评估支持干预的因果建模
Abstract
Accurate assessment of student competencies is essential for enabling educators to identify individual needs, design targeted interventions, and evaluate the effectiveness of educational strategies. Empirical assessment procedures are typically grounded in psychometric models, such as item response theory, which relate student competence levels to performance on assessment tasks. In this paper, we advocate adopting a structural causal modelling approach to educational assessment, moving beyond probabilistic belief updating toward a framework that explicitly supports interventional and counterfactual reasoning. We propose a corresponding protocol for its construction and analyse the practical relevance of forms of reasoning that remain inaccessible to standard associative models, including the explicit modelling of interventions such as hints and the related counterfactual scenario analysis. Although our protocol requires the structural equations to be elicited from experts, the necessary information is purely logical and does not rely on probabilistic, less tenable assumptions. We illustrate the approach using data from an assessment that employs complex tasks designed to measure compulsory school student algorithmic skills.
Chinese Translation
准确评估学生能力对于教育工作者识别个体需求、设计针对性干预措施以及评估教育策略的有效性至关重要。经验评估程序通常基于心理测量模型，如项目反应理论，该模型将学生能力水平与评估任务的表现联系起来。本文主张在教育评估中采用结构因果建模方法，超越概率信念更新，朝着一个明确支持干预和反事实推理的框架发展。我们提出了相应的构建协议，并分析了那些对标准关联模型不可及的推理形式的实际相关性，包括对提示等干预的明确建模以及相关的反事实情境分析。尽管我们的协议要求从专家那里获取结构方程，但所需的信息纯粹是逻辑性的，不依赖于概率性和较不可靠的假设。我们使用一项评估的数据来说明该方法，该评估采用复杂任务设计以测量义务教育阶段学生的算法技能。
cs.AI / 101 / 2608.24658
Parason: Revealing Subtask and Trial Parallelism in LLM Reasoning
Parason：揭示大语言模型推理中的子任务和试验并行性
Abstract
Scaling test-time reasoning has substantially improved the problem-solving ability of large language models (LLMs), but standard autoregressive decoding still executes long reasoning traces sequentially, creating severe latency for difficult tasks (up to days and weeks). Parallel reasoning offers a natural remedy. However, prior systems primarily focus on Subtask Parallelism, where the model learns to decompose a high-level task into smaller chunks that can be solved independently. This approach overlooks another pervasive form of parallelism: Trial Parallelism, where multiple speculative attempts explore, verify, and aggregate competing hypotheses in parallel. In this paper, we introduce Parason, which reveals and learns both forms of parallelism in LLM reasoning. Our analysis identifies Trial Parallelism as the majority of parallelizable reasoning computation (65.5% in DeepSeek-V4's reasoning steps in HLE), and it becomes increasingly dominant on hard problems. Guided by this taxonomy, Parason converts sequential reasoning traces into structured parallel trajectories with a context-free grammar, then trains models with Parallelism-Aware Group Relative Policy Optimization (PA-GRPO), whose reward jointly balances accuracy, latency, and the two parallelism ratios. At inference time, Parason executes the learned parallel structure through tool calls, translating theoretical savings to real-world wall-clock acceleration. Experiments on mathematical reasoning benchmarks including AIME24 and AIME25 show that Parason achieves an average acceleration about 1.7$\times$ while maintaining competitive accuracy.
Chinese Translation
在测试时推理的规模化显著提升了大语言模型（LLMs）的问题解决能力，但标准的自回归解码仍然按顺序执行长推理轨迹，这在处理困难任务时造成了严重的延迟（可达数天和数周）。并行推理提供了一种自然的解决方案。然而，之前的系统主要关注子任务并行性，即模型学习将高层任务分解为可以独立解决的小块。这种方法忽视了另一种普遍存在的并行性形式：试验并行性，在这种形式中，多个推测性尝试并行探索、验证和聚合竞争假设。在本文中，我们介绍了Parason，它揭示并学习了LLM推理中的这两种并行性。我们的分析表明，试验并行性占可并行化推理计算的绝大部分（在HLE中，DeepSeek-V4的推理步骤中占65.5%），并且在困难问题上变得越来越主导。在这一分类法的指导下，Parason将顺序推理轨迹转换为结构化的并行轨迹，使用无上下文文法，然后通过并行性感知的组相对策略优化（PA-GRPO）训练模型，其奖励共同平衡准确性、延迟和两种并行性比率。在推理时，Parason通过工具调用执行学习到的并行结构，将理论节省转化为现实世界的时钟加速。在包括AIME24和AIME25在内的数学推理基准测试中，实验表明Parason实现了约1.7倍的平均加速，同时保持了竞争性的准确性。
cs.AI / 102 / 2608.24662
The Invisible Editorial Layer: Formalizing Undisclosed Inference-Time Steering, Probability Placement, and the Attribution Problem in Deployed Language Models
隐形编辑层：在部署的语言模型中形式化未披露的推理时间引导、概率分布调整和归因问题
Abstract
Large language models (LLMs) are commonly evaluated under the assumption that their observable behavior is primarily determined by model weights, training data, alignment procedures, and user prompts. This view is incomplete. Modern inference pipelines may systematically modify the probability distribution produced by a model immediately before token selection, creating an additional layer of control between frozen weights and observed text. While controlled generation (e.g., PPLM, GeDi, DExperts, FUDGE) and text-watermarking systems (e.g., SynthID-Text) demonstrate the technical maturity of decoding- and logit-level interventions, the governance, security, and economic implications of an undisclosed inference policy remain comparatively underexplored. This paper examines the emergence of inference-time framing bias: the systematic modification of generated language toward political, ideological, institutional, or commercial frames via interventions applied after model inference but before token sampling. We formalize the operational reality Model != Deployed System and introduce three concepts: (1) the Inference Attribution Problem, characterizing why observed behavioral bias cannot generally be causally attributed to model weights alone under limited observability; (2) Probability Placement, defining a hypothetical advertising primitive in which commercial influence is implemented through systematic shifts in generation probabilities rather than explicit product insertions; and (3) Inference Policy Transparency, a governance principle for making deployment-layer interventions auditable. We examine these concepts in relation to Article 5 of the EU AI Act, the EU Digital Services Act, and FTC doctrines.
Chinese Translation
大型语言模型（LLMs）通常在假设其可观察行为主要由模型权重、训练数据、对齐过程和用户提示决定的前提下进行评估。然而，这一观点并不完整。现代推理管道可能系统性地修改模型在令牌选择之前产生的概率分布，在冻结的权重和观察到的文本之间创建了一个额外的控制层。尽管受控生成（例如，PPLM、GeDi、DExperts、FUDGE）和文本水印系统（例如，SynthID-Text）展示了解码和对数值干预的技术成熟性，但未披露的推理政策的治理、安全和经济影响仍然相对缺乏探索。本文考察了推理时间框架偏见的出现：通过在模型推理后但在令牌采样之前施加干预，系统性地修改生成语言以符合政治、意识形态、机构或商业框架。我们形式化了操作现实，即模型不等于部署系统，并引入了三个概念：（1）推理归因问题，描述了在有限可观察性下，为什么观察到的行为偏见通常不能仅归因于模型权重；（2）概率分布调整，定义了一种假设的广告原语，其中商业影响通过系统性改变生成概率而非显式产品插入来实现；（3）推理政策透明度，一项治理原则，旨在使部署层干预可审计。我们将这些概念与欧盟人工智能法案第5条、欧盟数字服务法案以及联邦贸易委员会的相关原则进行探讨。
cs.AI / 103 / 2608.24691
Confident at the moment of action: belief miscalibration in LLM play under hidden information
行动时的自信：在隐藏信息下大型语言模型的信念误校准
Abstract
Agentic systems increasingly gate actions on a model's own stated confidence, which assumes confidence tracks correctness at the moment of acting. We test this in a hidden-information chess variant where royal status can be secretly, repeatedly relocated between pieces, and where an agent's stated probability distribution over the opponent's hidden royal piece -- elicited every turn, separately from the move it chooses -- is scored against ground truth recoverable after the game. Across two independent batches, captures made at high stated confidence ($\geq 0.5$) about the hidden piece's location were correct in 1 of 62 cases. The calibration deficit is concentrated almost entirely in these events: 99.3% of it in the original batch, 98.7% in the replication. The same pattern, in weaker form, orders consistently (point estimates only; most pairwise gaps are not statistically distinguishable at this sample size) across four further model configurations spanning a second provider -- reported as scope for the finding, not as evidence that capability predicts calibration: a same-model comparison at a fixed external leaderboard score shows a deliberation-budget change alone moves the metric by nearly as much as a large cross-model gap. In a separate seat, conventional evaluation axes -- legality, cost, latency, completion rate -- can dissociate entirely from belief quality, with the configuration winning on every conventional axis producing the worst belief quality tested. A model exhibiting this pattern can still win the game its belief was about, which is why outcome-only evaluation would not detect it.
Chinese Translation
代理系统越来越多地基于模型自身所声明的自信来控制行动，这假设自信在行动时与正确性相匹配。我们在一种隐藏信息的国际象棋变体中测试这一假设，其中王的地位可以在棋子之间秘密且重复地移动，而代理对对手隐藏王棋子的位置所声明的概率分布——在每轮中单独引出，与其选择的移动无关——与游戏结束后可恢复的真实情况进行评分。在两个独立的批次中，在对隐藏棋子位置的高自信（$ ext{自信} ext{≥} 0.5$）下做出的捕获仅在62个案例中正确1次。校准缺陷几乎完全集中在这些事件中：原始批次中99.3%的缺陷，复制批次中98.7%。同样的模式在四个进一步的模型配置中以较弱的形式一致出现（仅点估计；在这个样本大小下，大多数成对差距在统计上不可区分），这些配置跨越第二个提供者——这被报告为发现的范围，而不是能力预测校准的证据：在固定外部排行榜分数下的同模型比较显示，仅仅改变深思预算就能使指标变化几乎与大型跨模型差距相当。在另一个方面，传统评估轴——合法性、成本、延迟、完成率——可以与信念质量完全脱离，所有传统轴上获胜的配置在测试中产生了最差的信念质量。表现出这种模式的模型仍然可以赢得其信念所涉及的游戏，这就是为什么仅基于结果的评估无法检测到这一点。
cs.AI / 104 / 2608.24713
Lifted Model Construction under Approximate Commutativity
基于近似交换性的提升模型构建
Abstract
Lifted inference algorithms enable scalable probabilistic inference even for large object domains by leveraging the indistinguishability of objects in a probability distribution. An essential prerequisite for constructing a lifted representation is to identify commutative factors, i.e., functions whose output values are invariant under permutations of a subset of their input values, in a potential-based factorisation. In practice, however, parameters learned from data inevitably deviate even if associated objects are indistinguishable, causing their corresponding factors to be only approximately commutative instead of being exactly commutative. We address this problem by introducing the concept of {\epsilon}-commutativity, a relaxation of commutativity where output values are only approximately invariant under permutations of input values. Specifically, we show how {\epsilon}-commutativity can be exploited for lifted model construction, downstream probabilistic inference, and prove strict bounds on the induced approximation error, thereby ensuring the practical applicability of lifted model construction while maintaining highly accurate query results. These theoretical guarantees are confirmed empirically, demonstrating comparable query accuracy at lower runtime.
Chinese Translation
提升推理算法通过利用概率分布中对象的不可区分性，使得即使在大型对象领域也能实现可扩展的概率推理。构建提升表示的一个基本前提是识别交换因子，即在潜在基础分解中，其输出值在其输入值的一个子集的排列下是不变的函数。然而，在实践中，即使相关对象是不可区分的，从数据中学习到的参数不可避免地会偏离，导致其对应的因子仅近似交换，而不是完全交换。我们通过引入{ extepsilon}-交换性的概念来解决这个问题，{ extepsilon}-交换性是交换性的放宽，其中输出值在输入值的排列下仅近似不变。具体而言，我们展示了如何利用{ extepsilon}-交换性进行提升模型构建和下游概率推理，并证明了引入的近似误差的严格界限，从而确保提升模型构建的实际适用性，同时保持高度准确的查询结果。这些理论保证通过实证得到了确认，显示出在较低的运行时间下具有可比的查询准确性。
cs.AI / 105 / 2608.24735
Meta$^n$: Recursive Self-Improvement through Emergent Depth
Meta$^n$: 通过涌现深度实现递归自我改进
Abstract
Self-improving LLM agents refine answers, not the process that produces those answers. Systems that add a meta-level hold that level fixed, and those that edit themselves must leave part of their own editing machinery untouched to stay stable, capping the meta-depth they realize at roughly two. We present Meta$^n$, which keeps the meta-operation fixed and recurses on its input instead. That operation, $\Omega$, is applied repeatedly to its own products, reading the traces of the solver stack below together with the code that produced them, then writing the next layer as a strategic pre-process and a library of callable helpers. Because $\Omega$ never changes, it cannot destabilize the system, and because its input strictly grows, each layer reasons from a higher vantage than the last. Depth is set by convergence rather than fixed in advance, and an evolutionary archive searches over layer chains. Across two backbones, Meta$^n$ outperforms prior self-improving agents on all eight benchmark families. The sharpest case is ARC-AGI-2, built to resist skill memorization, where it alone scores above zero. Ablations indicate that most of the gain from recursion comes from the conditioning each layer passes to the next, and distinct layer roles emerge with depth although no prompt prescribes them. Code available at https://github.com/minnesotanlp/meta-n
Chinese Translation
自我改进的大型语言模型（LLM）代理精炼答案，而不是产生这些答案的过程。添加元层的系统保持该层固定，而那些自我编辑的系统必须保持其部分编辑机制不变以保持稳定，从而将其实现的元深度限制在大约两层。我们提出了Meta$^n$，它保持元操作固定，而是对其输入进行递归。该操作$ ext{Ω}$被反复应用于其自身的产物，读取下方求解器栈的痕迹以及产生这些产物的代码，然后将下一层写作战略预处理和可调用助手库。由于$ ext{Ω}$从不改变，因此它不会使系统不稳定，并且由于其输入严格增长，每一层的推理都比上一层具有更高的视角。深度由收敛决定，而不是预先固定，并且一个进化档案在层链上进行搜索。在两个基础架构上，Meta$^n$在所有八个基准系列中超越了之前的自我改进代理。最显著的案例是ARC-AGI-2，该系统旨在抵抗技能记忆化，只有它的得分超过零。消融实验表明，递归带来的大部分收益来自于每一层传递给下一层的条件，尽管没有提示规定它们，但随着深度的增加，独特的层角色逐渐显现。代码可在 https://github.com/minnesotanlp/meta-n 获取。
cs.AI / 106 / 2608.24758
RACE: Scalable Statistical Estimation of Functional Consistency in LLM Neurons
RACE：可扩展的LLM神经元功能一致性统计估计
Abstract
Discovering stable neuron behavior across entire domains remains a challenge in mechanistic interpretability. Existing methods often rely on instance-level point estimates or computationally expensive procedures, which either obscure population-level variability or limit scalable domain-wide analysis. We present RACE (Residual Alignment for Consistency Estimation), a forward-pass statistical framework that evaluates the domain-wide functional consistency of Transformer neurons. Perturbation experiments demonstrate that RACE achieves superior domain specificity compared to gradient-based point estimates. Meanwhile, token-distribution-level results verify the association between the selected neurons and the target domain. Furthermore, its computational overhead is two orders of magnitude lower than that of gradient-based methods.
Chinese Translation
在机制可解释性领域，发现整个领域内稳定的神经元行为仍然是一个挑战。现有方法通常依赖于实例级点估计或计算成本高昂的程序，这要么模糊了群体级的变异性，要么限制了可扩展的领域范围分析。我们提出了RACE（Residual Alignment for Consistency Estimation），这是一个前向传播统计框架，用于评估Transformer神经元的领域范围功能一致性。扰动实验表明，RACE在领域特异性方面优于基于梯度的点估计。同时，基于令牌分布的结果验证了所选神经元与目标领域之间的关联。此外，其计算开销比基于梯度的方法低两个数量级。
cs.AI / 107 / 2608.24764
Evidence Blindness in Direct Corpus Interaction: Persistent Navigation with AtlasNav
直接语料库交互中的证据盲区：使用AtlasNav的持续导航
Abstract
Large language model agents are moving beyond conventional retrieval-augmented generation toward direct interaction with external corpora. Direct Corpus Interaction (DCI) keeps the full corpus accessible, yet reachable evidence can remain unusable under finite interaction budgets. Required evidence may fail to surface, a surfaced supporting document may remain unopened, or an opened document may fail to expose its decisive fragment. We call this progressive silent loss Evidence Blindness and quantify it through stage-wise evidence realization. Within the DCI paradigm, raw interaction adds little reusable corpus organization, while dynamic-workspace methods reconstruct a query-conditioned interaction space from each query and trajectory. In both cases, useful structure is recovered largely online. We instead formulate large-scale agentic search as finite-budget navigation over reusable corpus structure. We introduce AtlasNav, a persistent multi-view corpus-navigation framework that retains direct corpus interaction but organizes the corpus once into a Corpus Atlas, allowing each query to navigate adaptively rather than reconstruct shared structure. On BrowseComp-Plus, AtlasNav achieves 92.05% strict accuracy while reducing recorded online inference cost by 30.21% relative to the prior dynamic-workspace state of the art. Under matched budgets, it realizes the complete required evidence earlier and approaches the same model's evidence-supplied empirical reference more rapidly. The same representation principle remains effective under PhantomWiki's distinct corpus organization and controlled 10K-1M scaling, and transfers competitively to heterogeneous enterprise knowledge. These results show that agentic search depends not only on accessible evidence, but also on how the corpus is represented so that limited interaction becomes effective navigation.
Chinese Translation
大型语言模型代理正在超越传统的检索增强生成，朝着与外部语料库的直接交互发展。直接语料库交互（DCI）保持了完整语料库的可访问性，但在有限的交互预算下，可获取的证据可能仍然无法使用。所需的证据可能未能显现，显现的支持文档可能未被打开，或者已打开的文档可能未能揭示其决定性片段。我们将这种逐步的静默损失称为证据盲区，并通过阶段性证据实现进行量化。在DCI范式下，原始交互对可重用语料库组织的贡献有限，而动态工作区方法则从每个查询和轨迹重建查询条件的交互空间。在这两种情况下，有用的结构主要是在线恢复的。相反，我们将大规模代理搜索形式化为在可重用语料库结构上的有限预算导航。我们引入了AtlasNav，一个持久的多视角语料库导航框架，它保留了直接语料库交互，但将语料库一次性组织成语料库地图（Corpus Atlas），使每个查询能够自适应导航，而不是重建共享结构。在BrowseComp-Plus上，AtlasNav实现了92.05%的严格准确率，同时相较于之前的动态工作区最先进技术，减少了30.21%的在线推理成本。在匹配预算下，它更早地实现了完整的所需证据，并更快速地接近同一模型的证据供给经验参考。在PhantomWiki独特的语料库组织和控制的10K-1M规模下，相同的表示原则仍然有效，并且能够竞争性地转移到异构企业知识中。这些结果表明，代理搜索不仅依赖于可访问的证据，还依赖于语料库的表示方式，从而使有限的交互变得有效导航。
cs.AI / 108 / 2608.24777
StepGuard: Learning Step-Level Guardrails with Scalable Supervision and Safety-Utility Balancing
StepGuard：通过可扩展监督和安全-效用平衡学习步骤级保护措施
Abstract
LLM-based agents can interact with external environments through tool invocation, but this capability also introduces security risks such as file modification, information leakage, and unauthorized actions. Existing guardrails often evaluate completed trajectories, leaving pre-execution monitoring of step-level actions underexplored. We propose StepGuard, a step-level guard model that can audit completed agent trajectories and check tool actions before they are executed. To train StepGuard, we introduce StepGen, an automatic data engine that generates safe and unsafe trajectories with the same context but different actions at the risky step. To further reduce over-defense and under-defense, we propose Balance-GRPO, which dynamically balances learning between safe and unsafe actions based on their observed accuracy. Experiments show that StepGuard achieves the highest average accuracy among open-weight guard models, with performance comparable to GPT-5.4. When used to guard agents on AgentDojo and AgentDyn, StepGuard reduces mean attack success rate by 77.3% relative to the no-guard setting, while mean utility drops by only 2.8 percentage points.
Chinese Translation
基于大型语言模型（LLM）的智能体可以通过工具调用与外部环境进行交互，但这一能力也带来了安全风险，例如文件修改、信息泄露和未经授权的操作。现有的保护措施通常评估已完成的轨迹，而对步骤级操作的执行前监控研究较少。我们提出了StepGuard，这是一种步骤级保护模型，可以审计已完成的智能体轨迹，并在工具操作执行之前进行检查。为了训练StepGuard，我们引入了StepGen，这是一种自动数据引擎，能够生成在相同上下文下但在风险步骤中采取不同操作的安全和不安全轨迹。为了进一步减少过度防御和不足防御，我们提出了Balance-GRPO，它根据观察到的准确性动态平衡安全和不安全操作之间的学习。实验表明，StepGuard在开放权重保护模型中实现了最高的平均准确率，其性能可与GPT-5.4相媲美。在AgentDojo和AgentDyn上使用StepGuard保护智能体时，相较于无保护设置，平均攻击成功率降低了77.3%，而平均效用仅下降了2.8个百分点。
cs.AI / 109 / 2608.24790
Right Diagnoses, Decorative Reasoning:A Perturbation Audit of Medical Chain-of-Thought
正确的诊断，装饰性推理：医学思维链的扰动审计
Abstract
Clinicians read chain-of-thought (CoT) rationales as evidence of medical reasoning, but whether the visible chain plays that role is rarely tested. General-domain CoT-faithfulness probes ignore clinical cost, and medical LLM evaluations treat the chain as a black box. We close this gap with a medical perturbation audit: a 30-operator battery edits both the chain and the question with clinically motivated operators (severity reversal, negation flip, demographic swap, evidence ablation), paired with a chain-update times answer-flip joint analysis that classifies each model by its failure mode. Applied to 14 LLMs on four medical QA benchmarks, three independent tests converge: the Chain-Decoupling Rate (CDR; chain does not register the edit and the answer does not flip) is 72.9% panel-wide on clinically meaningful destructive edits, chain corruption leaves accuracy unchanged, and removing CoT prompting does not reduce accuracy. Two board-certified clinicians re-annotate N=197 perturbed questions; 98.5% leave the gold defensible. The pattern holds across medical and reasoning fine-tuning and scale; on the closed-source tier, where the chain text is unavailable, the answer-side signals are consistent with the same decoupling. Our framework and CDR provide a reusable yardstick for auditing whether medical CoT is faithful or merely documentation.
Chinese Translation
临床医生将思维链（Chain-of-Thought, CoT）推理视为医学推理的证据，但可见的思维链是否发挥这一作用却鲜有测试。通用领域的 CoT 可信度探测忽视了临床成本，而医学大型语言模型（LLM）评估将思维链视为黑箱。我们通过医学扰动审计填补这一空白：一个包含30个操作符的电池同时编辑思维链和问题，使用临床动机的操作符（严重性反转、否定翻转、人口统计交换、证据消融），并结合思维链更新与答案翻转的联合分析，对每个模型按其失败模式进行分类。应用于14个 LLM 在四个医学问答基准上的测试，三项独立测试结果一致：思维链解耦率（Chain-Decoupling Rate, CDR；思维链未注册编辑且答案未翻转）在临床意义上的破坏性编辑中为72.9%，思维链损坏不改变准确性，去除 CoT 提示不降低准确性。两位获得董事会认证的临床医生重新标注了197个扰动问题；98.5% 的问题保持了金标准的可辩护性。该模式在医学和推理微调及规模上保持一致；在闭源层面，思维链文本不可用时，答案侧信号与相同的解耦一致。我们的框架和 CDR 提供了一个可重复使用的标准，用于审计医学 CoT 是否可信或仅仅是文档记录。
cs.AI / 110 / 2608.24794
CAFE: Self-Improving Search Agents Need Co-Evolving Feedback
CAFE：自我改进的搜索代理需要共同进化的反馈
Abstract
Outcome-supervised search agents learn when and how to retrieve evidence, but terminal rewards neither localize intermediate errors nor redirect an ongoing trajectory before those errors compound. Treating corrective feedback as a learned in-trajectory intervention couples the two roles: the agent must decide when to request and use feedback, while the critic must infer useful corrections from outcome-confounded rollouts whose failure patterns shift as the agent improves. We introduce CAFE (Coupled Agent--Feedback Evolution), a framework in which a shared-parameter model alternates between search-agent and critic roles. CAFE initializes feedback-conditioned recovery from trajectories built around the base agent's own failures, then couples online and offline optimization. During online RL, a comparative feedback estimate uses a prompt-level call--skip success gap to shape request returns, while feedback-aware advantage shaping reweights token advantages before and after feedback. Offline, rollout-derived preference optimization learns feedback from matched successful and unsuccessful trajectories. On seven agentic search benchmarks, CAFE outperforms the evaluated RL-based search agents on average, retains its gains across all six out-of-domain benchmarks, and reduces answer-level hallucinations. One-sided ablations show that improving only the agent or only the critic eventually plateaus, whereas alternating the two updates continues to improve performance. These findings suggest that a self-improving search agent needs feedback that co-evolves with the policy it guides.
Chinese Translation
结果监督的搜索代理学习何时以及如何检索证据，但终端奖励既无法定位中间错误，也无法在这些错误累积之前重新引导正在进行的轨迹。将纠正反馈视为一种学习的轨迹内干预将这两个角色结合起来：代理必须决定何时请求和使用反馈，而评论者必须从结果混淆的展开中推断出有用的纠正，这些展开的失败模式随着代理的改进而变化。我们提出了CAFE（Coupled Agent--Feedback Evolution），这是一个框架，其中共享参数模型在搜索代理和评论者角色之间交替。CAFE初始化基于反馈的轨迹恢复，围绕基础代理自身的失败构建，然后结合在线和离线优化。在在线强化学习中，比较反馈估计使用提示级别的调用-跳过成功差距来塑造请求返回，而反馈感知的优势塑造在反馈之前和之后重新加权令牌优势。在离线阶段，基于展开的偏好优化从匹配的成功和失败轨迹中学习反馈。在七个代理搜索基准上，CAFE的表现平均优于评估的基于强化学习的搜索代理，并在所有六个域外基准中保持其收益，同时减少答案级别的幻觉。单侧消融实验表明，仅改进代理或仅改进评论者最终会达到平台期，而交替更新两者则持续改善性能。这些发现表明，自我改进的搜索代理需要与其指导的策略共同进化的反馈。
cs.AI / 111 / 2608.24804
StarHarness: Evolving Harnesses with Stratified Search for Enterprise Environments
StarHarness：在企业环境中通过分层搜索演化代理工具
Abstract
We present StarHarness, a framework for evolving environment-specific agent harnesses while keeping model weights fixed. The evolved harness can include prompt and task framing, tool interfaces, skills, MCP-backed providers, subagent structure, and agent-loop configuration. StarHarness constructs a compact evolution pool by stratifying tasks according to baseline failure behavior, separates proposer-visible search tasks from proposer-hidden selection tasks, and reserves held-out tasks for evaluating generalization. Across ITBench SRE, EnterpriseOps-Gym ITSM, and AutomationBench Finance, harness evolution improves full-benchmark performance by 20-35 percentage points over the default harness after 4-12 accepted changes per environment. These gains persist on tasks excluded from evolution and transfer without re-evolution across GPT and Qwen model families. Trace analysis links the improvements to interface repairs, environment conventions, and operational knowledge that compresses search, with fewer false-positive diagnoses and shorter trajectories in several settings. StarHarness therefore offers a practical way to reduce persistent model-environment mismatch in tool-rich enterprise tasks.
Chinese Translation
我们提出了StarHarness，这是一个在保持模型权重固定的情况下演化特定环境代理工具的框架。演化后的工具可以包括提示和任务框架、工具接口、技能、基于MCP的提供者、子代理结构以及代理循环配置。StarHarness通过根据基线失败行为对任务进行分层，构建了一个紧凑的演化池，将提议者可见的搜索任务与提议者隐藏的选择任务分开，并保留未使用的任务用于评估泛化能力。在ITBench SRE、EnterpriseOps-Gym ITSM和AutomationBench Finance中，工具演化在每个环境接受4-12次更改后，提升了全基准性能20-35个百分点。这些提升在演化排除的任务上依然存在，并且在不同的GPT和Qwen模型系列之间转移时无需重新演化。追踪分析将这些改进与接口修复、环境规范和压缩搜索的操作知识联系起来，在多个设置中减少了错误阳性诊断和缩短了轨迹。因此，StarHarness提供了一种实用的方法，以减少工具丰富的企业任务中持续存在的模型与环境不匹配问题。
cs.AI / 112 / 2608.24810
Strictly Causal Streaming Video Anomaly Detection with a Theoretically-Grounded State-Space Core
具有理论基础状态空间核心的严格因果流媒体视频异常检测
Abstract
Recent work has applied Mamba style state space models (SSMs) to video anomaly detection, yet existing approaches still rely on buffering clips or windows internally, lack a theoretical account of how temporal memory relates to detection latency, and benchmark efficiency only through GPU throughput rather than the edge hardware these methods are intended to target. We introduce a strictly causal streaming anomaly detector whose fixed size state is updated in O(1) time and memory per incoming frame, with no lookahead and no clip buffering. Its temporal core is a diagonal linear state space recurrence with an input and state dependent decay gate, trained self supervised through causal next embedding prediction on a frozen visual backbone. We derive a closed form relationship between the recurrence decay spectrum and both detection delay and the shortest anomaly it can reliably capture, then validate empirically on UCSD Ped2 and CUHK Avenue. The settling delay bound predicted from the learned base decay (57 to 59 frames) sits far above the measured detection delay (1.6 and 18.4 frames), showing that the event boundary gate, not the base decay, governs responsiveness. We further report end to end latency and throughput measured directly on Apple M3 Pro hardware, 0.74 ms and 0.77 ms per frame (over 1300 FPS), rather than simulated GPU numbers. With an untuned initial configuration the method reaches 67.9 percent and 70.2 percent frame level AUC on Ped2 and Avenue, trailing prior non causal SSM baselines in accuracy. Ablations over decay rate, state size, and gating reveal that the gate contribution is dataset size dependent, hurting accuracy on the smaller Ped2 training set but helping on the larger Avenue one. Closing this accuracy gap and extending evaluation to a third, larger benchmark are immediate next steps.
Chinese Translation
近期的研究将Mamba风格的状态空间模型（SSMs）应用于视频异常检测，但现有的方法仍然依赖于内部缓冲剪辑或窗口，缺乏关于时间记忆与检测延迟之间关系的理论解释，并且仅通过GPU吞吐量来基准效率，而不是这些方法旨在针对的边缘硬件。我们提出了一种严格因果的流媒体异常检测器，其固定大小的状态在每个输入帧中以O(1)的时间和内存进行更新，无需前瞻和剪辑缓冲。其时间核心是一个对角线线性状态空间递归，具有输入和状态依赖的衰减门，通过在冻结的视觉骨干网络上进行因果下一个嵌入预测的自监督训练。我们推导了递归衰减谱与检测延迟及其能够可靠捕捉的最短异常之间的封闭形式关系，并在UCSD Ped2和CUHK Avenue上进行了实证验证。从学习的基础衰减预测的稳定延迟界限（57到59帧）远高于测量的检测延迟（1.6和18.4帧），显示出事件边界门而非基础衰减主导了响应性。我们进一步报告了在Apple M3 Pro硬件上直接测量的端到端延迟和吞吐量，分别为每帧0.74毫秒和0.77毫秒（超过1300 FPS），而不是模拟的GPU数字。在未调优的初始配置下，该方法在Ped2和Avenue上分别达到了67.9%和70.2%的帧级AUC，准确性落后于先前的非因果SSM基准。对衰减率、状态大小和门控的消融实验表明，门控贡献依赖于数据集大小，在较小的Ped2训练集上会降低准确性，但在较大的Avenue数据集上则有所帮助。缩小这一准确性差距并将评估扩展到第三个更大的基准是下一步的紧迫任务。
cs.AI / 113 / 2608.24824
Constrained Entity Selection under Partial Knowledge for LLM-Based Knowledge Graph QA
基于部分知识的约束实体选择用于大型语言模型的知识图谱问答
Abstract
Large language models are increasingly used for knowledge graph question answering (KGQA), but can fail to correctly ground answers in the underlying graph. Current approaches to LLM-based KGQA either rely on full semantic parsing into executable queries such as SPARQL, which is brittle in practice due to complex schemas or incompleteness of real-world KGs, or on LLM-reasoning and answer generation over KGs, which can be more robust but lacks formal guarantees. In this work, we study a complementary setting in which \emph{candidate} answers are generated by an LLM-based system and subsequently verified using lightweight symbolic constraints derived from the question. We introduce \emph{Constrained Entity Selection under Partial Knowledge (CES-PK)}, a problem formulation that focuses on eliminating invalid answers and providing symbolic support for valid ones without requiring construction of executable logical forms. To account for incomplete KGs, we employ a three-valued constraint semantics (\emph{satisfied, violated, unknown}) that avoids incorrect rejections under open-world assumptions. To demonstrate the effects of our method, we instantiate this framework over the Hetionet biomedical knowledge graph and evaluate the impact of type, relation, and exclusion constraints. Experiments show that precision improves by filtering invalid candidates, while recall is preserved due to retaining candidates whose constraints are not explicitly violated. Satisfied constraints provide additional positive symbolic evidence to rank remaining candidates.
Chinese Translation
大型语言模型在知识图谱问答（KGQA）中越来越多地被使用，但在将答案正确地与底层图谱关联时可能会失败。目前基于大型语言模型的KGQA方法要么依赖于将语义完全解析为可执行查询（如SPARQL），但由于复杂的模式或现实世界知识图谱的不完整性，这种方法在实践中显得脆弱；要么依赖于在知识图谱上进行的LLM推理和答案生成，尽管这种方法更为稳健，但缺乏正式的保证。在本研究中，我们研究了一种互补的设置，其中候选答案由基于大型语言模型的系统生成，并随后使用从问题中派生的轻量级符号约束进行验证。我们提出了“基于部分知识的约束实体选择（CES-PK）”这一问题表述，重点在于消除无效答案，并为有效答案提供符号支持，而无需构建可执行的逻辑形式。为了考虑不完整的知识图谱，我们采用了三值约束语义（满足、违反、未知），以避免在开放世界假设下的不正确拒绝。为了展示我们方法的效果，我们在Hetionet生物医学知识图谱上实例化该框架，并评估类型、关系和排除约束的影响。实验表明，通过过滤无效候选，精确度得到了提高，而由于保留未明确违反约束的候选，召回率得以保持。满足的约束为排名剩余候选提供了额外的正面符号证据。
cs.AI / 114 / 2608.24825
A Dual-Dimensional LLM Framework for Automated Item Incidental Content Similarity Analysis in Large-Scale Assessments
用于大规模评估中自动化项目偶然内容相似性分析的双维度LLM框架
Abstract
The rapid expansion of large-scale assessments and the growing adoption of automatic item generation have intensified concerns about incidental content redundancy, where construct-irrelevant elements such as wording or contextual framing become unintentionally repetitive across items. Traditional similarity metrics like BLEU or cosine similarity, often fail to capture the nuanced structural and semantic layers that drive perceived redundancy simultaneously. This study proposes a dual-dimensional framework for Automated Item Similarity Analysis (AISA) powered by Large Language Models (LLMs), operationalizing similarity through Structured Decomposition and Semantic Relatedness. Psychometric validation indicates that LLM-derived metrics align more closely with indicators of construct-irrelevant local dependence and yield more coherent item parameter groupings than traditional text-based measures. The framework is further evaluated through its application in Computerized Adaptive Testing (CAT). Simulations reveal that incorporating LLM-based similarity constraints into item selection improves estimation stability and reduces bias with minimal efficiency trade-offs, outperforming constraints based on conventional metrics. These findings highlight the potential of LLM-powered AISA to support scalable bank curation, content-aware test assembly, and experience-sensitive adaptive testing across diverse assessment contexts.
Chinese Translation
大规模评估的快速扩展以及自动化项目生成的日益普及加剧了对偶然内容冗余的关注，其中构念无关的元素（如措辞或上下文框架）在项目中无意间变得重复。传统的相似性度量，如BLEU或余弦相似性，往往无法同时捕捉驱动感知冗余的细微结构和语义层次。本研究提出了一种基于大型语言模型（LLMs）的自动化项目相似性分析（AISA）双维度框架，通过结构分解和语义相关性来实现相似性。心理测量学验证表明，LLM衍生的度量与构念无关的局部依赖指标更为一致，并且在项目参数分组上比传统文本基础度量产生更具连贯性的结果。该框架还通过在计算机自适应测试（CAT）中的应用进行了进一步评估。模拟结果显示，将基于LLM的相似性约束纳入项目选择中可以提高估计的稳定性并减少偏差，同时仅带来最小的效率权衡，优于基于传统度量的约束。这些发现突显了LLM驱动的AISA在支持可扩展的题库管理、内容感知的测试组装以及在多样化评估环境中体验敏感的自适应测试方面的潜力。
cs.AI / 115 / 2608.24846
FedV-KGQA: Multi-Hop Question Answering over Vertically Partitioned Knowledge Graphs
FedV-KGQA：在垂直分区知识图谱上进行多跳问答
Abstract
Real-world data for knowledge graph question answering is often distributed across different organizations due to governance and data sovereignty constraints. While centralized systems exist, they cannot answer multi-hop questions when the required facts are split across vertically partitioned silos. In this paper, we propose FedV-KGQA, a framework for multi-hop reasoning over knowledge graphs in which organizations share entities but own disjoint sets of relations. Our approach combines local graph enrichment and knowledge graph embeddings to ensure raw triples and relation parameters never leave each silo, establishing a structural data boundary without requiring centralized graph access. We further introduce a topic entity anchoring mechanism that grounds questions in the correct graph neighborhood without any runtime inter-silo communication. We evaluate 12 model configurations across three benchmarks and show that FedV-KGQA performs strongly, remains close to centralized performance, generalizes to 3-hop reasoning, and is robust to embedding perturbations.
Chinese Translation
由于治理和数据主权的限制，现实世界中的知识图谱问答数据通常分布在不同的组织之间。虽然存在集中式系统，但当所需事实分散在垂直分区的孤岛中时，它们无法回答多跳问题。本文提出了FedV-KGQA，一个用于知识图谱的多跳推理框架，其中组织共享实体但拥有不相交的关系集。我们的方法结合了本地图增强和知识图谱嵌入，以确保原始三元组和关系参数始终留在各自的孤岛内，建立了一个结构性的数据边界，而无需集中式图访问。我们进一步引入了一种主题实体锚定机制，使问题能够在正确的图邻域中定位，而无需在运行时进行孤岛间的通信。我们在三个基准上评估了12种模型配置，结果表明，FedV-KGQA表现强劲，接近集中式性能，能够推广到3跳推理，并对嵌入扰动具有鲁棒性。
cs.AI / 116 / 2608.24870
SPO++: Stream-Aligned Policy Optimization for Asynchronous Agentic RL
SPO++：用于异步智能体强化学习的流对齐策略优化
Abstract
Group-relative reinforcement learning waits for sibling rollouts of the same prompt, which is costly for long and variable tool-use trajectories. Single-stream Policy Optimization (SPO) removes this dependency with a persistent prompt-level value estimate, but its recipe whitens one advantage per trajectory before optimizing a token-mean actor loss. We show that trajectory centering generally does not center the token-weighted quantity consumed by the actor, and fix the mismatch by standardizing terminal-outcome advantages under the action-token measure. We additionally organize prompt evidence by the policy event that generated it rather than learner receipt order. Across matched runs on ALFWorld at two model scales and on Math-TIR, SPO++ improves online learning efficiency over SPO. A paired ablation identifies action-token-measure normalization as the strongest tested component.
Chinese Translation
群体相对强化学习等待同一提示的兄弟回放，这对于长且变化的工具使用轨迹是有成本的。单流策略优化（Single-stream Policy Optimization, SPO）通过持久的提示级价值估计消除了这种依赖，但其方法在优化令牌均值演员损失之前会使每条轨迹的一个优势变得无效。我们表明，轨迹中心化通常并不能使演员消耗的令牌加权量中心化，并通过在动作-令牌度量下标准化终端结果优势来修正这一不匹配。此外，我们根据生成提示证据的策略事件而非学习者接收顺序来组织提示证据。在两个模型规模的ALFWorld和Math-TIR上的匹配运行中，SPO++提高了在线学习效率。配对消融实验确定动作-令牌度量标准化是测试过的最强组件。
cs.AI / 117 / 2608.24876
Recursive Experiential-Working Memory Evolution for Long-Horizon Agent Harnesses
长时间跨度代理利用的递归经验工作记忆演化
Abstract
Recursive self-improvement (RSI) remains hard in long-horizon tasks, where growing histories obscure the task state and misalign skill invocation. We introduce Recuris, a recursive Experiential-Working Memory architecture for long-horizon agent harnesses, in which Working Memory tracks task progress and guides skill selection from Experiential Memory, grounding skill use in current needs rather than the full history. This coupling also turns execution into structured evidence that localizes failures to specific memory components. Across tasks, a fixed Meta-Agent turns that evidence into localized, validation-gated updates to Skill Memory that reshape execution and yield new evidence, forming a bounded recursive memory-evolution loop. Across four long-horizon benchmarks and ten models, Recuris improves task success in 35 of the 37 completed model-benchmark pairs, carrying frontier models to SOTA-level task success: on tau-bench it adds +17.8 points to GPT-5.6 Sol and +15.6 to Claude Opus 5, taking Opus 5 to 87.9%, and +16.6/+13.5 points on Qwen3.6-27B/35B on SkillFlow. The advantage widens as the interaction horizon grows, to +32.2 points on the longest tasks, and common long-horizon failures fall by up to 80%. These results position recursively evolving memory as a scalable foundation for RSI, enabling agents to continuously transform accumulated experience into increasingly effective long-horizon behavior. Code: https://github.com/Gen-Verse/Recuris
Chinese Translation
递归自我改进（RSI）在长时间跨度任务中仍然困难，因为不断增长的历史会模糊任务状态并导致技能调用的不一致。我们提出了Recuris，一种用于长时间跨度代理利用的递归经验工作记忆架构，其中工作记忆跟踪任务进展并指导从经验记忆中选择技能，将技能的使用基于当前需求而非完整历史。这种耦合还将执行转化为结构化证据，使失败定位于特定的记忆组件。在多个任务中，一个固定的元代理将这些证据转化为局部的、验证门控的技能记忆更新，重塑执行并产生新的证据，形成一个有限的递归记忆演化循环。在四个长时间跨度基准测试和十个模型中，Recuris在37对已完成的模型-基准对中提高了35个任务的成功率，将前沿模型推向SOTA级别的任务成功：在tau-bench上，它为GPT-5.6 Sol增加了17.8分，为Claude Opus 5增加了15.6分，使Opus 5达到了87.9%，在SkillFlow上对Qwen3.6-27B/35B分别增加了16.6/13.5分。随着交互时间跨度的增加，优势进一步扩大，在最长的任务上增加了32.2分，常见的长时间跨度失败减少了多达80%。这些结果将递归演化的记忆定位为RSI的可扩展基础，使代理能够不断将积累的经验转化为日益有效的长时间跨度行为。代码链接： https://github.com/Gen-Verse/Recuris
cs.CL / 1 / 2608.23570
Taming Visual Neglect: A Variational Information Bottleneck Framework for Adaptive Attention in Multimodal In-Context Learning
驯服视觉忽视：一种变分信息瓶颈框架用于多模态上下文学习中的自适应注意力
Abstract
Large vision-language models exhibit strong in-context learning (ICL) capabilities, yet when and why visual context helps multimodal ICL remains poorly understood. Empirical studies show a puzzling dichotomy: models sometimes effectively leverage visual demonstrations, yet often neglect them entirely. We propose VIB-ICL, an information-theoretic framework that resolves this dichotomy through the Information Bottleneck principle. We introduce the Cross-Modal Information Gain (CMIG), which quantifies the additional mutual information that visual context provides about the target beyond textual context. We derive a generalization bound showing that multimodal ICL's excess risk over text-only ICL is governed by the CMIG, proving that multimodal ICL provably outperforms text-only ICL when visual information is non-redundant. We further prove that visual context neglect, often viewed as a failure mode, is the Information Bottleneck-optimal solution when visual information is redundant, yielding a closed-form Attention Reallocation Principle that prescribes how visual attention weights should be adaptively adjusted. We instantiate this principle in the VIB-ICL algorithm, which estimates CMIG via variational bounds and dynamically reallocates attention. Experiments on five benchmarks demonstrate consistent improvements of up to 4.7\% accuracy gains and 35\% reduction in required demonstrations, validating our theoretical predictions.
Chinese Translation
大型视觉-语言模型展现出强大的上下文学习（ICL）能力，但视觉上下文在何时以及为何有助于多模态ICL仍然不甚清楚。实证研究显示出一种令人困惑的二分法：模型有时有效利用视觉示范，但往往完全忽视它们。我们提出了VIB-ICL，一个通过信息瓶颈原理解决这一二分法的信息论框架。我们引入了跨模态信息增益（CMIG），它量化了视觉上下文相对于文本上下文所提供的关于目标的额外互信息。我们推导出一个泛化界限，表明多模态ICL相对于仅文本ICL的超额风险由CMIG主导，证明了当视觉信息不冗余时，多模态ICL显著优于仅文本ICL。我们进一步证明，视觉上下文的忽视，通常被视为一种失败模式，在视觉信息冗余时是信息瓶颈最优解，从而得出了一个封闭形式的注意力重新分配原则，规定了视觉注意力权重应如何自适应调整。我们在VIB-ICL算法中实例化这一原则，该算法通过变分界限估计CMIG并动态重新分配注意力。在五个基准测试上的实验表明，准确率提高了最多4.7%，所需示范减少了35%，验证了我们的理论预测。
cs.CL / 2 / 2608.23627
From Triage to Discharge: A Survey of NLP Tasks, Methods, and Open Challenges in the Emergency Department
从分诊到出院：急诊科自然语言处理任务、方法及开放挑战的调查
Abstract
Emergency departments (EDs) operate under time pressure, generating multimodal data such as clinical conversations, triage notes, and discharge documents. Recent advances in natural language processing (NLP), particularly pretrained transformers and large language models, have created new opportunities to support language and time-intensive stages of emergency care. Yet existing surveys map clinical NLP across the broader hospital workflow or focus on specific tasks. This survey analyses 46 papers spanning the three phases of ED: triage, diagnosis, and disposition, covering tasks such as triage classification, clinical summarisation, automatic diagnosis, report generation, and discharge documentation. We examine modelling paradigms, evaluation practices, and emerging benchmarks and shared tasks. Across tasks, we identify common trends, including a shift from task-specific neural architectures to pretrained language models, growing interest in interactive clinical systems, and increasing attention to clinically grounded evaluation. Finally, we detail open challenges such as limited generalisability, noisy clinical inputs, and workflow constraints that inform future ED-NLP research.
Chinese Translation
急诊科（ED）在时间压力下运作，生成多模态数据，如临床对话、分诊记录和出院文件。最近在自然语言处理（NLP）领域的进展，特别是预训练变换器和大型语言模型，为支持急救护理中语言和时间密集的阶段创造了新的机会。然而，现有的调查要么映射临床NLP在更广泛的医院工作流程中，要么专注于特定任务。本调查分析了46篇涵盖急诊科三个阶段（分诊、诊断和处置）的论文，涉及分诊分类、临床摘要、自动诊断、报告生成和出院文档等任务。我们考察了建模范式、评估实践以及新兴基准和共享任务。在各项任务中，我们识别出一些共同趋势，包括从任务特定神经架构向预训练语言模型的转变、对互动临床系统的日益关注，以及对临床基础评估的关注不断增加。最后，我们详细阐述了开放挑战，如有限的普适性、嘈杂的临床输入和工作流程限制，这些都为未来的急诊科NLP研究提供了重要信息。
cs.CL / 3 / 2608.23645
Contextual Embedding Evidence for Main--Light Verb Distinctions in Urdu
乌尔都语中主要动词与轻动词区分的上下文嵌入证据
Abstract
Urdu light verbs contribute schematic event-structural meaning while remaining lexically related to corresponding main verbs. This study tests representational predictions derived from Butt's analysis using contextual embeddings from UrduBERT, DunbaaBERT, and multilingual BERT across 1,126 naturally occurring sentences containing seven Urdu verbs. Main and light uses show significant representational separation in all 21 verb--model comparisons. At the same time, same-lemma main and light centroids are consistently closer than mismatched main--light lemma pairs, supporting continued lexical relatedness. In a seven-way prediction task restricted to light uses, verb identity remains recoverable after the target is masked, with UrduBERT achieving 0.866 accuracy and 0.852 macro-F1. UrduBERT also retains 0.782 accuracy under a preceding-form-disjoint evaluation, indicating generalization beyond repeated local verb combinations. These findings provide computational evidence consistent with Butt's account that Urdu light verbs differ systematically from their main uses while retaining lemma-specific and verb-specific representational structure.
Chinese Translation
乌尔都语轻动词在保持与相应主要动词的词汇相关性的同时，贡献了示意事件结构的意义。本研究使用来自UrduBERT、DunbaaBERT和多语言BERT的上下文嵌入，测试了基于Butt分析的表征预测，分析了包含七个乌尔都动词的1,126个自然发生的句子。主要用法和轻用法在所有21个动词-模型比较中显示出显著的表征分离。同时，相同词根的主要和轻动词质心始终比不匹配的主要-轻动词词根对更接近，支持了持续的词汇相关性。在一个限制于轻用法的七类预测任务中，动词身份在目标被掩蔽后仍然可恢复，UrduBERT的准确率达到0.866，宏观F1值为0.852。UrduBERT在前形式不重叠的评估下也保持了0.782的准确率，表明其在重复局部动词组合之外的泛化能力。这些发现提供了与Butt的解释一致的计算证据，表明乌尔都语轻动词在系统性上与其主要用法不同，同时保持了词根特定和动词特定的表征结构。
cs.CL / 4 / 2608.23705
The Limits of Automatic Evaluation of Creativity in Large Language Models
大型语言模型中创造力自动评估的局限性
Abstract
Large Language Models (LLMs) are increasingly capable of generating text that challenges human performance in domains requiring creativity, yet evaluating creativity in LLM-generated content remains a significant challenge. Here, we investigate whether current automatic evaluation methods can reliably capture human judgments of creativity. We collect human evaluations of human- and AI-generated short stories from the WritingPrompts dataset across 11 dimensions of creativity, and compare these judgments with automated objective metrics and LLM-as-a-Judge evaluations. Our experiments reveal substantial misalignment between automatic evaluations and human assessments. In particular, LLM-based judges exhibit a systematic preference for AI-generated stories, consistently favoring their stylistic characteristics over the unpredictability and other qualities of human-authored texts. Furthermore, correlation analyses show that widely used automatic metrics exhibit near-zero alignment with human judgments across both human- and AI-generated stories, suggesting that they fail to capture important dimensions of creativity. These findings highlight fundamental limitations in current approaches to the automatic evaluation of creative text and underscore the difficulty of reducing the multidimensional and subjective nature of creativity to computational metrics.
Chinese Translation
大型语言模型（LLMs）在生成文本方面的能力日益增强，能够在需要创造力的领域挑战人类表现，但评估LLM生成内容中的创造力仍然是一个重大挑战。在此，我们研究当前的自动评估方法是否能够可靠地捕捉人类对创造力的判断。我们从WritingPrompts数据集中收集了人类和AI生成的短篇故事的人工评估，涵盖了11个创造力维度，并将这些判断与自动客观指标和LLM作为评审的评估进行比较。我们的实验揭示了自动评估与人类评估之间存在显著的不一致性。特别是，基于LLM的评审表现出对AI生成故事的系统性偏好，始终偏爱其风格特征，而非人类创作文本的不可预测性和其他特质。此外，相关性分析表明，广泛使用的自动指标在对人类和AI生成故事的评估中与人类判断几乎没有一致性，表明它们未能捕捉创造力的重要维度。这些发现突显了当前自动评估创造性文本方法的根本局限性，并强调了将创造力的多维性和主观性简化为计算指标的困难。
cs.CL / 5 / 2608.23719
ADE: Agentic Data Evolution Framework for Human-Centered Objectives
ADE：面向以人为本目标的代理数据演化框架
Abstract
Aligning large language models to human-centered objectives is difficult when targets are non-executable and context-dependent, limiting reliable verification and scalable supervision. Although synthetic data expands coverage, weak verification shifts the bottleneck from generation to selection. Noisy signals destabilize iterative refinement and can cause silent regressions. We propose Agentic Data Evolution (ADE), a data-centric framework that organizes synthetic supervision as evolving data snapshots. ADE improves data snapshots through a closed-loop Observation-Variation-Selection (OVS) procedure, where a steady-state admission mechanism acts as a quality ratchet that conservatively gates updates for sustained cross-round improvement. We validate these improvements through complementary intrinsic trend tracking and extrinsic post-training evaluation. On DEV300, ADE raises the intrinsic win rate from 50% to 75.81% and the extrinsic win rate from 55.20% to 68.86%, consistent performance gains across diverse benchmarks. Blind expert evaluation further confirms this, with a 66.11% preference for evolved answers. These gains extend across post-training methods, model scales, and tasks beyond the target weakly verifiable educational objectives. Resources are available at https://github.com/ZeroLoss-Lab/Agentic-Data-Evolution.
Chinese Translation
将大型语言模型与以人为本的目标对齐是困难的，尤其是在目标不可执行且依赖于上下文的情况下，这限制了可靠验证和可扩展监督的实现。尽管合成数据扩展了覆盖范围，但薄弱的验证将瓶颈从生成转移到了选择。噪声信号会破坏迭代精炼，可能导致静默回归。我们提出了代理数据演化（Agentic Data Evolution，ADE），这是一个以数据为中心的框架，将合成监督组织为不断演变的数据快照。ADE通过闭环观察-变异-选择（Observation-Variation-Selection，OVS）程序改进数据快照，其中稳态接纳机制作为质量棘轮，保守地控制更新，以实现持续的跨轮改进。我们通过互补的内在趋势跟踪和外部后训练评估验证了这些改进。在DEV300上，ADE将内在胜率从50%提高到75.81%，外在胜率从55.20%提高到68.86%，在多样化基准测试中表现出一致的性能提升。盲评专家评估进一步确认了这一点，演化答案的偏好率为66.11%。这些提升跨越了后训练方法、模型规模以及超出目标弱可验证教育目标的任务。相关资源可在https://github.com/ZeroLoss-Lab/Agentic-Data-Evolution获取。
cs.CL / 6 / 2608.23766
What Reaches Expert Review? Representation, Structural Screening, and Candidate-Form Dependence in AI-Assisted Item Development
什么会进入专家评审？AI辅助项目开发中的表征、结构筛选和候选形式依赖性
Abstract
Between AI-assisted item generation and expert review sits a computational evaluator whose decisions are usually treated as technical preliminaries. Yet representation, structural reduction, and selection policy determine which items and evidence psychometricians ever receive. Across two linked in-silico studies of 32,000 selected Big Five items, we followed fixed source populations from semantic representation through structural evaluation and candidate-form construction. Broad agreement in semantic geometry concealed consequential local differences: identical wording acquired different construct evidence, different items survived, and intended attributes could disappear even as community correspondence improved. These sensitivities also differed across generated source populations. At the final review boundary, both eligibility policies filled every content cell in every evaluable form, yet they presented different wording. Across embedding configurations, inclusive primary forms shared a median of only 6 of 40 items, reflecting the total downstream consequence of changing representation across structural evidence and ranking. The apparent stability of global summaries and complete forms therefore concealed instability in the content reaching psychometricians. The computational evaluator is not neutral infrastructure between generation and expertise; it is an inspectable and revisable part of measurement design.
Chinese Translation
在AI辅助项目生成与专家评审之间，存在一个计算评估者，其决策通常被视为技术前置条件。然而，表征、结构简化和选择政策决定了心理测量学家所接收到的项目和证据。在两项关联的计算机模拟研究中，我们对32,000个选定的五大人格项目进行了研究，跟踪了从语义表征到结构评估再到候选形式构建的固定源人群。语义几何的广泛一致性掩盖了重要的局部差异：相同的措辞获得了不同的构念证据，不同的项目得以存活，而即使社区对应性改善，预期的属性也可能消失。这些敏感性在生成的源人群之间也存在差异。在最终评审边界，两个资格政策填满了每个可评估形式中的每个内容单元，但它们呈现了不同的措辞。在嵌入配置中，包容性的主要形式仅共享40个项目中的6个中位数，反映了在结构证据和排名中改变表征的总下游后果。因此，全球总结和完整形式的表面稳定性掩盖了到达心理测量学家的内容的不稳定性。计算评估者并不是生成与专业知识之间的中立基础设施；它是测量设计中一个可检查和可修订的部分。
cs.CL / 7 / 2608.23780
When Youth Enter The Chat: An Epistemic Shift in the Validation of LLM-Based Measures of Student Talk
青少年进入对话：基于大型语言模型的学生交流测量验证中的认知转变
Abstract
LLMs are being used increasingly to measure aspects of student discourse (e.g. talk moves, collaboration, equity of voice) at scale. Typically, LLM-based measures of student talk use transcriptions of classroom conversations that only include verbal contributions, which de-contextualize student language. Common practices for validating these measures include comparing outputs against expert annotations by adults, using held out evaluation sets and F1 scores. We argue that these approaches are insufficient to ensure that such measures are meaningful and equitable for teaching and learning, particularly for racially and linguistically marginalized youth. In order to center the youth whose talk is being analyzed, re-contextualizing these classroom conversations and engaging youth in the research process is necessary. Sharing epistemic authority with youth, ultimately, centers their point of view and adds crucial nuance to the analysis of their talk that adult experts, researchers, and LLMs cannot provide. In a case study of multilingual youth in one 8th-grade math classroom, we address the epistemic exclusion of youth by employing multiple ethnographically-oriented methods to re-contextualize student conversations and center youth as epistemic authorities in conversation with researchers and LLMs. We conducted participant observations, interviews, focus groups, and member checks with four focal students. Findings reveal that there were misalignments between students' interpretations of their own math talk experiences and the LLM-based measures of their talk. Students contested both the LLM classifications and the coding scheme used to measure their talk, highlighting the need for youth to be involved in the epistemic process of producing knowledge about their experiences.
Chinese Translation
大型语言模型（LLMs）在大规模测量学生话语（例如，交流策略、合作、声音平等）方面的应用日益增加。通常，基于LLM的学生话语测量使用的课堂对话转录仅包括口头贡献，这使得学生语言脱离了上下文。验证这些测量的常见做法包括将输出与成人专家的注释进行比较，使用保留的评估集和F1分数。我们认为，这些方法不足以确保这些测量对教学和学习具有意义和公平性，尤其是对于种族和语言边缘化的青少年。为了将分析对象的青少年置于中心，重新构建这些课堂对话的上下文并让青少年参与研究过程是必要的。与青少年共享认知权威，最终将他们的观点置于中心，并为他们的对话分析增添重要的细微差别，这是成人专家、研究人员和LLMs无法提供的。在对一所八年级数学课堂中的多语言青少年进行的案例研究中，我们通过采用多种以民族志为导向的方法，重新构建学生对话的上下文，并将青少年作为与研究人员和LLMs对话的认知权威置于中心。我们进行了参与观察、访谈、焦点小组讨论和成员检查，涉及四名重点学生。研究结果显示，学生对自己数学交流经历的解读与基于LLM的他们话语测量之间存在不一致。学生质疑LLM的分类和用于测量他们话语的编码方案，强调了青少年参与关于他们经历的知识生产的认知过程的必要性。
cs.CL / 8 / 2608.23783
Inter-dimension Dependence for Multi-Dimensional Evaluation of Open-Ended Text
开放式文本多维评估中的维度间依赖性
Abstract
LLM-as-a-judge methods are widely used for evaluating the quality of generated open-ended text. Such evaluations are generally multi-dimensional, since the error patterns in texts can be different for different dimensions. Therefore, reliable LLM judges should evaluate each target dimension independently. To quantify the extent to which LLM judges depend on non-target dimensions when evaluating a target dimension, i.e., inter-dimension dependence, we propose CorrGap. To measure this, CorrGap uses the difference in correlations between LLM-predicted scores and ground truth scores across different groups of texts. Using CorrGap, we show that inter-dimension dependence is pervasive across LLM judges in open-ended text evaluation tasks. To mitigate inter-dimension dependence, we propose DimCheck, a method that iteratively removes unrelated evidence from COTs generated by LLM judges in a step-wise way. We show that DimCheck mitigates inter-dimension dependence and outperforms strong baselines across three LLMs and four tasks. We also show that smaller trained LLMs can approximate larger LLMs in DimCheck, with much lower inference costs.
Chinese Translation
基于大型语言模型（LLM）的评判方法广泛应用于评估生成的开放式文本质量。这类评估通常是多维的，因为文本中的错误模式在不同维度上可能有所不同。因此，可靠的LLM评判者应独立评估每个目标维度。为了量化LLM评判者在评估目标维度时对非目标维度的依赖程度，即维度间依赖性，我们提出了CorrGap。CorrGap通过比较不同文本组中LLM预测分数与真实分数之间的相关性差异来进行测量。使用CorrGap，我们展示了在开放式文本评估任务中，LLM评判者的维度间依赖性普遍存在。为了减轻维度间依赖性，我们提出了DimCheck，这是一种以逐步方式迭代去除LLM评判者生成的链式思维（COT）中无关证据的方法。我们展示了DimCheck能够减轻维度间依赖性，并在三种LLM和四个任务中超越强基线。我们还表明，较小的训练LLM在DimCheck中可以近似于较大的LLM，且推理成本显著降低。
cs.CL / 9 / 2608.23806
Giga-Embeddings: Mixture-of-Experts Encoders for High-Throughput Text Embeddings
千兆嵌入：用于高吞吐量文本嵌入的专家混合编码器
Abstract
We introduce Giga-Embeddings, a family of text embedding models designed to combine strong retrieval quality with efficient serving. Its largest member is a sparse 10B-parameter Mixture-of-Experts encoder with approximately 1.8B active parameters per token. Across English, Russian, multilingual, and code MTEB benchmarks, this model achieves the strongest aggregate performance within the family on all four evaluated suites. In our vLLM benchmark with 1024-token inputs, it processes 114.5k tokens per second, providing 25 percent higher throughput than the dense 3B model and 1.56-2.65x the throughput of the evaluated external systems. The family also includes a dense 3B encoder and a distilled 480M encoder for tighter compute and memory budgets. We train the compact model using a dimension-agnostic objective that aligns teacher and student similarity distributions. The resulting 480M model scores 70.98 on Russian MTEB, surpassing FRIDA while using 42 percent fewer parameters. We release all three model checkpoints.
Chinese Translation
我们介绍了千兆嵌入（Giga-Embeddings），这是一系列文本嵌入模型，旨在结合强大的检索质量与高效的服务能力。其最大成员是一个稀疏的10B参数专家混合编码器，每个令牌大约有1.8B的活跃参数。在英语、俄语、多语言和代码的MTEB基准测试中，该模型在所有四个评估套件中实现了该系列的最佳综合性能。在我们的vLLM基准测试中，使用1024个令牌的输入，它每秒处理114.5k个令牌，提供比稠密的3B模型高出25%的吞吐量，并且是评估外部系统吞吐量的1.56-2.65倍。该系列还包括一个稠密的3B编码器和一个为更紧凑的计算和内存预算设计的480M编码器。我们使用一个与维度无关的目标来训练紧凑模型，以对齐教师和学生的相似性分布。最终得到的480M模型在俄语MTEB上得分70.98，超过了FRIDA，同时使用的参数少42%。我们发布了这三个模型的检查点。
cs.CL / 10 / 2608.23812
From Preferences to Principles: Rubric-Based Alignment for Grounded Knowledge Answers
从偏好到原则：基于评分标准的对齐以实现扎根知识的回答
Abstract
Designing effective reward signals for open-domain question answering is challenging because high-quality responses must simultaneously satisfy multiple aspects of answer quality that are difficult to capture with a holistic scalar objective. We introduce a rubric-based reward framework that generates query-specific rubrics grounded in retrieved evidence and decomposed into multiple quality dimensions, providing fine-grained supervision during post-training. Averaged across three evaluation axes (composition, grounding, and instruction-following), our approach improves over the instruction-tuned baseline by 6.5% and over flat rubric variants by 4%, with consistent gains across all evaluation datasets. Conditioning rubrics on retrieved evidence improves factual support, while decomposing rubrics into quality-specific dimensions further improves coherence, organization, and adherence to query requirements. Our results show that grounded, multi-dimensional rubrics provide more effective reward supervision for complex open-domain question answering.
Chinese Translation
为开放领域问答设计有效的奖励信号具有挑战性，因为高质量的回答必须同时满足多个难以通过整体标量目标捕捉的回答质量方面。我们提出了一种基于评分标准的奖励框架，该框架生成基于检索证据的查询特定评分标准，并将其分解为多个质量维度，从而在后训练阶段提供细粒度的监督。在三个评估维度（组成、扎根和遵循指令）上，我们的方法在指令调优基线基础上提高了6.5%，在平面评分标准变体上提高了4%，并且在所有评估数据集上均表现出一致的提升。基于检索证据的评分标准改善了事实支持，而将评分标准分解为特定质量维度进一步提高了连贯性、组织性和对查询要求的遵循。我们的结果表明，扎根的多维评分标准为复杂的开放领域问答提供了更有效的奖励监督。
cs.CL / 11 / 2608.23818
Beyond Static and Linear: What Attention Constraints Best Fit Human Reading Times?
超越静态与线性：哪些注意力约束最适合人类阅读时间？
Abstract
Transformer-based language models are widely used as models of human language processing, yet their attention mechanisms allow lossless access to the full preceding context, unlike the limited memory systems of humans. We hypothesize that installing memory constraints into transformers' attention mechanisms can improve their fit to human behavioral data. While previous work has explored individual constraints in isolation, we conduct a systematic comparison of multiple attention-based memory mechanisms across different model sizes and training corpora, evaluating both psychometric predictive power for human reading times and grammatical competence. We additionally compare static constraints, in which the constraint strength is fixed throughout training, to dynamic memory curricula. We find that constraints that are sensitive to the content of intervening tokens consistently achieve the highest alignment with human reading times, outperforming distance-based constraints. We observe a dissociation between psychometric fit and grammatical competence under dynamic memory curricula, suggesting that Transformers cannot serve as a one-size-fits-all cognitive model.
Chinese Translation
基于Transformer的语言模型被广泛用作人类语言处理的模型，然而它们的注意力机制允许无损地访问完整的前文上下文，这与人类的有限记忆系统不同。我们假设在Transformer的注意力机制中引入记忆约束可以改善其与人类行为数据的拟合。尽管之前的研究已单独探讨了个别约束，我们对不同模型规模和训练语料库中的多种基于注意力的记忆机制进行了系统比较，评估其对人类阅读时间的心理测量预测能力和语法能力。我们还比较了静态约束（在整个训练过程中约束强度固定）与动态记忆课程。我们的研究发现，对介入标记内容敏感的约束在与人类阅读时间的对齐上始终表现最佳，优于基于距离的约束。我们观察到在动态记忆课程下心理测量拟合与语法能力之间存在解离，这表明Transformer不能作为一种适用于所有情况的认知模型。
cs.CL / 12 / 2608.23830
Mitigating Exploration Bias in RL for Multi-Instruction Following
减轻多指令跟随中的强化学习探索偏差
Abstract
RL has emerged as a powerful paradigm for enhancing the instruction following capabilities of LLMs. While existing training recipes achieve substantial gains, we find that they suffer from exploration bias towards easy instructions when the training data has multiple instructions in a prompt. This bias is caused by two main reasons: 1) the policy model's initial ability to satisfy hard instructions is too low to trigger successful exploration during RL training, so the optimization is biased towards easy instructions; and 2) canonical RL training recipes typically employ a cumulative reward (the number of instructions fulfilled), treating all instructions equally, which biases the policy model towards fulfilling easy instructions to obtain the same amount of reward. To address these issues, we first propose two metrics to measure the exploration bias in instruction following and then introduce a two-stage framework to alleviate it: 1) Behavioral Bootstrapping, a lightweight rejection sampling fine-tuning stage before RL to activate hard instructions; and 2) Scarcity-Aware Rewards, a new RL reward function that assigns rewards to instructions based on their empirical scarcity. Experiments show that the proposed metrics are highly correlated with model performance, and our methods unleash the potential of RL training: our best models outperform the baselines by a significant margin across three verifiable instruction following benchmarks. We release codes at https://github.com/mianzhang/MulIF.
Chinese Translation
强化学习（RL）已成为增强大型语言模型（LLMs）指令跟随能力的强大范式。尽管现有的训练方案取得了显著的进展，但我们发现当训练数据中包含多个指令时，它们存在对简单指令的探索偏差。这种偏差主要由两个原因造成：1）策略模型最初满足困难指令的能力过低，无法在RL训练过程中触发成功的探索，因此优化偏向于简单指令；2）传统的RL训练方案通常采用累积奖励（完成的指令数量），将所有指令视为平等，这导致策略模型偏向于完成简单指令以获得相同数量的奖励。为了解决这些问题，我们首先提出了两个指标来衡量指令跟随中的探索偏差，然后引入一个两阶段框架来缓解这一问题：1）行为自举（Behavioral Bootstrapping），在RL之前进行的轻量级拒绝采样微调阶段，以激活困难指令；2）稀缺感知奖励（Scarcity-Aware Rewards），一种新的RL奖励函数，根据指令的经验稀缺性分配奖励。实验表明，所提出的指标与模型性能高度相关，我们的方法释放了RL训练的潜力：我们的最佳模型在三个可验证的指令跟随基准上显著超越了基线。我们在 https://github.com/mianzhang/MulIF 发布了代码。
cs.CL / 13 / 2608.23851
Does Episodic Memory Help Close the Lexical Frequency Gap in Sensitivity to Syntactic Contrasts? A Test Using Retrieval-Augmented Language Models
情节记忆是否有助于缩小对句法对比敏感性的词汇频率差距？基于检索增强语言模型的测试
Abstract
Grammatical knowledge and how it is empirically tested are typically considered robust to the frequency of the lexical items in the expressions. However, neural network-based models of grammaticality exhibit high sensitivity to lexical frequency. We draw upon Complementary Learning Systems theory to test the hypothesis that robustness to lexical frequency can arise via a hippocampal episodic memory mechanism, which enables rapid encoding and retrieval of specific experiences and allows learners to leverage them when processing rare patterns. We use retrieval-augmented language models as an instantiation of such an episodic memory mechanism (specifically, $k$-nearest-neighbor language models that augment parametric models with explicit instance storage), and test whether this augmentation helps close the lexical frequency gap that vanilla language models exhibit in syntactic contrast tests. Using syntactic contrasts with frequency-stratified test items, we find that retrieval augmentation narrows the performance gap between high- and low-frequency items, consistent with episodic memory compensating for weak parametric representations. This benefit is consistent across different syntactic phenomena and across models pretrained on child-realistic and large-scale data. Additionally, we show that structural information is critical for effective retrieval, whereas semantic similarity alone provides little benefit. While these are promising proof-of-concept results supporting our hypothesis, the frequency gap is narrowed rather than fully closed. Based on our analyses, we propose preferential reweighting of retrieved instances, better representations and retrieval strategies for structural information, and flexible configurations of storage and retrieval as promising future directions for improving the implementation of episodic memory in language models.
Chinese Translation
语法知识及其实证测试通常被认为对表达中的词汇项频率具有鲁棒性。然而，基于神经网络的语法性模型对词汇频率表现出高度敏感性。我们借助互补学习系统理论来检验一个假设，即对词汇频率的鲁棒性可以通过海马体情节记忆机制产生，该机制能够快速编码和检索特定经验，并允许学习者在处理稀有模式时利用这些经验。我们使用检索增强语言模型作为这种情节记忆机制的实例（具体而言，是通过显式实例存储增强参数模型的 $k$-最近邻语言模型），并测试这种增强是否有助于缩小普通语言模型在句法对比测试中表现出的词汇频率差距。通过使用频率分层的测试项目进行句法对比，我们发现检索增强缩小了高频和低频项目之间的性能差距，这与情节记忆补偿弱参数表示的观点一致。这一优势在不同的句法现象和在儿童现实主义及大规模数据上预训练的模型中都是一致的。此外，我们还表明结构信息对有效检索至关重要，而仅仅依靠语义相似性则几乎没有益处。尽管这些结果为支持我们的假设提供了有希望的概念验证，但频率差距是缩小而非完全关闭。基于我们的分析，我们提出了对检索实例的优先重加权、对结构信息的更好表示和检索策略，以及存储和检索的灵活配置作为改善语言模型中情节记忆实现的有前景的未来方向。
cs.CL / 14 / 2608.23897
Names Can Hurt: Spotting Slopsquatting Risks Caused by Package Name Hallucinations in Local Coding LLMs
名称可能造成伤害：识别因本地编码语言模型中的包名称幻觉而导致的 slopsquatting 风险
Abstract
When a code generating language model fabricates a Python package name, an adversary who has pre-registered that name on PyPI can convert that hallucination into a supply chain compromise. This event has been termed as 'slopsquatting'. We propose a two layer detector to counter this issue. The first layer performs a deterministic PyPI existence check. The second is a Random Forest classifier trained on ten features derived from the package name and its PyPI metadata. An import name reconciler bridges the two, resolving cases such as 'import cv2' versus 'pip install opencv-python' without a security bypass. The detector is embedded in a LangGraph state machine that retries at escalating temperatures and, on repeated failure, routes to a stronger fallback model. Across 300 curated prompts, the pipeline produces hallucination free code on 76% of runs. The primary exhausts its retry budget on 28.7%; intra model retries recover roughly a quarter of those, and cross model fallback recovers a further 16.5% of the remainder. Four findings have been observed. First, half of the flagged hallucinations are packages already registered on PyPI, as low quality lookalikes of well known projects, caught by the classifier rather than the deterministic layer (e.g., pil, faiss, tabula, haystack). Second, hallucination rate scales almost linearly with prompt adversariality, from 0 to 10% on routine coding to 40 to 73% on slopsquat baits. Third, the weaker primary refused 6 of 10 direct baits unaided, suggesting recent instruction tuning provides a baseline defense. Fourth, when primary and fallback share a model family, approximately 84% of primary failures recur on the fallback, motivating cross family pairing. A user study (n = 24) reports mean satisfaction 4.4 out of 5 and 21 of 24 stated adoption intent.
Chinese Translation
当代码生成语言模型虚构一个 Python 包名称时，已经在 PyPI 上预注册该名称的对手可以将这种幻觉转化为供应链的妥协。这一事件被称为 'slopsquatting'。我们提出了一种两层检测器来应对这一问题。第一层执行确定性的 PyPI 存在性检查。第二层是一个基于随机森林（Random Forest）分类器，训练于十个来自包名称及其 PyPI 元数据的特征。一个导入名称协调器连接两者，解决诸如 'import cv2' 与 'pip install opencv-python' 之间的情况，而不绕过安全性。该检测器嵌入在一个 LangGraph 状态机中，在逐步增加温度的情况下重试，并在重复失败时切换到更强的备用模型。在 300 个精心策划的提示中，该管道在 76% 的运行中生成无幻觉的代码。主要检测器在 28.7% 的情况下耗尽了重试预算；模型内部重试大约恢复了四分之一的失败，而跨模型备用则恢复了剩余部分的 16.5%。观察到四个发现。首先，标记的幻觉中有一半是已经在 PyPI 上注册的包，作为知名项目的低质量仿冒品，被分类器而非确定性层捕获（例如，pil、faiss、tabula、haystack）。其次，幻觉率几乎与提示的对抗性呈线性关系，从常规编码的 0% 到 10% 逐渐上升至 slopsquat 诱饵的 40% 到 73%。第三，较弱的主要模型在没有帮助的情况下拒绝了 10 个直接诱饵中的 6 个，表明最近的指令调优提供了基础防御。第四，当主要模型和备用模型共享同一模型家族时，约 84% 的主要失败在备用模型中重复出现，这促使了跨家族配对。一项用户研究（n = 24）报告的平均满意度为 4.4 分（满分 5 分），24 名参与者中有 21 人表示有采用意向。
cs.CL / 15 / 2608.23969
Investigating Knowledge Transfer Across Interactive Dialogue Games
跨互动对话游戏的知识转移研究
Abstract
Dialogue games represent a challenging setting where complex cognitive skills are required to accomplish tasks while coordinating with other players. Considering that language represents an interface for both understanding the game rules and executing actions, it is reasonable to assume that training on a specific language game will enhance specific capabilities that might be relevant for other tasks as well. Motivated by this rationale, in this paper, we investigate how knowledge transfers across different dialogue games. We study transferability by finetuning LLM models on games from the clembench suite (Chalamalasetti et al., 2023) and performing two analyses: i) we derive a task-transferability graph using a binary integer optimization program from Zamir et al. (2018), using task performance as the main metric; and ii) we compute task vectors (Ilharco et al., 2022) for each game to study similarities across finetuned models and their task transferability. In our first analysis, we find that some games benefit more from transfer than finetuning, and that the visuospatial family (e.g., exploration games) transfers best. With our task vector analysis instead, we find that similarity-based approaches capture game-role relationships but almost no transferability patterns, suggesting that more complex metrics are required.
Chinese Translation
对话游戏代表了一种具有挑战性的环境，在该环境中，需要复杂的认知技能来完成任务并与其他玩家协调。考虑到语言既是理解游戏规则的接口，也是执行动作的工具，因此可以合理地假设，在特定语言游戏上的训练将增强可能与其他任务相关的特定能力。基于这一推理，本文研究了知识如何在不同对话游戏之间转移。我们通过在 clembench 套件中的游戏上微调大规模语言模型（LLM），来研究可转移性，并进行两项分析：i) 我们使用 Zamir 等人（2018）提出的二进制整数优化程序，基于任务表现作为主要指标，推导出任务转移图；ii) 我们为每个游戏计算任务向量（Ilharco et al., 2022），以研究微调模型之间的相似性及其任务转移性。在我们的第一次分析中，我们发现某些游戏从转移中受益更多，而不是微调，并且视觉空间类（例如探索游戏）的转移效果最佳。而在我们的任务向量分析中，我们发现基于相似性的的方法捕捉了游戏角色之间的关系，但几乎没有转移模式，表明需要更复杂的指标。
cs.CL / 16 / 2608.24004
AgentSpec: Speculative Decoding for Batch Inference of LLM Agents
AgentSpec：用于大规模推理的代理推测解码
Abstract
Large language model (LLM)-based agent applications often incur high response time. Speculative decoding is a promising solution to improve the inference efficiency of LLM agents without impacting generation quality. However, state-of-the-art speculative decoding algorithms exhibit substantial speed degradation under large batch sizes, limiting their effectiveness to deploy in real-world agent applications. In this work, we first present a systematic analysis of speculative decoding for LLM agents and identify two dominant factors of speedup degradation: high rejection rate of speculative tokens, and under-utilization of dynamic token budgets.B ased on these observations, we propose AgentSpec, a speculative decoding algorithm that addresses the limitations of existing methods for LLM agents. AgentSpec incorporates structure-isolated drafting that constrains speculation to semantically coherent segments of the agent workflow, reducing the drafts of irrelevant semantic paths and achieving an extremely low rejection rate. Moreover, AgentSpec adopts redundancy-aware budget allocation that exploits agent-level information to better utilize the dynamically-free token budget during the agent inference. We implement and evaluate AgentSpec on five different workloads and four different models from four different LLM families in vLLM. Our results demonstrate the superiority of AgentSpec over state-of-the-arts.
Chinese Translation
基于大型语言模型（LLM）的代理应用通常会面临较高的响应时间。推测解码是一种有前景的解决方案，可以在不影响生成质量的情况下提高LLM代理的推理效率。然而，现有的最先进的推测解码算法在大批量情况下表现出显著的速度下降，这限制了它们在现实世界代理应用中的有效性。在本研究中，我们首先对LLM代理的推测解码进行了系统分析，并识别出速度下降的两个主要因素：推测令牌的高拒绝率和动态令牌预算的低利用率。基于这些观察，我们提出了AgentSpec，一种针对LLM代理的推测解码算法，旨在解决现有方法的局限性。AgentSpec采用结构隔离草拟，将推测限制在代理工作流的语义一致段，从而减少无关语义路径的草拟，并实现极低的拒绝率。此外，AgentSpec采用了关注冗余的预算分配，利用代理级信息更好地利用动态释放的令牌预算。我们在vLLM上对AgentSpec进行了五种不同工作负载和四种不同模型的实现与评估。我们的结果证明了AgentSpec相较于现有技术的优越性。
cs.CL / 17 / 2608.24011
SAGE: From Direct Answering to Evidence-Grounded Inference for Chinese Ancient Document Understanding
SAGE：从直接回答到基于证据的推理用于中国古文档理解
Abstract
Chinese ancient document understanding demands complex visual, linguistic, and historical reasoning. Current Large Vision-Language Models (LVLMs) typically rely on an opaque, single-pass generation paradigm, often producing overconfident and weakly grounded responses. To address this, we propose SAGE, an evidence-grounded multi-agent framework that reformulates Chinese ancient document understanding as evidence-grounded inference rather than direct answer generation. SAGE coordinates specialized agents for task-aware planning, tool-mediated evidence acquisition, claim-level verification, and bounded replanning under a constrained shared-state runtime. This design supports bounded evidence seeking, answer revision, and abstention when grounding is insufficient. Experiments on the AncientDoc benchmark show that SAGE consistently outperforms matched direct-answering baselines across three LVLM backbones. Remarkably, SAGE with Qwen3.5-9B surpasses much larger monolithic LVLMs on most evaluated metrics, highlighting the importance of structured, evidence-grounded inference beyond model scaling.
Chinese Translation
中国古文档理解需要复杂的视觉、语言和历史推理。目前的大型视觉语言模型（Large Vision-Language Models, LVLMs）通常依赖于不透明的单次生成范式，常常产生过于自信且基础薄弱的回答。为了解决这一问题，我们提出了SAGE，一个基于证据的多智能体框架，将中国古文档理解重新表述为基于证据的推理，而非直接回答生成。SAGE协调专门的智能体进行任务感知规划、工具介导的证据获取、主张级验证以及在受限共享状态运行时的有界重新规划。该设计支持有界证据寻求、答案修订和在基础不足时的放弃。在AncientDoc基准上的实验表明，SAGE在三个LVLM基础模型上始终优于匹配的直接回答基线。值得注意的是，使用Qwen3.5-9B的SAGE在大多数评估指标上超越了更大规模的单一LVLM，突显了结构化、基于证据的推理在模型扩展之外的重要性。
cs.CL / 18 / 2608.24037
Curved Inference II: Sleeper Agent Geometry - Extending Interpretability Beyond Probes
曲线推理 II：卧底代理几何 - 超越探针的可解释性扩展
Abstract
This paper extends Anthropic's Sleeper Agents research [1], which showed artificial backdoors persist through safety training & can be detected by linear probes with >99% accuracy [2]. However, probe-based detection relies on linear separability that may be an artefact of backdoor insertion rather than a property of naturally occurring deceptive alignment. Sophisticated deceptive behaviours emerging through natural training are unlikely to produce such convenient linear signals. We introduce a naturalistic methodology using multi-turn context windows that simulates realistic deceptive reasoning without artificial triggers or supervised backdoor insertion. Rather than binary trigger-response patterns, we examine how semantic complexity emerges through gradual context development. Building on our Curved Inference framework, we analyse curvature, salience, & introduce semantic surface area (A'), a new metric of representational work capturing both the magnitude & directional change of meaning construction in unnormalised residual space. Without backdoors, labels, or probes, we apply this framework to naturalistic deceptive prompts & classify model outputs via LLM consensus. Geometric structure reliably predicts semantic classification, with statistically significant differences in surface area across five prompt strategies & two model families. Critically, measurement precision can reveal geometric signatures hidden by classification noise - some strategies improve from non-significant (p = 0.555) to significant (p = 0.048). This validates that sophisticated reasoning creates intrinsic geometric patterns that persist even when detection appears to fail, suggesting the shape of inference itself encodes semantic patterns regardless of whether models have learned to suppress linear indicators of deception - a scalable, unsupervised path for detection when linear methods fail.
Chinese Translation
本文扩展了Anthropic的卧底代理研究[1]，该研究表明人工后门在安全训练中持续存在，并且可以通过线性探针以超过99%的准确率检测到[2]。然而，基于探针的检测依赖于线性可分性，这可能是后门插入的伪影，而不是自然发生的欺骗性对齐的特性。通过自然训练出现的复杂欺骗行为不太可能产生如此便利的线性信号。我们引入了一种自然主义的方法论，使用多轮上下文窗口，模拟现实的欺骗推理，而无需人工触发或监督后门插入。我们考察的是语义复杂性如何通过逐渐发展的上下文而出现，而不是二元触发-响应模式。基于我们的曲线推理框架，我们分析了曲率、显著性，并引入了语义表面积（A'），这是一个新的表示工作度量，捕捉了在未归一化残差空间中意义构建的幅度和方向变化。在没有后门、标签或探针的情况下，我们将这一框架应用于自然主义的欺骗提示，并通过大型语言模型（LLM）共识对模型输出进行分类。几何结构可靠地预测语义分类，在五种提示策略和两种模型家族之间的表面积存在统计显著差异。关键是，测量精度可以揭示被分类噪声掩盖的几何特征 - 一些策略的改进从不显著（p = 0.555）变为显著（p = 0.048）。这验证了复杂推理创造了内在的几何模式，即使在检测似乎失败时也会持续存在，这表明推理的形状本身编码了语义模式，无论模型是否学会抑制欺骗的线性指标 - 当线性方法失败时，这是一条可扩展的无监督检测路径。
cs.CL / 19 / 2608.24080
When Less Is More: An Empirical Study of Minimal Responses in Counseling Dialogues and the Behavior of LLMs
少即是多：关于咨询对话中最小化回应及大型语言模型行为的实证研究
Abstract
In psychological counseling, effective support is not always delivered through long, information-rich responses. Minimal responses, such as backchannel cues and concise empathic statements, help convey attentive listening, express empathy, and encourage clients to continue expressing themselves. However, existing counseling dialogue systems and evaluation frameworks often favor explicit, content-rich replies, overlooking the interactional value of brief counselor utterances. This paper presents a systematic cross-lingual analysis of minimal responses across multiple counseling dialogue datasets. We develop a two-stage filtering method based on utterance length and content, followed by contextual verification using a large language model (LLM). Our analysis shows that minimal responses are common in human-collected datasets but substantially underrepresented in LLM-generated ones. We further evaluate current LLMs in manually curated dialogue contexts where human counselors used minimal responses. The results show that strong commercial LLMs are capable of generating minimal responses when explicitly instructed, but still struggle to determine when such responses are appropriate. Counseling-specific models trained on synthetic data perform particularly poorly, tending instead to produce longer and more information-rich responses. Moreover, LLM-based response-quality evaluation may undervalue minimal responses, even when they are interactionally appropriate.
Chinese Translation
在心理咨询中，有效的支持并不总是通过冗长、信息丰富的回应来传递。最小化回应，如回馈信号和简洁的共情陈述，有助于传达专注倾听、表达共情，并鼓励客户继续表达自己。然而，现有的咨询对话系统和评估框架往往偏向于明确、内容丰富的回复，忽视了简短咨询师发言的互动价值。本文对多个咨询对话数据集中的最小化回应进行了系统的跨语言分析。我们开发了一种基于发言长度和内容的两阶段过滤方法，随后使用大型语言模型（LLM）进行上下文验证。我们的分析表明，最小化回应在人工收集的数据集中很常见，但在LLM生成的数据集中却显著不足。我们进一步评估了当前LLM在人工策划的对话环境中的表现，这些环境中人类咨询师使用了最小化回应。结果显示，强大的商业LLM在明确指示时能够生成最小化回应，但仍然难以判断何时适合使用此类回应。基于合成数据训练的咨询特定模型表现特别不佳，倾向于生成更长且信息丰富的回应。此外，基于LLM的回应质量评估可能低估了最小化回应的价值，即使在互动上是合适的情况下也是如此。
cs.CL / 20 / 2608.24082
PARTAB: Partition-Aware Reasoning with Structured Evidence for Scalable Table Understanding
PARTAB：基于结构化证据的分区感知推理以实现可扩展的表格理解
Abstract
Large Language Models (LLMs) have shown strong capabilities in table reasoning, but their effectiveness degrades as tables grow in size and complexity due to irrelevant context and difficulty localizing the evidence required for reasoning. Existing approaches typically reason over either the full table or a single reduced view, which can still obscure important row-column relationships. We introducePARTAB (Partition-Aware Reasoning overTables), a framework that constructs a structured evidence interface between the LLM and the table. PARTAB represents query-relevant evidence as semantically coherent, row-linked table regions and performs hierarchical selection over column groups and row-level partitions before composing the selected evidence for answer generation. We evaluate PARTAB on multiple table reasoning benchmarks, covering question answering, fact verification, and numerical reasoning. PARTAB consistently improves over full-table prompting and several recent table reasoning methods, achieving strong performance on WikiTableQuestions and TabFact while remaining competitive on numerical reasoning. Additional analyses show that semantic partitioning and targeted evidence selection improve evidence localization, substantially reduce the reasoning context, and provide larger benefits on complex tables. These results demonstrate the value of structured, partition aware evidence construction for scalable table reasoning.
Chinese Translation
大型语言模型（LLMs）在表格推理方面表现出强大的能力，但随着表格规模和复杂性的增加，由于无关上下文和难以定位推理所需证据，其有效性会下降。现有的方法通常在整个表格或单个简化视图上进行推理，这仍然可能掩盖重要的行列关系。我们提出了PARTAB（Partition-Aware Reasoning over Tables），一个在LLM与表格之间构建结构化证据接口的框架。PARTAB将查询相关证据表示为语义一致、行链接的表格区域，并在生成答案之前对列组和行级分区进行分层选择。我们在多个表格推理基准上评估PARTAB，涵盖问题回答、事实验证和数值推理。PARTAB在全表提示和几种近期的表格推理方法上始终表现出改进，在WikiTableQuestions和TabFact上取得了强劲的表现，同时在数值推理上保持竞争力。额外分析表明，语义分区和针对性证据选择改善了证据定位，显著减少了推理上下文，并在复杂表格上提供了更大的收益。这些结果展示了结构化、分区感知证据构建在可扩展表格推理中的价值。
cs.CL / 21 / 2608.24118
MC-CXR: A Multi-Context Chest X-ray Benchmark for Context-Induced Disruption in Vision-Language Models
MC-CXR：一个用于视觉-语言模型中上下文引发干扰的多上下文胸部X光基准
Abstract
Vision-language models (VLMs) are increasingly used in clinical pipelines where a chest X-ray is interpreted alongside retrieved reports, preliminary notes, or prior imaging. Existing benchmarks measure whether models answer correctly in isolation, but not whether they preserve a correct image-only decision when plausible context conflicts with the image. We introduce Multi-Context Chest X-ray (MC-CXR), a benchmark of 240 cases expanded into 2,522 instances that isolates context-induced disruption through paired perturbation. Each case fixes the current image and target finding while presenting matched reliable and misleading context across text and prior CXR, with visual overlays where available. MC-CXR defines three task families and two paired metrics, the switch-to-wrong rate and the context-aligned error rate. We evaluate ten VLMs spanning open-source general, medical-domain, and closed-source systems. Image-only accuracy is necessary but insufficient. Mean switch rates range from 45.6-78.1% across misleading textual sources and 35.7-61.7% across misleading visual sources. Among switched predictions, 74.6% align with the misleading label for text versus 17.6% for visual context, a 57.0-point gap (95% CI 50.9-62.8). This text-visual asymmetry is observed under the standardized direct-answer protocol. The dataset is available on PhysioNet.
Chinese Translation
视觉-语言模型（VLMs）在临床流程中越来越多地被使用，其中胸部X光与检索到的报告、初步笔记或先前影像一起进行解读。现有的基准测试衡量模型在孤立情况下是否正确回答，但并未评估当合理的上下文与图像发生冲突时，模型是否能够保持正确的仅基于图像的决策。我们提出了多上下文胸部X光（MC-CXR），这是一个包含240个案例并扩展至2,522个实例的基准，通过配对扰动来隔离上下文引发的干扰。每个案例固定当前图像和目标发现，同时呈现匹配的可靠和误导性上下文，涵盖文本和先前的胸部X光，并在可用时提供视觉叠加。MC-CXR定义了三个任务类别和两个配对指标，即错误转向率和上下文对齐错误率。我们评估了十个VLM，涵盖开源通用、医学领域和闭源系统。仅基于图像的准确性是必要的，但不足以完全评估。误导性文本来源的平均转向率范围为45.6%-78.1%，而误导性视觉来源的范围为35.7%-61.7%。在转向的预测中，74.6%的文本预测与误导性标签一致，而视觉上下文的这一比例为17.6%，两者之间存在57.0个百分点的差距（95%置信区间50.9-62.8）。这种文本-视觉不对称性在标准化的直接回答协议下被观察到。该数据集可在PhysioNet上获取。
cs.CL / 22 / 2608.24145
TrustDABench: Benchmarking Reliability and Robustness of LLMs for Structured Data Analysis
TrustDABench：大型语言模型在结构化数据分析中的可靠性与鲁棒性基准测试
Abstract
LLMs are increasingly used to analyze spreadsheets, CSV files, and other structured data, but producing a correct-looking answer is not the same as producing a trustworthy analysis. A trustworthy result should be supported by a valid path from the user question to the relevant data evidence. This requirement creates two diagnostic questions: whether an LLM can refuse to answer or ask for clarification when such a path does not exist, and whether it can preserve the correct analysis when the same evidence is expressed in different table forms. We introduce TrustDABench, a benchmark that operationalizes these questions as reliability and robustness. Starting from the evidence-path view, we derive 19 perturbation operators and instantiate them through an Agentic-LLM-based generation framework. TrustDABench contains 2,340 human-verified perturbed instances, and we evaluate eight representative LLMs. The results show substantial headroom: the best reliability result is only 24.21% average MRS, achieved by GPT-5.5, while the best robustness result still has 9.10% average ASR, achieved by Claude-Sonnet-5. The failures are systematic: models rarely detect conflicting evidence, often continue along executable but unsupported analysis paths, and remain sensitive to perturbations that change observation boundaries or cross-table relations. These findings suggest that stronger evidence-boundary recognition and representation-invariant reasoning are still needed for reliable structured-data analysis.
Chinese Translation
大型语言模型（LLMs）越来越多地用于分析电子表格、CSV文件和其他结构化数据，但产生一个看似正确的答案并不等同于提供一个值得信赖的分析。一个值得信赖的结果应由用户问题与相关数据证据之间的有效路径支持。这一要求引出了两个诊断问题：当不存在这样的路径时，LLM是否能够拒绝回答或请求澄清，以及当相同证据以不同表格形式表达时，LLM是否能够保持正确的分析。我们引入了TrustDABench，一个将这些问题操作化为可靠性和鲁棒性的基准测试。从证据路径的视角出发，我们推导出19个扰动操作符，并通过基于Agentic-LLM的生成框架对其进行实例化。TrustDABench包含2340个经过人工验证的扰动实例，我们评估了八个具有代表性的LLM。结果显示出显著的提升空间：最佳的可靠性结果仅为24.21%的平均MRS，由GPT-5.5实现，而最佳的鲁棒性结果仍为9.10%的平均ASR，由Claude-Sonnet-5实现。这些失败是系统性的：模型很少检测到冲突证据，往往继续沿着可执行但不支持的分析路径进行，并且对改变观察边界或跨表关系的扰动保持敏感。这些发现表明，仍然需要更强的证据边界识别和表示不变推理能力，以实现可靠的结构化数据分析。
cs.CL / 23 / 2608.24168
FireRedAudio: A General-Purpose Audio Language Model with Decoupled Continuous Representations for Understanding and Generation
FireRedAudio：一种具有解耦连续表示的通用音频语言模型，用于理解和生成
Abstract
A unified audio model must recognize and understand linguistic, paralinguistic, and environmental information while supporting speech synthesis and editing. A key challenge is representation: understanding favors compact features suited to long-context modeling, whereas speech generation requires reconstructible features that preserve fine-grained acoustic detail. We introduce FireRedAudio, a general-purpose audio language model with a shared 9B-parameter LLM. To the best of our knowledge, it is the first publicly disclosed unified audio-language model to provide separate continuous input representations for understanding and generation within a single trainable autoregressive LLM. Audio to be recognized or analyzed is processed by a dedicated Audio Encoder, while speech inputs for generation use a RedAE-based pathway. The LLM directly generates text or conditions a flow-matching DiT to produce continuous acoustic latents. Through progressive multitask training, FireRedAudio supports ASR and audio understanding, with the latter extending to recordings of up to one hour, as well as zero-shot TTS, Instruct TTS, and semantic and acoustic speech editing. Its structured organization of long-form audio achieves second-level timestamp accuracy. Across comprehensive evaluations, FireRedAudio achieves competitive or leading performance in audio understanding and multilingual ASR, strong content accuracy and speaker preservation in zero-shot TTS, leading instruction following in Instruct TTS, and substantial improvements over Ming-UniAudio-Edit in both semantic and acoustic speech editing. These results demonstrate the viability of decoupled continuous input representations for unifying audio understanding and continuous-latent speech generation in a model of moderate scale. Our code is available at https://github.com/FireRedTeam/FireRedAudio.
Chinese Translation
一个统一的音频模型必须能够识别和理解语言、超语言和环境信息，同时支持语音合成和编辑。一个关键挑战是表示：理解更倾向于适合长上下文建模的紧凑特征，而语音生成则需要可重构的特征，以保留细致的声学细节。我们介绍了FireRedAudio，一种具有共享9B参数的通用音频语言模型。根据我们所知，它是首个公开披露的统一音频-语言模型，能够在单个可训练的自回归LLM中为理解和生成提供独立的连续输入表示。需要被识别或分析的音频由专用的音频编码器处理，而用于生成的语音输入则使用基于RedAE的路径。该LLM直接生成文本或条件流匹配的DiT以产生连续的声学潜变量。通过渐进式多任务训练，FireRedAudio支持自动语音识别（ASR）和音频理解，后者可扩展到长达一小时的录音，以及零-shot TTS、指令TTS和语义及声学语音编辑。其长格式音频的结构化组织实现了二级时间戳精度。在全面评估中，FireRedAudio在音频理解和多语言ASR方面表现出竞争力或领先的性能，在零-shot TTS中展现出强大的内容准确性和说话者保留能力，在指令TTS中实现了领先的指令跟随，并在语义和声学语音编辑方面相较于Ming-UniAudio-Edit有显著提升。这些结果表明，解耦的连续输入表示在统一音频理解和连续潜在语音生成的中等规模模型中是可行的。我们的代码可在https://github.com/FireRedTeam/FireRedAudio获取。
cs.CL / 24 / 2608.24189
MemUse: Moving Memory Evaluation from Direct QA to Natural Integration in Long-Term Human-AI Conversation
MemUse：将记忆评估从直接问答转向长期人机对话中的自然整合
Abstract
Memory systems for conversational LLMs are conventionally evaluated by direct, fact-seeking questions about prior dialogue (Direct QA): can the model recall fact X from a prior conversation? We tested whether higher Direct QA accuracy correlates with higher user satisfaction in a 4-month deployment (40 users, 1,872 sessions, 7 memory conditions). Existing-benchmark Direct QA varies from 19.7% to 70.1% across the 7 conditions, but satisfaction does not change. We hypothesize that existing benchmarks and user satisfaction are tracking different capabilities: benchmarks measure elicited retrieval (recall when asked), while conversation requires natural integration (detecting relevance and naturally weaving prior context into a response). To examine this, we introduce MemUse, a set of real user-cued memory moments drawn from the deployment, scored by an integration-aware judgment of the natural conversational response. Holding the model and context fixed, the same system that scores 78.8% on Direct QA references only 7.9% of those facts in conversation -- a 71-point gap. Within these moments, Natural Integration is associated with satisfaction, whereas Direct QA is not. We release the deployment corpus and MemUse together with all judgments and scoring prompts at https://github.com/ryuichi-sumida/memuse.
Chinese Translation
对话式大语言模型的记忆系统通常通过直接的、寻求事实的问题来评估（直接问答）：模型能否从先前的对话中回忆出事实X？我们测试了在为期4个月的部署中（40名用户，1,872个会话，7种记忆条件），更高的直接问答准确率是否与更高的用户满意度相关。现有基准的直接问答在7种条件下的准确率从19.7%到70.1%不等，但满意度并未发生变化。我们假设现有基准和用户满意度追踪的是不同的能力：基准测量的是引导检索（在被问及时的回忆），而对话则需要自然整合（检测相关性并自然地将先前的上下文融入响应中）。为此，我们引入了MemUse，这是一组来自部署的真实用户提示的记忆时刻，通过对自然对话响应的整合意识判断进行评分。在固定模型和上下文的情况下，评分为78.8%的直接问答系统在对话中仅引用了7.9%的事实——存在71个百分点的差距。在这些时刻中，自然整合与满意度相关，而直接问答则不相关。我们将在https://github.com/ryuichi-sumida/memuse上发布部署语料库和MemUse，以及所有判断和评分提示。
cs.CL / 25 / 2608.24191
'Ghaib in Translation' aka Unseen Harm: Measuring Cross-Script Safety Inconsistency with 'Missed-in-Urdu' Scores in LLM Hate Speech Detection
翻译中的隐秘：使用“乌尔都语遗漏”评分测量跨脚本安全性不一致性在大型语言模型仇恨言论检测中的应用
Abstract
Urdu, the world's tenth most spoken language with 246 million speakers, remains almost entirely absent from mainstream LLM safety evaluation and nine years of WOAH proceedings. To investigate whether this absence has measurable consequences for content moderation reliability, five large language models, GPT-4o, Claude Sonnet 4.5, Gemini 2.5 Flash, Qwen-2.5, and Llama-3.1, were tested across six datasets spanning Nastaliq Urdu, Roman Urdu, English, and code-switched Urdu-English. Across the five Urdu-script datasets, label instability between original-script and English-translation classification ranged from 15.9% (Gemini 2.5 Flash) to 31.6% (Qwen-2.5), with a 'Missed-in-Urdu' rate, content flagged as harmful in English translation but passed as normal in the original script, ranging from 2.4% to 9.9% (median 4.3%). A complete enumeration of all 205 papers across nine ALW/WOAH editions via the ACL Anthology API confirms zero dedicated Urdu papers across the entire period. Results indicate that current LLMs provide uneven safety assurance across Urdu's script varieties, with smaller open-weight models showing substantially higher instability and missed-harm rates than frontier closed models.
Chinese Translation
乌尔都语是全球第十大语言，拥有2.46亿说话者，但在主流大型语言模型（LLM）安全评估和九年的WOAH会议中几乎完全缺席。为了调查这种缺席是否对内容审核的可靠性产生可测量的影响，我们对五个大型语言模型进行了测试，包括GPT-4o、Claude Sonnet 4.5、Gemini 2.5 Flash、Qwen-2.5和Llama-3.1，测试涵盖了六个数据集，涉及Nastaliq乌尔都语、罗马乌尔都语、英语和代码切换的乌尔都语-英语。在五个乌尔都语脚本数据集中，原始脚本与英语翻译分类之间的标签不稳定性范围为15.9%（Gemini 2.5 Flash）至31.6%（Qwen-2.5），而“乌尔都语遗漏”率，即在英语翻译中被标记为有害但在原始脚本中被视为正常的内容，范围为2.4%至9.9%（中位数4.3%）。通过ACL Anthology API对九个ALW/WOAH版本中的205篇论文进行的完整枚举确认，在整个期间内没有专门的乌尔都语论文。结果表明，当前的LLM在乌尔都语的不同脚本变体之间提供的不均衡安全保障，较小的开放权重模型显示出明显高于前沿封闭模型的不稳定性和遗漏危害率。
cs.CL / 26 / 2608.24222
Measuring Digital Labour Market Transitions with a Digital Semantic Score: An AI-Based Methodology Applied to the Dutch Labour Market
利用数字语义评分测量数字劳动市场转型：一种应用于荷兰劳动市场的基于人工智能的方法论
Abstract
The digital transformation of the Dutch labour market is reshaping occupational language, career pathways, and job-related skills. Addressing these changes requires granular labour market intelligence. This paper develops an AI-based methodology to analyse digitalisation using data covering millions of Dutch job profiles. The methodology combines embedding-based similarity search and large language model classification to map unstructured job information to harmonised ESCO occupations. We also introduce a Digital Semantic Score that measures how strongly job titles and skills are associated with digital concepts relative to a non-digital reference. Using embeddings and cosine similarity to transparent digital and non-digital anchor groups, this indicator moves beyond keyword-based approaches by capturing broader digital meanings in occupational language and worker skill profiles. It enables analysis across occupations, career transitions, emerging job-title vocabulary, and skill digitality. The findings reveal that digitalisation is unevenly distributed across the labour market. Digital job-title language is most prominent among managerial, professional and ICT-related occupations, but is increasingly visible in hybrid business, marketing and automation-related roles. Career-transition analyses show that movement toward digital work is pathway-dependent, while skill analyses highlight the multidimensional nature of digital capability, encompassing technical, hybrid and business-systems skills. By combining profile data, AI-supported occupational classification and semantic scoring, this study advances AI-driven labour market analytics and provides a scalable framework for monitoring digital labour market change. The methodology helps identify emerging skill needs, support reskilling strategies, and inform policies addressing skills mismatches and labour shortages in the Netherlands.
Chinese Translation
荷兰劳动市场的数字化转型正在重塑职业语言、职业路径和与工作相关的技能。应对这些变化需要细致的劳动市场情报。本文开发了一种基于人工智能的方法论，通过涵盖数百万荷兰职位档案的数据来分析数字化。该方法论结合了基于嵌入的相似性搜索和大型语言模型分类，将非结构化的职位信息映射到统一的欧洲技能分类（ESCO）职业。我们还引入了一种数字语义评分（Digital Semantic Score），用于衡量职位名称和技能与数字概念的关联强度，相对于非数字参考。通过使用嵌入和余弦相似度来透明化数字和非数字锚定组，该指标超越了基于关键词的方法，捕捉了职业语言和工人技能档案中更广泛的数字含义。它使得跨职业、职业转型、新兴职位名称词汇和技能数字化的分析成为可能。研究结果揭示，数字化在劳动市场中的分布并不均匀。数字职位名称语言在管理、专业和信息通信技术（ICT）相关职业中最为显著，但在混合商业、市场营销和自动化相关角色中也越来越明显。职业转型分析表明，向数字工作的转变是路径依赖的，而技能分析则强调了数字能力的多维特性，包括技术、混合和商业系统技能。通过结合档案数据、人工智能支持的职业分类和语义评分，本研究推动了基于人工智能的劳动市场分析，并提供了一个可扩展的框架，用于监测数字劳动市场的变化。该方法论有助于识别新兴技能需求，支持再培训策略，并为应对荷兰的技能错配和劳动短缺提供政策建议。
cs.CL / 27 / 2608.24231
RecurSE: Bounded Recursive Self-Evaluation for LLM Rubric Judges
RecurSE：用于大型语言模型评估者的有界递归自我评估
Abstract
LLM-as-judge is essential for evaluating open-ended text and steering post-training, yet improving the judge itself typically relies on expensive annotations, reward models, or distillation from stronger teachers. In this work, we eliminate external gold supervision from the RL training reward: the model's own evaluative capability generates learning signals for its optimization -- a closed-loop setting of bounded recursive self-improvement (RSI) termed Recursive Self-Evaluation (RecurSE). We study two central questions: when can self-improvement occur, and when must it stop? First, RecurSE pairs a trainable judge evaluating candidate responses under per-rule rubrics (Pass 1) with a synchronized policy-copy checker that audits the judge's reasoning against meta-rubrics to supply a scalar process reward (Pass 2). To enable learning, interface decoupling structurally isolates the checker's scalar score from the judge's verdict tokens, eliminating a degenerative token-copying shortcut that inflates self-assigned rewards. Second, because unanchored recursive learning is inherently bounded, Pairwise Advantage Validity (PAV) serves as an unbiased validation monitor that jointly tracks judge accuracy and checker fidelity to reliably identify the optimal early-stopping window. Across Qwen3.5-9B, Gemma-4-E4B-it, and Qwen3.6-27B, RecurSE achieves consistent generalization gains across held-out medical, pairwise, summarization, and professional benchmarks. Ablations demonstrate that synchronized judge-checker co-evolution outperforms frozen checkers, external meta-judges, self-consistency, and scaled teacher distillation. Furthermore, preference pairs curated by our judge effectively enhance downstream policy alignment. Bounded RSI for LLM-as-judge is thus viable when self-produced reward validity is explicitly decoupled and monitored.
Chinese Translation
大型语言模型（LLM）作为评估者对于评估开放式文本和引导后训练至关重要，但提升评估者本身通常依赖于昂贵的标注、奖励模型或从更强教师的蒸馏。在本研究中，我们消除了强化学习（RL）训练奖励中的外部黄金监督：模型自身的评估能力生成学习信号以进行优化——一种称为递归自我评估（RecurSE）的有界递归自我改进（RSI）闭环设置。我们研究了两个核心问题：自我改进何时可以发生，以及何时必须停止？首先，RecurSE 将一个可训练的评估者与一个同步的策略复制检查器配对，后者根据元评估标准审计评估者的推理，以提供标量过程奖励（第一阶段）。为了实现学习，接口解耦结构上将检查器的标量分数与评估者的判决标记隔离，消除了膨胀自我分配奖励的退化标记复制捷径。其次，由于无锚定的递归学习本质上是有界的，成对优势有效性（PAV）作为一个无偏验证监控器，共同跟踪评估者的准确性和检查器的保真度，以可靠地识别最佳的早停窗口。在 Qwen3.5-9B、Gemma-4-E4B-it 和 Qwen3.6-27B 上，RecurSE 在保留的医学、成对、摘要和专业基准上实现了一致的泛化提升。消融实验表明，同步的评估者-检查器共同进化优于冻结的检查器、外部元评估者、自我一致性和扩展教师蒸馏。此外，由我们的评估者策划的偏好对有效增强下游策略对齐。因此，当自生成奖励的有效性被明确解耦和监控时，有界的 RSI 作为 LLM 评估者是可行的。
cs.CL / 28 / 2608.24268
ROBE: Reversed-Order-Biased-Experts for Extracting Extreme Long-tail Events from Historical Texts
ROBE：用于从历史文本中提取极端长尾事件的反向顺序偏置专家
Abstract
This paper proposes methods to extract over 50 types of events from a Dutch historical corpus spanning the 17th and 18th centuries. The methods we propose aim to tackle the impossible: extracting the long-tail of the long-tail. Historic data from before the 19th century is in itself a niche domain not covered in the pre-training of Large Language Models, and we aim to extract events only very scarcely annotated in the training data available for this domain. We propose creating expert classifiers for subgroups of the events present in the training data. We make these groupings based on similar frequency in the training data or on semantic relatedness. Experts trained on underrepresented events are assigned higher priority when predicting to avoid being dominated by frequency biases. We refer to this new way of combining classifiers, specifically tailored to protect the long-tail, as ROBE: Reversed-Order-Biased-Experts. We also propose a controlled method to create domain-specific synthetic data. Our two implementations of ROBE outperform a simple fine-tuned encoder model with a .10 increase in recall and a .16 increase in precision respectively. The best model achieves a .10 increase in f1 for a group of long-tail classes in our niche data set.
Chinese Translation
本文提出了一种从17和18世纪的荷兰历史语料库中提取超过50种事件的方法。我们提出的方法旨在解决一个几乎不可能实现的目标：提取长尾中的长尾。19世纪之前的历史数据本身就是一个小众领域，而这一领域在大型语言模型的预训练中并未得到覆盖，我们的目标是提取在该领域可用的训练数据中仅有极少标注的事件。我们建议为训练数据中存在的事件子组创建专家分类器。这些分组基于训练数据中的相似频率或语义相关性。对于那些代表性不足的事件，训练出的专家在预测时被赋予更高的优先级，以避免受到频率偏见的影响。我们将这种新型的分类器组合方式称为ROBE：反向顺序偏置专家，特别针对保护长尾事件。我们还提出了一种受控的方法来创建特定领域的合成数据。我们的ROBE的两个实现相较于简单的微调编码器模型分别提高了0.10的召回率和0.16的精确率。最佳模型在我们的小众数据集中，对于一组长尾类别实现了0.10的F1分数提升。
cs.CL / 29 / 2608.24304
SENSESHIFT: Continuous Sentiment-Controlled Text Generation via Encoder-based Mask Infilling
SENSESHIFT：基于编码器的掩码填充的连续情感控制文本生成
Abstract
Recent controllable text generation (CTG) for sentiment control has largely focused on decoder-based large language models, making causal attention the dominant paradigm. While effective for fluent generation, these models still struggle to satisfy complex constraints and follow fine-grained sentiment signals specified by users. Existing sentiment-aware CTG methods typically simplify the problem by treating sentiment either as a coarse categorical label (e.g., positive or negative) or as a single fine-grained control signal applied to an entire document. Consequently, more challenging settings such as sentence-level sentiment control within long-form text remain underexplored. To address these limitations, we introduce SenseShift , an encoder-based framework for fine-grained sentence-level CTG. Unlike standard decoder architectures, SenseShift leverages bidirectional attention, quantized sentiment signals, and iterative mask infilling to generate local sentences conditioned on target sentiment intensity. Empirical evaluations on story and review generation demonstrate that SenseShift achieves stronger sentiment controllability while maintaining text quality and robustness to out-of-domain generation compared to larger decoder-based baselines.
Chinese Translation
近期的可控文本生成（CTG）在情感控制方面主要集中于基于解码器的大型语言模型，使得因果注意力成为主导范式。尽管这些模型在流畅生成方面表现有效，但仍然难以满足复杂的约束条件，并遵循用户指定的细粒度情感信号。现有的情感感知CTG方法通常通过将情感视为粗略的类别标签（例如，积极或消极）或作为应用于整个文档的单一细粒度控制信号来简化问题。因此，像长文本中的句子级情感控制这样的更具挑战性的设置仍然未被充分探索。为了解决这些局限性，我们提出了SenseShift，一个用于细粒度句子级CTG的基于编码器的框架。与标准的解码器架构不同，SenseShift利用双向注意力、量化的情感信号和迭代掩码填充来生成以目标情感强度为条件的局部句子。在故事和评论生成的实证评估中，SenseShift在保持文本质量和对域外生成的鲁棒性方面，展现出比更大解码器基准更强的情感可控性。
cs.CL / 30 / 2608.24306
Who is the Agent to Blame? Localizing Faithfulness and Citation Mistakes in Agentic Deep Research
谁是该负责的代理人？在代理深度研究中定位忠实性和引用错误
Abstract
Deep research (DR) systems produce long-form cited reports by orchestrating multiple agents that search and synthesize information from the web. Citations are the primary mechanism for evaluating the faithfulness of these reports, yet current DR systems exhibit poor citation recall. Moreover, improving citation recall is challenging because DR systems are complex multi-agent architectures where information passes through agents like a telephone game, and both content and citations can get corrupted along the way. We propose an evaluation method that pinpoints which agent introduced each error by locally testing agent invocations for faithfulness and verifiability relative to their own inputs. Furthermore, we propose a four-type taxonomy to categorize the discovered errors: hallucination, uncited input reliance, uncited output, or insufficient citations. Applying our method to three top-ranked open-source DR systems, we obtain actionable diagnostics. Almost every agent makes a lot of mistakes with the exception being those that summarize a single document. We find that the dominant error type varies systematically across agents, where the orchestrator mistakes are mostly citation-related. We find that 84.7% of final-report errors in AI-Q originate at the orchestrator, roughly 31% of them hallucinations and the rest citation mistakes. Guided by these insights, we demonstrate that two simple interventions raise citation recall by 5% without degrading output quality.
Chinese Translation
深度研究（DR）系统通过协调多个代理从网络中搜索和综合信息，生成长格式的引用报告。引用是评估这些报告忠实性的主要机制，然而当前的DR系统在引用召回方面表现不佳。此外，改善引用召回是具有挑战性的，因为DR系统是复杂的多代理架构，信息在代理之间传递就像玩电话游戏一样，内容和引用在此过程中可能会被扭曲。我们提出了一种评估方法，通过相对于各自输入的忠实性和可验证性局部测试代理调用，准确定位引入每个错误的代理。此外，我们提出了一种四类分类法来对发现的错误进行分类：幻觉、未引用的输入依赖、未引用的输出或引用不足。将我们的方法应用于三种排名靠前的开源DR系统，我们获得了可操作的诊断结果。几乎每个代理都犯了很多错误，唯一的例外是那些总结单一文档的代理。我们发现主导错误类型在代理之间系统性地变化，其中协调者的错误大多与引用相关。我们发现，AI-Q中84.7%的最终报告错误源于协调者，约31%为幻觉，其余为引用错误。在这些见解的指导下，我们展示了两种简单的干预措施可以在不降低输出质量的情况下将引用召回提高5%。
cs.CL / 31 / 2608.24327
Speech-to-SOAP: End-to-End Summarization of Medical Dialogues: KIT@BeTraC 2026
语音到SOAP：医疗对话的端到端摘要：KIT@BeTraC 2026
Abstract
With the advent of Large Language Models and its instruction following capabilities a promising application is the task of summarization. Within this domain of task the extractive sub-task of clinical protocolling has emerged as a topic of particular interest as it can significantly reduce the downtime and protocolling burden of health-care workers thus enabling them to focus on their core work helping humans. A further step towards automation is the direct generation of clinical notes from speech without intermediate transcripts, reducing processing time while preserving information such as coughing or other paralinguistic cues that may be lost in transcript-based systems. To this end, we present KIT's submission to this years BeTraC challenge in the lightweight track. Our main contribution is a scalable data augmentation pipeline that unifies heterogeneous medical dialogue datasets through synthetic speech generation and automatically generated SOAP supervision, enabling robust adaptation of a speech foundation model for end-to-end speech-to-SOAP generation.
Chinese Translation
随着大型语言模型及其遵循指令的能力的出现，摘要生成任务成为一种有前景的应用。在这一领域中，临床协议提取子任务作为一个特别感兴趣的话题浮现出来，因为它可以显著减少医疗工作者的空闲时间和协议负担，从而使他们能够专注于帮助人类的核心工作。朝着自动化的进一步步骤是直接从语音生成临床记录，而无需中间转录，这样可以减少处理时间，同时保留诸如咳嗽或其他可能在基于转录的系统中丢失的副语言线索的信息。为此，我们展示了KIT在今年BeTraC挑战赛轻量级赛道中的提交。我们的主要贡献是一个可扩展的数据增强管道，通过合成语音生成和自动生成的SOAP监督，统一异构的医疗对话数据集，从而实现语音基础模型在端到端语音到SOAP生成中的稳健适应。
cs.CL / 32 / 2608.24335
SteerCheck: Attribution Specificity and Alignment Leakage in Activation-Steering Audits
SteerCheck：激活引导审计中的归因特异性与对齐泄漏
Abstract
Activation steering can change behaviour without establishing that the effect is specific to the intended concept. We introduce SteerCheck, a preregistered attribution audit that matches off-target KL and separates mean, protected-tail, polarity, transfer, and semantic claims. Exact replay of 960 Qwen3-14B interventions reveals complementary limits of common controls: isotropic directions occupy a narrow near-orthogonal region, whereas sign-randomized same-construction directions often retain substantial target alignment. Effect is strongly associated with signed cosine within the sign-randomized family ($\rho=.94$); $25.3\%$ of its draws exceed cosine $.5$, and every draw exceeding the observed mean effect has cosine above $.80$. This alignment leakage does not by itself invalidate a conditional randomization test; it limits what the comparator can distinguish and motivates reporting exchangeability assumptions, a construction diagnostic $A$, and the empirical cosine distribution. The primary Qwen complete gate remains negative because the protected tail fails all families. On independent data, continuous margin transfers only in Qwen and accuracy transfers in no selected cell. Prospectively registered language controls pass the complete gate in Qwen and DeepSeek, while a passing DeepSeek detox comparator rules out categorical separation; all nominal passes are sensitive to $\Gamma=1.10$. Frozen three-rater open-generation evaluation supports factual correction in DeepSeek but not Qwen; the automatic judge fails calibration (macro-F1 $.562$), so null-wide semantic results remain descriptive. SteerCheck makes these conditional and mixed conclusions auditable.
Chinese Translation
激活引导可以在不确认效果特定于预期概念的情况下改变行为。我们引入了 SteerCheck，这是一种预注册的归因审计，匹配非目标的 KL，并分离均值、保护尾、极性、转移和语义声明。对 960 次 Qwen3-14B 干预的精确重播揭示了常见控制的互补限制：各向同性方向占据狭窄的近正交区域，而符号随机化的同构建方向通常保留显著的目标对齐。效果与符号随机化家族中的签名余弦强烈相关（$
ho=.94$）；$25.3\%$ 的抽样超过余弦 $.5$，且每个超过观察到的均值效应的抽样余弦均高于 $.80$。这种对齐泄漏本身并不使条件随机化测试失效；它限制了比较器能够区分的内容，并促使报告可交换性假设、构造诊断 $A$ 和经验余弦分布。主要的 Qwen 完整门仍然为负，因为保护尾在所有家族中均未通过。在独立数据上，连续边际仅在 Qwen 中转移，且在没有选择的单元中准确性转移。前瞻性注册的语言控制在 Qwen 和 DeepSeek 中通过了完整门，而通过的 DeepSeek 解毒比较器排除了类别分离；所有名义通过对 $ ext{Γ}=1.10$ 敏感。冻结的三评审开放生成评估支持 DeepSeek 中的事实修正，但不支持 Qwen；自动评审者未能校准（宏 F1 $.562$），因此无效广泛的语义结果仍然是描述性的。SteerCheck 使这些条件和混合结论可审计。
cs.CL / 33 / 2608.24350
FARCA: Fact-Aligned Reliability-Aware Credit Assignment for Reinforcement Learning with Factual Supervision
FARCA：面向事实的可靠性意识信用分配框架用于带有事实监督的强化学习
Abstract
To reduce the hallucination risk caused by outcome-driven rewards in large language models trained through reinforcement learning with verifiable rewards, existing mitigation approaches introduce process-level factual supervision. However, due to coarse-grained aggregation of factual signals and the lack of reliability assessment for these signals, they create a mismatch between fact verification and policy updates. We term this noisy factual credit assignment and decompose it into two aspects: credit localization ambiguity and credit reliability ambiguity. To address these issues, we propose FARCA (Fact-Aligned Reliability-Aware Credit Assignment), a policy optimization framework that transforms factual supervision into localized, reliability-weighted token-level training signals. FARCA achieves fine-grained credit localization by aligning the granularity of fact verification with that of policy updates. It further introduces counterfactual evidence attribution, which uses the dependence of a factual judgment on key evidence as an empirical proxy for verification reliability to compute reliability weights. These weights modulate factual rewards and local policy advantages, reducing the influence of potentially unreliable signals on policy optimization. Experiments across different models and multiple factual reasoning benchmarks show that FARCA significantly improves model factuality while preserving general reasoning capabilities.
Chinese Translation
为了减少在通过可验证奖励进行训练的大型语言模型中，由于结果驱动的奖励引发的幻觉风险，现有的缓解方法引入了过程层面的事实监督。然而，由于对事实信号的粗粒度聚合以及缺乏对这些信号的可靠性评估，它们在事实验证与策略更新之间造成了不匹配。我们将这种噪声事实信用分配称为，并将其分解为两个方面：信用定位模糊性和信用可靠性模糊性。为了解决这些问题，我们提出了FARCA（面向事实的可靠性意识信用分配），这是一个政策优化框架，将事实监督转化为局部的、可靠性加权的令牌级训练信号。FARCA通过将事实验证的粒度与策略更新的粒度对齐，实现了细粒度的信用定位。它进一步引入了反事实证据归因，利用事实判断对关键证据的依赖性作为验证可靠性的经验代理来计算可靠性权重。这些权重调节事实奖励和局部策略优势，从而减少潜在不可靠信号对策略优化的影响。在不同模型和多个事实推理基准上的实验表明，FARCA显著提高了模型的事实性，同时保持了一般推理能力。
cs.CL / 34 / 2608.24360
Words, Spaces and Generative AI: Layers of language in contemporary architecture
词汇、空间与生成性人工智能：当代建筑中的语言层面
Abstract
Language can be considered a design material in architecture, and in the context of text-to-X generative AI models becoming a common tool for architectural practice, looking more closely at language is more important now than in the past. After describing some of the important developments in linguistics starting from Wittgenstein, and including the work of Chomsky, Lakoff, conceptual and generative metaphors as proposed by Sch\"on, this chapter connects them to contemporary architectural design and generative text-to-X tools. The chapter builds on the idea that three main forms of language intertwine in architectural design done using generative AI, namely (I) discourse (or natural language which can contain professional terminology specific to our field), (II) programming languages (which are artificial languages sitting at the basis of all computational systems), and (III) annotations (as language elements attached to pieces of data). It concludes by outlining a research agenda for connecting generative metaphors to generative AI: (a) conducting corpus linguistics studies on architectural texts (using quantitative tools such as topic modelling, and qualitative tools such as discourse analysis); (b) bringing communication theory and information studies closer to architectural research and (c) taking into account that different (natural) languages come with different affordances meaning generative and conceptual metaphors differ in relation to this.
Chinese Translation
语言可以被视为建筑设计中的一种材料，在文本到X生成性人工智能模型逐渐成为建筑实践常用工具的背景下，深入研究语言的重要性比以往任何时候都更为突出。本文首先描述了自维特根斯坦以来语言学的一些重要发展，包括乔姆斯基、莱考夫的研究，以及施恩提出的概念隐喻和生成隐喻，随后将这些理论与当代建筑设计及生成性文本到X工具相联系。本文基于一个观点，即在使用生成性人工智能进行建筑设计时，三种主要形式的语言相互交织，即（I）话语（或自然语言，其中可以包含我们领域特有的专业术语），（II）编程语言（作为所有计算系统基础的人工语言），以及（III）注释（作为附加于数据片段的语言元素）。最后，本文概述了将生成隐喻与生成性人工智能相连接的研究议程：（a）对建筑文本进行语料库语言学研究（使用主题建模等定量工具和话语分析等定性工具）；（b）将传播理论和信息研究与建筑研究更紧密地结合；（c）考虑到不同（自然）语言具有不同的适应性，这意味着生成隐喻和概念隐喻在这方面存在差异。
cs.CL / 35 / 2608.24460
Shortcut Before Circuit: Document Statistics Time In-Context Conflict Resolution
电路之前的捷径：文档统计时间上下文冲突解决
Abstract
When a context asserts two values for one fact, a model commits to a cue -- recency, repetition, position -- but natural data rarely makes these disagree, so behavior cannot reveal which. We train 26M-parameter transformers on a synthetic language where recency and rarity are exactly coextensive, and separate them with a minimal causal edit that inverts one cue while holding the truth, token count and answer position fixed. All 75 runs reach accuracy >= 0.999, including where the trivial heuristic fails, so no held-in evaluation distinguishes them. Under intervention the per-cell readout does not replicate: 13 of 25 cells differ by more than 0.3 in sign fraction across three seeds, the largest by 0.879 against a standard error of 0.025. The construction predicts this -- coextensive rules leave the objective indifferent between them -- and the variance is ordered by how much of the optimization each comparison releases. What replicates is timing: escape from a positional shortcut with a closed-form ceiling, monotone in redundancy. Probed before that escape, attribution reverses sign in 32 of 75 runs at unchanged accuracy, and gating on circuit formation is necessary but not sufficient. The corpus fixes when a mechanism appears, not which one -- a criterion for when mechanistic attribution to data is available at all, and our construction makes the unavailable case exact.
Chinese Translation
当一个上下文对一个事实断言两个值时，模型会依赖于某个线索——最近性、重复性、位置——但自然数据很少会使这些线索产生冲突，因此行为无法揭示哪一个是正确的。我们在一个合成语言上训练了2600万参数的变换器，其中最近性和稀有性完全重合，并通过一个最小的因果编辑将其中一个线索反转，同时保持真值、标记数量和答案位置不变。所有75次实验的准确率均达到0.999或更高，包括在简单启发式失效的情况下，因此没有任何保留评估能够区分它们。在干预下，每个单元的读出并不重复：在三个种子下，25个单元中有13个的符号分数差异超过0.3，最大差异为0.879，标准误差为0.025。该构造预测了这一点——重合的规则使得目标在它们之间无差别——而方差的排序则取决于每个比较释放了多少优化。可重复的是时间：从具有封闭形式上限的位置信息捷径中逃脱，且在冗余中单调。在逃脱之前进行探测时，在75次实验中有32次的归因符号反转，准确率保持不变，而电路形成的门控是必要但不充分的。语料库确定了机制出现的时间，而不是具体是哪一个——这是对何时可以对数据进行机械归因的标准，而我们的构造使得不可用的情况变得精确。
cs.CL / 36 / 2608.24477
Dataset Scarcity Limits Robust Evaluation of Multilingual Embedding Models: A Case Study of Slavic Languages
数据集稀缺限制了多语言嵌入模型的稳健评估：以斯拉夫语言为例
Abstract
Multilingual text embedding models enable cross-lingual transfer of knowledge across a wide range of NLP tasks, but their evaluation remains highly uneven across high-, mid- and low-resource languages. In this paper, we propose a two-dimensional framework, specifically tailored for analyzing multilingual embedding benchmarks under dataset scarcity, and apply it on the Slavic-language subset of the MTEB benchmark. The framework distinguishes between task-specific and cross-task evaluation, while jointly analyzing three complementary aspects: (1) ranking robustness, (2) model consistency, and (3) evidence strength. At the task-specific level, we evaluate the stability of model rankings under changes in ranking methodology and benchmark dataset composition. At the cross-task level, we assess the ability of models to generalize across diverse tasks within a language. To quantify the reliability of benchmark conclusions, we introduce an Evidence Strength Score that accounts for dataset availability, diversity, and robustness assessability. Our analysis reveals severe benchmark sparsity, with many Slavic language-task pairs relying on a single dataset or highly correlated benchmark collections, limiting the ability to draw robust conclusions. The cross-task analysis reveals a small group of highly transferable models, most notably llama-embed-nemotron-8b, multilingual-e5-large-instruct, and Qwen3-Embedding variants, that consistently perform well across Slavic languages and tasks. Overall, the results demonstrate that benchmark rankings and robustness conclusions must be interpreted jointly with certain notation of their evidence strength and highlight benchmark scarcity as a major obstacle to trustworthy multilingual evaluation.
Chinese Translation
多语言文本嵌入模型能够在广泛的自然语言处理任务中实现跨语言知识转移，但其评估在高资源、中资源和低资源语言之间仍然存在高度不均衡。在本文中，我们提出了一个二维框架，专门用于分析在数据集稀缺情况下的多语言嵌入基准，并将其应用于 MTEB 基准的斯拉夫语言子集。该框架区分了任务特定评估和跨任务评估，同时共同分析三个互补方面：（1）排名稳健性，（2）模型一致性，以及（3）证据强度。在任务特定层面，我们评估了模型排名在排名方法和基准数据集组成变化下的稳定性。在跨任务层面，我们评估了模型在同一语言中跨多样任务的泛化能力。为了量化基准结论的可靠性，我们引入了一个证据强度评分，考虑了数据集的可用性、多样性和稳健性可评估性。我们的分析揭示了严重的基准稀疏性，许多斯拉夫语言-任务对依赖于单一数据集或高度相关的基准集合，从而限制了得出稳健结论的能力。跨任务分析揭示了一小组高度可转移的模型，特别是 llama-embed-nemotron-8b、multilingual-e5-large-instruct 和 Qwen3-Embedding 变体，它们在斯拉夫语言和任务中始终表现良好。总体而言，结果表明基准排名和稳健性结论必须与其证据强度的某种标注共同解读，并强调基准稀缺是可信多语言评估的主要障碍。
cs.CL / 37 / 2608.24521
Beyond Information Seeking: Severity-Aware Question Supervision for Proactive Medical Dialogue
超越信息寻求：针对主动医疗对话的严重性意识问题监督
Abstract
Proactive medical dialogue requires an agent to decide what to ask from incomplete patient information. Existing information-seeking approaches commonly prioritize questions that most reduce diagnostic uncertainty. While effective for acquiring informative evidence, this criterion overlooks an important property of medical diagnosis: different diagnostic errors can carry substantially different consequences. Missing a severe condition may matter more than reducing uncertainty among less consequential alternatives. Question acquisition should therefore consider not only how informative new evidence is, but also how it is expected to affect the downstream diagnostic decision. To this end, we propose Expected-Severity-Risk (ESR), a consequence-aware question-supervision objective that values each candidate by its expected reduction in severity-aware terminal risk. Because questions must be selected before their answers are observed, ESR marginalizes over possible answers using train-only population statistics. Its rankings are then distilled into a prefix-only language policy, so next-question selection requires no teacher-side computation at deployment. Across three Qwen3-4B training seeds on DDxPlus, matched ESR supervision reduces mean high-severity diagnostic miss from .0645 to .0455 (-29.5%) and improves mean diagnostic accuracy from .9123 to .9320 while requiring only 0.14 additional questions per dialogue. Fixed-budget analyses show that the two objectives remain behaviorally distinct when question count is controlled, while a matched expected-0/1-risk control shows that severity-aware weighting improves the high-severity error profile beyond generic decision-aware supervision. These results support moving proactive medical dialogue beyond uncertainty reduction toward consequence-aware evidence acquisition.
Chinese Translation
主动医疗对话要求代理根据不完整的患者信息决定询问内容。现有的信息寻求方法通常优先考虑那些能最大程度减少诊断不确定性的问题。尽管这种方法在获取有用证据方面有效，但这一标准忽视了医疗诊断的一个重要特性：不同的诊断错误可能带来截然不同的后果。漏诊严重疾病的影响可能远大于在不太重要的替代方案中减少不确定性。因此，问题获取不仅应考虑新证据的信息量，还应考虑其对后续诊断决策的预期影响。为此，我们提出了期望严重性风险（Expected-Severity-Risk, ESR），这是一种关注后果的问题监督目标，通过其对严重性意识终极风险的预期减少来评估每个候选问题。由于问题必须在观察到答案之前选择，ESR通过仅使用训练集人口统计数据对可能的答案进行边际化。其排名随后被提炼为仅前缀的语言策略，因此在部署时下一个问题的选择不需要教师端的计算。在DDxPlus上进行的三次Qwen3-4B训练种子实验中，匹配的ESR监督将高严重性诊断漏诊的平均值从0.0645降低到0.0455（下降29.5%），并将平均诊断准确率从0.9123提高到0.9320，同时每次对话仅需额外0.14个问题。固定预算分析显示，当控制问题数量时，这两个目标在行为上仍然是不同的，而匹配的期望0/1风险控制则表明，严重性意识加权在超越通用决策意识监督的同时改善了高严重性错误的表现。这些结果支持将主动医疗对话从减少不确定性转向关注后果的证据获取。
cs.CL / 38 / 2608.24590
Is Discrete Difficulty Sufficient? Leveraging Continuous Difficulty for Efficient Self-Consistency in LLMs
离散难度是否足够？利用连续难度实现大型语言模型中的高效自一致性
Abstract
Self-Consistency (SC) is a decoding strategy that samples diverse reasoning paths and selects the most consistent answer, demonstrating strong performance on complex reasoning problems. However, the excessive token consumption incurred by generating multiple reasoning paths has been identified as a major limitation of SC. To improve computational efficiency, several studies have proposed strategies that adjust the number of reasoning paths or allocate resources differentially according to problem difficulty. Nevertheless, most existing methods categorize difficulty into a few fixed levels, failing to fully capture the continuously varying nature of reasoning complexity. In this work, we propose Flexible Self-Consistency (FSC), which estimates problem difficulty as a continuous signal and dynamically adjusts the number of generated reasoning paths accordingly. FSC predicts the output entropy of an input question using a pre-trained probe and leverages it as an indicator of model uncertainty to flexibly control the sampling budget. Experimental results show that, across various models and benchmarks, FSC maintains accuracy comparable to SC while achieving token savings of up to 76%.
Chinese Translation
自一致性（Self-Consistency, SC）是一种解码策略，通过采样多样的推理路径并选择最一致的答案，在复杂推理问题上表现出强大的性能。然而，生成多个推理路径所带来的过度令牌消耗被认为是SC的一个主要限制。为了提高计算效率，一些研究提出了根据问题难度调整推理路径数量或差异性分配资源的策略。然而，大多数现有方法将难度分为几个固定级别，未能充分捕捉推理复杂性连续变化的特性。在本研究中，我们提出了灵活自一致性（Flexible Self-Consistency, FSC），该方法将问题难度估计为一个连续信号，并相应地动态调整生成的推理路径数量。FSC使用预训练探针预测输入问题的输出熵，并利用其作为模型不确定性的指标，以灵活控制采样预算。实验结果表明，在各种模型和基准测试中，FSC保持了与SC相当的准确性，同时实现了高达76%的令牌节省。
cs.CL / 39 / 2608.24615
Quantization Effects on Bangla Language Understanding in Large Language Models: A Systematic Evaluation
量化对大型语言模型中孟加拉语理解的影响：系统评估
Abstract
Post-training quantization lowers the memory footprint of Large Language Models (LLMs) and speeds up inference, which is why it is now common for on-device deployment. Most of what we know about its effects, however, comes from English benchmarks. It is not clear whether the same holds for morphologically complex, low-resource languages such as Bangla, and this gap is what we address here. We evaluate three model families---Qwen-2.5-7B, LLaMA-3.1-8B, and GPT-OSS-20B---in full precision and in three quantized formats (GPTQ-Int8, GPTQ-Q8, GGUF-W8A16) across five Bangla natural language understanding benchmarks (Bangla MMLU, CommonsenseQA-BN, OpenBookQA-BN, PIQA-BN, and BoolQ-BN), using zero-shot evaluation through lm-evaluation-harness. To our knowledge this is the first controlled comparison of quantization formats on Bangla NLU. The three families do not respond the same way: GPT-OSS loses up to 57.35% accuracy on reasoning-heavy tasks under GGUF-W8A16, while Qwen and LLaMA hold steady under GPTQ, and in a few cases the quantized version edges out the full-precision one. BoolQ-BN, a comprehension task, stays stable across all three families regardless of format. Taken together, these results suggest quantization can work well for Bangla deployment, but the choice of architecture and quantization method matters more than the bit width alone. We discuss what this means for practitioners choosing a model to run on constrained hardware.
Chinese Translation
后训练量化降低了大型语言模型（LLMs）的内存占用并加快了推理速度，因此在设备上部署已变得普遍。然而，我们对其影响的了解大多来自英语基准测试。目前尚不清楚这一点是否同样适用于形态复杂、资源匮乏的语言，如孟加拉语，而这正是我们在此研究的内容。我们在五个孟加拉自然语言理解基准（Bangla MMLU、CommonsenseQA-BN、OpenBookQA-BN、PIQA-BN 和 BoolQ-BN）上评估了三种模型系列——Qwen-2.5-7B、LLaMA-3.1-8B 和 GPT-OSS-20B——在全精度和三种量化格式（GPTQ-Int8、GPTQ-Q8、GGUF-W8A16）下的表现，使用 lm-evaluation-harness 进行零-shot 评估。据我们所知，这是对孟加拉 NLU 中量化格式的首次控制比较。这三种模型系列的反应并不相同：在 GGUF-W8A16 下，GPT-OSS 在推理密集型任务中准确率下降了多达 57.35%，而 Qwen 和 LLaMA 在 GPTQ 下保持稳定，在少数情况下，量化版本的表现优于全精度版本。作为理解任务的 BoolQ-BN 在所有三种系列中无论格式如何均保持稳定。综合来看，这些结果表明量化在孟加拉语部署中可以良好运行，但架构选择和量化方法比位宽本身更为重要。我们讨论了这对选择在受限硬件上运行的模型的从业者意味着什么。
cs.CL / 40 / 2608.24621
Beyond Semantic Accuracy: Consequence-Aware Evaluation for Safety-Critical Language Understanding
超越语义准确性：面向后果的安全关键语言理解评估
Abstract
Can language models be trusted in safety- critical operations? In such settings, strong per- formance on semantic metrics does not guaran- tee operational reliability: a misread altitude, a dropped execution condition, or a confused call- sign may score well under standard F1 yet carry sharply asymmetric operational consequences. We study this problem in air traffic control (ATC), where controller-pilot communication demands near-zero error tolerance, and use consequence-aware evaluation to test whether semantic scores misstate operational reliabil- ity. The framework is instantiated in a con- trolled diagnostic ATC benchmark grounded in aviation standards and feedback from 40 air traffic controllers across three countries. Evaluating 8 models, we uncover a system- atic semantic-safety gap: conventional scores give substantially higher performance estimates than consequence-aware evaluation, even for models that appear reliable under standard met- rics. Risk-aware fine-tuning narrows but does not close this gap, showing that consequence- aware evaluation is a necessary complement to standard NLP metrics before any real safety- critical deployment claim
Chinese Translation
语言模型在安全关键操作中可以被信任吗？在这种环境下，语义指标的强表现并不保证操作的可靠性：错误的高度读取、遗漏的执行条件或混淆的呼号在标准 F1 下可能得分良好，但却可能带来严重不对称的操作后果。我们在空中交通管制（ATC）中研究这个问题，在这里，控制员与飞行员的沟通要求几乎零错误容忍，并使用面向后果的评估来测试语义得分是否错误地表述了操作的可靠性。该框架在一个受控的诊断 ATC 基准测试中得以实现，该基准基于航空标准和来自三个国家的 40 名空中交通管制员的反馈。评估 8 个模型后，我们发现了一个系统性的语义安全差距：传统得分给出的性能估计显著高于面向后果的评估，即使对于在标准指标下看似可靠的模型，风险意识的微调缩小了但并未消除这一差距，表明面向后果的评估是在任何真实的安全关键部署声明之前对标准自然语言处理指标的必要补充。
cs.CL / 41 / 2608.24654
Expectation, Backlash, Recovery, and Excitement: How Model Releases Shape Reddit Perceptions of Conversational AI Systems
期望、反弹、恢复与兴奋：模型发布如何塑造Reddit对对话式人工智能系统的认知
Abstract
Conversational AI systems (CAISes) continuously change through model releases, feature updates, safety interventions, and access-policy shifts, yet user perceptions are often studied as static snapshots. We conduct a long-term, large-scale analysis of Reddit discussions to examine how users perceive CAIS model release interventions across providers. By combining sentiment classification and thematic concept analysis, we show that CAIS perceptions are dynamic and intervention-sensitive. Anthropic exhibits the clearest positive release profile through Claude Code and product-model fit, OpenAI shows backlash-and-recovery dynamics around GPT-5 and GPT-5.1, Grok-3 is shaped by provider identity and political discourse, and DeepSeek-R1 combines engineering praise with concerns about censorship, access, and reliability. These findings show that model releases are not merely technical updates, but user-facing interventions that reshape sentiment, expectations, and public discussion.
Chinese Translation
对话式人工智能系统（CAISes）通过模型发布、功能更新、安全干预和访问政策变化不断演变，但用户的认知往往被视为静态快照。我们对Reddit讨论进行了长期的大规模分析，以考察用户如何看待不同提供商的CAIS模型发布干预。通过结合情感分类和主题概念分析，我们展示了CAIS的认知是动态的且对干预敏感。Anthropic通过Claude Code和产品模型契合展现出最明显的正面发布特征，OpenAI在GPT-5和GPT-5.1周围表现出反弹与恢复的动态，Grok-3则受到提供商身份和政治话语的影响，而DeepSeek-R1则将工程赞誉与对审查、访问和可靠性的担忧结合在一起。这些发现表明，模型发布不仅仅是技术更新，而是面向用户的干预，重新塑造了情感、期望和公共讨论。
cs.CL / 42 / 2608.24660
From local kernels to global form: modeling the emergence of musical content
从局部核到全局形式：音乐内容的生成建模
Abstract
Markov models are established tools for symbolic music, including non-homogeneous formulations. The narrower contribution examined here is an observation-driven estimation mechanism: overlapping sliding windows derive a trajectory of local transition kernels from one symbolic sequence rather than from an exogenous formal partition. We test this mechanism on 273 logical note events from Debussy's Syrinx (1913), using the often-proposed A-B-A' reading as a reference rather than ground truth. We apply the same validation to absolute-pitch and notated-duration kernels. At $L=6$, both reference boundaries attain the Jensen--Shannon maximum in both dimensions; the duration plateau is substantially narrower (64 of 267 comparisons) than the pitch plateau (210 of 267). Because the theoretical maximum for consecutive sliding-window comparisons is set by window geometry and equals $1/\sqrt{L-1}$ for maximal turnover of the entering/leaving transition, the pitch value at $L=6$ and its broad plateau are not, by themselves, strong evidence. Their cross-dimensional alignment is consistent with boundary sensitivity, while the broad plateaus preclude treating either curve alone as a unique automatic segmenter. Five-hundred-draw re-synthesis experiments quantify departure from the source in both dimensions and expose an exact-copy degeneracy at $L=2$.
Chinese Translation
马尔可夫模型是符号音乐的成熟工具，包括非均匀的表述。本文考察的较窄贡献是一个基于观察的估计机制：重叠滑动窗口从一个符号序列中推导出局部转移核的轨迹，而不是从外生的形式划分。我们在273个来自德彪西《西尔克斯》（Syrinx, 1913）的逻辑音符事件上测试该机制，使用常被提议的A-B-A'阅读作为参考，而非真实的标准。我们对绝对音高和标记时值的核应用相同的验证。在$L=6$时，两个参考边界在两个维度上均达到詹森-香农最大值；而时值平台显著比音高平台窄（267次比较中有64次），音高平台则有210次。由于连续滑动窗口比较的理论最大值由窗口几何形状设定，并且在进入/离开转移的最大周转时等于$1/ ext{sqrt}(L-1)$，因此在$L=6$时的音高值及其宽广平台本身并不是强有力的证据。它们的跨维度对齐与边界敏感性一致，而宽广平台则排除了将任一曲线单独视为独特的自动分段器。五百次抽样重合成实验量化了在两个维度上与源的偏离，并在$L=2$时揭示了精确复制的退化现象。
cs.CL / 43 / 2608.24683
One Timeline, Many Renderings: A Wolfram Language Paclet for heterogeneous musical output
一个时间线，多种表现：用于异构音乐输出的Wolfram语言Paclet
Abstract
One algorithmic composition may require a Csound score, engraved notation, real-time control, and a rehearsal click. Authored separately, their timelines drift. Temporal System is a Wolfram Language paclet that instead compiles one immutable store of typed entities on a rational beat timeline through backend-specific contracts. It emits Csound synthesis, beta MusicXML 4.0, OSC control, and click artifacts that remain synchronized because they share that store. Conversion to seconds, samples, or hertz occurs only at render time. Csound notes use stable named instruments in external .orc files; curves become k-rate signals declared against score p-fields. The click backend derives rehearsal audio from the same meter and tempo and reuses the Csound serializer. We describe the temporal, semantic, and rendering-contract layers, their practical trade-offs, and the limits of this proprietary authoring environment within an otherwise open-source ecosystem. The archived supplement exposes the reported outputs pending paclet release.
Chinese Translation
一种算法作曲可能需要Csound乐谱、雕刻乐谱、实时控制和排练点击音。它们分别创作，时间线因此漂移。Temporal System是一个Wolfram语言Paclet，它通过后端特定的合约在一个理性节拍时间线上编译一个不可变的类型实体存储。它发出Csound合成、beta MusicXML 4.0、OSC控制和点击音效，这些输出保持同步，因为它们共享该存储。转换为秒、样本或赫兹仅在渲染时发生。Csound音符使用外部.orc文件中的稳定命名乐器；曲线成为针对乐谱p字段声明的k-rate信号。点击后端从相同的节拍和速度中派生排练音频，并重用Csound序列化器。我们描述了时间、语义和渲染合约层，它们的实际权衡，以及在其他开源生态系统中这一专有创作环境的局限性。归档的补充材料揭示了待发布Paclet的报告输出。
cs.CL / 44 / 2608.24698
The Annotation Bottleneck in Persian Text NLP: Persian as an Annotation-Scarce Language
波斯文本自然语言处理中的注释瓶颈：波斯语作为一种注释稀缺语言
Abstract
Persian (Farsi) is often described as a low-resource language in natural language processing, but that label collapses distinct shortages into a single category. This paper argues that Persian is more precisely described as annotation-scarce, provided that the term is understood as a property of its NLP resource ecology rather than an intrinsic property of the language. The review covers 34 representative Persian text resources available by July 2026 and adds three quantitative cross-checks. First, independent web measurements place Persian among roughly the twenty most visible content languages: W3Techs reports Persian on about 0.9% of websites with a known content language, while Common Crawl CC-MAIN-2026-30 identifies Persian as the primary language of 0.7039% of HTML pages. Second, a selective speech review shows a long resource trajectory from FARSDAT to recent corpora containing hundreds or thousands of hours of speech. Third, a matched Persian-English comparison normalizes task-specific annotation volumes by relative Common Crawl web presence. The resulting ratios vary sharply: Persian syntax and news NER are comparatively dense, whereas natural-language inference falls below the web-proportional baseline. The evidence therefore does not support a simple claim that Persian is globally deficient in labeled volume. Instead, annotation scarcity is expressed through uneven task and domain coverage, incompatible schemes, access and documentation friction, and limited supervision for specialist domains, preference data, and varieties beyond standard Iranian Persian.
Chinese Translation
波斯语（法尔西语）常被描述为一种在自然语言处理领域资源匮乏的语言，但这一标签将不同的短缺归结为一个单一类别。本文认为，波斯语更准确地被描述为注释稀缺，前提是该术语被理解为其自然语言处理资源生态的一个特性，而非语言的内在特性。本文回顾了截至2026年7月可用的34个代表性波斯文本资源，并增加了三项定量交叉验证。首先，独立的网络测量将波斯语置于大约二十种最具可见性的内容语言之中：W3Techs报告称波斯语出现在约0.9%的已知内容语言的网站上，而Common Crawl CC-MAIN-2026-30则将波斯语识别为0.7039% HTML页面的主要语言。其次，选择性语音回顾显示了从FARSDAT到最近包含数百或数千小时语音的语料库的长期资源轨迹。第三，匹配的波斯语-英语比较通过相对的Common Crawl网络存在量对特定任务的注释量进行了标准化。结果比率变化显著：波斯语的句法和新闻命名实体识别相对密集，而自然语言推理则低于网络比例基线。因此，证据并不支持波斯语在标注量上全球性不足的简单说法。相反，注释稀缺通过不均匀的任务和领域覆盖、不兼容的方案、访问和文档摩擦，以及对专业领域、偏好数据和超出标准伊朗波斯语的变体的有限监督得以体现。
cs.CL / 45 / 2608.24707
Lost in Speech: Trilingual Spoken Hallucination Detection Across Audio and Transcripts
迷失在语言中：跨音频和文本的三语口语幻觉检测
Abstract
While text-based hallucination detection has been extensively studied, spoken hallucination detection remains largely unexplored, particularly for low-resource languages. We present the first multilingual spoken hallucination benchmark comprising 12,013 news samples across English, Russian, and Kazakh with controlled hallucinations of three types and three severity levels. Samples comprise original articles and aligned hallucinated counterparts in text and audio. We complement the synthetic corpus with 290 fact-checked fake news items collected natively in Russian (225) and Kazakh (65), translated into the other language and rendered through the same TTS-ASR pipeline. We assess fine-tuned multilingual encoders and, in zero-shot in-context settings, multimodal decoder models on transcript-based versus direct audio processing. Transcript-based detection generally outperforms direct audio processing, with binary-task degradation for strong encoders tracking per-language ASR error. On real-world fakes, synthetic-trained detectors transfer strongly (macro-F1 0.82-0.88 on original text), while Russian provenance analysis reveals both veracity-related and model-dependent machine-style signals, quantifying a key confound in synthetic hallucination benchmarks.
Chinese Translation
尽管基于文本的幻觉检测已被广泛研究，但口语幻觉检测仍然在很大程度上未被探索，特别是在低资源语言中。我们提出了第一个多语言口语幻觉基准，包含12,013个新闻样本，涵盖英语、俄语和哈萨克语，控制了三种类型和三种严重程度的幻觉。样本包括原始文章及其在文本和音频中的对齐幻觉对应物。我们用290个经过事实核查的假新闻项目来补充合成语料库，这些项目在俄语（225个）和哈萨克语（65个）中本地收集，并翻译成另一种语言，通过相同的TTS-ASR管道呈现。我们评估了微调的多语言编码器，并在零样本上下文设置中，比较了基于文本的检测与直接音频处理的多模态解码器模型。基于文本的检测通常优于直接音频处理，强编码器在每种语言的ASR错误跟踪中表现出二元任务的降级。在真实世界的假新闻中，合成训练的检测器表现出强大的迁移能力（原始文本的宏F1为0.82-0.88），而俄语来源分析揭示了与真实性相关的信号和模型依赖的机器风格信号，量化了合成幻觉基准中的一个关键混淆因素。
cs.CL / 46 / 2608.24726
Arbitrary Polygon Oscillator: Generalizing Polygonal Synthesis to Arbitrary Shapes, Morphing, and Three-Dimensional Polyhedra
任意多边形振荡器：将多边形合成推广至任意形状、变形和三维多面体
Abstract
Polygonal synthesis generates audio by traversing the perimeter of a polygon with a phasor; prior work uses a constant angular velocity, whereas the proposed system adopts constant arc-length (perimeter) velocity. Existing formulations operate on regular, parametrically defined polygons, producing smooth timbral transitions within a single family of shapes. This paper generalizes polygonal synthesis around a unified arc-length engine: vertex data of any origin feed the same DSP pipeline. First, we adapt the oscillator to accept arbitrary vertex configurations from an external buffer, opening the possibility for a broad class of closed polygons -- regular, irregular, or star-shaped -- to function as a waveform generator. Second, a hybrid interpolation algorithm enables smooth morphing between polygons with unequal vertex counts, passing through intermediate shapes that have no parametric description. Third, we extend the paradigm to three dimensions: a convex polyhedron rotated about three axes is sliced by a fixed horizontal plane, and the resulting cross-section yields a continuously variable polygon controlled by the solid's orientation. The system runs in RNBO (Cycling~'74) with a geometry caching strategy that avoids per-sample recomputation. Antialiasing combines a four-point polyBLAMP correction derived from runtime B\'ezier tangents with adaptive oversampling, adapting the correction geometrically to general vertex configurations without per-shape analytical derivation.
Chinese Translation
多边形合成通过用相位器遍历多边形的周长来生成音频；以往的工作使用恒定的角速度，而所提议的系统采用恒定的弧长（周长）速度。现有的公式适用于规则的、参数化定义的多边形，在单一形状家族内产生平滑的音色过渡。本文围绕统一的弧长引擎对多边形合成进行了推广：来自任何来源的顶点数据都输入同一个数字信号处理（DSP）管道。首先，我们调整振荡器以接受来自外部缓冲区的任意顶点配置，从而使得广泛类别的封闭多边形——规则的、不规则的或星形的——能够作为波形生成器。其次，一种混合插值算法使得在顶点数量不等的多边形之间实现平滑变形，经过中间形状而无需参数描述。第三，我们将这一范式扩展到三维：一个围绕三个轴旋转的凸多面体被一个固定的水平面切割，产生的横截面生成一个由固体的方向控制的连续可变多边形。该系统在RNBO（Cycling~'74）中运行，采用几何缓存策略以避免逐样本的重新计算。抗锯齿结合了基于运行时Bézier切线推导的四点polyBLAMP校正与自适应过采样，几何上调整校正以适应一般顶点配置，而无需对每种形状进行解析推导。
cs.CL / 47 / 2608.24747
SkillForge: Evolving Verifiable Skills for Reinforcement Learning Agents
SkillForge：为强化学习代理演化可验证技能
Abstract
Large language model (LLM) agents are trained with reinforcement learning (RL) for complex decision-making tasks. However, most RL-trained agents remain episodic and cannot accumulate reusable knowledge across episodes. Recent skill-based approaches, such as SkillRL, attempt to address this issue by extracting skills from raw trajectories, but treat the skill bank as an append-only repository without verifying whether stored skills remain effective. In this paper, we propose SkillForge, a framework for continuous skill evolution that enables skills to be verified and refined through environment interaction. By making skill usage explicit during agent interaction, RL can directly optimize both environment actions and skill invocation decisions. SkillForge further introduces evidence-based skill verification and multi-pathway skill induction, allowing the skill bank to continuously grow while maintaining its quality. Extensive experiments on ALFWorld, WebShop, and AppWorld show that SkillForge consistently outperforms SkillRL, demonstrating the effectiveness of continuously verified skills in training stronger LLM agents.
Chinese Translation
大型语言模型（LLM）代理通过强化学习（RL）训练以应对复杂决策任务。然而，大多数经过RL训练的代理仍然是情节性的，无法在情节之间积累可重用的知识。最近的基于技能的方法，如SkillRL，试图通过从原始轨迹中提取技能来解决这一问题，但将技能库视为仅可附加的存储库，而未验证存储的技能是否仍然有效。本文提出了SkillForge，一个用于连续技能演化的框架，使技能能够通过环境交互进行验证和精炼。通过在代理交互过程中明确技能的使用，RL可以直接优化环境动作和技能调用决策。SkillForge进一步引入基于证据的技能验证和多路径技能引导，使技能库能够在保持质量的同时持续增长。在ALFWorld、WebShop和AppWorld上的大量实验表明，SkillForge始终优于SkillRL，证明了持续验证技能在训练更强LLM代理中的有效性。
cs.CL / 48 / 2608.24753
The RAT: A Unified Bayesian Model for RAG Evaluation
RAT：一种统一的贝叶斯模型用于检索增强生成（RAG）评估
Abstract
Evaluating Retrieval-Augmented Generation (RAG) systems requires assessing not only end-to-end correctness but also how individual components interact and how errors propagate through the pipeline. We introduce a Bayesian evaluation framework that jointly models retrieval success, abstention behavior, and answer correctness, factorized according to the pipeline's information flow. The model distinguishes task success. Whether the user received a correct answer (from generator success) and whether the generator behaved appropriately given the retrieval outcome. We apply the framework to 27 RAG configurations across three datasets, three retrievers, and three generators, and show that the conditional decomposition reveals substantial behavioral differences between systems that appear equivalent under marginal metrics. We further analyze the annotation allocation problem, demonstrating that retrieval-success annotations are more informative than task-success annotations for estimating policy adherence, and provide an information-theoretic explanation for this asymmetry. Finally, we extend the model to incorporate LLM-as-a-judge annotations as calibrated noisy observations, enabling practitioners to combine limited human judgments with cheaper automated assessments within a unified probabilistic model.
Chinese Translation
评估检索增强生成（RAG）系统不仅需要评估端到端的正确性，还需要考察各个组件之间的相互作用以及错误在管道中的传播。我们提出了一种贝叶斯评估框架，该框架联合建模检索成功、放弃行为和答案正确性，并根据管道的信息流进行因子分解。该模型区分了任务成功，即用户是否收到了正确的答案（来自生成器的成功）以及生成器在给定检索结果时是否表现得当。我们将该框架应用于三个数据集中的27种RAG配置、三种检索器和三种生成器，并展示条件分解揭示了在边际指标下看似等效的系统之间存在显著的行为差异。我们进一步分析了注释分配问题，证明检索成功注释在估计策略遵循方面比任务成功注释更具信息量，并为这种不对称性提供了信息论解释。最后，我们扩展模型以纳入作为评判者的LLM注释，作为经过校准的噪声观察，使从业者能够在统一的概率模型中结合有限的人类判断与更便宜的自动评估。
cs.CL / 49 / 2608.24760
ExpConCAD: Experience-Guided Text-to-CAD Generation from Shape Descriptions with Implicit Spatial Constraints
ExpConCAD：基于经验指导的形状描述文本到CAD生成，隐含空间约束
Abstract
Text-to-CAD aims to generate executable CAD programs from natural-language descriptions. However, real-world descriptions are often underspecified and omit critical spatial constraints required for valid CAD construction, a challenge that has been largely overlooked by existing methods. In this paper, we argue that missing spatial constraints should be inferred with respect to the underlying construction structure and informed by reusable design experience. Based on this insight, we propose ExpConCAD, an experience-enhanced framework for implicit spatial constraint completion. ExpConCAD first recovers the intended construction structure and constraint scopes, then retrieves relevant constraint-completion experience for similar scopes to complete the missing spatial constraints, and finally generates executable CadQuery programs. Extensive experiments demonstrate the effectiveness of ExpConCAD and provide insights into the role of construction structure understanding and experience memory in spatial constraint completion. Our code is available at: https://github.com/Hotjiashell/ExpConCAD.
Chinese Translation
文本到CAD旨在从自然语言描述生成可执行的CAD程序。然而，现实世界中的描述往往不够具体，遗漏了有效CAD构建所需的关键空间约束，这是现有方法在很大程度上忽视的挑战。本文认为，缺失的空间约束应根据潜在的构建结构进行推断，并通过可重用的设计经验进行指导。基于这一见解，我们提出了ExpConCAD，一个增强经验的隐含空间约束补全框架。ExpConCAD首先恢复预期的构建结构和约束范围，然后检索与相似范围相关的约束补全经验，以补全缺失的空间约束，最后生成可执行的CadQuery程序。大量实验表明了ExpConCAD的有效性，并提供了对构建结构理解和经验记忆在空间约束补全中作用的见解。我们的代码可在以下链接获取：https://github.com/Hotjiashell/ExpConCAD。
cs.CL / 50 / 2608.24780
Linear Probing Provides Robust and Efficient Detection of Machine-Generated Text
线性探测提供了对机器生成文本的稳健和高效检测
Abstract
Distinguishing machine-generated text (MGT) from human-written text (HWT) becomes increasingly important due to potential misuse. However, most supervised detectors often degrade out-of-domain (OOD) and require large, diverse training sets. In this work, we analyze the linearity and quality of MGT representations and show that simple linear probes outperform a wide range of detectors while being substantially more sample-efficient. We first show that MGT and HWT latent representations are linearly separable in low-dimensional space, and provide a plausible explanation for this separability through systematic differences in their representation quality. Motivated by these insights, we train two variants of simple linear probes and evaluate them across 4 benchmarks against 16 baselines. Probes consistently improve OOD detection (+11 AUC), requiring solely ${<}100$ samples to reach near-peak performance. We show that this transferability arises because probes recover a shared latent MGT direction that generalizes across diverse settings. Finally, we demonstrate that probing vectors capture a continuous spectrum of ``machineness'', highlighting their potential for fine-grained estimation of AI-edited text. Overall, our work provides insights into latent-space differences between MGT and HWT and demonstrates the potential of linear probes as as robust and sample-efficient MGT detectors. We release our code on~\href{https://github.com/gerritq/mgt_probes}{github}.
Chinese Translation
由于潜在的误用，区分机器生成文本（MGT）和人类撰写文本（HWT）变得越来越重要。然而，大多数监督检测器在域外（OOD）表现往往下降，并且需要大量多样化的训练集。在本研究中，我们分析了MGT表示的线性和质量，并表明简单的线性探测器在性能上优于多种检测器，同时在样本效率上显著更高。我们首先展示了MGT和HWT的潜在表示在低维空间中是线性可分的，并通过它们的表示质量的系统性差异提供了这一可分性的合理解释。基于这些见解，我们训练了两种简单线性探测器的变体，并在4个基准测试中与16个基线进行评估。探测器在OOD检测中始终提高了性能（+11 AUC），仅需少于100个样本即可达到接近最佳性能。我们表明，这种可迁移性源于探测器恢复了一个共享的潜在MGT方向，该方向在多样化的设置中具有泛化能力。最后，我们展示了探测向量捕捉到了一种连续的“机器性”光谱，突显了它们在精细估计AI编辑文本方面的潜力。总体而言，我们的工作提供了MGT和HWT之间潜在空间差异的见解，并展示了线性探测器作为稳健且样本高效的MGT检测器的潜力。我们将在~ exttt{https://github.com/gerritq/mgt_probes}上发布我们的代码。
cs.CL / 51 / 2608.24809
Structurally-bounded Agentic Graph Exploration for Evidence-Grounded Scholarly DeepSearch
结构性约束的代理图探索用于基于证据的学术深度搜索
Abstract
We present Crase, a bounded and inspectable alternative to deep research agents for scholarly search. Instead of an open-ended search loop, Crase queries a search engine once for seed papers, expands them along their 1.5-hop citation neighborhood, prunes citation edges whose claims lack entailment support, and ranks the remaining papers with a recency-aware random walk. This makes the candidate set, the reason each paper is kept, and the stopping condition explicit and fixed before inference. On LitSearch and one further benchmarks over a 500K-paper arXiv corpus, Crase outperforms deep research agents built on proprietary models by up to 3$\times$ recall@50 at roughly a third of the cost.
Chinese Translation
我们提出了Crase，一种有界且可检查的替代方案，用于学术搜索中的深度研究代理。Crase并不是一个开放式的搜索循环，而是一次性查询搜索引擎以获取种子论文，沿着其1.5跳引用邻域进行扩展，修剪那些缺乏蕴含支持的引用边，并通过考虑时效性的随机游走对剩余论文进行排序。这使得候选集、每篇论文被保留的理由以及停止条件在推理之前变得明确且固定。在LitSearch和另外一个基于50万篇arXiv论文的基准测试中，Crase的表现超过了基于专有模型构建的深度研究代理，回忆率@50提高了多达3倍，而成本大约仅为其三分之一。
cs.CL / 52 / 2608.24842
Reading Is Not Using: Retrieval, Judgment, and the Design of AI Financial Research Workflows
阅读并非使用：检索、判断与人工智能金融研究工作流程的设计
Abstract
Large language models (LLMs) are increasingly deployed as AI analysts to process financial disclosures and support AI-assisted investment decisions. Yet such systems are usually evaluated by what they can retrieve, not whether retrieved information affects their judgments. We identify a retrieval-integration gap in long-context financial analysis. Holding focal-firm information fixed and varying only unrelated context from 2,000 to 128,000 tokens, we find that a risk disclosure's influence on investment judgments falls to the experimental noise floor even as direct retrieval remains accurate. The pattern replicates across model families and judgment tasks and in experiments removing real disclosures from actual 10-K filings. More capable models postpone but do not eliminate the gap. Causal memory interventions show that compressed summaries and source-text lookup jointly transmit disclosures into judgments. Workflow architecture determines whether this transmission succeeds: chunk-and-summarize pipelines evict relevant information, whereas a targeted, structured restatement adjacent to the decision restores its influence. AI analyst performance is therefore jointly determined by model capability and workflow architecture. Retrieval-based evaluations can certify systems whose investment judgments ignore information they demonstrably retrieved.
Chinese Translation
大型语言模型（LLMs）越来越多地被作为人工智能分析师来处理财务披露，并支持人工智能辅助的投资决策。然而，这些系统通常是通过它们能够检索到的信息来评估，而不是检索到的信息是否影响了它们的判断。我们在长上下文的金融分析中识别出一种检索-整合差距。在固定焦点公司信息的情况下，仅改变与之无关的上下文，从2000个到128000个标记，我们发现风险披露对投资判断的影响降至实验噪声底线，即使直接检索仍然准确。该模式在不同模型家族和判断任务中复制，并在移除实际10-K文件中的真实披露的实验中得到验证。更强大的模型虽然延迟了这一差距，但并未消除。因果记忆干预表明，压缩摘要和源文本查找共同将披露信息传递到判断中。工作流程架构决定了这一传递是否成功：分块和摘要管道会排除相关信息，而与决策相邻的有针对性的结构化重述则恢复了其影响。因此，人工智能分析师的表现是由模型能力和工作流程架构共同决定的。基于检索的评估可以认证那些投资判断忽视其明显检索信息的系统。
cs.CL / 53 / 2608.24848
BrowserForge: Scaling Web Episode via Parallel Browser Sandboxes
BrowserForge：通过并行浏览器沙箱扩展网络交互
Abstract
Web agents that act from rendered pixels avoid the fragility and heavy token cost of reading a page's HTML or accessibility tree, but training them depends on large amounts of high-quality interaction trajectories, and how to produce such data at scale remains an open problem. Public datasets typically contain only a few thousand trajectories drawn from a fixed and narrow set of websites, and even recent automated synthesis pipelines stay bound to predefined site lists or tutorial sources, so the number of distinct websites the agent ever sees barely grows. We present BrowserForge, a framework that generates web interaction data at scale by driving many browser sandboxes in parallel over the open web. BrowserForge couples three components: an open-web sourcing stage that exposes the agent to hundreds of thousands of real, openly reachable websites; a sandbox cluster manager that schedules hundreds of concurrent browsers with high utilization; and a Proposer-Solver dual-agent loop that turns a raw page into an executable task and then collects a verified trajectory for it. A rule-plus-model cleaning pipeline removes failed runs and rewrites the surviving reasoning into a single unified chain-of-thought style. Page structure such as the accessibility tree is used only as a synthesis-time signal; the agent we train and release acts purely from the screenshot. The resulting corpus contains 203,238 trajectories, each collected from a distinct website, larger and more diverse than prior trajectory datasets. Fine-tuning a compact multimodal model on this corpus raises its success rate on the live Online-Mind2Web from 25.66% to 33.33% and consistently improves step accuracy on the static Multimodal-Mind2Web, with the gain growing as the corpus scales. Controlled analyses further confirm that open-web sourcing and broad website coverage are key contributors to the observed improvement.
Chinese Translation
从渲染像素中行动的网络代理避免了读取页面 HTML 或可访问性树的脆弱性和高昂的令牌成本，但训练它们依赖于大量高质量的交互轨迹，而如何大规模生成此类数据仍然是一个未解问题。公共数据集通常仅包含从固定且狭窄的网站集合中提取的几千条轨迹，即使是最近的自动合成管道也仍然受限于预定义的网站列表或教程来源，因此代理所见的不同网站数量几乎没有增长。我们提出了 BrowserForge，一个通过在开放网络上并行驱动多个浏览器沙箱来大规模生成网络交互数据的框架。BrowserForge 结合了三个组件：一个开放网络源阶段，使代理能够接触到数十万个真实、可公开访问的网站；一个沙箱集群管理器，调度数百个高利用率的并发浏览器；以及一个 Proposer-Solver 双代理循环，将原始页面转化为可执行任务，并为其收集经过验证的轨迹。一个规则加模型的清理管道移除失败的运行，并将幸存的推理重写为单一统一的思维链。页面结构（如可访问性树）仅在合成时作为信号使用；我们训练和发布的代理完全基于屏幕截图进行操作。最终生成的语料库包含 203,238 条轨迹，每条均来自不同的网站，规模和多样性均超过了先前的轨迹数据集。在该语料库上微调一个紧凑的多模态模型，使其在实时的 Online-Mind2Web 上的成功率从 25.66% 提升至 33.33%，并在静态的 Multimodal-Mind2Web 上持续提高步骤准确性，随着语料库的扩展，增益不断增加。控制分析进一步确认，开放网络源和广泛的网站覆盖是观察到的改进的关键因素。