cs.RO / 1 / 2608.26239
WALL-SS: Scaling Long-horizon World Models via Next-Scale Autoregression
WALL-SS：通过下一尺度自回归扩展长时间视野世界模型
Zhang, Maeve, Sun, Rain, Wang, Xiang, Zhang, Cyril, Li, Shalfun, Cao, Meng, Lu, Howard, Chen, Ethan, Jhou, Harry, Zheng, KZ, Shi, Lights, Cheng, Regis, Lorenzin, Wang, Robert, Yao, Victor, Li, Gody, Mon, Elise, Tang, Yohann, Yu, Ryan, Zhang, PS, Chen, Vincent, Su, Hang, Gan, Roy, Wang, Hao, Wang, Qian
Abstract
Generative world models provide robots with predictive models of how the world evolves under interaction, with growing potential for simulation, planning, policy evaluation, and robot learning. Beyond clip-level future prediction, a unified generative formulation should relate actions to consequences, support flexible horizons and continuous interaction, and enable reward-driven optimization. We introduce WALL-SS, a world model that generates visual futures through Scale-wise autoregressive Scaling, enabling action-controllable and long-horizon robotic simulation. WALL-SS represents embodied trajectories as causal sequences of temporally interleaved observations and actions, making action-dependent state transitions explicit while naturally supporting variable-length generation, streaming extension through reusable causal states, and direct optimization through sequence probabilities. To make this formulation effective over long horizons, we generate each future observation in a coarse-to-fine manner and develop three complementary components within the same hierarchy. Action-conditioned next-scale prediction injects scale-aligned action representations to improve action-future coupling and model both successful and failed behaviors. Scale-compressed long-horizon memory retains recent interactions at fine resolution while compressing distant observations and actions, with scale-wise dream forcing enhancing robustness to self-generated context. Finally, on-policy alignment optimizes autoregressive visual dynamics with action-following and long-term consistency rewards while preserving the pretrained visual distribution. Experiments show that WALL-SS improves action following and trajectory accuracy, supports coherent minute-long streaming rollout under bounded memory, and consistently benefits from on-policy alignment in reducing action drift and long-horizon inconsistency.
Chinese Translation
生成性世界模型为机器人提供了预测模型，描述了在交互下世界如何演变，具有越来越大的潜力用于模拟、规划、策略评估和机器人学习。超越剪辑级别的未来预测，统一的生成性公式应将动作与后果联系起来，支持灵活的时间视野和持续的交互，并实现基于奖励的优化。我们引入了WALL-SS，这是一种通过尺度自回归扩展生成视觉未来的世界模型，使得机器人模拟可以控制动作并具备长时间视野。WALL-SS将具身轨迹表示为因果序列，这些序列由时间交错的观察和动作组成，使得依赖于动作的状态转变变得明确，同时自然支持可变长度生成、通过可重用因果状态进行流式扩展，以及通过序列概率进行直接优化。为了使这一公式在长时间视野上有效，我们以粗到细的方式生成每个未来观察，并在同一层次结构中开发三个互补组件。基于动作的下一尺度预测注入了与尺度对齐的动作表示，以改善动作与未来的耦合，并建模成功和失败的行为。尺度压缩的长时间视野记忆以细致的分辨率保留近期交互，同时压缩远期观察和动作，尺度自回归的梦境强制增强了对自生成上下文的鲁棒性。最后，基于策略的对齐优化了自回归视觉动态，结合了跟随动作和长期一致性奖励，同时保持预训练的视觉分布。实验表明，WALL-SS改善了动作跟随和轨迹准确性，支持在有限内存下进行连贯的分钟级流式展开，并在减少动作漂移和长时间视野不一致性方面持续受益于基于策略的对齐。
cs.RO / 2 / 2608.26273
Constraint-Aware Physics-Informed Neural Networks for Static Shape Estimation of Co-Manipulative Continuum Robots
考虑约束的物理信息神经网络用于共操控连续机器人静态形状估计
Abstract
Static shape estimation of co-manipulative continuum robots (CCRs) is challenging because the continuum arms and manipulated flexible object form a closed chain that must satisfy both static equilibrium and geometric loop-closure constraints. This paper presents a constraint-aware physics-informed neural network (PINN) for static shape estimation of a tendon-driven CCR modeled using the geometric variable strain formulation. The proposed method incorporates a projected static equilibrium residual and a configuration-level geometric residual to enforce the governing mechanics and closed-chain geometry. In simulation, the PINN is compared with a purely data-driven artificial neural network (ANN) under limited and noisy training data. With 140 samples and 50% label noise, the PINN reduces the relative configuration error, equilibrium residual, and closed-chain residual by 67.88%, 67.35%, and 88.06%, respectively. Using the full dataset, the PINN achieves 0.1597% relative configuration error with an inference time of 0.1773 ms, compared with 17.97 s for an iterative nonlinear solver. Experimental fine-tuning reduces the marker RMSE from 2.657 mm to 0.497 mm and increases R2 from -0.788 to 0.937. These results demonstrate accurate, physically consistent, and computationally efficient static shape estimation of closed-chain CCRs.
Chinese Translation
共操控连续机器人（CCRs）的静态形状估计具有挑战性，因为连续臂和被操控的柔性物体形成一个闭合链条，必须同时满足静态平衡和几何闭合约束。本文提出了一种考虑约束的物理信息神经网络（PINN），用于基于几何变量应变模型的腱驱动CCR的静态形状估计。所提出的方法结合了投影静态平衡残差和配置级几何残差，以强制执行主导力学和闭合链几何。在仿真中，PINN与在有限和噪声训练数据下的纯数据驱动人工神经网络（ANN）进行了比较。在140个样本和50%的标签噪声下，PINN分别将相对配置误差、平衡残差和闭合链残差降低了67.88%、67.35%和88.06%。使用完整数据集时，PINN实现了0.1597%的相对配置误差，推理时间为0.1773毫秒，而迭代非线性求解器的推理时间为17.97秒。实验微调将标记均方根误差（RMSE）从2.657毫米降低到0.497毫米，R²从-0.788提高到0.937。这些结果展示了闭合链CCR的静态形状估计在准确性、物理一致性和计算效率方面的优越性。
cs.RO / 3 / 2608.26314
Dispersive Forward Tree Search for Optimal Control: Coverage, Complexity, and Computation
用于最优控制的分散前向树搜索：覆盖、复杂性与计算
Abstract
Steering-based planners require solutions to state-to-state boundary value problems, which can be inaccessible for nonlinear platforms. Forward propagation evades the steering requirement, but the finite-sample behavior of the associated planners remains uncharacterized and their implementations underperform in practice. This paper develops a propagation-based kinodynamic planner with deterministic finite-sample near-optimality guarantees. We work within the large class of differentially flat nonlinear systems and show that a forward tree of locally dispersive control commands contains a near-optimal trajectory at a certified tree size. We provide a general mechanism to construct dispersive command sets for control-affine systems, which are necessary to implement the search algorithm prescribed by the theory. We show that covering the certified trajectory class irrespective of cost provably demands a tree exponentially sized in the problem horizon, and present a cost-conditioned dominance pruning procedure that retains near-optimality at a tree size polynomial in the horizon. We implement the resulting search algorithm, Dispersive Forward Tree search (DFT*), as breadth-first expansion of the forward tree, which maps naturally onto parallel hardware. We design efficient dispersive samplers for the unicycle, the trailer car, and the quadrotor and evaluate challenging planning tasks for these platforms. DFT* delivers consistently competitive and often substantially better solution quality than state-of-the-art kinodynamic planners at comparable solution times on embedded-tier processors, accelerating further as parallel compute is scaled. We also implement DFT* in a receding-horizon loop to demonstrate real-time planning in dynamic environments at embedded-tier compute budgets.
Chinese Translation
基于引导的规划器需要解决状态到状态的边界值问题，这对于非线性平台可能是不可达的。前向传播避免了引导要求，但相关规划器的有限样本行为仍未被表征，其实现效果在实践中表现不佳。本文开发了一种基于传播的运动动力学规划器，具有确定性的有限样本近似最优性保证。我们在一类大规模的微分平坦非线性系统中工作，并展示了一个局部分散控制命令的前向树在经过认证的树大小下包含一个近似最优的轨迹。我们提供了一种通用机制来构建控制仿射系统的分散命令集，这是实现理论所规定的搜索算法所必需的。我们表明，无论成本如何，覆盖经过认证的轨迹类在理论上需要一个在问题视野中呈指数级增长的树，并提出了一种基于成本条件的主导剪枝程序，该程序在树的大小上保持多项式近似最优性。我们实现了结果搜索算法——分散前向树搜索（Dispersive Forward Tree search，DFT*），作为前向树的广度优先扩展，自然映射到并行硬件上。我们为独轮车、拖车和四旋翼设计了高效的分散采样器，并评估了这些平台的挑战性规划任务。DFT*在嵌入式处理器上提供了一致的竞争力，且在可比解决时间内通常显著优于最先进的运动动力学规划器，随着并行计算的扩展，性能进一步提升。我们还在递归视野循环中实现了DFT*，以展示在嵌入式计算预算下的动态环境中的实时规划。
cs.RO / 4 / 2608.26383
Cross-Platform Benchmark of Neural 3D Reconstruction for Autonomous Laboratory Robots
跨平台神经3D重建基准测试用于自主实验室机器人
Abstract
Autonomous robots performing laboratory tasks depend on 3D reconstruction pipelines that can turn raw camera streams into actionable object representations within the latency budget of a physical control loop. Neural 3D reconstruction methods have demonstrated high-quality view synthesis, but their real-time viability across the compute platforms on which laboratory robots actually run remains poorly characterized. In this work, we present a systematic compute-platform benchmark of neural 3D reconstruction methods, evaluating NeRF and 3D Gaussian Splatting training and rendering on GPU-enabled computing devices ranging from single-board computers to server-class nodes, and place Meta's SAM3D single-image reconstruction on the same axes to quantify its latency and fidelity gap relative to per-scene optimization. Our results show that Gaussian Splatting yields higher rendering quality than NeRF at greater GPU cost, and that onboard compute is insufficient for full per-scene optimization at interactive rates. Our preliminary assessment on SAM3D indicates that it delivers plausible object geometry within seconds, but with detail mismatches that can compromise downstream manipulation. Together, these findings motivate tiered pipelines in which lightweight feed-forward reconstruction sustains the real-time perception-and-tracking loop for laboratory robots, while heavier neural reconstruction is scheduled selectively on suitable compute.
Chinese Translation
自主机器人在实验室任务中依赖于3D重建流程，将原始摄像头流转换为可操作的物体表示，且需在物理控制循环的延迟预算内完成。神经3D重建方法已展示出高质量的视图合成，但其在实验室机器人实际运行的计算平台上的实时可行性仍然缺乏充分的表征。在本研究中，我们对神经3D重建方法进行了系统的计算平台基准测试，评估了NeRF和3D高斯溅射在从单板计算机到服务器级节点的GPU支持计算设备上的训练和渲染，并将Meta的SAM3D单图像重建放置在同一坐标轴上，以量化其相对于每场景优化的延迟和保真度差距。我们的结果表明，高斯溅射在更高的GPU成本下提供了比NeRF更高的渲染质量，并且板载计算不足以在交互速率下进行全面的每场景优化。我们对SAM3D的初步评估表明，它在几秒钟内提供了合理的物体几何形状，但细节不匹配可能会影响后续操作。综合来看，这些发现促使我们提出分层管道，其中轻量级前馈重建支持实验室机器人的实时感知和跟踪循环，而更重的神经重建则在合适的计算资源上进行选择性调度。
cs.RO / 5 / 2608.26496
RTNav: Towards Real-Time Zero-Shot Object Navigation
RTNav：面向实时零-shot物体导航
Abstract
Navigation in unknown environments to find unforeseen objects has become increasingly feasible with capable vision and language foundation models. However, these models also introduce non-negligible inference latency, which becomes an important concern when agents must operate continuously in the real world. Most state-of-the-art methods are still developed in synchronous simulators, where the environment waits for the agent to act and inference time is effectively free. As a result, agents are often designed around the sequential execution of perception, reasoning, and action, with little regard for time constraints. Under real-time execution, where wall-clock time counts towards the task budget, the inefficiencies of these architectures become clear. We show that recent zero-shot object navigation methods suffer consistent performance degradation under such realistic timing conditions. Motivated by this observation, we propose RTNav, a simple but effective architecture that treats inference latency, asynchronous environment stepping, and bounded compute as explicit design considerations. Evaluated on real-time variants of HM3D-v1, HM3D-v2, and HM3D-OVON, RTNav improves the success rate by up to 11% and the Success weighted by Completion Time by up to 5.1 points over prior work.
Chinese Translation
在未知环境中寻找未预见物体的导航随着强大的视觉和语言基础模型的出现变得越来越可行。然而，这些模型也引入了不可忽视的推理延迟，当代理必须在现实世界中持续操作时，这成为一个重要问题。大多数最先进的方法仍然是在同步模拟器中开发的，在这种情况下，环境等待代理的行动，推理时间实际上是免费的。因此，代理通常围绕感知、推理和行动的顺序执行进行设计，而很少考虑时间限制。在实时执行的情况下，墙钟时间计入任务预算，这些架构的低效性变得显而易见。我们表明，最近的零-shot物体导航方法在这样的现实时间条件下表现出一致的性能下降。受到这一观察的启发，我们提出了RTNav，这是一种简单但有效的架构，将推理延迟、异步环境步进和有限计算作为明确的设计考虑。经过在HM3D-v1、HM3D-v2和HM3D-OVON的实时变体上的评估，RTNav的成功率提高了最多11%，成功率加权完成时间提高了最多5.1分，优于之前的工作。
cs.RO / 6 / 2608.26505
Closing the Loop on the Poppy Humanoid: Bipedal Locomotion with Linear-Quadratic Control and Learned Cost Functions
闭环控制在Poppy人形机器人上的应用：基于线性二次控制和学习成本函数的双足运动
Abstract
The Poppy Humanoid is an open-source, low-cost robot suitable for research and education in artificial intelligence. However, we are unaware of any published methodology that achieves reliable, unassisted bipedal locomotion on the standard Poppy hardware. This paper contributes a functional closed-loop walking controller for Poppy, based on the linear-quadratic regulator (LQR) framework for trajectory tracking. Starting with data collected from open-loop playback of a nominal walking trajectory, our proposed method learns a quadratic cost function for an LQR controller that substantially improves the reliability of the motion. The closed-loop controller is validated empirically, demonstrating statistically significant improvements in walking performance compared to open-loop trajectory playback.
Chinese Translation
Poppy人形机器人是一款开源、低成本的机器人，适用于人工智能领域的研究和教育。然而，目前尚未有任何已发表的方法能够在标准Poppy硬件上实现可靠的无辅助双足运动。本文提出了一种基于线性二次调节器（LQR）框架的功能性闭环行走控制器，用于Poppy的轨迹跟踪。我们的提议方法从开放循环回放的名义行走轨迹中收集数据，学习LQR控制器的二次成本函数，从而显著提高运动的可靠性。闭环控制器经过实证验证，与开放循环轨迹回放相比，行走性能有统计学显著改善。
cs.RO / 7 / 2608.26545
Memory Anchors for Continual Robot Learning
持续机器人学习的记忆锚点
Abstract
Robot policies deployed in the wild should have the capability to continually learn new tasks without forgetting existing behaviors. A common approach to combat such catastrophic forgetting is to train on new task data with a replay buffer of previously learned task data. Although this buffer is commonly sampled randomly from all prior experiences, we show that a small set of these experiences contributes greatly in anchoring past performance. We call these experiences Memory Anchors. We identify Memory Anchors in regions where representations of new-task observations collapse onto those of old-task observations even though the tasks require conflicting actions, like when a familiar object must be manipulated in a new way. Rehearsing old data in this region plays a key role in preventing destructive overwriting of past task knowledge, serving as this critical Memory Anchor role. Excluding only 10% Memory Anchors before sampling the buffer leads to more than a 4.5x increase in catastrophic forgetting on the LIBERO benchmark suites. Conversely, enriching the replay buffer with Memory Anchors can decrease high-conflict task forgetting by 63% and enables successful continual learning of two task sequences on a real robot. Videos and additional visualizations can be found at https://robot-adaptation.github.io/MemoryAnchors
Chinese Translation
在实际应用中部署的机器人策略应具备持续学习新任务而不遗忘现有行为的能力。应对这种灾难性遗忘的常见方法是利用重放缓冲区，在新任务数据上进行训练，同时保留以前学习的任务数据。尽管该缓冲区通常是从所有先前经验中随机抽样的，但我们表明，这些经验中的一小部分在锚定过去的表现方面具有重要贡献。我们称这些经验为记忆锚点。我们在新任务观察的表示与旧任务观察的表示重叠的区域中识别记忆锚点，即使这些任务要求相互冲突的动作，例如当一个熟悉的物体必须以新的方式进行操作时。在这一区域中重演旧数据在防止对过去任务知识的破坏性覆盖方面发挥了关键作用，充当了这一重要的记忆锚点角色。在采样缓冲区之前排除仅10%的记忆锚点会导致LIBERO基准测试套件上的灾难性遗忘增加超过4.5倍。相反，通过记忆锚点丰富重放缓冲区可以减少高冲突任务的遗忘率63%，并使得在真实机器人上成功持续学习两个任务序列成为可能。视频和其他可视化内容可以在 https://robot-adaptation.github.io/MemoryAnchors 找到。
cs.RO / 8 / 2608.26578
TrapVLA: Trapping Vision-Language-Action Models in Configured Failure Modes
TrapVLA：在配置失败模式中捕获视觉-语言-动作模型
Abstract
This work introduces Configured Failure Trapping, a novel backdoor attack task against Vision-Language-Action (VLA) models, which aims to activate attacks through stealthy textual triggers and induce configured failure modes. Unlike prior backdoor attacks that treat any task failure as a successful attack, Configured Failure Trapping requires the attacker to control how the robot fails (e.g., causing the robot to grasp with a specified positional offset), making it substantially more challenging and hard to detect. To support the new task, we propose an effective data engine for synthesizing high-quality target trajectories and an automated suite for measuring configured-failure fidelity. Then, based on this foundation, we construct two new benchmarks, namely Trap-LIBERO and Trap-RoboTwin, that instantiate Configured Failure Trapping across four representative failure modes. To address this task, we identify sparse action deviation as a critical challenge and accordingly propose a novel method named TrapVLA, which explicitly learns trigger-induced action residuals to steer the policy toward the configured failure behavior. Extensive experiments across simulation benchmarks and real-world robotic settings show that TrapVLA effectively injects configured failure modes into VLA models while largely preserving performance on clean data. Project page: https://john-liua.github.io/TrapVLA/
Chinese Translation
本研究介绍了配置失败捕获（Configured Failure Trapping），这是一种针对视觉-语言-动作（VLA）模型的新型后门攻击任务，旨在通过隐蔽的文本触发器激活攻击并诱导配置的失败模式。与之前将任何任务失败视为成功攻击的后门攻击不同，配置失败捕获要求攻击者控制机器人失败的方式（例如，导致机器人以特定位置偏移进行抓取），这使得攻击变得更加具有挑战性且难以检测。为了支持这一新任务，我们提出了一种有效的数据引擎，用于合成高质量的目标轨迹，以及一个自动化套件，用于测量配置失败的保真度。在此基础上，我们构建了两个新的基准，即Trap-LIBERO和Trap-RoboTwin，分别在四种代表性的失败模式下实例化配置失败捕获。为了解决这一任务，我们将稀疏动作偏差识别为一个关键挑战，并相应地提出了一种新方法，命名为TrapVLA，该方法明确学习触发器引起的动作残差，以引导策略朝向配置的失败行为。在模拟基准和真实机器人环境中的广泛实验表明，TrapVLA能够有效地将配置失败模式注入到VLA模型中，同时在干净数据上大幅保持性能。项目页面：https://john-liua.github.io/TrapVLA/
cs.RO / 9 / 2608.26583
SOLO: Stable Omni-terrain Long-Horizon Perceptive Humanoid Locomotion
SOLO：稳定的全地形长视距感知类人步态
Abstract
Humans traverse complex terrain over long distances without losing balance, whereas perceptive humanoid policies become fragile as perception and control errors accumulate. We present SOLO, a unified framework addressing two compounding causes of this long-horizon fragility: dense terrain reconstruction smooths action-critical details, and pointwise imitation lacks temporal credit assignment. Its Query Reconstructor (QR) uses Fourier-encoded cell queries to retrieve spatially specific evidence from depth-proprioception tokens, preserving sharp terrain boundaries. Trajectory-Aware MSE (TA-MSE) Distillation adds next-state teacher-student disagreement to the PPO reward, enabling Generalized Advantage Estimation to propagate future disagreement penalties to preceding actions. In simulation, QR reduces height-map L1 error by factors of 3.3-4.0, while TA-MSE surpasses PPO and MSE+PPO in curriculum progression. On stress-test terrains, SOLO achieves 97.5% mean traversal success and 96% stepping-stone success, versus 75.0-75.6% and 0-3% for dense-reconstructor variants. Deployed zero-shot with only a chest-mounted depth camera and proprioception, SOLO completes a continuous 1.5-km outdoor route and an indoor mixed-terrain course. Project page: https://sunpihai-up.github.io/solo/
Chinese Translation
人类能够在复杂地形上长距离行走而不失去平衡，而感知类人策略在感知和控制误差累积时变得脆弱。我们提出了SOLO，一个统一的框架，解决这种长视距脆弱性的两个复合原因：密集的地形重建平滑了对动作至关重要的细节，而逐点模仿缺乏时间信用分配。其查询重构器（Query Reconstructor，QR）使用傅里叶编码的单元查询从深度-本体感知标记中检索空间特定证据，保留了清晰的地形边界。轨迹感知均方误差（Trajectory-Aware MSE，TA-MSE）蒸馏将下一状态教师-学生的不一致性添加到PPO奖励中，使广义优势估计能够将未来的不一致惩罚传播到先前的动作。在仿真中，QR将高度图L1误差减少了3.3-4.0倍，而TA-MSE在课程进展中超越了PPO和MSE+PPO。在压力测试地形上，SOLO实现了97.5%的平均通过成功率和96%的踏脚石成功率，而密集重构变体的成功率为75.0-75.6%和0-3%。在仅使用胸部安装的深度相机和本体感知的情况下，SOLO成功完成了1.5公里的连续户外路线和一个室内混合地形课程。项目页面：https://sunpihai-up.github.io/solo/
cs.RO / 10 / 2608.26622
Relaxation-Aware Multimodal Sensing of Soft Gripper Driven by Structure-Perception-Learning
基于结构-感知-学习的放松感知多模态软抓手
Abstract
Achieving stable, sustained grasping with soft robotic hands remains a fundamental challenge. Compliance enables safe and adaptive contact, yet the intrinsic viscoelasticity of soft polymers leads to stress relaxation and a continuous decay of grasping force during holding. Inspired by human grasping, which combines phase-dependent stiffness regulation with continuous sensing and feedback, this paper presents an integrated structure--perception--learning framework. We develop a variable-stiffness soft gripper that uses onboard vision and infrared thermography to track deformation and the temperature field in real time, preserving continuous tracking of the interaction state. To mitigate relaxation-induced force decay, we propose a temperature-coupled viscoelastic force representation, together with a physics-informed learning model, to reconstruct the force trend and provide explicit compensation during holding. Experiments show that, in a 280s force-controlled grasp-and-hold task, the proposed method maintains the desired force with a mean absolute error of 0.066N, outperforming fixed-aperture and instantaneous-only baselines by 80% and 95%, respectively. Overall, the results support a mechanism--AI co-design view: mechanisms shape feasible interactions, while learning compensates remaining uncertainty in viscoelastic dynamics, together enabling stable, sustained grasping.
Chinese Translation
实现软机器人手的稳定、持续抓取仍然是一个基本挑战。柔性使得安全和自适应接触成为可能，但软聚合物的内在粘弹性导致在保持过程中应力放松和抓取力的持续衰减。受到人类抓取的启发，人类抓取结合了相位依赖的刚度调节与持续的感知和反馈，本文提出了一种集成的结构-感知-学习框架。我们开发了一种可变刚度的软抓手，利用车载视觉和红外热成像实时跟踪变形和温度场，保持对交互状态的持续跟踪。为了减轻因放松引起的力衰减，我们提出了一种温度耦合的粘弹性力表示，并结合物理信息学习模型，重建力的趋势并在保持过程中提供明确的补偿。实验表明，在一个280秒的力控制抓取保持任务中，所提出的方法以0.066N的平均绝对误差维持了所需的力，分别比固定孔径和瞬时基线提高了80%和95%。总体而言，结果支持一种机制-人工智能协同设计的观点：机制塑造可行的交互，而学习补偿粘弹性动态中的剩余不确定性，从而共同实现稳定、持续的抓取。
cs.RO / 11 / 2608.26645
FLARE: A Failure-Aware Framework for Autonomous Correction and Recovery in Visual-Language Robotic Manipulation
FLARE：一种面向故障的自主纠正和恢复框架用于视觉-语言机器人操作
Abstract
Vision-Language-Action Models~(VLAs) have demonstrated significant promise in generalizing to complex, long-horizon robotic manipulation tasks. However, their performance remains brittle, as they are typically trained on trajectory-monotonic, failure-free demonstrations. This reliance on ``perfect" data leaves them unable to recover from common execution errors, such as a missed grasp, a dropped object, or an unexpected collision. In this paper, we propose FLARE, a novel framework that endows VLAs with robust error recovery capabilities through a ``Retry" and ``Reset" paradigm. First, we introduce a ``Retry" mechanism by injecting perturbation and bridging segments that decouple robot pose from environment state into demonstrations, enabling the policy to autonomously handle execution deviations. Second, to address critical, state-breaking (OOD) failures, we introduce a ``Reset" pipeline. We leverage an MLLM for offline failure analysis to automatically identify OOD states from execution videos. This analysis enables the efficient, targeted collection of a small library of object-centric ``Reset" skills, which are trained to restore the environment to a task-valid state. Our full framework integrates these learned policies. At inference, an online MLLM monitor arbitrates between task execution and ``Reset" skills. Experiments on challenging, contact-rich manipulation tasks show our approach significantly improves task success and robustness.
Chinese Translation
视觉-语言-动作模型（VLAs）在推广到复杂的长时间机器人操作任务方面展现了显著的潜力。然而，由于它们通常是在轨迹单调、无故障的演示上进行训练，其性能仍然脆弱。这种对“完美”数据的依赖使得它们无法从常见的执行错误中恢复，例如错过抓取、物体掉落或意外碰撞。本文提出了FLARE，一个新颖的框架，通过“重试”（Retry）和“重置”（Reset）范式赋予VLAs强大的错误恢复能力。首先，我们通过在演示中注入扰动和桥接段，将机器人姿态与环境状态解耦，引入了“重试”机制，使策略能够自主处理执行偏差。其次，为了解决关键的、状态破坏的（OOD）故障，我们引入了“重置”管道。我们利用MLLM进行离线故障分析，自动识别执行视频中的OOD状态。这一分析使得我们能够高效、有针对性地收集一小组以物体为中心的“重置”技能库，这些技能经过训练以将环境恢复到任务有效状态。我们的完整框架整合了这些学习到的策略。在推理时，在线MLLM监控器在任务执行和“重置”技能之间进行裁决。在具有挑战性的、接触丰富的操作任务上的实验表明，我们的方法显著提高了任务成功率和鲁棒性。
cs.RO / 12 / 2608.26669
Beyond the Proving Ground: Independent Public-Road Testing of Assisted Lane Change Systems using LiDAR
超越试验场：基于LiDAR的辅助变道系统在公共道路上的独立测试
Abstract
Testing of commercial Advanced Driver Assistance Systems is essential to ensure safety and compliance during type approval and in service operation. However, proving ground scenarios may not reflect real world driving complexity, while geo fencing can require manufacturer collaboration and limit assessment independence. This work presents a methodology for independently testing Assisted Lane Change systems on public roads. A campaign on the A31 French motorway used a test vehicle equipped with a LiDAR based vehicle detection and tracking system. Tests covered combinations of inter vehicle distance and speed between the test vehicle and the take over vehicle. Real time kinematic global navigation satellite system receivers assessed detection and tracking performance. Recorded lane change trajectories were compared with the lane change suppression requirements of UNECE Regulation Number 79. Of 27 predefined lane change manoeuvres, 18 were completed and 9 suppressed. In 6 cases, the system allowed manoeuvres that did not meet regulatory minimum distance requirements. In 3 cases, the deviation remained statistically significant after accounting for measurement uncertainty. To the authors knowledge, this is the first public road campaign designed to assess Assisted Lane Change compliance with Regulation Number 79 safety distance requirements. The results demonstrate the suitability of LiDAR based sensing for this purpose. The methodology can support market surveillance and future regulatory revisions by revealing real world behaviours not covered by approval procedures.
Chinese Translation
对商业高级驾驶辅助系统的测试对于确保安全性和合规性在类型批准和实际运营中至关重要。然而，试验场场景可能无法反映现实世界驾驶的复杂性，而地理围栏可能需要制造商的合作并限制评估的独立性。本研究提出了一种在公共道路上独立测试辅助变道系统的方法。我们在法国A31高速公路上进行了一项测试活动，使用了一辆配备基于LiDAR的车辆检测和跟踪系统的测试车辆。测试涵盖了测试车辆与接管车辆之间的车辆间距和速度的组合。实时动态全球导航卫星系统接收器评估了检测和跟踪性能。记录的变道轨迹与联合国欧洲经济委员会第79号法规的变道抑制要求进行了比较。在27个预定义的变道操作中，完成了18个，抑制了9个。在6个案例中，系统允许了未满足监管最低距离要求的操作。在3个案例中，考虑测量不确定性后，偏差仍然具有统计显著性。据作者所知，这是第一个旨在评估辅助变道系统是否符合第79号法规安全距离要求的公共道路测试活动。结果表明，基于LiDAR的传感技术适合此目的。该方法可以通过揭示未涵盖在批准程序中的现实世界行为，支持市场监管和未来的法规修订。
cs.RO / 13 / 2608.26673
PredVLA: A Sub-Million-Parameter Predictive-Coding Policy for Robot Manipulation
PredVLA：一种具有不到百万参数的预测编码策略用于机器人操作
Abstract
Large pretrained vision-language-action models dominate modern robot-manipulation benchmarks, but it remains unclear how much model scale is necessary for strong language-conditioned control, or whether fundamentally different control architectures can remain competitive at much smaller parameter budgets. We present PredVLA, a language-conditioned predictive-coding policy with only 0.68 million trainable network parameters and no robot-data pretraining, whose hierarchical generative recurrent dynamics predict visual features and proprioception while observations influence latent state only through online inference from the resulting sensory prediction errors. On LIBERO, PredVLA achieves an 86.9% mean success rate across the three short-horizon suites and 75.4% when the long-horizon suite is included. Under a controlled comparison using the same frozen front end, demonstrations, action decoder, and evaluation protocol, PredVLA achieves 3.7x and 7.4x mean success rates of parameter-matched Transformer and LSTM policies, respectively. The predictive-coding formulation also makes the contribution of observation-driven correction directly measurable: because observations influence the recurrent state only through prediction-error-based latent inference, disabling this inference yields an exact open-loop control condition. Together, these results show that a sub-million-parameter recurrent generative policy can achieve strong performance on modern language-conditioned manipulation benchmarks while providing an explicit mechanism for prediction-error-driven online state correction.
Chinese Translation
大型预训练视觉-语言-动作模型在现代机器人操作基准中占据主导地位，但尚不清楚强语言条件控制所需的模型规模有多大，或者根本不同的控制架构是否能够在更小的参数预算下保持竞争力。我们提出了PredVLA，这是一种语言条件的预测编码策略，仅具有68万可训练的网络参数，并且没有机器人数据的预训练，其层次生成递归动态预测视觉特征和本体感知，同时观察仅通过对结果感知预测误差的在线推断影响潜在状态。在LIBERO上，PredVLA在三个短期任务套件中实现了86.9%的平均成功率，在包含长期任务套件时为75.4%。在使用相同的冻结前端、演示、动作解码器和评估协议的受控比较中，PredVLA的平均成功率分别是参数匹配的Transformer和LSTM策略的3.7倍和7.4倍。预测编码的形式化还使得观察驱动的修正贡献可以直接测量：因为观察仅通过基于预测误差的潜在推断影响递归状态，禁用这种推断将产生一个精确的开环控制条件。这些结果表明，一个不到百万参数的递归生成策略可以在现代语言条件操作基准上实现强大的性能，同时提供了一个明确的机制用于基于预测误差的在线状态修正。
cs.RO / 14 / 2608.26739
Residual Deep Reinforcement Learning-Based Computed Torque Control for a Cable-Driven Lower-Limb Rehabilitation Robot under Disturbances and Parametric Uncertainties
基于残差深度强化学习的电缆驱动下肢康复机器人在干扰和参数不确定性下的计算力矩控制
Abstract
Accurate trajectory tracking in cable-driven lower-limb rehabilitation robots is challenging because model uncertainty, external disturbances, joint constraints, and pull-only cable actuation can degrade nominal control performance. Conventional model-based controllers provide an interpretable control structure but remain sensitive to model mismatch, whereas fully learning-based control can reduce transparency and complicate constraint-aware operation. This study proposes a residual deep reinforcement learning-enhanced computed torque control framework in which computed torque control generates the nominal command and a bounded Deep Deterministic Policy Gradient policy supplies only an additional compensating torque. The approach is evaluated in simulation under nominal, uncertain, disturbed, combined, and generalization conditions, together with trajectory-tracking, joint-limit, cable-demand, workspace-feasibility, and cable-Jacobian diagnostics. Across the evaluated conditions, the residual controller improves tracking and disturbance rejection relative to computed torque control while preserving the interpretable model-based command structure and satisfying the reported feasibility checks in the representative evaluation. Broader tests indicate that tracking improvements can persist beyond the representative case while also exposing trajectory-dependent constraint limitations. These results support bounded residual learning as a practical robustness-enhancement strategy for simulation-based rehabilitation robot control and motivate further constraint-aware and experimental validation.
Chinese Translation
在电缆驱动的下肢康复机器人中，准确的轨迹跟踪面临挑战，因为模型不确定性、外部干扰、关节约束和仅拉动的电缆驱动可能会降低名义控制性能。传统的基于模型的控制器提供了可解释的控制结构，但对模型不匹配仍然敏感，而完全基于学习的控制则可能降低透明度并使约束感知操作复杂化。本研究提出了一种增强残差深度强化学习的计算力矩控制框架，其中计算力矩控制生成名义指令，而有界的深度确定性策略梯度（Deep Deterministic Policy Gradient）策略仅提供额外的补偿力矩。该方法在名义、不确定、干扰、组合和泛化条件下进行仿真评估，同时进行轨迹跟踪、关节极限、电缆需求、工作空间可行性和电缆雅可比诊断。在评估的条件下，残差控制器相较于计算力矩控制改善了跟踪和干扰拒绝，同时保持了可解释的基于模型的指令结构，并满足了代表性评估中的可行性检查。更广泛的测试表明，跟踪改进可以超出代表性案例持续存在，同时暴露出轨迹依赖的约束限制。这些结果支持有界残差学习作为一种实用的鲁棒性增强策略，用于基于仿真的康复机器人控制，并激励进一步的约束感知和实验验证。
cs.RO / 15 / 2608.26766
MeshPriorDiT: Hierarchical Modeling for Action-Conditioned Cloth Dynamics
MeshPriorDiT：用于动作条件布料动力学的层次建模
Abstract
Action-conditioned cloth dynamics prediction requires both locally plausible deformation and long-range coordination. Existing approaches largely follow two paradigms. Mesh-based GNNs capture local physical responses through material connectivity. However, their finite message-passing range limits coordination between topologically distant regions, while autoregressive rollouts tend to accumulate prediction errors. Transformer-based dynamics models capture long-range interactions through global attention, but often operate without explicit material connectivity and must learn local topological responses directly from data. We propose MeshPriorDiT, a hierarchical dynamics model that decomposes future cloth motion into a structured mesh prior and a generative residual. An action-conditioned mesh GNN first predicts multi-step vertex displacements, yielding a reference trajectory that respects material topology and grasp constraints. Conditioned on historical states, planned actions, and the mesh prior, a Residual DiT then uses conditional flow matching to jointly generate the residual motion not captured by the prior. The generated residual is further rescaled and decoded using material adjacency to coordinate corrections across neighboring vertices. We evaluate MeshPriorDiT on 15-step autoregressive rollouts across three cloth manipulation tasks. Averaged over the three tasks, MeshPriorDiT reduces average Global MSE by 43.42% relative to the GNN-Only baseline and by 75.03% relative to the DiT-DDPM baseline, while maintaining a favorable Edge-strain MSE comparable to that of GNN-Only.
Chinese Translation
动作条件下的布料动力学预测需要局部合理的变形和长程协调。现有方法主要遵循两种范式。基于网格的图神经网络（GNN）通过材料连接捕捉局部物理响应。然而，它们有限的消息传递范围限制了拓扑上远离区域之间的协调，而自回归展开往往会累积预测误差。基于变换器的动力学模型通过全局注意力捕捉长程交互，但通常在没有明确材料连接的情况下操作，必须直接从数据中学习局部拓扑响应。我们提出了MeshPriorDiT，一种层次动力学模型，将未来布料运动分解为结构化的网格先验和生成残差。一个动作条件的网格GNN首先预测多步顶点位移，生成一个尊重材料拓扑和抓取约束的参考轨迹。在历史状态、计划动作和网格先验的条件下，残差DiT使用条件流匹配共同生成先验未捕捉的残余运动。生成的残差进一步通过材料邻接进行重新缩放和解码，以协调相邻顶点之间的修正。我们在三个布料操作任务上评估了MeshPriorDiT的15步自回归展开。平均而言，MeshPriorDiT相较于仅使用GNN的基线减少了43.42%的平均全局均方误差（Global MSE），相较于DiT-DDPM基线减少了75.03%，同时保持了与GNN-Only相当的有利边缘应变均方误差（Edge-strain MSE）。
cs.RO / 16 / 2608.26789
Online Joint Calibration of Steering Offset and Planar LiDAR Extrinsics for Wheeled Mobile Robots
轮式移动机器人在线联合标定转向偏移和平面LiDAR外参
Abstract
Accurate steering sensing and LiDAR-to-vehicle extrinsics are crucial for reliable path tracking in warehouse mobile robots (WMRs); miscalibration often leads to snaking, weaving, and elevated cross-track error (CTE). In practice, steering ``zero'' is commonly set manually (e.g., eyeballing straightness via a PS4 joystick), while LiDAR extrinsics are assumed from CAD and may drift after maintenance. Such static, manual procedures frequently cause miscalibration in safety-critical environments. This paper presents an Extended Kalman Filter (EKF)--based method for online estimation of steering offset and planar LiDAR extrinsics within a bicycle-kinematics model, providing a principled alternative to manual calibration. Experiments on real datasets show that correcting steering offset reduces CTE substantially, validating the effectiveness of the proposed approach.
Chinese Translation
准确的转向传感和LiDAR与车辆的外参对于仓库移动机器人（WMRs）可靠的路径跟踪至关重要；标定错误常常导致蛇形行驶、编织行驶以及较高的横向误差（CTE）。在实际操作中，转向“零”通常通过手动设置（例如，通过PS4手柄目测直线度），而LiDAR外参则通常基于CAD假设，并可能在维护后发生漂移。这种静态的手动程序在安全关键环境中常常导致标定错误。本文提出了一种基于扩展卡尔曼滤波器（EKF）的方法，用于在自行车运动学模型内在线估计转向偏移和平面LiDAR外参，为手动标定提供了一种有原则的替代方案。对真实数据集的实验表明，纠正转向偏移显著减少了CTE，验证了所提方法的有效性。
cs.RO / 17 / 2608.26800
Rapid On-Robot Learning for Dynamic Manipulation Skills: Robot Juggling
动态操作技能的快速机器人在线学习：机器人杂耍
Abstract
We present an online learning framework that enables a bimanual robot to acquire diverse juggling patterns directly on physical hardware within minutes, even with a significant sim2real gap. One of the most important lessons from this work is that a model, even when far from reality, can be extremely useful for learning. This motivates a central philosophy of our approach: learning should build upon the robot's current knowledge rather than replace it. Our regularized memory-based learning puts this principle into practice by learning a local model from accumulated experience while retaining the global prior model to extrapolate where experience is sparse. This enables efficient and stable online learning from each new experience without resorting to uninformed exploration over a vast space of possible behaviors. Equally important to continual on-robot learning is safety, allowing the robot to repeatedly practice and improve in the real world. We construct a mutually reachable set that allows safe transitions between successive throws and catches, without driving either arm into a state from which its next action would require violating the robot's joint or actuator limits. Together, these ideas enable a bimanual robot with multi-fingered hands and onboard vision to safely learn and compose five canonical three-ball juggling patterns, including cascade, tennis, half-shower, shower, and box, within less than 5 minutes of real-world interaction. More broadly, this work points toward robots that build upon imperfect prior knowledge and continually refine their behavior through their own real-world experience.
Chinese Translation
我们提出了一种在线学习框架，使双手机器人能够在物理硬件上直接在数分钟内习得多样的杂耍模式，即使存在显著的仿真到现实（sim2real）差距。本工作的一个重要启示是，即使模型与现实相差甚远，也能在学习过程中发挥极大作用。这激发了我们方法的核心理念：学习应建立在机器人当前知识的基础上，而非取代它。我们采用正则化的基于记忆的学习方法，通过从积累的经验中学习局部模型，同时保留全局先验模型，以在经验稀缺的区域进行外推。这使得机器人能够高效且稳定地从每次新经验中在线学习，而无需在庞大的可能行为空间中进行无指导的探索。同样重要的是持续的机器人在线学习必须保证安全，使机器人能够在真实环境中反复练习和改进。我们构建了一个相互可达集，允许连续投掷和接球之间的安全转换，避免任一机械臂进入其下一动作需违反关节或执行器限制的状态。综合这些理念，使得配备多指手和车载视觉的双手机器人能够在不到5分钟的真实交互中安全学习并组合五种典型的三球杂耍模式，包括级联（cascade）、网球（tennis）、半淋浴（half-shower）、淋浴（shower）和盒式（box）。更广泛地说，本工作指向了能够基于不完美先验知识并通过自身真实世界经验不断优化行为的机器人。
cs.RO / 18 / 2608.26819
CLIPPER: Replayable Shortlisted Optimization for Repeated Spatial Coverage Planning
CLIPPER：可重放的短名单优化用于重复空间覆盖规划
Abstract
Operational requirements developed with the City of Braunschweig frame municipal micromobility planning under geofenced exclusions, mandatory retained sites, spacing rules, and area-level caps. Each policy edit requires a new feasible plan; full-set greedy takes tens of seconds per alternative at city scale. We present CLIPPER (Constraint-exact Low-latency Iterative Planning with Pooled Evaluation and Replay). It forms bounded candidate pools but recomputes exact current gains and checks every active constraint before selection. Coverage from each candidate alone sets the initial order. Offline full-set scans measure gains omitted by the pool; online, a conservative bound triggers expansion or audit. CLIPPER-F gives each proposal group the same number of candidate slots. Across Braunschweig, Munich, and Berlin, its mean coverage over complete chains stays within 0.245 percentage points of full-set greedy under the same policy, with 13.6--28.9 times lower mean rollout time. CLIPPER-A instead distributes one shared candidate budget across the groups. Under its coverage-prioritized policy, it uses 9--15% of full-set greedy's rollout time under the same policy, with mean gaps of 1.82 percentage points in Braunschweig, 0.12 in Munich, and 0.27 in Berlin. Together, CLIPPER enables rapid, replayable comparison of recorded city-scale planning states while enforcing every encoded model constraint.
Chinese Translation
与布伦瑞克市（City of Braunschweig）制定的运营要求框架下，市政微移动性规划受到地理围栏排除、强制保留地点、间距规则和区域级上限的影响。每次政策编辑都需要一个新的可行计划；在城市规模下，完整集合贪婪算法每个替代方案需要数十秒。我们提出了CLIPPER（Constraint-exact Low-latency Iterative Planning with Pooled Evaluation and Replay）。它形成有限的候选池，但在选择之前重新计算当前的确切收益并检查每个活动约束。每个候选者单独的覆盖范围设定了初始顺序。离线完整集合扫描测量池中遗漏的收益；在线时，一个保守的界限触发扩展或审计。CLIPPER-F为每个提案组提供相同数量的候选槽。在布伦瑞克、慕尼黑和柏林，其完整链的平均覆盖率在相同政策下保持在完整集合贪婪算法的0.245个百分点以内，平均推出时间低于13.6至28.9倍。CLIPPER-A则在各组之间分配一个共享的候选预算。在其覆盖优先政策下，它在相同政策下使用的推出时间为完整集合贪婪算法的9%至15%，在布伦瑞克的平均差距为1.82个百分点，慕尼黑为0.12，柏林为0.27。总之，CLIPPER能够快速、可重放地比较记录的城市规模规划状态，同时强制执行每个编码模型约束。
cs.RO / 19 / 2608.26821
TemporalFlow-VLA: Learning Physically Grounded Execution History for Long-Horizon Robot Manipulation
TemporalFlow-VLA：学习物理基础的执行历史以实现长时间范围的机器人操作
Abstract
Vision-language-action (VLA) models leverage pretrained vision-language representations for robot control, yet simply adding historical frames does not reliably capture recent physical change. This is especially problematic in multi-stage manipulation, where visually similar states may require different actions depending on prior execution. To address this challenge, we present TemporalFlow-VLA, which learns compact execution history through physically grounded temporal supervision. Using recorded robot states, robot geometry, and calibrated cameras, we construct robot-surface temporal flow as a training-only target and supervise two execution-aligned temporal queries that provide structured history to the action expert. The geometric supervision path is not evaluated at deployment. TemporalFlow-VLA achieves 97.63 +/- 0.26% average success on LIBERO, including 96.60 +/- 0.87% on LIBERO Long, and 85.5%/84.2% Clean/Randomized success across 12 RoboTwin tasks. It shows its clearest advantage over prior methods on longer-horizon, multi-stage manipulation. Controlled history interventions show that action prediction depends on both historical content and temporal order. With asynchronous feature caching, temporal conditioning maintains single-frame-level server-side sampling latency without additional historical-encoding overhead. Overall, TemporalFlow-VLA provides a compact, physically grounded interface for exploiting ordered execution history without explicit motion estimation or geometric processing at deployment.
Chinese Translation
视觉-语言-动作（VLA）模型利用预训练的视觉-语言表示进行机器人控制，但仅仅添加历史帧并不能可靠地捕捉最近的物理变化。这在多阶段操作中尤为突出，因为视觉上相似的状态可能需要根据先前的执行采取不同的行动。为了解决这一挑战，我们提出了TemporalFlow-VLA，它通过物理基础的时间监督学习紧凑的执行历史。利用记录的机器人状态、机器人几何形状和校准相机，我们构建了机器人-表面时间流作为仅用于训练的目标，并监督两个与执行对齐的时间查询，为动作专家提供结构化的历史信息。几何监督路径在部署时不进行评估。TemporalFlow-VLA在LIBERO上实现了97.63 +/- 0.26%的平均成功率，其中LIBERO Long的成功率为96.60 +/- 0.87%，在12个RoboTwin任务中，Clean和Randomized的成功率分别为85.5%和84.2%。它在长时间范围的多阶段操作中显示出相对于先前方法的明显优势。受控历史干预表明，动作预测依赖于历史内容和时间顺序。通过异步特征缓存，时间条件保持单帧级服务器端采样延迟，而无需额外的历史编码开销。总体而言，TemporalFlow-VLA提供了一个紧凑的、物理基础的接口，用于利用有序的执行历史，而无需在部署时进行显式的运动估计或几何处理。
cs.RO / 20 / 2608.26883
Active Surface-Driven Reconfigurable Gripper: Robust Grasping and Sequential Manipulation of Thin Objects
主动表面驱动的可重构夹持器：薄物体的稳健抓取与顺序操作
Abstract
Robotic grippers face substantial challenges in grasping and manipulating thin objects. Most existing grippers rely on highly precise approach and grasp motions, which limits robustness and reduces applicability. This paper explores thin-object grasping using books as a representative example. Here, we propose a novel solution that integrates an active surface with underactuated compliance to achieve stable grasping of thin objects without complex control. First, an underactuated gripper with an active surface is designed. The active-surface thumb performs in-hand repositioning of the target book without requiring adjustments of the robot arm or the other fingers, while the underactuated fingers establish compliant contact conditions with the environment, and the reconfigurable structure enables reliable grasping of books under different configurations. Second, we establish a kinematic model of the gripper, and determine the initial grasp postures for two representative scenarios (books lying flat on a desktop and books vertically packed in a shelf). Third, by analyzing the physical model of a book lying on a table and its interaction with the gripper and the environment, we systematically optimize the structural parameters and grasping strategy. Finally, extensive experiments validate the effectiveness of the proposed gripper and strategy. The results demonstrate strong robustness and adaptability when grasping thin objects placed flat (including books, paper, fabric, plastic film, and mouse pad), as well as a high success rate when grasping vertically packed books. Moreover, the proposed gripper can reliably complete long sequential "grasp-place" tasks.
Chinese Translation
机器人夹持器在抓取和操作薄物体时面临重大挑战。大多数现有夹持器依赖于高度精确的接近和抓取动作，这限制了其稳健性并降低了适用性。本文以书籍作为代表性例子，探讨薄物体的抓取。我们提出了一种新颖的解决方案，将主动表面与欠驱动的柔顺性相结合，以实现对薄物体的稳定抓取，而无需复杂的控制。首先，设计了一种具有主动表面的欠驱动夹持器。主动表面拇指在不需要调整机器人手臂或其他手指的情况下，能够在手中重新定位目标书籍，而欠驱动手指则与环境建立柔顺接触条件，且可重构结构使得在不同配置下可靠抓取书籍。其次，我们建立了夹持器的运动学模型，并确定了两种代表性场景（书籍平放在桌面上和书籍垂直放置在书架上）的初始抓取姿态。第三，通过分析平放在桌子上的书籍的物理模型及其与夹持器和环境的相互作用，我们系统地优化了结构参数和抓取策略。最后，广泛的实验验证了所提夹持器和策略的有效性。结果表明，在抓取平放的薄物体（包括书籍、纸张、织物、塑料薄膜和鼠标垫）时具有强大的稳健性和适应性，同时在抓取垂直放置的书籍时也具有较高的成功率。此外，所提夹持器能够可靠地完成长时间的顺序“抓取-放置”任务。
cs.RO / 21 / 2608.26888
Beyond Shallow-Water Photorealism: Physically and Sensor-Grounded Simulation for Deep-Sea Robotics
超越浅水区的真实感：深海机器人物理与传感器基础的仿真
Abstract
Many recent underwater simulators emphasize visual realism at the expense of physical fidelity, focusing on shallow-water effects with limited relevance in deep-water environments and high computational cost. In this work, we shift the focus toward deep-sea physical and sensor realism. We present a physics- and sensor-grounded extension of the Stonefish simulator that augments its hydrodynamic models with stochastic IMU and DVL drift, magnetometer disturbances, higher-order hydrodynamics, terramechanics, pressure-driven environmental variability, and physically based underwater optics. These additions are designed to better capture the forces and measurements shaping the behavior of deep-ocean AUVs, ROVs, landers, ASVs, and gliders, while remaining compatible with real-time simulation. This work advances underwater simulation toward more representative deep-sea operating conditions, which is particularly relevant for long-duration navigation and learning-based autonomy, where inaccurate sensor and environmental models introduce non-physical artifacts and overly optimistic performance. While challenges remain, including complex fluid-structure interactions and full environmental stochasticity, the proposed framework provides a practical foundation for navigation, perception, and autonomy research under deep-sea conditions.
Chinese Translation
许多近期的水下仿真器强调视觉真实感，但以牺牲物理真实性为代价，主要关注浅水区效应，这在深水环境中相关性有限且计算成本高昂。在本研究中，我们将重点转向深海的物理和传感器真实感。我们提出了一种基于物理和传感器的Stonefish仿真器扩展，通过随机惯性测量单元（IMU）和多普勒速度计（DVL）漂移、磁力计干扰、高阶水动力学、土壤力学、压力驱动的环境变化以及基于物理的水下光学来增强其水动力模型。这些补充旨在更好地捕捉影响深海自主水下航行器（AUVs）、遥控水下机器人（ROVs）、着陆器（landers）、自主水面船（ASVs）和滑翔机（gliders）行为的力和测量，同时保持与实时仿真的兼容性。本研究推动水下仿真向更具代表性的深海操作条件发展，这对于长时间导航和基于学习的自主性尤为重要，因为不准确的传感器和环境模型会引入非物理伪影和过于乐观的性能。尽管仍面临复杂的流体-结构相互作用和完全的环境随机性等挑战，所提出的框架为深海条件下的导航、感知和自主性研究提供了实用基础。
cs.RO / 22 / 2608.26932
Contact-Aided Factor-Graph Localization for Underwater Sampling
基于接触辅助的因子图定位用于水下采样
Abstract
Accurate state estimation for autonomous underwater vehicles performing close-range seafloor sampling remains challenging. In low-altitude operation, down-looking cameras over featureless planar seabeds produce scale ambiguity, lateral degeneracy, and inconsistent feature tracking. Meanwhile, inertial-Doppler Velocity Log (DVL) fusion alone provides no mechanism for structural drift correction. We propose a Contact-Aided Factor-Graph Localization framework that treats physical interaction as an informative geometric constraint within a smoothing-based localization formulation. The method tightly fuses suction-based manipulator contact events with adaptive visual odometry, learned object detections, and on-board sensors. Visual odometry relative-pose factors and landmark bearing-range factors are uncertainty-scaled according to inlier statistics to prevent visually weak frames from destabilizing the estimator, while contact events are modeled as high-confidence factors that induce implicit loop closures without appearance-based place recognition. Furthermore, the system can fully initialize online during motion. Experimental evaluation in tanks, harbor, and simulation environments demonstrates that contact-induced constraints significantly reduce trajectory drift and improve object revisit accuracy compared to filtering-based navigation and contact-free graph formulations. These results highlight the role of embodied physical interaction as a localization primitive in perception-degraded underwater environments
Chinese Translation
对于执行近距离海底采样的自主水下航行器，准确的状态估计仍然具有挑战性。在低空操作中，面向特征缺乏的平面海床的向下摄像头会产生尺度模糊、横向退化和不一致的特征跟踪。同时，仅依靠惯性-多普勒速度计（DVL）融合无法提供结构漂移校正的机制。我们提出了一种接触辅助的因子图定位框架，将物理交互视为平滑型定位公式中的一种信息几何约束。该方法紧密融合了基于吸力的操控器接触事件、自适应视觉里程计、学习的物体检测和机载传感器。视觉里程计的相对位姿因子和地标方位-距离因子根据内点统计进行不确定性缩放，以防止视觉上较弱的帧使估计器不稳定，而接触事件则被建模为高置信度因子，能够在没有基于外观的地点识别的情况下引入隐式回环闭合。此外，该系统可以在运动过程中完全在线初始化。在水池、港口和仿真环境中的实验评估表明，与基于滤波的导航和无接触图形公式相比，接触引起的约束显著减少了轨迹漂移并提高了物体重访的准确性。这些结果突显了在感知退化的水下环境中，具身物理交互作为定位原语的重要性。
cs.RO / 23 / 2608.26947
4DSynth: Controllable Procedural World Synthesis for Dynamic Embodied Simulation
4DSynth：用于动态具身仿真的可控程序化世界合成
Abstract
Embodied agents need environments that are visually diverse, physically interactive, and changing over time. Procedural simulators can generate large interactive scene collections, and recent 4D generators produce compelling visual dynamics. Combining these properties in one environment, however, still demands extensive manual effort, and the result is rarely editable or controllable enough to reuse at scale. We present 4DSynth, a controllable procedural system that turns a natural-language description, a blueprint mask, or a single photograph into an editable 4D environment with explicit geometry, animated actors, collision-free trajectories, and physics-ready simulation state. Multiple scene routes share one geometry-grounded representation, so the same pipeline handles animation, camera planning, rendering, and task generation. To validate the full pipeline, we construct 4DSynth-Nav, an interactive navigation benchmark generated entirely from 4DSynth's procedural scenes. Two vision-language models evaluated across three difficulty tiers both fail the majority of tasks and stall after early subtasks. The same procedural controllability that produces these environments also makes each failure reproducible and each difficulty axis independently tunable. This paper presents both a controllable generation pipeline and the scalable benchmark it enables, offering a practical foundation for developing and evaluating embodied agents.
Chinese Translation
具身代理需要视觉多样、物理互动且随时间变化的环境。程序化模拟器能够生成大量互动场景集合，而最近的4D生成器则能够产生引人注目的视觉动态。然而，将这些特性结合在一个环境中仍然需要大量的手动努力，且结果往往不够可编辑或可控，难以大规模重用。我们提出了4DSynth，一个可控的程序化系统，它能够将自然语言描述、蓝图掩模或单张照片转化为一个可编辑的4D环境，具备明确的几何形状、动画角色、无碰撞轨迹和物理准备的仿真状态。多个场景路径共享一个基于几何的表示，因此同一管道处理动画、相机规划、渲染和任务生成。为了验证整个管道，我们构建了4DSynth-Nav，这是一个完全由4DSynth的程序化场景生成的互动导航基准。在三个难度等级中评估的两个视觉-语言模型在大多数任务中均未能成功，并在早期子任务后停滞。产生这些环境的相同程序化可控性也使得每次失败都可以重现，并且每个难度轴都可以独立调节。本文展示了一个可控的生成管道及其所支持的可扩展基准，为开发和评估具身代理提供了实用的基础。
cs.RO / 24 / 2608.27000
Arbitrary-Order Hermite Interpolation of Rigid-Motion Jets via Hyper-Multidual Quaternions
通过超多重四元数进行任意阶Hermite插值的刚体运动喷流
Abstract
We study bilateral interpolation of finite-order rigid-motion jets represented by unit dual quaternions. An order-$n$ multidual (MD) algebra is the truncated polynomial algebra $\mathbb{R}[\varepsilon]/(\varepsilon^{n+1})$; hyper-multidual (HMD) quaternions are dual quaternions with coefficients in this algebra. Temporal HMD transforms encode a pose and its derivatives, whereas a generic HMD curve need not be the temporal jet of its pose projection; we call this requirement holonomicity. We show that a temporal transform and its relative descriptor are unitary and derive recursive coefficient constraints, together with a local realizability converse in an admissible logarithm chart. We then extend screw linear interpolation (ScLERP) algebraically to unit HMD quaternions. Although it matches complete endpoint transforms, direct HMD--ScLERP is generically non-holonomic for arbitrary endpoint jets. We give a coefficient criterion and explicit endpoint and first-order interior contact defects. A holonomic alternative is obtained by mapping endpoint transforms to logarithmic dual-quaternion coordinates, applying the degree-$(2n+1)$ Hermite polynomial that matches derivatives through order $n$, and lifting by the exponential. HMD arithmetic also recovers higher-order rigid-motion acceleration fields without explicit differentiation of $\mathrm{dexp}$. Rotation and full $\mathrm{SE}(3)$ tests through second order, with an additional third-order polynomial check, reproduce the stated defects and endpoint jets.
Chinese Translation
我们研究了由单位双四元数表示的有限阶刚体运动喷流的双边插值。阶数为$n$的多重（MD）代数是截断的多项式代数$ ext{R}[oldsymbol{ ext{ε}}]/(oldsymbol{ ext{ε}}^{n+1})$；超多重（HMD）四元数是具有该代数系数的双四元数。时间HMD变换编码了一个姿态及其导数，而一般的HMD曲线不必是其姿态投影的时间喷流；我们称这种要求为全约束性（holonomicity）。我们证明了时间变换及其相对描述符是单位的，并推导出递归系数约束，以及在可接受对数图表中的局部可实现性对偶。然后，我们在代数上将螺旋线性插值（ScLERP）扩展到单位HMD四元数。尽管它匹配完整的端点变换，但直接的HMD--ScLERP对于任意端点喷流通常是非全约束的。我们给出了一个系数标准和明确的端点及一阶内部接触缺陷。通过将端点变换映射到对数双四元数坐标，应用通过阶数$n$匹配导数的阶数为$(2n+1)$的Hermite多项式，并通过指数提升，获得了一种全约束的替代方案。HMD算术还可以在不显式微分$ ext{dexp}$的情况下恢复高阶刚体运动加速度场。通过二阶的旋转和完整$ ext{SE}(3)$测试，以及额外的三阶多项式检查，重现了所述缺陷和端点喷流。
cs.RO / 25 / 2608.27033
Riemann-1.0: An Embodied World Action Model for Physical AI
Riemann-1.0：一种用于物理人工智能的具身世界动作模型
Abstract
We introduce Riemann-1.0, a fully causal autoregressive World Action Model for embodied intelligence. Riemann-1.0 jointly models multi-view visual observations, robot states, and embodiment-specific actions within a unified causal autoregressive sequence, representing robot actions and world evolution as causal state transitions. Unlike existing WAMs based on joint generation, video-first prediction, or decoupled modeling paradigms, Riemann-1.0 unifies online robot policy execution and action-conditioned world simulation within a single model, enabling it to function as both an executable robot policy and a multi-embodiment visual world simulator. To scale embodied experience across heterogeneous data sources, we further develop a progressive embodied pretraining framework that unifies learning from egocentric human videos, handheld-gripper demonstrations, and heterogeneous robot trajectories under a shared World Action Modeling objective. Built upon 200K+ hours of interaction data, Riemann-1.0 progressively transfers large-scale embodied experience into executable robot manipulation capabilities. Riemann-1.0 achieves state-of-the-art performance across both simulation benchmarks and real-world manipulation tasks. It achieves success rates of 94.3% on RoboTwin2.0, 99.0% on LIBERO, and 62.6% on the long-horizon compositional benchmark RoboCasa-365, outperforming the previous best method by 8.4% On long-horizon real-world manipulation tasks, Riemann-1.0 achieves a Success Rate (SR) of 85.0% and a Progress Success Rate (PSR) of 94.4%, exceeding the strongest open-source baseline by 15% in SR. These results demonstrate that unified World Action Modeling together with progressive embodied pretraining effectively transforms large-scale embodied experience into generalizable robot manipulation capabilities.
Chinese Translation
我们介绍了Riemann-1.0，一种完全因果的自回归具身智能世界动作模型。Riemann-1.0在统一的因果自回归序列中共同建模多视角视觉观测、机器人状态和具身特定动作，将机器人动作和世界演变表示为因果状态转移。与现有基于联合生成、视频优先预测或解耦建模范式的世界动作模型（WAMs）不同，Riemann-1.0在单一模型中统一了在线机器人策略执行和动作条件的世界模拟，使其既能作为可执行的机器人策略，又能作为多具身视觉世界模拟器。为了在异构数据源之间扩展具身经验，我们进一步开发了一种渐进式具身预训练框架，该框架在共享的世界动作建模目标下统一了从自我中心人类视频、手持抓取演示和异构机器人轨迹的学习。基于超过20万小时的交互数据，Riemann-1.0逐步将大规模具身经验转化为可执行的机器人操控能力。Riemann-1.0在模拟基准和现实世界操控任务中均达到了最先进的性能。在RoboTwin2.0上成功率为94.3%，在LIBERO上为99.0%，在长时间跨度的组合基准RoboCasa-365上为62.6%，超越了之前最佳方法8.4%。在长时间跨度的现实世界操控任务中，Riemann-1.0的成功率（SR）为85.0%，进展成功率（PSR）为94.4%，在成功率方面超越了最强的开源基线15%。这些结果表明，统一的世界动作建模与渐进式具身预训练有效地将大规模具身经验转化为可推广的机器人操控能力。
cs.RO / 26 / 2608.27079
GRAFT: Grounded and Efficient Online Reinforcement Adaptation for Fine-Grained Robot Manipulation
GRAFT：用于细粒度机器人操作的基础和高效在线强化适应
Abstract
Pretrained vision-language-action (VLA) policies provide strong priors for robot manipulation, yet adapting them online to fine-grained biomedical tasks remains challenging. Task success often hinges on subtle, view-dependent visual cues, while task-level rewards provide little guidance about which regions matter, making it difficult to learn task-relevant visual grounding from limited real-robot interaction. Online adaptation is further constrained by the computational cost of VLA inference and replay-based updates. We introduce GRAFT (Grounded Reinforcement Adaptation for Fast Task Learning), a framework for efficient online VLA adaptation through grounded perception. GRAFT uses region-level supervision to learn view-specific visual anchors that focus perception on task-relevant local cues without requiring region proposals at deployment. It further combines single-step action generation with cached visual-language prefix reuse to accelerate online learning. Across four biomedical manipulation tasks, GRAFT improves success rates by 25 percentage points under matched adaptation budgets, while reducing the computational overhead of online policy updates.
Chinese Translation
预训练的视觉-语言-动作（VLA）策略为机器人操作提供了强有力的先验知识，但将其在线适应于细粒度生物医学任务仍然具有挑战性。任务的成功往往依赖于微妙的、视角依赖的视觉线索，而任务级奖励对哪些区域重要提供的指导有限，这使得从有限的真实机器人交互中学习与任务相关的视觉基础变得困难。在线适应进一步受到VLA推理和基于重放的更新的计算成本的限制。我们提出了GRAFT（快速任务学习的基础强化适应），这是一个通过基础感知实现高效在线VLA适应的框架。GRAFT使用区域级监督来学习视角特定的视觉锚点，专注于与任务相关的局部线索，而无需在部署时进行区域提议。它还结合了单步动作生成与缓存的视觉-语言前缀重用，以加速在线学习。在四个生物医学操作任务中，GRAFT在匹配的适应预算下提高了成功率25个百分点，同时减少了在线策略更新的计算开销。
cs.RO / 27 / 2608.27085
Pass the Bucket: Efficient, Robust, Local Load Balancing for Teams of Heterogeneous Robots
传递负载：异构机器人团队的高效、鲁棒、本地负载均衡
Abstract
We study the problem of decentralized, self-organized task sharing for a swarm of heterogeneous robots that collaborate in transportation or other objectives that require coordinated motion planning. To this end, we present theoretical and practical results for the simple but effective mechanism of \emph{bucket brigades} for load balancing, in which a team of heterogenous robots share a spatial task in a confined, one-dimensional space, while only being able to sense collisions with neighbors or walls. The goal is to optimize throughput of the overall system, without central control or information, aiming at an interval partition proportional to robot velocities. We address possible chaotic system behavior by developing a stabilization mechanism based on simple local aid, a ``token'', that temporarily decelerates robots after an encounter. This purely local change eliminates persistent oscillations, resulting in convergence towards a stable system state. We accelerate system convergence by comparing a single boundary token to ubiquitous two-directional tokens and optimizing the deceleration factor. Event-driven simulations report convergence times and robustness: For a large variety of perturbations (such as robot deletion, position or velocity jittering), the system reliably re-converges. The results suggest a local, practical mechanism for robust load balancing for heterogeneous teams of robots that promises an effective tool as basis for more complex scenarios.
Chinese Translation
我们研究了异构机器人群体在运输或其他需要协调运动规划的目标下进行去中心化、自组织任务共享的问题。为此，我们提出了简单但有效的负载均衡机制—— extit{桶接力}的理论和实践结果，其中一组异构机器人在一个有限的一维空间内共享空间任务，同时只能感知与邻居或墙壁的碰撞。我们的目标是优化整个系统的吞吐量，而无需中央控制或信息，旨在实现与机器人速度成比例的区间划分。我们通过开发基于简单本地援助的稳定机制来解决可能的混沌系统行为，该机制使用一个“令牌”，在相遇后暂时减速机器人。这一纯粹的本地变化消除了持续的振荡，导致系统状态向稳定状态收敛。我们通过将单一边界令牌与普遍存在的双向令牌进行比较并优化减速因子，加速了系统的收敛。事件驱动的模拟报告了收敛时间和鲁棒性：对于各种扰动（如机器人删除、位置或速度抖动），系统可靠地重新收敛。结果表明，这是一种本地、实用的机制，为异构机器人团队提供了鲁棒的负载均衡，预示着作为更复杂场景基础的有效工具。
cs.RO / 28 / 2608.27088
Active sensing to characterize the heterogeneity of plant stress
主动感知以表征植物胁迫的异质性
Abstract
While most phenotyping platforms rely primarily on image-based measurements, advanced plant characterization requires the integration of active physiological sensing modali- ties such as chlorophyll fluorescence. We present an autonomous robotic platform designed to perform targeted fluorescence measurements on plant leaves. The system combines 3D plant reconstruction, geometric analysis, and motion planning to localize suitable measurement points and generate collision-free trajectories for a robotic manipulator. A dense 3D model of the plant is reconstructed from multi-view data and used to extract candidate leaf surfaces based on orientation, accessibility, and sensing constraints. These targets are then integrated into a task-level planning framework that guides the end-effector to precise contact or near-contact configurations required for point-based fluorescence acquisition. The platform enables automated, repeatable, and spatially resolved physiological measurements that go beyond passive imaging. By tightly coupling perception, geometric reasoning, and manipulation, the proposed system provides a robotics-driven approach to high-resolution plant phenotyping and opens new directions for autonomous agricultural inspection and plant-aware manipulation.
Chinese Translation
尽管大多数表型分析平台主要依赖于基于图像的测量，但先进的植物表征需要整合主动生理感知模式，例如叶绿素荧光。我们提出了一种自主机器人平台，旨在对植物叶片进行定向荧光测量。该系统结合了三维植物重建、几何分析和运动规划，以定位合适的测量点并为机器人操纵器生成无碰撞的轨迹。通过多视角数据重建出植物的密集三维模型，并根据方向、可达性和感知约束提取候选叶面。这些目标随后被整合到一个任务级规划框架中，引导末端执行器达到进行点基荧光采集所需的精确接触或近接触配置。该平台实现了自动化、可重复和空间分辨的生理测量，超越了被动成像。通过紧密结合感知、几何推理和操作，该系统提供了一种以机器人驱动的高分辨率植物表型分析方法，并为自主农业检查和植物感知操作开辟了新的方向。
cs.RO / 29 / 2608.27151
Planning a Shared Modular Fixture Layout Across Robotic Disassembly Stages
跨机器人拆解阶段的共享模块化夹具布局规划
Abstract
Stable support remains challenging in robotic disassembly of irregularly shaped products. As components are progressively removed, the available support surfaces, mass distribution, and task loads change throughout the process. A fixture layout designed for one workpiece state may therefore become infeasible at later stages, motivating unified support planning over the complete disassembly sequence. This paper presents a modular vacuum-based fixturing system that plans one shared support configuration for the complete disassembly sequence of a screwdriver or shaver, allowing each sequence to proceed without fixture reconfiguration. To search the mixed continuous--discrete layout space under repeated cross-stage evaluation, a denoising diffusion probabilistic model generates physics-informed initial configurations that are refined through Bayesian optimization. Robotic screw and component-removal experiments verified the disassembly feasibility of the planned layouts, while 11 directional-load tests quantified their stability. Comparisons between the measured operational loads and directional responses yielded mean empirical stability margins of 66.9% for the screwdriver and 81.6% for the shaver. These results demonstrate that a product-specific shared layout can provide stable support throughout the tested robotic disassembly sequence.
Chinese Translation
在不规则形状产品的机器人拆解过程中，稳定支撑仍然是一个挑战。随着组件的逐步移除，可用的支撑面、质量分布和任务负载在整个过程中不断变化。因此，为某一工件状态设计的夹具布局在后期阶段可能变得不可行，这促使我们在整个拆解序列中进行统一的支撑规划。本文提出了一种基于模块化真空夹具系统的方案，该系统为螺丝刀或剃须刀的完整拆解序列规划一个共享的支撑配置，使每个序列能够在不重新配置夹具的情况下进行。为了在重复的跨阶段评估下搜索混合的连续-离散布局空间，采用去噪扩散概率模型生成物理信息驱动的初始配置，并通过贝叶斯优化进行精细化。机器人螺丝和组件移除实验验证了规划布局的拆解可行性，而11个方向载荷测试量化了其稳定性。测量的操作载荷与方向响应之间的比较显示，螺丝刀的平均经验稳定裕度为66.9%，剃须刀为81.6%。这些结果表明，特定产品的共享布局能够在整个测试的机器人拆解序列中提供稳定的支撑。
cs.RO / 30 / 2608.27186
Task-space model-based control of pneumatic soft actuators
基于任务空间模型的气动软驱动器控制
Abstract
Soft actuators enable dexterous and compliant interaction, but closed-loop task-space control remains challenging due to strong nonlinearities, distributed deformation, and uncertainty in their dynamics. This paper presents a real-time dynamic-model-based task-space feedback and estimation framework based on a non-minimal coordinate discrete elastic rod model formulated in absolute coordinates with holonomic constraints. The resulting structure preserves distributed mechanics while maintaining computational efficiency through sparse system matrices, enabling real-time control with up to 10 discretized rods. A quasi-static feedforward inverse model is combined with a task-space PI controller and a dynamic observer that fuses measurement residuals as virtual forces, enabling full-state estimation from sparse sensing. The approach is experimentally validated on three planar pneumatic soft actuators with varying geometries. Across five tasks, including drawing the digits 0-9 across the workspace (3-18 mm/s tip speed), tracking periodic motion (up to 37 cm/s), cross-platform generalization, reduced sensing conditions, and real-time user-defined references, our method achieves 1.5-2.3 mm root mean square error (RMSE) for precision motions and 5.5-12.4 mm RMSE at 1-2 Hz. Results demonstrate that structured, non-minimal dynamic models can enable real-time, high-precision, moderate-bandwidth task-space control of planar soft pneumatic actuators in free space.
Chinese Translation
软驱动器能够实现灵活和顺应的交互，但由于其强非线性、分布变形和动态的不确定性，闭环任务空间控制仍然具有挑战性。本文提出了一种基于实时动态模型的任务空间反馈和估计框架，该框架基于在绝对坐标下构建的非最小坐标离散弹性杆模型，并包含全约束。所得到的结构在保持分布力学特性的同时，通过稀疏系统矩阵维持计算效率，使得能够实现对多达10根离散杆的实时控制。将准静态前馈逆模型与任务空间比例-积分（PI）控制器和动态观察器相结合，后者将测量残差融合为虚拟力，从而实现从稀疏传感中获得全状态估计。该方法在三种不同几何形状的平面气动软驱动器上进行了实验验证。在五个任务中，包括在工作空间内绘制数字0-9（尖端速度为3-18 mm/s）、跟踪周期性运动（最高可达37 cm/s）、跨平台泛化、降低传感条件和实时用户定义参考，我们的方法在精确运动中实现了1.5-2.3 mm的均方根误差（RMSE），在1-2 Hz时实现了5.5-12.4 mm的RMSE。结果表明，结构化的非最小动态模型能够实现平面气动软驱动器在自由空间中的实时、高精度、中等带宽任务空间控制。
cs.RO / 31 / 2608.27221
Tensegrity Continuum Robots Enable Task-Adaptive Morphologies for Cooperative Behaviors
张力完整体机器人实现任务自适应形态以支持协作行为
Abstract
Robots that can change their morphologies and behaviors for different tasks and environments hold great promise for adaptable, multifunctional systems. Modular reconfigurable robots (MRRs) can achieve such functionalities by docking and rearranging individual units, but most rely on rigid modules that lack structural compliance, resulting in limited capabilities. Continuum robots offer compliance through flexible backbones, yet they cannot self-reconfigure into task-adaptive multi-robot configurations. Here, we introduce an MRR that unifies the advantages of both architectures by combining a tensegrity-based compliant body with claw-based connection mechanisms. Each robot can manipulate and locomote independently, and multiple robots can self-reconfigure into different morphologies (e.g., chains, loops, branches) for cooperative manipulation and locomotion. We demonstrate the robots' capability across diverse tasks and environments, including coordinated object manipulation and transport, multimodal locomotion, and loco-manipulation in real-world scenarios. These results lay a foundation for adaptable and multifunctional robotic collectives, with broad potential applications in manufacturing, space exploration, and search-and-rescue operations.
Chinese Translation
能够根据不同任务和环境改变形态和行为的机器人在适应性和多功能系统中具有巨大的潜力。模块化可重构机器人（MRRs）通过对接和重新排列单个单元实现这种功能，但大多数依赖于缺乏结构柔性的刚性模块，导致能力有限。连续体机器人通过灵活的主干提供了柔性，但它们无法自我重构为任务自适应的多机器人配置。在此，我们介绍了一种模块化可重构机器人（MRR），它通过结合基于张力完整体的柔性主体和爪形连接机制，统一了两种架构的优点。每个机器人可以独立操控和移动，多个机器人可以自我重构为不同的形态（例如，链条、环、分支），以实现协作操控和移动。我们展示了机器人在多种任务和环境中的能力，包括协调物体操控和运输、多模态移动，以及在现实场景中的运动操控。这些结果为适应性和多功能的机器人集体奠定了基础，具有广泛的应用潜力，如制造、太空探索和搜救行动。
cs.RO / 32 / 2608.27225
STEP: State-Aware Task Estimation and Planning with Multi-Modal LLMs for Human-Robot Collaboration
STEP：基于状态感知的任务估计与规划，结合多模态大型语言模型用于人机协作
Abstract
Effective human-robot collaboration in industrial settings requires robots to understand human intentions and assist with task planning, reducing workload. Recent works have explored the use of Multi-modal Large Language Models (MM-LLMs) for task planning in such data-scarce scenarios, leveraging in-context learning to interpret user actions and generate long-horizon action plans in natural language. However, MM-LLMs inherently lack an understanding of system states and do not track state transitions, often leading to hallucinated actions that deviate from the intended goal. Additionally, generating action plans in natural language tends to limit the generated plans to a high level, introducing ambiguity in action execution. To address these limitations, we propose the State-aware Task Estimator and Planner (STEP), which prompts a MM-LLM to explicitly estimate the state of the system and predict the state transitions resulting from executed actions. By forecasting future states alongside actions, STEP ensures task-convergent planning while also providing additional assistance parameters necessary for executing the predicted actions. We evaluate STEP in a simulated environment using a robot assembly task. Our approach outperforms the state-of-the-art by 32.8% in action executability and 14.8% in final-state error.
Chinese Translation
在工业环境中，有效的人机协作要求机器人理解人类意图并协助任务规划，从而减轻工作负担。近期的研究探讨了在数据稀缺场景中使用多模态大型语言模型（MM-LLMs）进行任务规划，利用上下文学习来解释用户行为并生成自然语言中的长期行动计划。然而，MM-LLMs 本质上缺乏对系统状态的理解，并且不跟踪状态转变，常常导致产生偏离预期目标的虚假行动。此外，以自然语言生成行动计划往往将生成的计划限制在较高层次，导致行动执行中的模糊性。为了解决这些局限性，我们提出了状态感知任务估计与规划器（STEP），该方法促使 MM-LLM 明确估计系统状态并预测执行动作所导致的状态转变。通过预测未来状态与动作，STEP 确保任务收敛规划，同时提供执行预测动作所需的额外辅助参数。我们在模拟环境中使用机器人组装任务对 STEP 进行了评估。我们的方案在行动可执行性上比最先进的方法提高了 32.8%，在最终状态误差上提高了 14.8%。
cs.RO / 33 / 2608.27371
Embodied Scene Rearrangement Planning
具身场景重排规划
Abstract
This paper introduces Embodied Scene Rearrangement Planning (ESRP), a novel task requiring embodied agents to rearrange furniture in 3D scenes to match a target configuration using only egocentric observations and a top-down target layout. Unlike prior rearrangement tasks, ESRP precludes global state access and introduces mutual object occlusions, reflecting the practical constraints of real-world robotic deployment. These factors make aligning partial egocentric observations with the global target layout particularly challenging for long-horizon planning. To facilitate research, we present ESRP-Bench, a comprehensive benchmark built on OmniGibson featuring over 5,400 scene pairs and 8,200 objects. We define three multi-level metrics to evaluate rearrangement quality and provide four baselines: a hierarchical task-and-motion planning method, a vision-language-model-based method, and two learning-based approaches (IL and RL). Experimental results demonstrate that current methods struggle to complete the task efficiently, highlighting ESRP as a challenging frontier for embodied agents in scene understanding and long-horizon task planning. This work serves as a stepping stone toward deploying intelligent agents in real-world scenarios. Project page: https://pie-lab.cn/ESRP/.
Chinese Translation
本文介绍了具身场景重排规划（Embodied Scene Rearrangement Planning, ESRP），这是一项新颖的任务，要求具身智能体在三维场景中重新排列家具，以匹配目标配置，仅使用自我中心的观察和自上而下的目标布局。与以往的重排任务不同，ESRP 不允许访问全局状态，并引入了相互遮挡的物体，反映了现实世界机器人部署的实际限制。这些因素使得将部分自我中心的观察与全局目标布局对齐在长时间规划中尤为具有挑战性。为了促进研究，我们提出了 ESRP-Bench，这是一个基于 OmniGibson 的综合基准，包含超过 5400 对场景和 8200 个物体。我们定义了三个多层次指标来评估重排质量，并提供了四个基线：一种分层任务与运动规划方法、一种基于视觉语言模型的方法，以及两种基于学习的方法（IL 和 RL）。实验结果表明，目前的方法在高效完成任务方面存在困难，突显了 ESRP 在场景理解和长时间任务规划中对具身智能体的挑战性。这项工作为在现实场景中部署智能体奠定了基础。项目页面：https://pie-lab.cn/ESRP/
cs.RO / 34 / 2608.27384
FlashVLA: Streaming Action Decoding for Fast and Asynchronous VLA Inference
FlashVLA：快速和异步VLA推理的流式动作解码
Abstract
Vision-Language-Action (VLA) models are increasingly promising for robotic manipulation, yet their real-world deployment remains bottlenecked by high inference latency and unstable asynchronous execution. This challenge is particularly pronounced in flow-matching-based VLA models, where action decoding requires multiple iterative steps conditioned on the VLM context. While efficient inference methods improve control frequency and asynchronous methods reduce execution idle time, existing approaches often fail to jointly achieve low-latency inference and accurate, temporally consistent asynchronous execution. We introduce \textbf{FlashVLA}, a streaming action decoding framework that addresses both challenges in a unified formulation. FlashVLA maintains a streaming action buffer with multiple chunks at different noise levels and decodes them using chunk-wise causal attention. This design allows FlashVLA to produce one executable action chunk per inference step. Moreover, its chunk-wise autoregressive formulation implicitly preserves action continuity, enabling smooth asynchronous execution without extra future-state conditioning. Across extensive simulated and real-world experiments, FlashVLA substantially improves inference speed while maintaining strong task performance. It can achieve $\geq$30\,Hz control frequency on a single GPU with smooth asynchronous inference in real-world deployment.
Chinese Translation
视觉-语言-动作（VLA）模型在机器人操作中展现出越来越大的潜力，但其在实际应用中的部署仍然受到高推理延迟和不稳定的异步执行的瓶颈。这个挑战在基于流匹配的VLA模型中尤为明显，因为动作解码需要多个基于视觉语言模型（VLM）上下文的迭代步骤。尽管高效的推理方法提高了控制频率，而异步方法减少了执行空闲时间，但现有的方法往往无法同时实现低延迟推理和准确、时间一致的异步执行。我们提出了 extbf{FlashVLA}，一个流式动作解码框架，统一解决这两个挑战。FlashVLA维护一个具有多个不同噪声水平的流式动作缓冲区，并使用块级因果注意力进行解码。这一设计使得FlashVLA能够在每个推理步骤中生成一个可执行的动作块。此外，其块级自回归的形式隐式地保持了动作的连续性，从而实现平滑的异步执行，而无需额外的未来状态条件。在广泛的模拟和实际实验中，FlashVLA显著提高了推理速度，同时保持了强大的任务性能。它能够在单个GPU上实现$ extgeq$30 Hz的控制频率，并在实际部署中实现平滑的异步推理。
cs.RO / 35 / 2608.27406
CLAP: Cross-Embodiment Video World Models are Zero-Shot Physical Simulators
CLAP：跨体现视频世界模型是零-shot物理模拟器
Abstract
State-of-the-art action-conditioned video models are typically restricted to a single robot embodiment, preventing them from leveraging the vast corpus of heterogeneous video data that contains rich signals for learning generalizable physics. To bridge this gap, we introduce CLAP, a framework for cross-embodiment action-conditioned video generation capable of being trained on diverse, internet-scale videos across human and robotic agents. CLAP is grounded in the insight that universal physical laws govern spatiotemporal dynamics regardless of the actor. However, cross-embodiment learning is non-trivial because action representations vary sharply across robot platforms and are typically absent in human videos. CLAP addresses this fundamental challenge through the following core contributions. First, CLAP reconciles disparate action spaces using end-effector poses, language instructions, and latent actions. Second, to resolve their individual limitations, CLAP introduces a curriculum-based cross-embodiment learning recipe that first learns foundational physical priors across unlabeled video data using latent actions and subsequently grounds them in end-effector action spaces for zero-shot deployment to real-world tasks. Crucially, CLAP approaches or surpasses state-of-the-art single-embodiment video models in challenging environments like DROID. These performance advantages compound via few-shot adaptation to establish a novel paradigm for training single-embodiment video world models. Ultimately, CLAP delivers the most comprehensive suite of action-conditioned video world models to date - spanning diverse action-conditioning spaces (end-effector, language, and latent) and robot morphologies (including cross-embodiment, DROID, Bridge, bimanual YAM robots, and G1 humanoids). We open-source all code and models. Project Website at https://omni-clap.github.io .
Chinese Translation
最先进的基于动作的视频模型通常限制于单一的机器人体现，这使得它们无法利用包含丰富信号以学习可推广物理的异构视频数据的庞大语料库。为了解决这一问题，我们提出了CLAP，一个能够在跨人类和机器人代理的多样化互联网规模视频上进行训练的跨体现动作条件视频生成框架。CLAP的基础在于一个洞察，即普遍的物理法则支配着时空动态，无论参与者是谁。然而，跨体现学习并非易事，因为动作表示在不同机器人平台之间差异显著，并且在人体视频中通常缺失。CLAP通过以下核心贡献解决了这一基本挑战。首先，CLAP使用末端执行器姿态、语言指令和潜在动作来调和不同的动作空间。其次，为了克服各自的局限性，CLAP引入了一种基于课程的跨体现学习方案，该方案首先利用潜在动作在未标记的视频数据中学习基础物理先验，然后将其固化在末端执行器动作空间中，以实现对现实任务的零-shot部署。至关重要的是，CLAP在像DROID这样的挑战性环境中接近或超越了最先进的单体现视频模型。这些性能优势通过少量样本适应进一步增强，建立了一个训练单体现视频世界模型的新范式。最终，CLAP提供了迄今为止最全面的动作条件视频世界模型套件——涵盖多样的动作条件空间（末端执行器、语言和潜在）和机器人形态（包括跨体现、DROID、Bridge、双手YAM机器人和G1人形机器人）。我们开源了所有代码和模型。项目网站：https://omni-clap.github.io 。
cs.CV / 1 / 2608.26214
Surgical Video Generation From Diffusion to World Models: A Survey
从扩散到世界模型的外科视频生成：一项综述
Abstract
Surgical video data provides the primary training resource for models of intraoperative perception, surgical workflow understanding, and robotic decision-making. However, clinical data acquisition remains constrained by privacy, cost, and class imbalance. Surgical video generation has emerged as a transformative approach to addressing data scarcity and as a foundation for surgical simulation, training, and robotic policy learning. The field has developed rapidly without a clear conceptual framework. This survey organizes the 2024-2026 literature into three categories: unconditional generation, conditional generation, and world modeling generation, revealing a fundamental shift in how the task is defined from synthesizing visually plausible frames to modeling the causal dynamics of surgical scenes. We examine the persistent gap between pixel-level fidelity and clinical plausibility, and identify generalization, physical realism, controllability, and interpretability as bottlenecks. We further summarize experimental results of representative methods on public datasets to provide a quantitative reference for the field. This survey provides a structured overview of the current state and open challenges, offering a reference for researchers working at the intersection of intelligent perception, multi-modal fusion, generative AI, and surgical data science.
Chinese Translation
外科视频数据为术中感知模型、外科工作流程理解和机器人决策提供了主要的训练资源。然而，临床数据的获取仍受到隐私、成本和类别不平衡的限制。外科视频生成作为一种变革性的方法，已成为解决数据稀缺问题的途径，并为外科模拟、培训和机器人策略学习奠定了基础。该领域迅速发展，但缺乏明确的概念框架。本综述将2024-2026年的文献组织为三类：无条件生成、有条件生成和世界建模生成，揭示了任务定义从合成视觉上可信的帧到建模外科场景因果动态的根本转变。我们考察了像素级保真度与临床可信度之间的持续差距，并识别出泛化、物理现实性、可控性和可解释性作为瓶颈。此外，我们总结了代表性方法在公共数据集上的实验结果，以提供该领域的定量参考。本综述为当前状态和开放挑战提供了结构化的概述，为在智能感知、多模态融合、生成性人工智能和外科数据科学交叉领域工作的研究人员提供了参考。
cs.CV / 2 / 2608.26238
Procedura: Agentic 3D Modeling with Procedural Control
Procedura：具有过程控制的自主3D建模
Abstract
Native 3D generators now recover impressive mesh geometry from a single image. However, a dense mesh stays soft where a machined object should be sharp, it carries no part decomposition, and it exposes no parameter a user could edit. To address this, we explore the paradigm of 3D shape as code, leveraging and scaling the coding ability of an LLM for 3D modeling. We introduce Procedura, a novel 3D modeling agent framework that writes an object as a procedural assembly, a parametric program whose named parts are joined by typed, machine-checkable mates. From a text prompt, the agent plans the object as an assembly graph and writes the program part by part, solving each placement from the mated frames rather than guessing it, and admitting a part only once compile, mate, and connectivity checks pass. A decoupled vision critic then refines the assembly one diagnosed fix at a time. Moreover, the same graph carries per-part materials and a simulator-validated articulation. We evaluate on P3D-Bench under its assembly judge, and with the same judge on MechBench-36, our hard-surface benchmark. On both, Procedura outperforms state-of-the-art native 3D generators and every prior 3D-code agent on judged quality, produces the sharpest edges of any method we evaluate, and is the only one whose output is an editable, part-structured program.
Chinese Translation
当前的原生3D生成器能够从单幅图像中恢复出令人印象深刻的网格几何形状。然而，密集的网格在应当锋利的机械物体处显得模糊，且没有部件分解，也没有用户可以编辑的参数。为了解决这个问题，我们探索了将3D形状视为代码的范式，利用并扩展大型语言模型（LLM）在3D建模中的编码能力。我们引入了Procedura，一个新颖的3D建模代理框架，它将对象写作过程装配，即一个参数化程序，其命名部件通过类型化、机器可检查的连接件连接。代理从文本提示中规划对象的装配图，并逐部分编写程序，通过配对框架解决每个部件的位置，而不是进行猜测，并且仅在编译、配对和连通性检查通过后才允许一个部件的加入。一个解耦的视觉评估器随后逐个修正装配中的问题。此外，同一图形还携带每个部件的材料和经过模拟验证的关节运动。我们在P3D-Bench上进行评估，并在MechBench-36（我们的硬表面基准测试）上使用相同的评估标准。在这两个测试中，Procedura在评估质量上超越了最先进的原生3D生成器和所有先前的3D代码代理，产生了我们评估的任何方法中最锋利的边缘，并且是唯一一个输出为可编辑、部件结构化程序的方法。
cs.CV / 3 / 2608.26317
Modality Maturity Index: A benchmark for assessing multimodal capabilities of omni models
模态成熟度指数：评估全能模型多模态能力的基准
Abstract
Frontier language models are increasingly marketed as omni systems that can perceive and respond across modalities. Existing evaluation frameworks, however, focus almost exclusively on bimodal understanding, typically text plus one other modality. We propose the Modality Maturity Index (MMI), a benchmark designed to evaluate the multimodal capabilities of large language models across five modalities (text, image, audio, video and document) and combinations of up to three modalities in both inputs and outputs. MMI consists of 893 questions, each carefully crafted to require the model to demonstrate its understanding of multiple input modalities and to generate responses that incorporate various output formats. The questions are designed to be self-contained, with clear expectations for the correct modality or mix of modalities required for an accurate response. Every MMI prompt carries human-authored rubric criteria for each output modality expected in the response; a model's MMI Value expresses the average of the per-modality scores for each prompt. Because low scores can reflect either failure to generate a modality (lack of presence) or failure to generate correct content, we introduce also a supplementary Modality Presence Score (MPS), a per-prompt F1 over the expected output modalities. Applying MMI to five frontier multimodal models, we find that the MPS ranges from only 15.6 (Claude Opus 4.6) to 34.9 (GPT-5.4). Given the low availability of returned modalities to even grade, we report MPS as our main result pending model improvements. To assess the viability of judging output correctness with LLM judges and rubrics, we run a separate experiment with custom generation tools. On the assets that generates, we find that an LLM judge applying the rubrics agrees with rubric-blind human annotators (who score the outputs directly and never see the criteria) on 70.8% of judgments.
Chinese Translation
前沿语言模型越来越多地被宣传为能够跨模态感知和响应的全能系统。然而，现有的评估框架几乎专注于双模态理解，通常是文本加上另一种模态。我们提出了模态成熟度指数（Modality Maturity Index, MMI），这是一个旨在评估大型语言模型在五种模态（文本、图像、音频、视频和文档）及最多三种模态组合的多模态能力的基准。MMI包含893个问题，每个问题都经过精心设计，以要求模型展示其对多种输入模态的理解，并生成包含各种输出格式的响应。这些问题旨在自成体系，明确期望正确响应所需的模态或模态组合。每个MMI提示都有针对每种预期输出模态的人类编写的评分标准；模型的MMI值表示每个提示的每模态得分的平均值。由于低得分可能反映出未能生成某种模态（缺乏存在）或未能生成正确内容，我们还引入了补充的模态存在分数（Modality Presence Score, MPS），这是针对预期输出模态的每提示F1得分。将MMI应用于五个前沿多模态模型，我们发现MPS的范围仅为15.6（Claude Opus 4.6）到34.9（GPT-5.4）。考虑到返回模态的低可用性以至于无法评分，我们将MPS报告为我们的主要结果，待模型改进。为了评估使用大型语言模型（LLM）评判输出正确性的可行性，我们进行了一项使用自定义生成工具的独立实验。在生成的资产上，我们发现应用评分标准的LLM评审者与盲评分的人类注释者（直接评分输出且从未看到标准）在70.8%的判断上达成一致。
cs.CV / 4 / 2608.26355
Finding the Right Evidence: Factor-Guided Coarse-to-Fine Reasoning for Long Videos
寻找正确证据：针对长视频的因素引导粗到细推理
Abstract
While LVLMs rapidly improve, long-video question answering still remains challenging: relevant evidence is sparse, and question-relevant context often fails to provide cues that discriminate the correct answer from plausible alternatives. Diagnostic analysis on a manually annotated subset of MMR-V shows that prior agentic systems substantially improve cue retrieval over direct VLM inference yet fail to achieve a corresponding gain in answer accuracy, indicating that the bottleneck lies in option-discriminative evidence rather than topical relevance alone. We propose PACE (Progressive Acquisition of Critical Evidence), a factor-guided framework for long-video evidence acquisition. PACE proceeds in two stages: it first indexes clip-level descriptions guided by question-derived factors without observing the candidate answers; it then uses the candidate answers to derive contrastive cues and queries the index for verification. On MMR-V with the open-source Qwen3-VL backbone, PACE achieves 42.6% accuracy, outperforming direct inference and prior agentic baselines including Deep Video Discovery (DVD). On the same diagnostic subset, PACE recovers 66.9% of the annotated cues, providing empirical evidence that its gains are associated with improved evidence recovery rather than stronger answer-side priors alone. Consistent gains over DVD on LVBench, Video-MME, EgoSchema, and LongVideoBench suggest that option-aware evidence acquisition transfers beyond MMR-V. Code is available at https://github.com/HKUST-KnowComp/PACE.
Chinese Translation
尽管长视频语言模型（LVLMs）迅速发展，长视频问答仍然面临挑战：相关证据稀缺，且与问题相关的上下文往往无法提供区分正确答案与合理替代答案的线索。对手动标注的MMR-V子集的诊断分析表明，先前的代理系统在线索检索方面显著优于直接的VLM推理，但未能在答案准确性上实现相应的提升，这表明瓶颈在于选项区分证据而不仅仅是主题相关性。我们提出了PACE（关键证据的渐进获取），这是一个针对长视频证据获取的因素引导框架。PACE分为两个阶段：首先，它在不观察候选答案的情况下，根据问题导出的因素对剪辑级描述进行索引；然后，它利用候选答案推导对比线索，并查询索引进行验证。在使用开源Qwen3-VL骨干网络的MMR-V上，PACE实现了42.6%的准确率，超越了直接推理和先前的代理基线，包括深度视频发现（Deep Video Discovery，DVD）。在同一诊断子集上，PACE恢复了66.9%的标注线索，提供了实证证据，表明其提升与改善的证据恢复相关，而不仅仅是更强的答案侧先验。在LVBench、Video-MME、EgoSchema和LongVideoBench上相对于DVD的一致性提升表明，关注选项的证据获取超越了MMR-V。代码可在https://github.com/HKUST-KnowComp/PACE获取。
cs.CV / 5 / 2608.26363
A Unified Framework for the Mechanics of Information in Convolutional Neural Network Image Space
卷积神经网络图像空间中信息力学的统一框架
Abstract
This paper introduces a unified mathematical framework for modeling information propagation through convolutional neural networks (CNNs), with the aim of connecting descriptions of physical space and information space. A correspondence is presented linking discrete filter symmetry and the relativistic energy--momentum relation under the widely used nonlinear rectified convolution operation. Specifically, symmetric filter components (e.g. the sum $\Sigma = [1,1]$) operate analogously to rest energy $mc^2$ in preserving the image centre of mass (e.g. isotropic diffusion), whereas antisymmetric components (e.g. the gradient $\nabla = [-1,1]$) operate analogously to the momentum term $pc$ in generally inducing a displacement (e.g. vibration or translation). For typical small discrete filters, this displacement is determined by the ratio of antisymmetric to total filter energy, analogously to how the displacement of a relativistic particle relates to a Lorentz transform with beta parameter $\beta = \frac{v}{c}=\frac{pc}{E}$ equal to the ratio of momentum $pc$ to total energy $E$. Repeated filtering leads to the Gaussian scale-space and emergent scale-invariant features. These constructions share a Laplacian-driven structure with the classical heat (diffusion) equation and, via standard mathematical correspondences, with the Schr\"odinger equation and aspects of the Friedmann equations, together with emergent Morse topological structure. Demonstrations in 3D images reveal blob-like, scale-invariant Morse critical points in images spanning a wide range of physical scales, including organic sugar molecules and inorganic silicon crystals, human and primate brains in magnetic resonance images (MRI), galaxies and the cosmic microwave background (CMB).
Chinese Translation
本文介绍了一个统一的数学框架，用于建模信息在卷积神经网络（CNNs）中的传播，旨在连接物理空间和信息空间的描述。提出了一种对应关系，将离散滤波器对称性与广泛使用的非线性整流卷积操作下的相对论能量-动量关系联系起来。具体而言，对称滤波器组件（例如，和 $ ext{Σ} = [1,1]$）在保持图像质心（例如，各向同性扩散）方面类似于静止能量 $mc^2$ 的作用，而反对称组件（例如，梯度 $
abla = [-1,1]$）在一般性诱导位移（例如，振动或平移）方面类似于动量项 $pc$ 的作用。对于典型的小离散滤波器，这种位移由反对称能量与总滤波器能量的比率决定，类似于相对论粒子的位移与洛伦兹变换的关系，其中贝塔参数 $eta = rac{v}{c}= rac{pc}{E}$ 等于动量 $pc$ 与总能量 $E$ 的比率。重复滤波导致高斯尺度空间和涌现的尺度不变特征。这些构造与经典热（扩散）方程共享拉普拉斯驱动结构，并通过标准数学对应关系与薛定谔方程及弗里德曼方程的某些方面相联系，同时涌现出莫尔斯拓扑结构。在三维图像中的演示揭示了在跨越广泛物理尺度的图像中出现的类斑点、尺度不变的莫尔斯临界点，包括有机糖分子和无机硅晶体、人类和灵长类动物的大脑在磁共振成像（MRI）中的表现，以及星系和宇宙微波背景（CMB）。
cs.CV / 6 / 2608.26382
VIPER: An Expert-Curated Benchmark for Vision-Language Models in Veterinary Pathology
VIPER：兽医病理学中视觉-语言模型的专家策划基准
Abstract
Pathology vision-language models are advancing rapidly, yet existing benchmarks remain focused on human tissue, particularly oncology, leaving non-human pathology largely unaddressed. This gap is especially important in toxicologic pathology, where microscopic tissue examination of laboratory animals is a core component of preclinical drug safety assessment. To address it, we introduce VIPER, the first expert-curated benchmark for vision-language model evaluation in toxicologic pathology. VIPER contains 1,251 questions associated with 419 H&E-stained rat histology images across seven organ systems, covering multiple-choice, KPrim, and free-text formats. All questions were curated and validated by board-certified veterinary pathologists. In total, we benchmarked 16 models, including two newly introduced veterinary-pathology models, seven human pathology-specialized models, and seven general-purpose frontier models. The results identify a substantial domain gap between veterinary and human pathology, expose the risk of over-diagnosis of normal tissue in frontier models, and show that domain-specific training remains critical for visually grounded predictions. VIPER data and evaluation code are available at https://github.com/mahmoodlab/viper.
Chinese Translation
病理视觉-语言模型正在迅速发展，但现有基准仍然集中于人类组织，特别是肿瘤学，导致非人类病理学在很大程度上未得到关注。这一差距在毒理病理学中尤为重要，因为实验动物的显微组织检查是临床前药物安全性评估的核心组成部分。为了解决这一问题，我们推出了VIPER，这是第一个针对毒理病理学中视觉-语言模型评估的专家策划基准。VIPER包含与419张H&E染色的鼠类组织学图像相关的1,251个问题，涵盖七个器官系统，问题形式包括多项选择、KPrim和自由文本。所有问题均由获得认证的兽医病理学专家策划和验证。总的来说，我们对16个模型进行了基准测试，包括两个新引入的兽医病理模型、七个人类病理专用模型和七个通用前沿模型。结果表明，兽医病理与人类病理之间存在显著的领域差距，揭示了前沿模型在正常组织过度诊断的风险，并显示领域特定的训练对于视觉基础预测仍然至关重要。VIPER数据和评估代码可在https://github.com/mahmoodlab/viper获取。
cs.CV / 7 / 2608.26471
Mapping Woody Vegetation from Multi-Source Imagery and Prediction Fusion for Enhanced Data Efficiency and Accuracy
基于多源影像和预测融合的木本植被制图方法以提高数据效率和准确性
Abstract
Tree cover maps are a fundamental remote sensing product, used to derive ecological insights about the landscape and are essential to change detection, vegetation mapping and fire monitoring programs. However, comprehensive tree cover mapping requires reliable and high-quality imagery, free of cloud and weather defects to ensure accurate model outputs. Deep learning approaches can generate high quality maps with minimal human intervention but require large amounts of human annotated data to be successful. In this work we propose a framework consisting of methods that aim to improve the data efficiency and robustness of deep learning models using data fusion techniques to segment woody vegetation defined as vegetation over the height of 2m across the state of New South Wales, Australia. To improve robustness against varying image quality, we propose an image composition method that normalizes the imagery and removes defects, whilst also minimizing the reliance on individual image quality by proposing a prediction fusion method. The two methods resulted in an error reduction of 38.2% and 53.6% respectively compared to single-source imagery. To address deep learning approaches' limitation of requiring large amounts of data, we apply label transfer to multiple sources of imagery as a form of data augmentation to improve data efficiency. Learning from multiple image sources was shown to be the biggest improvement in performance, resulting in an error reduction between 28.1% to 76.2% across the different validation experiments, whilst reducing the standard deviation of performance across image dates by a factor of 13.
Chinese Translation
树冠覆盖图是基本的遥感产品，用于推导关于景观的生态洞察，并且对变化检测、植被制图和火灾监测程序至关重要。然而，全面的树冠覆盖制图需要可靠且高质量的影像，必须无云和天气缺陷，以确保模型输出的准确性。深度学习方法能够在最小人工干预的情况下生成高质量的地图，但成功实施需要大量人工标注的数据。在本研究中，我们提出了一种框架，旨在通过数据融合技术提高深度学习模型的数据效率和稳健性，以对澳大利亚新南威尔士州的木本植被（定义为高度超过2米的植被）进行分割。为了提高对不同影像质量的稳健性，我们提出了一种影像组合方法，该方法规范化影像并去除缺陷，同时通过提出预测融合方法来最小化对单个影像质量的依赖。这两种方法分别使得错误率较单源影像降低了38.2%和53.6%。为了应对深度学习方法需要大量数据的限制，我们将标签转移应用于多源影像，作为一种数据增强形式，以提高数据效率。从多个影像源学习被证明是性能提升的最大因素，在不同的验证实验中，错误率降低了28.1%至76.2%，同时将不同影像日期的性能标准差降低了13倍。
cs.CV / 8 / 2608.26476
Zero-Shot Video Restoration and Enhancement with Text-to-Image Latent Diffusion Models and Multi-Modal References
基于文本到图像潜在扩散模型和多模态参考的零样本视频修复与增强
Abstract
Zero-shot image restoration methods with text-to-image latent diffusion models have achieved great success in universal image restoration tasks without training. However, applying them to video restoration will result in severe temporal flickering. In this paper, we propose a novel framework for zero-shot video restoration and enhancement which uses a text-to-image latent diffusion model and multi-modal references. Through the proposed dual prompt tuning inversion and sampling, the inference time can be reduced to nearly 1/3 of the original. The performance and temporal consistency can be also significantly stregthened. By using the proposed texture-aware video token merging, the temporal correlation between frames can be further utilized to improve the temporal consistency. We futher propose the referenced self-attention and referenced token merging to support image reference. Experimental results demonstrate the superiority of the proposed method in restoring and enhancing temporally consistent videos.
Chinese Translation
基于文本到图像潜在扩散模型的零样本图像修复方法在无需训练的情况下已在通用图像修复任务中取得了显著成功。然而，将其应用于视频修复会导致严重的时间闪烁。本文提出了一种新颖的零样本视频修复与增强框架，该框架利用文本到图像潜在扩散模型和多模态参考。通过提出的双提示调优反演和采样，推理时间可以缩短至原来的近1/3。同时，性能和时间一致性也得到了显著增强。通过使用提出的纹理感知视频标记合并，可以进一步利用帧之间的时间相关性来改善时间一致性。我们进一步提出了参考自注意力和参考标记合并，以支持图像参考。实验结果表明，所提出的方法在修复和增强时间一致性视频方面具有优越性。
cs.CV / 9 / 2608.26489
Learning Woody Clearing With Loss Alignment for Zero-Shot Regrowth and Woody Segmentation
基于损失对齐的木质清理学习：零样本再生与木质分割
Abstract
Detecting woody clearing is vital for managing biodiversity. Deep learning models can detect change in woody vegetation from bitemporal remote sensing imagery, however generated products may not meet end-user specifications due to unaligned loss definitions. Further limitations of deep learning models are the reliance on large datasets which can be difficult to attain for spatially rare and ambiguous events such as regrowth detection. In this work we train a model to detect woody change using bitemporal Sentinel-2 imagery consisting of 7 years' worth of annual imagery across the state of New South Wales, Australia. To align the objective of the model with end-user metrics, we introduce the loss scaling coefficient $\alpha$ which transforms the objective to optimize for specific $F_{\beta}$ scores. Introducing $\alpha$ was found to increase precision by 1.85x or recall by 1.12x. We propose input imagery augmentation and generation techniques that allow the woody change detection model to zero-shot transfer to regrowth and woody segmentation tasks. For woody segmentation, image generation techniques using activation maximization with low $\alpha$ values for stability and image generation techniques derived from handcrafted features utilizing a mosaic of clearing patches and artificial trees for contextual grounding were found to outperform prior woody segmentation works of the study area, reducing the overall error by up to 18.2%. For zero-shot woody regrowth, creating pseudo-post and prior images resulted in the model achieving an F1 score of 0.845, creating a foundation for future regrowth detection work.
Chinese Translation
检测木质清理对于管理生物多样性至关重要。深度学习模型能够从双时相遥感影像中检测木质植被的变化，然而，由于损失定义未对齐，生成的产品可能无法满足最终用户的规格。深度学习模型的进一步局限在于对大规模数据集的依赖，而这些数据集对于空间稀缺和模糊事件（如再生检测）往往难以获得。在本研究中，我们训练了一个模型，利用来自澳大利亚新南威尔士州7年年度影像的双时相Sentinel-2影像来检测木质变化。为了将模型的目标与最终用户的指标对齐，我们引入了损失缩放系数$eta$，该系数将目标转化为优化特定的$F_{eta}$分数。引入$eta$后，发现精确度提高了1.85倍，召回率提高了1.12倍。我们提出了输入影像增强和生成技术，使木质变化检测模型能够零样本迁移到再生和木质分割任务。在木质分割方面，使用低$eta$值的激活最大化图像生成技术和利用清理斑块和人工树木的手工特征构建的图像生成技术被发现优于该研究区域之前的木质分割工作，整体误差降低了多达18.2%。对于零样本木质再生，通过创建伪后期和前期影像，模型实现了0.845的F1分数，为未来的再生检测工作奠定了基础。
cs.CV / 10 / 2608.26495
Video-FLAIR: Not Whether to Reason, But How
视频-FLAIR：不是是否推理，而是如何推理
Abstract
Multimodal queries can require different types of reasoning. Some can be answered via perceptual reasoning, extracting information directly from the visual signal, while others require compositional reasoning that combines observations or deliberative reasoning that evaluates competing hypotheses. However, many existing methods apply a uniform reasoning strategy across queries, leading to unnecessary computation on simple tasks and insufficient reasoning on complex ones. We introduce Video-FLAIR, a training framework that learns to select the appropriate reasoning mode for each query using reinforcement learning. During training, the model generates responses under all three modes for the same prompt, enabling direct comparison. A composite reward compares these responses to favor the most effective one based on correctness, grounding, and cost, while discouraging unsupported or misaligned deliberation. This yields a supervision signal for learning adaptive reasoning without per-query annotations. Video-FLAIR improves accuracy over the Qwen2.5-VL base model by +5.4 on MathVista, +4.8 on Video-Holmes, and +4.8 on Video-MMMU, while reducing average token usage to 95 compared to 417 for always-thinking baselines.
Chinese Translation
多模态查询可能需要不同类型的推理。有些可以通过感知推理来回答，直接从视觉信号中提取信息，而另一些则需要组合推理，结合观察结果，或深思熟虑的推理，评估竞争假设。然而，许多现有方法在查询中应用统一的推理策略，导致在简单任务上产生不必要的计算，在复杂任务上则推理不足。我们提出了视频-FLAIR，这是一种训练框架，利用强化学习学习为每个查询选择适当的推理模式。在训练过程中，模型针对同一提示生成三种模式下的响应，从而实现直接比较。复合奖励将这些响应进行比较，优先考虑基于正确性、基础性和成本的最有效响应，同时抑制不支持或不一致的深思熟虑。这为学习自适应推理提供了监督信号，而无需逐查询注释。视频-FLAIR在MathVista上比Qwen2.5-VL基础模型提高了5.4的准确率，在Video-Holmes上提高了4.8，在Video-MMMU上提高了4.8，同时将平均标记使用量减少到95，相较于始终思考基线的417。
cs.CV / 11 / 2608.26500
Systematic Literature Review of Machine Learning Models and Applications for Text Recognition
机器学习模型及其在文本识别中的应用的系统文献综述
Abstract
Optical Character Recognition (OCR) for text recognition using machine vision has significantly improved, particularly when handling heterogeneous textual data. Traditional OCR models struggle with script variations, writing styles, and degraded documents. Advancements in technology are leading to new AI models with improved architecture for handling multiple languages and complex data formats. Despite this progress, a comprehensive evaluation of OCR advancements remains limited. Based on the established preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines, this literature review presents an extensive assessment of OCR research to trace the evolution of AI models over the past decade. It explores the transition in AI models, application domains, data types, linguistic coverage, and challenges. Through a detailed analysis of 97 selected studies published during January 2015 - January 2025, key OCR models are identified, and their performance, strengths, and limitations are analyzed. The findings highlight how OCR technologies have evolved to address structured and unstructured text, scene text recognition, and multilingual processing. Unresolved challenges include limited resources for underrepresented languages, high variability in handwritten text, visual similarity among characters, and constraints in real-time OCR applications. To address these issues, several promising approaches are proposed. Key suggestions include self-supervised learning, multimodal AI, automated machine learning (AutoML), AI-assisted postprocessing, tiny machine learning (TinyML), and the creation of joint corpora for script matching. The future recommendations aim to enhance OCR accuracy and tackle the challenges identified for real-time industrial applications. This study will guide future research and establish a foundation for OCR field.
Chinese Translation
光学字符识别（OCR）在利用机器视觉进行文本识别方面取得了显著进展，尤其是在处理异构文本数据时。传统的OCR模型在应对脚本变体、书写风格和退化文档时表现不佳。技术的进步催生了新的人工智能（AI）模型，这些模型在处理多种语言和复杂数据格式方面具有更好的架构。尽管取得了这些进展，但对OCR进展的全面评估仍然有限。基于已建立的系统评价和荟萃分析的优选报告项目（PRISMA）指南，本次文献综述对OCR研究进行了广泛评估，以追踪过去十年AI模型的演变。它探讨了AI模型的转变、应用领域、数据类型、语言覆盖范围和面临的挑战。通过对2015年1月至2025年1月期间发表的97项选定研究的详细分析，识别出关键的OCR模型，并分析了它们的性能、优势和局限性。研究结果突显了OCR技术如何演变以应对结构化和非结构化文本、场景文本识别以及多语言处理。尚未解决的挑战包括对代表性不足语言的资源有限、手写文本的高变异性、字符之间的视觉相似性以及实时OCR应用的限制。为解决这些问题，提出了几种有前景的方法。主要建议包括自监督学习、多模态AI、自动化机器学习（AutoML）、AI辅助后处理、微型机器学习（TinyML）以及创建用于脚本匹配的联合语料库。未来的建议旨在提高OCR的准确性，并解决为实时工业应用识别的挑战。本研究将为未来的研究提供指导，并为OCR领域奠定基础。
cs.CV / 12 / 2608.26504
NeuDonatello: Uncertainty-Aware Framework for Accurate Neural SDF Learning
NeuDonatello：一种关注不确定性的准确神经SDF学习框架
Abstract
Neural surface reconstruction has emerged as a powerful paradigm for recovering high-quality 3D surfaces from multi-view images. However, recovering accurate geometry solely from RGB images remains challenging due to uncertainties arising from textureless regions, occlusions, and inherent scene ambiguities. Existing methods often overlook such uncertainties, leading to inaccurate estimates of the signed distance function (SDF). We introduce NeuDonatello, a novel framework that models and leverages SDF uncertainty to improve surface reconstruction. Central to our approach is to model spatially varying uncertainty using a Monte Carlo sampling strategy. Using this uncertainty, we develop an adaptive regularization that selectively strengthens geometric constraints where RGB supervision is unreliable, avoiding incorrect surface reconstruction. We further introduce an uncertainty-aware scale parameter for the SDF-to-density conversion. Conditioned on uncertainty, this design enables more accurate modeling of spatially varying densities. Extensive experiments demonstrate that NeuDonatello achieves state-of-the-art reconstruction accuracy, with robust performance across diverse scenes using only posed RGB images.
Chinese Translation
神经表面重建已成为从多视角图像恢复高质量3D表面的强大范式。然而，仅从RGB图像恢复准确几何形状仍然具有挑战性，因为无纹理区域、遮挡和固有场景模糊性带来了不确定性。现有方法往往忽视这些不确定性，导致对有符号距离函数（SDF）的不准确估计。我们提出了NeuDonatello，这是一种新颖的框架，旨在建模和利用SDF不确定性以改善表面重建。我们方法的核心是使用蒙特卡洛采样策略建模空间变化的不确定性。利用这种不确定性，我们开发了一种自适应正则化，选择性地加强几何约束，以应对RGB监督不可靠的情况，从而避免错误的表面重建。我们进一步引入了一种关注不确定性的尺度参数，用于SDF到密度的转换。在不确定性的条件下，这一设计使得空间变化密度的建模更加准确。大量实验表明，NeuDonatello在仅使用已定位的RGB图像的情况下，达到了最先进的重建精度，并在多样化场景中表现出强大的鲁棒性。
cs.CV / 13 / 2608.26517
HUG-VIS: A Multimodal Benchmark for Human-centered Understanding and Generation in Visual Intelligence
HUG-VIS：一个以人为中心的视觉智能理解与生成的多模态基准
Abstract
Visual intelligence seeks to perceive, interpret, and synthesize the visual world and is central to modern computer vision. Human-centered visual intelligence is especially demanding because it studies people as expressive, socially situated subjects whose meaning is rarely conveyed by appearance alone. It couples vision with audio and language across four representative tasks: human emotion recognition, human video generation, human voice cloning, and human video matting. Yet existing resources remain task-specific, providing modalities and annotations for individual problems rather than a shared foundation coordinating understanding and generation. This limits multimodal signal use and broader research. We address this gap with HUG-VIS, a unified benchmark for Human-centered Understanding and Generation in Visual Intelligence. It contains 8,400 seated half-body videos of 30 professional actors, each performing the same 280 emotion-action-prompt assignments under a controlled Mandarin studio protocol, with synchronized video, audio, text, and alpha mattes. We evaluate diverse open- and closed-source models across the four tasks under a unified zero-shot protocol using automatic metrics, criterion-specific mean opinion scores, and multiple cross-task analyses. Results show that (i) linguistic content dominates current emotion recognition, while purely visual affect recognition is weakest; (ii) in video generation and voice cloning, automatic metrics and human judgment agree overall but differ in their top rankings, requiring joint reporting; (iii) boundary fidelity under motion is the main remaining obstacle for human matting; and (iv) task difficulty varies across emotions, models, and metrics, with notable cross-task correlations. The dataset and results are available at https://github.com/GML-MMGroup/HUG-VIS.
Chinese Translation
视觉智能旨在感知、解释和合成视觉世界，是现代计算机视觉的核心。以人为中心的视觉智能尤其具有挑战性，因为它研究的是作为富有表现力的、社会情境中的主体的人，其意义往往无法仅通过外表传达。它将视觉与音频和语言结合，涵盖四个代表性任务：人类情感识别、人类视频生成、人类声音克隆和人类视频抠图。然而，现有资源仍然是任务特定的，为单个问题提供模态和注释，而不是协调理解与生成的共享基础。这限制了多模态信号的使用和更广泛的研究。我们通过HUG-VIS来填补这一空白，这是一个统一的以人为中心的视觉智能理解与生成基准。它包含了30名专业演员的8400个坐姿半身视频，每个演员在受控的普通话录音室协议下执行相同的280个情感-动作-提示任务，并提供同步的视频、音频、文本和alpha抠图。我们在统一的零样本协议下，使用自动度量、特定标准的平均意见分数和多项跨任务分析，评估不同的开源和闭源模型在这四个任务中的表现。结果显示：(i) 语言内容主导了当前的情感识别，而纯视觉的情感识别最弱；(ii) 在视频生成和声音克隆中，自动度量和人类判断总体一致，但在排名上存在差异，需要联合报告；(iii) 运动下的边界保真度是人类抠图的主要障碍；(iv) 任务难度在情感、模型和度量之间有所不同，且存在显著的跨任务相关性。数据集和结果可在https://github.com/GML-MMGroup/HUG-VIS获取。
cs.CV / 14 / 2608.26531
FAN-LoRA: A Fourier-Adaptive Nonlinear Low-Rank Adaptor for Medical Foundation Model Domain Adaptation
FAN-LoRA：一种用于医学基础模型领域适应的傅里叶自适应非线性低秩适配器
Abstract
The advent of vision foundation models, notably the Segment Anything Model (SAM), has catalyzed significant advancements in natural image segmentation. However, their direct transfer to medical imaging remains severely bottlenecked by profound domain gaps, such as cross-modality and cross-center shifts. Existing Parameter-Efficient Fine-Tuning (PEFT) methods facilitate the adaptation of SAM to medical domains; nevertheless, they frequently suffer from performance degradation under severe distribution shifts. This vulnerability primarily stems from the implicit entanglement of heterogeneous frequency components within a shared low-rank subspace, which directly exacerbates sub-optimal structural alignment and localized boundary blurring. To overcome this representational bottleneck, we propose the Fourier-Adaptive Nonlinear Low-Rank Adaptor (FAN-LoRA), a novel frequency-decoupled fine-tuning architecture. FAN-LoRA explicitly separates the optimization space by employing a B-spline-driven low-pass branch for global structural alignment, synergistically coupled with a discrete Fourier high-pass branch for local textural compensation. Extensive experiments across three challenging cross-modality and cross-center benchmarks demonstrate that FAN-LoRA consistently outperforms state-of-the-art PEFT baselines. Compared to the strongest competitors, our method achieves consistent improvements in average Dice scores and notable reductions in boundary errors, while maintaining a compact module size without compromising computational efficiency.
Chinese Translation
视觉基础模型的出现，特别是Segment Anything Model (SAM)，催生了自然图像分割领域的重大进展。然而，这些模型在医学成像中的直接迁移仍然受到严重的领域差距的制约，例如跨模态和跨中心的偏移。现有的参数高效微调（PEFT）方法虽然能够促进SAM在医学领域的适应，但在严重的分布偏移下往往表现出性能下降。这种脆弱性主要源于异构频率成分在共享低秩子空间中的隐性纠缠，这直接加剧了次优结构对齐和局部边界模糊。为了解决这一表征瓶颈，我们提出了傅里叶自适应非线性低秩适配器（FAN-LoRA），这是一种新颖的频率解耦微调架构。FAN-LoRA通过采用基样条驱动的低通分支进行全局结构对齐，明确分离优化空间，并与离散傅里叶高通分支协同耦合，以实现局部纹理补偿。在三个具有挑战性的跨模态和跨中心基准测试中，广泛的实验表明，FAN-LoRA始终优于最先进的PEFT基线。与最强的竞争对手相比，我们的方法在平均Dice分数上实现了一致的提升，并显著减少了边界错误，同时保持了紧凑的模块大小而不影响计算效率。
cs.CV / 15 / 2608.26580
Visual Information-Guided Parallel Decoding for Diffusion Multimodal Large Language Models
视觉信息引导的扩散多模态大语言模型并行解码
Abstract
Diffusion multimodal large language models (dMLLMs) have recently emerged as a new decoding paradigm for multimodal generation. Starting from a fully masked sequence, dMLLMs progressively decode the sequence by unmasking a subset of the remaining masked positions at each step. Since the selected tokens serve as the prediction context for subsequent steps, deciding which tokens to decode is crucial to the quality of the final output. The most common strategy prioritizes tokens based on a certainty measure that tends to favor tokens frequently observed in the training data. Recent approaches instead order tokens according to their influence on subsequent predictions, but do not explicitly account for the input image. We propose the Visual Information-Guided Sampler (VIG-Sampler), which prioritizes tokens based on their attention to image tokens. We further impose a constraint that penalizes candidate tokens whose image-attention distributions are similar to those of previously selected tokens, thereby increasing the information gain of the decoded subset. Extensive experiments on 7 captioning and VQA benchmarks with 3 open-source dMLLMs demonstrate the effectiveness of VIG-Sampler, which outperforms the Info-Gain Sampler by an average of 19.3 CIDEr points across the captioning benchmarks and surpasses it on COCO Caption while using only half as many decoding steps.
Chinese Translation
扩散多模态大语言模型（dMLLMs）最近作为多模态生成的新解码范式出现。从一个完全被掩蔽的序列开始，dMLLMs通过在每一步解开一部分剩余的掩蔽位置，逐步解码序列。由于所选的标记作为后续步骤的预测上下文，因此决定解码哪些标记对最终输出的质量至关重要。最常见的策略是根据一种倾向于优先考虑在训练数据中频繁出现的标记的确定性度量来优先选择标记。最近的方法则根据标记对后续预测的影响来排序标记，但没有明确考虑输入图像。我们提出了视觉信息引导采样器（Visual Information-Guided Sampler, VIG-Sampler），该采样器根据标记对图像标记的关注程度来优先选择标记。我们进一步施加了一项约束，惩罚那些图像关注分布与之前选择的标记相似的候选标记，从而增加解码子集的信息增益。在7个图像描述和视觉问答（VQA）基准上进行的广泛实验表明，VIG-Sampler的有效性，其在图像描述基准上平均比信息增益采样器（Info-Gain Sampler）高出19.3 CIDEr点，并在COCO Caption上超过了它，同时仅使用了一半的解码步骤。
cs.CV / 16 / 2608.26589
DPA-I2P: Depth-Guided Projective Alignment for Image-to-Point-Cloud Registration in Autonomous Driving
DPA-I2P：用于自主驾驶中图像与点云配准的深度引导投影对齐
Abstract
Image-to-Point Cloud Registration aims to estimate the camera pose of a given image within a 3D scene point cloud, which is a fundamental task in autonomous driving and large-scale outdoor localization. Recent implicit correspondence learning methods have improved registration performance by learning cross-modal alignment in an end-to-end framework, leading to more accurate camera pose estimation. However, due to the inherent modality discrepancy between images and sparse LiDAR point clouds, reliable cross-modal correspondence learning remains challenging. To address this issue, we propose Depth-Guided Projective Alignment for Image-to-Point-Cloud Registration (DPA-I2P). Unlike naive depth or feature concatenation, Ray-Conditioned Metric Depth Encoding (RMDE) and Projection-Consistent Vision Lifting (PVL) exploit depth and visual cues in a structured, geometry-aware manner. In addition, Cross-Modal Query Pruning (CQP) suppresses unreliable queries during early refinement to improve matching stability. Experiments on KITTI and nuScenes demonstrate the effectiveness of the proposed method. On KITTI, DPA-I2P reduces RTE and RRE by 45.0% and 55.6% over the strongest implicit baseline, respectively. On nuScenes, DPA-I2P also improves registration accuracy over the evaluated baselines, suggesting better transferability to different driving scenes.
Chinese Translation
图像与点云配准旨在估计给定图像在三维场景点云中的相机位姿，这是自主驾驶和大规模户外定位中的一项基础任务。最近的隐式对应学习方法通过在端到端框架中学习跨模态对齐，提升了配准性能，从而实现了更准确的相机位姿估计。然而，由于图像与稀疏激光雷达点云之间固有的模态差异，可靠的跨模态对应学习仍然具有挑战性。为了解决这一问题，我们提出了用于图像与点云配准的深度引导投影对齐方法（DPA-I2P）。与简单的深度或特征拼接不同，射线条件度量深度编码（Ray-Conditioned Metric Depth Encoding, RMDE）和投影一致视觉提升（Projection-Consistent Vision Lifting, PVL）以结构化和几何感知的方式利用深度和视觉线索。此外，跨模态查询修剪（Cross-Modal Query Pruning, CQP）在早期精化过程中抑制不可靠的查询，以提高匹配的稳定性。在KITTI和nuScenes上的实验表明了所提方法的有效性。在KITTI数据集上，DPA-I2P分别比最强的隐式基线降低了45.0%的重投影误差（RTE）和55.6%的旋转重投影误差（RRE）。在nuScenes上，DPA-I2P也提高了与评估基线相比的配准精度，表明其在不同驾驶场景中的更好迁移性。
cs.CV / 17 / 2608.26607
FU-Mamba: A Frequency-Enhanced Dynamic Scanning Framework for Oralscan Image Segmentation
FU-Mamba：一种用于Oralscan图像分割的频率增强动态扫描框架
Abstract
Oralscan image segmentation is essential for computer-aided diagnosis and treatment planning in digital dentistry. However, existing visual state space models (SSMs) often rely on manually designed scanning orders to flatten image patches into sequences, which disrupts the semantic spatial continuity and hinders coherent feature extraction from key foreground regions. Moreover, elements such as inconsistent lighting, reflective surfaces, and noise during data acquisition disrupt the frequency distribution by diminishing high-frequency details while enhancing low-frequency components, consequently hindering the accurate localization of boundaries. In response to these challenges, we introduce FU-Mamba, an innovative framework that incorporates dynamic scanning and frequency domain enhancement within the SSM architecture. Specifically, the Dynamic Mamba Block (DMB) adaptively learns sampling offsets via a trainable offset prediction network and performs flexible bilinear interpolation, enabling content-aware scanning that preserves spatial coherence. Furthermore, a frequency domain enhancement block balances spectral components through wavelet-guided decomposition and spectrum pooling, improving robustness under adverse imaging conditions. Experimental findings indicate that FU-Mamba attains a notable enhancement in segmentation accuracy, evidenced by a 1.1% increase in the mean intersection over union (mIoU) metric when evaluated on the dental segmentation dataset. Project page: https://byte2bite.github.io/FU-Mamba/
Chinese Translation
Oralscan图像分割对于数字牙科中的计算机辅助诊断和治疗规划至关重要。然而，现有的视觉状态空间模型（SSMs）通常依赖于手动设计的扫描顺序，将图像块展平为序列，这破坏了语义空间的连续性，并妨碍了从关键前景区域提取一致特征。此外，数据采集过程中不一致的照明、反射表面和噪声等因素通过降低高频细节而增强低频成分，干扰了频率分布，从而妨碍了边界的准确定位。针对这些挑战，我们提出了FU-Mamba，一个创新框架，结合了动态扫描和频域增强于SSM架构中。具体而言，动态Mamba块（DMB）通过可训练的偏移预测网络自适应地学习采样偏移，并执行灵活的双线性插值，实现了内容感知的扫描，保持空间一致性。此外，频域增强块通过小波引导的分解和频谱池化平衡光谱成分，提高了在不利成像条件下的鲁棒性。实验结果表明，FU-Mamba在分割精度上取得了显著提升，在牙科分割数据集上的平均交并比（mIoU）指标提高了1.1%。项目页面：https://byte2bite.github.io/FU-Mamba/
cs.CV / 18 / 2608.26624
Text-to-seed generation: Training-free open-vocabulary seeded semantic segmentation via re-purposing diffusion as text-guided seed generator
文本到种子生成：通过重新利用扩散作为文本引导的种子生成器实现无训练开放词汇种子语义分割
Abstract
Open-vocabulary semantic segmentation (OVSS) aims to segment image regions corresponding to arbitrary text queries. Although the Segment Anything Model (SAM) is a powerful foundation model for segmentation, its standalone performance on OVSS remains limited. Existing methods therefore often use SAM to refine coarse masks predicted by other models, but this strategy is unreliable when the initial masks are inaccurate. In this work, we argue that more reliable segmentation can be achieved by exploiting SAM as a region expansion module guided by accurate object points (i.e., seeds) rather than inaccurate coarse masks. Inspired by classical seeded segmentation, we reformulate OVSS as text-guided seed localization followed by seed-based region expansion. To realize this idea, we propose Text-to-Seed (T2S), a training-free framework that leverages the text-to-region correspondence of Stable Diffusion to generate attention-based seed points for target categories described by text. These sparse seeds are then used as point prompts for SAM to produce full object masks. Without task-specific training or additional annotations, T2S achieves strong performance on standard OVSS benchmarks, demonstrating the effectiveness of combining semantic grounding with seed-driven spatial segmentation.
Chinese Translation
开放词汇语义分割（OVSS）旨在对与任意文本查询相对应的图像区域进行分割。尽管Segment Anything Model（SAM）是一个强大的分割基础模型，但其在OVSS上的独立性能仍然有限。因此，现有方法通常使用SAM来细化其他模型预测的粗略掩膜，但当初始掩膜不准确时，这种策略并不可靠。在本研究中，我们认为通过利用SAM作为一个由准确的物体点（即种子）引导的区域扩展模块，可以实现更可靠的分割，而不是依赖不准确的粗略掩膜。受到经典种子分割的启发，我们将OVSS重新表述为文本引导的种子定位，随后进行基于种子的区域扩展。为了实现这一构想，我们提出了无训练框架Text-to-Seed（T2S），该框架利用Stable Diffusion的文本到区域对应关系生成针对文本描述的目标类别的基于注意力的种子点。这些稀疏种子随后被用作SAM的点提示，以生成完整的物体掩膜。在没有任务特定训练或额外注释的情况下，T2S在标准OVSS基准测试中取得了强劲的表现，展示了将语义基础与种子驱动的空间分割相结合的有效性。
cs.CV / 19 / 2608.26632
Who Remains, What Changes: Identity Anchored Composed Gait Retrieval
谁留下，什么改变：基于身份锚定的组合步态检索
Abstract
Gait recognition has achieved remarkable progress, yet existing methods remain confined to rigid visual matching and often overlook the potential of natural language instructions for interactive retrieval. In this paper, we introduce Composed Gait Retrieval (CoGR), a novel task that retrieves a target gait sequence based on a reference sequence and a natural language modification query. To address the absence of existing datasets for this task, we design an automated annotation pipeline powered by large vision-language models (VLMs) to construct the first gait-language datasets: Language-Augmented CCPG and Language-Augmented CASIA-B. Building on this, we propose ComposeGait, an identity-anchored composition framework designed to prevent the identity drift that arises when generic composed retrieval follows the instruction but returns the wrong person. Its Part-aware Identity Adapter (PIA) aggregates multi-frame, part-aware identity evidence into a sample-specific ID token. We inject the ID tokens into both branches of a shared Q-Former to preserve identity, while excluding the ID-token outputs from the final retrieval embeddings. Joint identity and task-adapted composed-retrieval objectives optimize this space end to end. We evaluate ComposeGait on both benchmarks and show that it achieves the best R@1 among the compared methods, reaching 72.38% on Language-Augmented CCPG and 83.61% on Language-Augmented CASIA-B. These results establish ComposeGait as a strong baseline for CoGR. The datasets and code will be made publicly available.
Chinese Translation
步态识别取得了显著进展，但现有方法仍然局限于刚性的视觉匹配，往往忽视了自然语言指令在交互式检索中的潜力。本文介绍了一种新任务——组合步态检索（Composed Gait Retrieval, CoGR），该任务基于参考步态序列和自然语言修改查询来检索目标步态序列。为了解决这一任务缺乏现有数据集的问题，我们设计了一种由大型视觉-语言模型（Vision-Language Models, VLMs）驱动的自动化标注管道，以构建首个步态-语言数据集：语言增强的CCPG和语言增强的CASIA-B。在此基础上，我们提出了ComposeGait，一种基于身份锚定的组合框架，旨在防止在遵循指令但返回错误人物时出现的身份漂移。其部分感知身份适配器（Part-aware Identity Adapter, PIA）将多帧、部分感知的身份证据聚合成样本特定的ID令牌。我们将ID令牌注入共享Q-Former的两个分支，以保持身份，同时排除ID令牌输出对最终检索嵌入的影响。联合身份和任务适应的组合检索目标端到端优化这一空间。我们在两个基准上评估了ComposeGait，并显示其在比较方法中实现了最佳的R@1，语言增强的CCPG达到72.38%，语言增强的CASIA-B达到83.61%。这些结果确立了ComposeGait作为CoGR的强基线。数据集和代码将公开发布。
cs.CV / 20 / 2608.26644
Real-time Unsupervised Object Discovery from Asynchronous Event Streams
基于异步事件流的实时无监督物体发现
Abstract
Event cameras capture pixel-level intensity changes with microsecond resolution to produce highly sparse asynchronous data streams. For visual perception in latency-critical environments, we propose a lightweight, training-free framework for discovery of moving objects based on spatio-temporal clustering. This framework is driven by two core contributions. First, a linear-time Spatio-temporal Probabilistic Event Filter (SPEF) that introduces an adaptive event acceptance threshold to distinguish salient motion structures from background noise. Second, an Event Morton Code Clustering (EMCC) module that bypasses expensive distance matrix computation to efficiently group events for unsupervised discovery of moving objects. On the E-MLB dataset benchmark, SPEF achieves the best denoising performance among classical filtering methods and remains competitive with learning-based approaches without requiring any offline training. On object discovery, EMCC achieves the highest overall accuracy and lowest execution time across the FRED and eTraM datasets, outperforming established density-based clustering baselines by a substantial margin. Overall, this work establishes a new performance benchmark for classical object discovery in event data, providing a highly scalable, training-free solution for resource-constrained visual perception. The code is available at https://github.com/PrathamShenwai/SPEF_EMCC
Chinese Translation
事件相机以微秒级分辨率捕捉像素级强度变化，从而生成高度稀疏的异步数据流。为了在对延迟敏感的环境中实现视觉感知，我们提出了一种轻量级、无训练的框架，用于基于时空聚类发现移动物体。该框架由两个核心贡献驱动。首先，提出了一种线性时间的时空概率事件滤波器（Spatio-temporal Probabilistic Event Filter, SPEF），该滤波器引入自适应事件接受阈值，以区分显著运动结构与背景噪声。其次，提出了一种事件莫顿码聚类（Event Morton Code Clustering, EMCC）模块，该模块绕过昂贵的距离矩阵计算，有效地对事件进行分组，以实现无监督的移动物体发现。在E-MLB数据集基准测试中，SPEF在经典滤波方法中实现了最佳去噪性能，并且在不需要任何离线训练的情况下与基于学习的方法保持竞争力。在物体发现方面，EMCC在FRED和eTraM数据集中实现了最高的整体准确率和最低的执行时间，显著超越了已建立的基于密度的聚类基线。总体而言，本研究为事件数据中的经典物体发现建立了新的性能基准，提供了一种高度可扩展、无训练的解决方案，以满足资源受限的视觉感知需求。代码可在 https://github.com/PrathamShenwai/SPEF_EMCC 获取。
cs.CV / 21 / 2608.26647
Tissue-Mixture Entropy-Weighted Reconstruction for Partial-Volume-Aware Brain MRI Super-Resolution
基于组织混合熵加权重建的部分体积感知脑MRI超分辨率
Abstract
Full-image objectives in brain magnetic resonance imaging (MRI) super-resolution (SR) can underweight tissue-transition regions affected by the partial-volume effect (PVE), as these regions occupy only a small fraction of the image. Binary boundaries also do not capture the continuous mixture of cerebrospinal fluid, gray matter, and white matter within a voxel. We propose Anatomy-Guided Gaussian-Parameter Warping with PVE-Balanced Reconstruction (AGW-PBR), which combines a low-resolution (LR)-only reconstruction backbone with a training-time objective that emphasizes tissue transitions. The backbone integrates LR-derived Sobel guidance, soft latent-basis assignment, and bounded grid-anchored residual warping. Fixed, quality-controlled tissue fractions derived from registered T1/T2/PD IXI images are converted into tissue-mixture entropy, which defines mean-normalized reconstruction weights within validated PVE support. These sidecars are used only during training, and inference requires only the LR image. AGW-PBR is evaluated on T2-weighted IXI images at 2x, 4x, and 6x using three seeds and subject-level paired analyses. At 4x, test-only SynthSeg masks independently assess reconstruction in tissue-interface and non-interface regions. Targeted ablations examine valid-support supervision, spatially aligned entropy weighting, and soft latent assignment. The AGW-backbone is also trained from scratch on fastMRI at 4x without PVE supervision. AGW-PBR improves full-image reconstruction across the tested IXI scales and regional fidelity at 4x, while the PVE-free backbone retains strong performance on fastMRI. These findings support tissue-mixture entropy weighting for partial-volume-aware brain MRI SR.
Chinese Translation
在脑磁共振成像（MRI）超分辨率（SR）中，完整图像目标可能会低估受到部分体积效应（PVE）影响的组织过渡区域，因为这些区域仅占图像的一小部分。二元边界也无法捕捉到体素内脑脊液、灰质和白质的连续混合。我们提出了一种解剖引导的高斯参数扭曲与PVE平衡重建（AGW-PBR），该方法将仅低分辨率（LR）重建主干与强调组织过渡的训练目标相结合。该主干集成了基于LR的Sobel引导、软潜在基分配和有界网格锚定残差扭曲。从注册的T1/T2/PD IXI图像中获得的固定质量控制组织分数被转换为组织混合熵，从而定义了在验证的PVE支持下的均值归一化重建权重。这些辅助信息仅在训练期间使用，推理时只需要LR图像。AGW-PBR在T2加权IXI图像上进行了2倍、4倍和6倍的评估，使用了三个种子和受试者级别的配对分析。在4倍时，测试专用的SynthSeg掩膜独立评估组织界面和非界面区域的重建。针对有效支持监督、空间对齐熵加权和软潜在分配的定向消融实验也进行了研究。AGW主干还在没有PVE监督的情况下从头开始在fastMRI上进行4倍训练。AGW-PBR在测试的IXI尺度和4倍的区域保真度上改善了完整图像重建，而无PVE的主干在fastMRI上保持了强劲的性能。这些发现支持用于部分体积感知脑MRI超分辨率的组织混合熵加权。
cs.CV / 22 / 2608.26648
Hierarchical Channel Stacking: A Structured Decision Framework for AI-Generated Image Detection
层级通道堆叠：用于AI生成图像检测的结构化决策框架
Abstract
Many synthetic-image detectors produce accurate predictions but offer limited insight into how those decisions are formed. This paper introduces Hierarchical Channel Stacking (HCS), a compact framework for AI-generated image detection that converts intermediate CNN activations into a structured 60-dimensional representation organized across three progressively deeper backbone stages. HCS uses per-channel Level-1 classifiers and a Level-2 aggregator to produce image-level predictions while preserving explicit hierarchical structure for analysis. On a benchmark spanning GAN and diffusion generators, HCS achieves 86.7% accuracy and 86.7% macro-F1 on the held-out test set. Stage ablation shows that the full three-stage system outperforms reduced single-stage and two-stage variants, indicating that the hierarchy carries complementary predictive information. Stage-level contribution analysis further shows that, in the analyzed detector setting, fake GAN and fake diffusion images exhibit distinct stage-level contribution profiles. These results position HCS not simply as a compact detector, but as a structured framework for studying how synthetic-image detectors assemble evidence across representation levels.
Chinese Translation
许多合成图像检测器能够产生准确的预测，但对这些决策形成的过程提供的洞见有限。本文介绍了层级通道堆叠（Hierarchical Channel Stacking, HCS），这是一个紧凑的AI生成图像检测框架，它将中间卷积神经网络（CNN）激活转换为一个结构化的60维表示，该表示在三个逐渐深入的主干阶段中组织。HCS使用每个通道的一级分类器和二级聚合器来生成图像级预测，同时保留显式的层级结构以便于分析。在涵盖生成对抗网络（GAN）和扩散生成器的基准测试中，HCS在保留的测试集上达到了86.7%的准确率和86.7%的宏观F1分数。阶段消融实验表明，完整的三阶段系统优于简化的单阶段和双阶段变体，表明层级结构携带了互补的预测信息。阶段级贡献分析进一步显示，在分析的检测器设置中，假GAN图像和假扩散图像表现出不同的阶段级贡献特征。这些结果使HCS不仅仅是一个紧凑的检测器，而是一个结构化框架，用于研究合成图像检测器如何在表示层级中汇集证据。
cs.CV / 23 / 2608.26656
CoGeo-GS: Concept-Driven and Geometry-Aware Multi-Object Removal in 3D Scenes
CoGeo-GS：基于概念驱动和几何感知的三维场景多物体去除
Abstract
Multi-object removal in 3D scenes is challenging due to severe occlusions, semantic entanglement, and the difficulty of maintaining geometric and multi-view consistency. Existing 3D Gaussian Splatting (3DGS) methods perform well for single-object editing but scale poorly to multi-object scenarios, often requiring repetitive optimization and yielding unstable geometry in removed regions. We propose CoGeo-GS, a concept-driven framework for controllable multi-object removal in 3D scenes. CoGeo-GS assigns concept-aware semantic tags to Gaussians, enabling flexible object selection and reducing interference between foreground objects and background structures within a single optimization stage. To recover plausible geometry, we introduce a geometry-aware completion pipeline that combines monocular depth priors with diffusion-based refinement and boundary-aligned blending. A geometry-regularized refinement strategy further stabilizes reconstruction and preserves multi-view consistency. Experiments demonstrate that CoGeo-GS outperforms existing methods in visual quality and reconstruction fidelity.
Chinese Translation
在三维场景中，多物体去除面临严重的遮挡、语义纠缠以及保持几何和多视图一致性的困难。现有的三维高斯点云（3D Gaussian Splatting, 3DGS）方法在单物体编辑方面表现良好，但在多物体场景中扩展性较差，通常需要重复优化，并在去除区域产生不稳定的几何形状。我们提出了CoGeo-GS，这是一个用于可控多物体去除的基于概念驱动的框架。CoGeo-GS为高斯分布分配了基于概念的语义标签，使得在单次优化阶段内能够灵活选择物体，并减少前景物体与背景结构之间的干扰。为了恢复合理的几何形状，我们引入了一个几何感知的补全管道，该管道结合了单目深度先验、基于扩散的细化和边界对齐的混合。几何正则化的细化策略进一步稳定了重建过程，并保持了多视图一致性。实验表明，CoGeo-GS在视觉质量和重建保真度方面优于现有方法。
cs.CV / 24 / 2608.26658
PailitaoGR: Latent Think-with-Images for Generative Image Retrieval
PailitaoGR：用于生成图像检索的潜在图像思维
Abstract
Generative retrieval has demonstrated strong performance by directly generating product semantic identifiers (SIDs). Extending this paradigm to image search, however, is nontrivial because real-world query images contain diverse information, including the search target, useful auxiliary evidence, and irrelevant visual content. This requires the model to identify and focus on the search target while selectively utilizing auxiliary evidence. In this paper, we propose \textbf{PailitaoGR}, a \emph{Latent Think-with-Images} method for generative image retrieval, which internalizes target-focused perception and selective auxiliary-evidence utilization into a the generative retrieval model, enabling \textit{Zooming without Cropping} and \textit{Reading without OCR}. Specifically, we design a target-focused perception mechanism that identifies and enhances visual tokens of the search target, consisting of a target Enhancer and a learning strategy based on on-policy distillation and attention guidance loss, enabling the model to focus on search-target regions. We also design a selective auxiliary-evidence utilization mechanism that identifies and enhances visual tokens of auxiliary evidence, including an auxiliary enhancer and an in-capacity incremental contrastive distillation strategy, enabling the model to exploit auxiliary evidence. We construct training and validation sets sampled from real-world online image-search logs. Experiments show that our method outperforms existing baselines by an average of 13.8\%, validating its effectiveness.
Chinese Translation
生成检索通过直接生成产品语义标识符（SIDs）展现了强大的性能。然而，将这一范式扩展到图像搜索并非易事，因为现实世界中的查询图像包含多样的信息，包括搜索目标、有用的辅助证据和无关的视觉内容。这要求模型在识别和聚焦于搜索目标的同时，选择性地利用辅助证据。本文提出了 extbf{PailitaoGR}，一种用于生成图像检索的 extit{潜在图像思维}方法，该方法将目标聚焦感知和选择性辅助证据利用内化到生成检索模型中，实现了 extit{无裁剪缩放}和 extit{无OCR阅读}。具体而言，我们设计了一种目标聚焦感知机制，该机制识别并增强搜索目标的视觉标记，包括一个目标增强器和基于策略蒸馏和注意力引导损失的学习策略，使模型能够聚焦于搜索目标区域。我们还设计了一种选择性辅助证据利用机制，该机制识别并增强辅助证据的视觉标记，包括一个辅助增强器和一种容量增量对比蒸馏策略，使模型能够有效利用辅助证据。我们构建了从现实世界在线图像搜索日志中抽样的训练和验证集。实验表明，我们的方法在现有基线之上平均提升了13.8 ext{%}，验证了其有效性。
cs.CV / 25 / 2608.26665
Hull First, Wake Second: Wake-Reliance Suppression for Robust Maritime Vessel Detection
船体优先，尾流次之：针对鲁棒海洋船舶检测的尾流依赖抑制
Abstract
Maritime vessel detectors often face scenes where hulls are small, low-contrast, or blurred, while wakes are longer and easier to detect. This creates a wake-reliance problem: detectors may miss slow or stationary vessels with weak wakes, or produce false positives on wake-like water clutter. We propose HullWake, a hull-first wake-second framework for robust maritime vessel detection. HullWake separates proposal-centered hull evidence from directional wake context, extracts wake cues with bidirectional proposal-anchored corridors, and suppresses wake-dominant predictions through wake response supervision, wake-attenuated consistency, wake-only confidence suppression, and hull--wake decorrelation. We also introduce a wake-oriented evaluation protocol covering weak/no-wake vessels, wake-like hard negatives, worst-group AP, and confidence drop after wake attenuation. Experiments are conducted on Curated-Wake, a wake-oriented maritime dataset of about 10,000 images curated from Ships/Vessels in Aerial Images, the SMD benchmark, and SeaDronesSee, with newly added detection- and segmentation-level wake annotations. Compared with box-only detectors and mask-supervised segmentation baselines, HullWake improves overall AP, weak/no-wake robustness, wake-like false positives, worst-group AP, and confidence stability after wake attenuation.
Chinese Translation
海洋船舶检测器常常面临船体小、对比度低或模糊的场景，而尾流则较长且更易于检测。这导致了尾流依赖问题：检测器可能会错过尾流弱的慢速或静止船舶，或在类似尾流的水面杂波上产生误报。我们提出了HullWake，一个船体优先、尾流次之的鲁棒海洋船舶检测框架。HullWake将以提议为中心的船体证据与方向性尾流上下文分离，通过双向提议锚定的走廊提取尾流线索，并通过尾流响应监督、尾流衰减一致性、仅尾流置信度抑制和船体-尾流去相关化来抑制以尾流为主的预测。我们还引入了一种面向尾流的评估协议，涵盖弱/无尾流船舶、类似尾流的困难负样本、最差组平均精度（worst-group AP）以及尾流衰减后的置信度下降。实验在Curated-Wake上进行，这是一个约10,000张图像的面向尾流的海洋数据集，图像来源于《Ships/Vessels in Aerial Images》、《SMD基准》和《SeaDronesSee》，并新增了检测和分割级别的尾流注释。与仅使用边框的检测器和掩膜监督分割基线相比，HullWake在整体平均精度（AP）、弱/无尾流鲁棒性、类似尾流的误报、最差组AP以及尾流衰减后的置信度稳定性方面都有所提升。
cs.CV / 26 / 2608.26671
RECAP-Forcing: Retaining Content Appearances for Long Video Generation
RECAP-Forcing：保持内容外观以实现长视频生成
Abstract
Long autoregressive video generation faces a fundamental memory challenge: with a finite attention window, a model must decide which information from an ever-expanding history to retain. Existing methods organize memory temporally, preserving recent frames while compressing or discarding older ones. We instead propose RECAP-Forcing, organizing memory by appearance novelty. A long video is not merely a sequence of frames, but an evolving cast of subjects, objects, and scenes whose identities must remain consistent over time. We organize memory by retaining the KV cache associated with newly appearing content--such as entering subjects, disoccluded regions, and newly introduced scenes--at the moment it first becomes visible, prioritizing novelty over recency. Memory should scale with the amount of newly introduced content, rather than with video length. This appearance-indexed memory makes long-range consistency an explicit property of the memory structure. Our framework unifies two mechanisms under this single principle. At the beginning of a video, when all visible content is novel, an attention sink preserves the initial scene. As the video evolves, an optical-flow-based novelty bank extends the same principle by selectively retaining newly revealed content. As a training-free inference method with no additional learnable parameters, RECAP-Forcing consistently improves visual quality and semantic fidelity across multiple strong baselines and outperforms existing memory methods.
Chinese Translation
长序列自回归视频生成面临根本性的记忆挑战：由于注意力窗口有限，模型必须决定从不断扩展的历史信息中保留哪些内容。现有方法按时间顺序组织记忆，保留最近的帧，同时压缩或丢弃较早的帧。我们提出了RECAP-Forcing，通过外观新颖性来组织记忆。长视频不仅是帧的序列，更是一个不断演变的主体、物体和场景的集合，其身份必须随时间保持一致。我们通过保留与新出现内容（如进入的主体、显露区域和新引入场景）相关的KV缓存来组织记忆，且在内容首次可见时即予以保留，优先考虑新颖性而非时间先后。记忆应随新引入内容的数量扩展，而非视频长度。基于外观索引的记忆使长距离一致性成为记忆结构的显式属性。我们的框架将两种机制统一于这一原则：视频开始时，所有可见内容均为新颖，注意力汇聚机制保存初始场景；随着视频发展，基于光流的“新颖性库”通过选择性保留新揭示内容来延续该原则。作为一种无需训练且无额外可学习参数的推理方法，RECAP-Forcing在多个强基线模型上持续提升视觉质量和语义一致性，且优于现有记忆方法。
cs.CV / 27 / 2608.26684
Reason in the Words You Speak: Idiolectal Paraphrasing Off-Policy Traces for Reasoning Distillation in VideoLLMs
你所说话语中的推理：用于视频大语言模型推理蒸馏的个体化释义离线轨迹
Abstract
Recent large language models achieve strong performance on complex reasoning tasks, where reinforcement learning with Group Relative Policy Optimization (GRPO) has emerged as a leading paradigm for optimizing models on self-generated trajectories. However, the on-policy nature of GRPO bounds the model to the reasoning skills it can already produce, restricting to learn more advanced capabilities. Prior works inject privileged reasoning traces from a stronger teacher policy to guide training, yet these traces are inherently out of distribution with respect to the student policy. We observe that this mismatch between on-policy and off-policy causes gradient clipping on semantically critical reasoning tokens, ultimately rewarding correct answers while leaving the reasoning that justifies them unlearned. Hence, we propose \textbf{Echo-GRPO}, a framework that lets the model reason in the words it speaks. Rather than imitating low-probability privileged traces from the teacher model, Echo-GRPO rewrites them into the student policy's own \textit{idiolect}, that is, its own characteristic vocabulary and expression patterns, while preserving their semantics via Dual-Reference Decoding. We instantiate this framework as \textbf{VideoEcho-R1} for video reasoning distillation, achieving consistent improvements across three multimodal LLM backbones and five benchmarks. Finally, we show that our idiolectal paraphrasing is a plug-in module that consistently improves both RL and supervised fine-tuning frameworks for reasoning distillation, demonstrating that policy-aligned supervision extends beyond GRPO.
Chinese Translation
近期的大型语言模型在复杂推理任务上表现出色，其中使用群体相对策略优化（Group Relative Policy Optimization, GRPO）的强化学习已成为优化自生成轨迹模型的主要范式。然而，GRPO的在线策略特性限制了模型只能利用其已经能够产生的推理能力，阻碍了更高级能力的学习。之前的研究通过注入来自更强教师策略的特权推理轨迹来指导训练，但这些轨迹在本质上与学生策略的分布不一致。我们观察到，在线策略与离线策略之间的这种不匹配导致了对语义关键推理标记的梯度裁剪，最终奖励了正确答案，而未能学习支持这些答案的推理。因此，我们提出了 extbf{Echo-GRPO}，一个让模型在其所说的话语中进行推理的框架。Echo-GRPO并不是模仿教师模型中低概率的特权轨迹，而是将其重写为学生策略自身的 extit{个体化语言}，即其特有的词汇和表达模式，同时通过双重参考解码（Dual-Reference Decoding）保持其语义。我们将该框架实例化为 extbf{VideoEcho-R1}，用于视频推理蒸馏，在三个多模态大语言模型基础和五个基准测试中实现了一致的改进。最后，我们展示了我们的个体化释义是一个插件模块，能够持续改善推理蒸馏的强化学习和监督微调框架，证明了策略对齐的监督超越了GRPO。
cs.CV / 28 / 2608.26686
Domain-Specific Self-Supervised Representation Learning for Retinal Fundus Classification
面向视网膜眼底分类的领域特定自监督表示学习
Abstract
Despite the growing number of public datasets, annotated medical images remain scarce. Supervised learning methods achieve strong performance on many benchmarks, however require large amounts of labeled data, which are costly and time-consuming to obtain in the medical domain. To address this limitation, contrastive self-supervised learning (SSL) has emerged as a promising alternative for learning useful representations from unlabeled data. In this work, we investigate two SSL frameworks, SimSiam and SimCLR, for retinal disease classification from fundus images. We focus on understanding how augmentation strategies and training parameters influence representation learning under resource-constrained settings. Given limited data and computational capacity, we explore the feasibility of training SSL models with small batch sizes incorporated with retinal-specific augmentation techniques. Through a series of experiments, we assess the quality of learned representations via linear evaluation and fine-tuning across downstream tasks, including multi-disease classification and diabetic retinopathy grading. Our results show that tailoring augmentation strategies to the characteristics of retinal images plays a critical role in improving performance. Even under constrained settings, lightweight SSL frameworks can learn transferable representations that reduce dependence on large annotated datasets and achieve competitive results.
Chinese Translation
尽管公共数据集数量不断增加，带注释的医学图像仍然稀缺。监督学习方法在许多基准测试中表现出色，但需要大量标注数据，而在医学领域获取这些数据既昂贵又耗时。为了解决这一限制，对比自监督学习（SSL）作为从未标注数据中学习有用表示的有前景的替代方法应运而生。在本研究中，我们探讨了两种SSL框架SimSiam和SimCLR在视网膜疾病分类（基于眼底图像）中的应用。我们重点研究了增强策略和训练参数在资源受限条件下对表示学习的影响。在数据和计算能力有限的情况下，我们探索了结合视网膜特定增强技术的小批量训练SSL模型的可行性。通过一系列实验，我们通过线性评估和微调在多疾病分类和糖尿病视网膜病变分级等下游任务中评估了所学表示的质量。结果表明，针对视网膜图像特性定制增强策略在提升性能方面起着关键作用。即使在受限条件下，轻量级SSL框架也能学习到可迁移的表示，减少对大规模标注数据集的依赖，并取得具有竞争力的结果。
cs.CV / 29 / 2608.26713
AesCanvas: A Large-Scale Dataset and Benchmark for Aesthetic Critique and Contextual Suitability
AesCanvas：一个用于美学批评和情境适宜性的大规模数据集与基准
Abstract
Recent advances in Multimodal Large Language Models (MLLMs) have extended Image Aesthetic Assessment (IAA) beyond scalar scores toward interpretable critique and guidance. Yet existing benchmarks mainly assess intrinsic visual quality or fixed domain criteria, leaving open whether an appealing image is appropriate for a specific purpose, audience, cultural setting, or domain convention. We introduce AesCanvas, a unified suite with two complementary components: CritiqueCanvas with 519,136 instruction-response pairs from 54,300 images supports long-form, multi-dimensional critique across photography, painting, and virtual imagery, whereas ContextCanvas with 301 expert-reviewed use scenarios evaluates contextual aesthetic suitability in realistic use scenarios. Under a unified protocol, we evaluate closed-source frontier, open-weight general, and aesthetic-specific MLLMs. Results reveal a clear separation between critique generation and context-sensitive judgment: reference-based lexical and semantic metrics only partially capture critique quality, while aesthetic specialists remain competitive on selected critique metrics yet substantially lag strong general-purpose MLLMs on ContextCanvas. Further analyses show that aesthetic specialization does not reliably transfer to contextual suitability and that model decisions may fail to track or ground themselves in decisive contextual visual cues. These findings establish culturally situated, evidence-grounded suitability as a distinct objective for aesthetic modeling.
Chinese Translation
近年来，多模态大型语言模型（MLLMs）的进展使得图像美学评估（IAA）超越了标量评分，朝着可解释的批评和指导方向发展。然而，现有基准主要评估内在视觉质量或固定领域标准，尚未探讨一个吸引人的图像是否适合特定的目的、受众、文化背景或领域惯例。我们推出了AesCanvas，一个统一的套件，包含两个互补的组成部分：CritiqueCanvas，包含来自54,300幅图像的519,136对指令-响应对，支持在摄影、绘画和虚拟图像领域进行长篇、多维度的批评；而ContextCanvas则包含301个专家审查的使用场景，评估在现实使用场景中的情境美学适宜性。在统一的协议下，我们评估了闭源前沿、开放权重通用和美学特定的MLLMs。结果显示，批评生成与情境敏感判断之间存在明显的分离：基于参考的词汇和语义指标仅部分捕捉批评质量，而美学专家在选定的批评指标上仍具竞争力，但在ContextCanvas上显著落后于强大的通用MLLMs。进一步分析表明，美学专业化并不可靠地转移到情境适宜性上，模型决策可能未能追踪或基于决定性的情境视觉线索。这些发现确立了文化背景下的、基于证据的适宜性作为美学建模的一个独特目标。
cs.CV / 30 / 2608.26714
LiveVVT: High-Fidelity Video Virtual Try-On in Real Time
LiveVVT：实时高保真视频虚拟试穿
Abstract
Diffusion-based Video Virtual Try-On (VVT) achieves high visual fidelity through bidirectional spatio-temporal modeling, but complete-clip dependence incurs prohibitive latency and computational overhead in practical continuous deployment. Naively enforcing causality disrupts pretrained bidirectional priors and substantially degrades synthesis quality. We introduce LiveVVT, a rolling streaming diffusion framework that preserves bounded bidirectional modeling within causal recurrent generation. Within a fixed-size window, LiveVVT jointly denoises multiple video chunks under bounded look-ahead, preserving local bidirectional interactions while emitting one clean chunk per iteration. Beyond the window, two complementary memories sustain long-term consistency: a bounded temporal memory propagates recent dynamics and occlusion context, whereas a persistent global appearance memory, constructed once from the target garment and a frontal try-on keyframe, anchors garment details and dressed appearance throughout the stream. We further introduce a progressive distillation framework integrating bidirectional VVT learning, teacher-trajectory regression for causal few-step adaptation, and Collaborative Matching Distillation, which couples teacher-distribution matching with rolling flow matching on real videos to align optimization with recurrent inference. Experiments on paired and unpaired long-sequence benchmarks demonstrate superior generation quality over similarly sized models, with $26\times$ lower latency and $11\times$ higher throughput, enabling high-fidelity real-time streaming VVT.
Chinese Translation
基于扩散的视频虚拟试穿（VVT）通过双向时空建模实现了高视觉保真度，但完整片段的依赖在实际连续部署中导致了不可接受的延迟和计算开销。简单地强制因果关系会破坏预训练的双向先验，并显著降低合成质量。我们提出了LiveVVT，这是一种滚动流扩散框架，在因果递归生成中保持有界的双向建模。在固定大小的窗口内，LiveVVT在有界的前瞻下联合去噪多个视频片段，同时每次迭代输出一个干净的片段。在窗口之外，两个互补的记忆维持长期一致性：有界的时间记忆传播最近的动态和遮挡上下文，而持久的全局外观记忆则从目标服装和正面试穿关键帧中构建，贯穿整个流保持服装细节和穿着外观。我们进一步引入了一种渐进蒸馏框架，整合了双向VVT学习、因果少步适应的教师轨迹回归和协作匹配蒸馏，将教师分布匹配与真实视频上的滚动流匹配结合，以使优化与递归推理对齐。在配对和非配对的长序列基准测试中的实验表明，与同等规模的模型相比，生成质量更优，延迟降低了$26 imes$，吞吐量提高了$11 imes$，实现了高保真的实时流媒体VVT。
cs.CV / 31 / 2608.26716
Beyond Atomic Layouts: Compositional Design Understanding with Vision-Language Models
超越原子布局：基于视觉-语言模型的组合设计理解
Abstract
Layout understanding, or the interpretation of element organization, is essential for document analysis, user interface (UI) creation, and graphic design. While recent vision-language models (VLMs) excel at interpreting atomic layouts composed of independent elements, they struggle with compositional layouts that require reasoning over visually entangled elements within hierarchical multi-layer structures. In this paper, we introduce a new task, compositional layout understanding, and present CoDeLayout, a VQA dataset of ~20K real-world multi-layer layouts annotated with compositional element pairs and design intent. Through empirical analysis on CoDeLayout, we identify two key challenges for existing VLMs: semantic drift between textual metadata and visual content, and structural ambiguity in hierarchical inter-element relationships. To address these challenges, we propose MASON, a post-training paradigm that integrates multimodal alignment (MA) and structural perception (SP). MA enhances element interpretation by grounding metadata-defined elements to their visual counterparts, mitigating semantic drift, while SP models layer-aware inter-element spatial relationships to improve hierarchical understanding and reduce structural ambiguity. Experiments reveal substantial gaps in existing VLMs: even the strongest baseline, GPT-o3, achieves only 79.68% accuracy, whereas Qwen2.5-VL 7B with MASON reaches 91.66%. Notably, MASON surpasses full-data Direct Finetune using only 30% of the training data and scales better with additional data.
Chinese Translation
布局理解，即元素组织的解释，对于文档分析、用户界面（UI）创建和图形设计至关重要。尽管最近的视觉-语言模型（VLMs）在解释由独立元素组成的原子布局方面表现出色，但它们在处理需要对层次多层结构中视觉交织元素进行推理的组合布局时却面临困难。本文介绍了一项新任务——组合布局理解，并提出了CoDeLayout，这是一个包含约2万种真实世界多层布局的视觉问答（VQA）数据集，数据集中的布局被注释为组合元素对和设计意图。通过对CoDeLayout的实证分析，我们识别出现有VLMs面临的两个主要挑战：文本元数据与视觉内容之间的语义漂移，以及层次间元素关系中的结构模糊性。为了解决这些挑战，我们提出了MASON，这是一种后训练范式，结合了多模态对齐（MA）和结构感知（SP）。MA通过将元数据定义的元素与其视觉对应物对齐，增强了元素解释，减轻了语义漂移，而SP则建模层感知的元素间空间关系，以改善层次理解并减少结构模糊性。实验结果显示现有VLMs存在显著差距：即使是最强的基线模型GPT-o3，其准确率也仅为79.68%，而使用MASON的Qwen2.5-VL 7B达到了91.66%。值得注意的是，MASON在仅使用30%的训练数据的情况下超越了全数据的直接微调，并且在增加数据时表现出更好的扩展性。
cs.CV / 32 / 2608.26720
Parameter Efficient Continual Learning for Sparse Event-Based Transformers
稀疏事件驱动变换器的参数高效持续学习
Abstract
Robotic and edge intelligence systems operate in dynamic environments where data arrives continuously, requiring models to adapt while preserving previously learned knowledge under strict memory and energy constraints. While parameter-efficient fine-tuning has shown promise for continual learning with vision transformers, conventional architectures rely on dense computation and remain costly for real-world deployment. Sparse event-based vision transformers provide energy-efficient event-driven computation, yet their continual learning capabilities remain largely unexplored. We here introduce sLoTh, a parameter-efficient continual learning framework for pretrained sparse event-based (spiking) vision transformers. sLoTh freezes the backbone and restricts plasticity to scalable-efficient low-rank attention updates (seLoRA) and shared neuronal threshold modulation, enabling adaptation without replay buffers by updating less than 1% of model parameters. Experiments across CIFAR-100, Tiny-ImageNet, ImageNet-100, and ImageNet-R with up to 100 tasks demonstrate competitive rehearsal-free performance in class-incremental learning and online continual learning, while enabling approximately 6.5x lower energy consumption than conventional dense vision transformers.
Chinese Translation
机器人和边缘智能系统在动态环境中运行，数据持续不断地到达，这要求模型在严格的内存和能量限制下适应并保留之前学习的知识。尽管参数高效的微调在视觉变换器的持续学习中显示出前景，但传统架构依赖于密集计算，且在实际部署中仍然成本高昂。稀疏事件驱动视觉变换器提供了能量高效的事件驱动计算，但其持续学习能力仍然在很大程度上未被探索。我们在此介绍了 sLoTh，一个针对预训练稀疏事件驱动（脉冲）视觉变换器的参数高效持续学习框架。sLoTh 冻结了主干网络，并将可塑性限制在可扩展高效的低秩注意力更新（seLoRA）和共享神经元阈值调制上，从而实现无需重放缓冲区的适应，仅更新不到 1% 的模型参数。在 CIFAR-100、Tiny-ImageNet、ImageNet-100 和 ImageNet-R 上进行的多达 100 个任务的实验表明，在类别增量学习和在线持续学习中，sLoTh 展现出竞争力的无重放性能，同时使能量消耗比传统的密集视觉变换器降低约 6.5 倍。
cs.CV / 33 / 2608.26722
UniGeo: A Multi-modal Large Language Model for Text-Guided Cross-View Geo-Localization
UniGeo：一种用于文本引导的跨视角地理定位的多模态大型语言模型
Abstract
Text-guided drone geo-localization aims to identify a target region in a large-scale image gallery from a natural-language description. Existing methods mainly formulate this task as direct matching between an open-ended text query and candidate images. However, incomplete queries and highly similar candidates often make global cross-modal matching insufficient for reliable fine-grained localization. We propose UniGeo, a unified multimodal large language model (MLLM) for text-guided drone geo-localization. Built on a shared vision-language framework, UniGeo jointly supports geo-semantic understanding, cross-view semantic generation, and candidate-level verification. Specifically, it establishes stable correspondences among local scene elements, spatial relations, and language descriptions through geo-semantic learning, and further models semantic mappings between drone and satellite views through cross-view generation. Based on these capabilities, a plug-and-play verification module performs fine-grained discrimination among highly confusable candidates. We further introduce a multi-stage training strategy that progressively learns geo-semantic understanding, cross-view generation, and candidate verification, improving adaptation to text-guided geo-localization. Experiments demonstrate consistent improvements across multiple retrieval backbones. On GeoText-1652, UniGeo improves R@10 and mAP by 13.59 and 2.83 percentage points, respectively, validating its effectiveness for fine-grained text-guided drone geo-localization.
Chinese Translation
文本引导的无人机地理定位旨在从自然语言描述中识别大型图像库中的目标区域。现有方法主要将此任务表述为开放式文本查询与候选图像之间的直接匹配。然而，不完整的查询和高度相似的候选者常常使得全局跨模态匹配不足以实现可靠的细粒度定位。我们提出了UniGeo，一种用于文本引导无人机地理定位的统一多模态大型语言模型（MLLM）。UniGeo基于共享的视觉-语言框架，联合支持地理语义理解、跨视角语义生成和候选级验证。具体而言，它通过地理语义学习在局部场景元素、空间关系和语言描述之间建立稳定的对应关系，并通过跨视角生成进一步建模无人机视图和卫星视图之间的语义映射。基于这些能力，一个即插即用的验证模块在高度混淆的候选者之间执行细粒度的区分。我们进一步引入了一种多阶段训练策略，逐步学习地理语义理解、跨视角生成和候选验证，从而提高对文本引导地理定位的适应性。实验表明，在多个检索骨干网络上均取得了一致的改进。在GeoText-1652数据集上，UniGeo分别提高了R@10和mAP指标13.59和2.83个百分点，验证了其在细粒度文本引导无人机地理定位中的有效性。
cs.CV / 34 / 2608.26724
GeoMAD: Geometry-Aware Multi-View Anomaly Detection via Deformable Fusion and Distributional Alignment
GeoMAD：通过可变形融合和分布对齐的几何感知多视角异常检测
Abstract
Multi-view anomaly detection (MvAD) detects defects by exploiting complementary observations from multiple camera viewpoints. The central challenge is to fuse views with sufficient geometric awareness while remaining scalable to multi-class industrial settings. Existing methods typically fall into two extremes: voxel-based fusion provides explicit geometric alignment but requires costly 3D construction and class-specific assumptions, whereas lightweight patch-based fusion is efficient but relies on discrete candidate matching and lacks continuous cross-view correspondence. In this paper, we propose GeoMAD, a unified multi-view, multi-class AD framework that addresses both geometric correspondence deficiency and distributional inconsistency. Our \textit{Cross-view Deformable Fusion Module} (CDFM) learns content-adaptive, view-pair-specific sampling offsets directly on 2D feature maps and arranges them across a multi-scale window pyramid with image-global reference sampling, enabling hierarchical cross-view correspondence without camera calibration, voxel construction, or class-specific 3D supervision. We further introduce \textit{Distributional View Alignment} (DVA), a self-supervised cross-view regularization loss that aligns each view's bottleneck distribution against a per-instance view-centric target, enforcing global consistency without pixel-level correspondence. Together, CDFM and DVA bridge local geometric correspondence and global distributional consistency, providing geometry-aware and distribution-consistent fusion while preserving the efficiency of 2D feature-space learning. Extensive experiments on Real-IAD and MANTA-Tiny show that GeoMAD achieves strong detection and localization performance in unified MvAD.
Chinese Translation
多视角异常检测（MvAD）通过利用来自多个摄像头视角的互补观察来检测缺陷。核心挑战在于在保持对多类工业环境的可扩展性的同时，融合具有足够几何感知的视角。现有方法通常落入两个极端：基于体素的融合提供明确的几何对齐，但需要昂贵的三维构建和特定类别的假设；而轻量级的基于补丁的融合则高效，但依赖于离散候选匹配，缺乏连续的跨视角对应关系。本文提出了GeoMAD，一个统一的多视角、多类别异常检测框架，解决了几何对应缺陷和分布不一致的问题。我们的 extit{跨视角可变形融合模块}（CDFM）直接在二维特征图上学习内容自适应、视角对特定的采样偏移，并在具有图像全局参考采样的多尺度窗口金字塔中排列，从而实现层次化的跨视角对应，而无需相机标定、体素构建或特定类别的三维监督。我们进一步引入 extit{分布视角对齐}（DVA），一种自监督的跨视角正则化损失，旨在将每个视角的瓶颈分布与每个实例的视角中心目标对齐，强制实现全局一致性而无需像素级对应。CDFM和DVA共同弥合了局部几何对应和全局分布一致性，提供了几何感知和分布一致的融合，同时保持二维特征空间学习的高效性。在Real-IAD和MANTA-Tiny上的大量实验表明，GeoMAD在统一的多视角异常检测中实现了强大的检测和定位性能。
cs.CV / 35 / 2608.26737
Generative Semantic Scene Completion
生成语义场景补全
Abstract
Outdoor LiDAR semantic scene completion (SSC) recovers a dense semantic voxel grid from a scan observing 1% of the target volume, under class imbalance beyond 7,000x. We recast SSC as generative semantic scene completion (GSSC): a single discrete-diffusion formulation in three roles. First, paired sparse-dense scene synthesis (PS$^3$) generates matched sparse LiDAR observations with their dense semantic completions, addressing the long tail at its source and yielding the PS$^3$-SemanticKITTI corpus we train on alongside SemanticKITTI. Second, semantic-guided generative scene completion (SGSC) generates the scene from noise with multinomial discrete diffusion, conditioned on the sparse scan through a bird's-eye-view semantic map and a sparse 3D feature stream. Third, the same framework instead refines an existing completion in one flow-matching step: structured source discrete diffusion (S$^2$D$^2$). S$^2$D$^2$ improves the mIoU of SGSC's own output and every external SSC base tested, without base retraining or test-time adaptation. On the strongest base, one step without test-time augmentation reaches 38.8% mIoU on the SemanticKITTI hidden test. To our knowledge that is the best causal, single-sweep, single-sample result on that leaderboard, +2.1 pp over the previous best published score under the same restriction. Four correction steps with eight-view test-time augmentation reach 39.2%, outside that restriction.
Chinese Translation
户外激光雷达语义场景补全（SSC）从观察目标体积1%的扫描中恢复出密集的语义体素网格，且类别不平衡超过7000倍。我们将SSC重新表述为生成语义场景补全（GSSC）：一种在三个角色中使用的单一离散扩散公式。首先，配对稀疏-密集场景合成（PS$^3$）生成与其密集语义补全相匹配的稀疏激光雷达观测，解决了长尾问题的根源，并生成了我们在SemanticKITTI上训练的PS$^3$-SemanticKITTI数据集。其次，语义引导的生成场景补全（SGSC）通过多项式离散扩散从噪声中生成场景，条件是通过鸟瞰视角的语义图和稀疏的3D特征流进行稀疏扫描。第三，同一框架在一个流匹配步骤中精炼现有的补全：结构化源离散扩散（S$^2$D$^2$）。S$^2$D$^2$提高了SGSC自身输出的mIoU，以及每个测试的外部SSC基准，而无需基准重训练或测试时适应。在最强的基准上，未进行测试时增强的一步达到了SemanticKITTI隐藏测试的38.8% mIoU。据我们所知，这是在该排行榜上最佳的因果、单次扫描、单样本结果，比在相同限制下之前发布的最佳分数高出2.1个百分点。经过四个修正步骤和八视图的测试时增强，达到了39.2%，超出了该限制。
cs.CV / 36 / 2608.26744
G2D: Generative-to-Discriminative Collaborative Inference for Zero-Shot Image Classification
G2D：用于零样本图像分类的生成到判别协同推理
Abstract
Zero-shot classification needs efficient label retrieval and fine-grained visual reasoning, yet discriminative and generative vision-language models fail in complementary ways.When CLIP's top-1 prediction is wrong, the correct label often remains in its top-$K$ shortlist, making disambiguation rather than recall the key challenge.Standalone generative models, however, are hindered by large label spaces and unconstrained outputs.This complementarity motivates separating broad candidate retrieval from fine-grained, image-grounded verification.We propose G2D, a training-free framework that uses a generative VLM to verify CLIP-retrieved candidates against the image.Candidate names and CLIP probabilities provide a structured prior for resolving visually similar classes.Fixed confidence routing, entropy-adaptive candidate sizing, and trie-constrained decoding focus generative reasoning on uncertain samples and ensure one valid output for each input at test time.Across eight benchmarks, G2D achieves 68.85% average accuracy, versus 59.35% for CLIP and 63.11% for the standalone VLM.Across seven generator configurations, candidate-set verification improves average accuracy by 1.08--27.42 percentage points.G2D also transfers to DCLIP, WaffleCLIP, and CuPL, supporting a practical interface between discriminative proposal and generative visual reasoning. Code: https://github.com/Harzva/G2D
Chinese Translation
零样本分类需要高效的标签检索和细粒度的视觉推理，但判别性和生成性视觉语言模型在互补方面存在不足。当CLIP的top-1预测错误时，正确标签通常仍然在其top-$K$候选列表中，这使得消歧而非召回成为关键挑战。然而，独立的生成模型受到大标签空间和不受限制输出的限制。这种互补性促使我们将广泛的候选检索与细粒度的图像基础验证分开。我们提出了G2D，这是一个无训练的框架，利用生成性视觉语言模型（VLM）对CLIP检索的候选进行图像验证。候选名称和CLIP概率为解决视觉相似类别提供了结构化的先验。固定置信度路由、熵自适应候选大小和前缀树约束解码将生成推理集中在不确定样本上，并确保每个输入在测试时有一个有效输出。在八个基准测试中，G2D实现了68.85%的平均准确率，而CLIP为59.35%，独立的VLM为63.11%。在七个生成器配置中，候选集验证将平均准确率提高了1.08到27.42个百分点。G2D还可以迁移到DCLIP、WaffleCLIP和CuPL，支持判别性提议与生成性视觉推理之间的实用接口。代码：https://github.com/Harzva/G2D
cs.CV / 37 / 2608.26752
Glass Surface Detection Grounded in 3D Visual Geometry
基于3D视觉几何的玻璃表面检测
Abstract
Glass surface detection (GSD) is critical for scene understanding and reconstruction, and yet remains challenging due to the transparency and reflectivity of glass surfaces. Existing GSD methods typically rely on 2D appearance cues, which may fail in geometrically ambiguous scenes. In this paper, we propose a paradigm shift: grounding GSD in 3D visual geometry to explicitly model the physical existence of glass surfaces. Our method first distills rich 3D priors from the visual geometry grounded transformer (VGGT) and generates glass-aware 3D representations. It then exploits multi-tasking learning with a novel glass detection head, consisting of two core modules: a Frequency Self-Attention Module (FSAM) that identifies glass-specific spectral features for glass surface localization, and a Geometry Grounding Block (GeGB) that selectively grounds 2D features in 3D geometry for glass surface segmentation. Extensive experiments demonstrate that our method achieves state-of-the-art performance across seven standard GSD benchmarks, generalizes well to video/multi-modal data, and substantially improves reconstruction in glass scenes. Code is available in https://github.com/YT3DVision/VGGT_GLASS.
Chinese Translation
玻璃表面检测（GSD）对于场景理解和重建至关重要，但由于玻璃表面的透明性和反射性，这一任务仍然具有挑战性。现有的GSD方法通常依赖于2D外观线索，这在几何模糊的场景中可能会失效。本文提出了一种范式转变：将GSD基于3D视觉几何，以明确建模玻璃表面的物理存在。我们的方法首先从视觉几何基础变换器（VGGT）中提取丰富的3D先验知识，并生成玻璃感知的3D表示。接着，它利用多任务学习，采用一种新颖的玻璃检测头，包含两个核心模块：频率自注意模块（Frequency Self-Attention Module, FSAM），用于识别特定于玻璃的光谱特征以实现玻璃表面定位；几何基础块（Geometry Grounding Block, GeGB），用于选择性地将2D特征与3D几何相结合，以实现玻璃表面分割。大量实验表明，我们的方法在七个标准GSD基准上实现了最先进的性能，能够良好地推广到视频/多模态数据，并显著改善玻璃场景的重建效果。代码可在 https://github.com/YT3DVision/VGGT_GLASS 获取。
cs.CV / 38 / 2608.26771
Cross-Architecture Knowledge Distillation from a Vision Foundation Model to a Lightweight Visual State Space Model for Tea Leaf Disease Classification
从视觉基础模型到轻量级视觉状态空间模型的跨架构知识蒸馏用于茶叶病害分类
Abstract
Automated tea leaf disease classification supports precision agriculture, yet deploying accurate models on edge devices remains challenging under tight compute budgets. Self-supervised vision foundation models such as DINOv2 provide strong features but are too large for field deployment, while lightweight models trained from scratch on small agricultural datasets often underfit. We study cross-architecture knowledge distillation (KD) from a fine-tuned DINOv2 teacher (Vision Transformer) to a compact bidirectional Visual State Space Model (LVSSM) student, an underexplored direction because the architectures use fundamentally different token-mixing mechanisms. We identify and fix two training-stability problems that prevent the from-scratch SSM student from learning on limited data: a single large patch-embedding convolution and a fusion layer that severs the residual path. With a progressive convolutional stem and gated bidirectional selective-scan block, the 4.45M-parameter student trains stably. Across three seeds, temperature-scaled logit distillation raises test accuracy from 92.32+/-2.14% to 95.41+/-1.17% (best single run: 96.20%; macro-F1: 94.45%), a +3.09 percentage-point mean gain. The student uses 5.0 times fewer parameters than the 22M-parameter teacher while retaining 98.3% of its accuracy. Ablations show that intermediate feature-alignment losses reduce accuracy, making simple logit-level KD the strongest configuration. A fair from-scratch comparison shows the gain is specific to students that start below the teacher. We report per-class metrics, confusion matrices, bootstrap confidence intervals, and FLOPs/latency measurements, and discuss limitations including the single-dataset scope and simplified non-official SSM implementation.
Chinese Translation
自动化茶叶病害分类支持精准农业，但在计算预算紧张的情况下，将准确模型部署到边缘设备上仍然具有挑战性。自监督视觉基础模型如 DINOv2 提供了强大的特征，但对于现场部署来说过于庞大，而从小型农业数据集上从头训练的轻量级模型往往存在欠拟合问题。我们研究了从微调后的 DINOv2 教师（视觉变换器）到紧凑的双向视觉状态空间模型（LVSSM）学生的跨架构知识蒸馏（KD），这一方向尚未得到充分探索，因为这两种架构使用了根本不同的令牌混合机制。我们识别并修复了两个训练稳定性问题，这些问题阻止了从头开始的 SSM 学生在有限数据上学习：一个大型补丁嵌入卷积和一个切断残差路径的融合层。通过渐进式卷积干和门控双向选择扫描块，4.45M 参数的学生模型训练稳定。在三个种子下，温度缩放的对数蒸馏将测试准确率从 92.32+/-2.14% 提升至 95.41+/-1.17%（最佳单次运行：96.20%；宏观 F1：94.45%），平均增益为 +3.09 个百分点。该学生模型使用的参数比 22M 参数的教师模型少 5.0 倍，同时保留了 98.3% 的准确率。消融实验表明，中间特征对齐损失会降低准确率，使得简单的对数级 KD 成为最强配置。公平的从头比较显示，增益特定于那些起始于教师模型之下的学生。我们报告了每类指标、混淆矩阵、自助置信区间以及 FLOPs/延迟测量，并讨论了局限性，包括单数据集范围和简化的非官方 SSM 实现。
cs.CV / 39 / 2608.26794
Ring Forcing: Towards Precise Long-Term Memory for Autoregressive Video Diffusion
环形强制：朝着精确的自回归视频扩散长期记忆迈进
Abstract
Scaling video generation to long durations reveals a critical bottleneck: current models lack robust long-term memory. This deficiency can be studied along two critical aspects: object permanence, the ability to precisely reproduce the appearance of objects upon re-entry; and memory capacity, the ability to process ultra-long context and use information from distant history. Robust long-term memory requires both: object permanence without sufficient context handling limits the temporal scope, while long context length without permanence fails to maintain identity. To address this, we present Ring Forcing, an autoregressive video diffusion framework designed to robustly construct and precisely utilize long-term memory. Our ring-structured training strategy enforces retrieval from distant history, effectively reconciling the trade-off between strict historical adherence and generative diversity. To expand memory capacity, we introduce a compression and timestep composition strategy. Under fixed sequence length constraints, this method extends the effective historical span to minutes-long durations and achieves a comprehensive receptive field over the entire history. Furthermore, we present a sparse RoPE mechanism to enable flexible, scalable memory adaptation while fully exploiting pre-trained priors. Extensive experiments demonstrate that Ring Forcing achieves superior minutes-long coherence and object permanence, significantly outperforming state-of-the-art methods.
Chinese Translation
将视频生成扩展到较长时长揭示了一个关键瓶颈：当前模型缺乏稳健的长期记忆。这一缺陷可以从两个关键方面进行研究：物体持久性，即在重新出现时精确再现物体外观的能力；以及记忆容量，即处理超长上下文并利用远程历史信息的能力。稳健的长期记忆需要两者的结合：在缺乏足够上下文处理的情况下，物体持久性限制了时间范围，而在缺乏持久性的情况下，长上下文长度无法维持身份。为了解决这一问题，我们提出了环形强制（Ring Forcing），一种旨在稳健构建和精确利用长期记忆的自回归视频扩散框架。我们的环形结构训练策略强制从远程历史中检索，有效调和了严格历史遵循与生成多样性之间的权衡。为了扩展记忆容量，我们引入了一种压缩和时间步组合策略。在固定序列长度约束下，该方法将有效历史跨度扩展到数分钟的时长，并在整个历史上实现了全面的感受野。此外，我们提出了一种稀疏的RoPE机制，以实现灵活、可扩展的记忆适应，同时充分利用预训练的先验知识。大量实验表明，环形强制在数分钟的连贯性和物体持久性方面表现优越，显著超越了最先进的方法。
cs.CV / 40 / 2608.26806
Multi-Image Visual Token Pruning in Large Visual Language Models
大型视觉语言模型中的多图像视觉标记剪枝
Abstract
With the growing demand for processing multiple image sequences in real-world applications, various visual token pruning methods have emerged to mitigate the computational and context length constraints faced by Large Vision Language Models (LVLMs). However, most existing pruning approaches rely on static strategies that struggle to adapt across different architectural LVLMs and multi-image scenarios, and are additionally constrained by their dependence on attention computations that are incompatible with efficient techniques like FlashAttention. To address these limitations, we propose a training-free, Adaptive Visual Token Pruning (AVTP) framework, applicable to diverse LVLM architectures. We strategically determine pruning layers based on empirical analysis of visual attention distributions across various LVLMs, and implement adaptive pruning ratios in multi-image contexts where images of higher importance retain proportionally more tokens. We conduct extensive experiments across different LVLMs to demonstrate the effectiveness and robustness of AVTP. Specifically, Qwen3VL-8B achieves 2 times inference speedup while maintaining 96.1\% of its original accuracy on multiple multi-image benchmarks, InternVL3.5-8B retains 94.1\% accuracy, and LLaVA-OV-7B even exceeds its original baseline performance. Our code is available at \href{https://github.com/zry13/AVTP}{this link}.
Chinese Translation
随着现实应用中对处理多图像序列需求的增长，各种视觉标记剪枝方法应运而生，以缓解大型视觉语言模型（Large Vision Language Models, LVLMs）面临的计算和上下文长度限制。然而，大多数现有的剪枝方法依赖于静态策略，难以适应不同架构的LVLM和多图像场景，并且受到与高效技术（如FlashAttention）不兼容的注意力计算依赖的限制。为了解决这些局限性，我们提出了一种无训练的自适应视觉标记剪枝（Adaptive Visual Token Pruning, AVTP）框架，适用于多种LVLM架构。我们基于对不同LVLM的视觉注意力分布的实证分析，战略性地确定剪枝层，并在多图像上下文中实施自适应剪枝比例，使得更重要的图像保留相对更多的标记。我们在不同的LVLM上进行了广泛的实验，以证明AVTP的有效性和鲁棒性。具体而言，Qwen3VL-8B在多个多图像基准测试中实现了2倍的推理加速，同时保持了96.1%的原始准确率，InternVL3.5-8B保持了94.1%的准确率，而LLaVA-OV-7B甚至超越了其原始基准性能。我们的代码可在[此链接](https://github.com/zry13/AVTP)获取。
cs.CV / 41 / 2608.26809
Thinking on Shots: Consistent Multi-Shot Video Editing with Agentic Reasoning
镜头思考：基于自主推理的一致性多镜头视频编辑
Abstract
While generative AI has significantly advanced video editing, existing methods primarily focus on single-shot or short video clips. Editing long videos with multiple instructions remains a formidable challenge. Naive chunking strategies, e.g., fixed-duration segmentation, often lead to entity fragmentation, severe editing hallucinations, and disrupted temporal continuity. To bridge this gap, we introduce the Multi-Instruction Multi-Shot Long-Video Editing (MMLVE) task, which is structured around three core objectives: Cross-Shot Editing Consistency (CSEC), Multi-Instruction Decoupling (MID), and Zero-Destruction on Spatiotemporal Structure (ZDSS). To tackle these three unique challenges, we introduce an agentic editing framework that leverages the synergy of Large Language Models (LLMs) and Vision-Language Models (VLMs) to achieve shot-level video decoupling and precise instruction parsing. Furthermore, to comprehensively evaluate this task, we construct MMLVE-Bench, which is an MMLVE-focused dataset characterized by complex real-world spatiotemporal dynamics, high-density heterogeneous instructions, and sparse, random entity distributions. Three MMLVE-focused evaluation metrics are further exploited to assess the quality of the editing results. Extensive experiments demonstrate that our MMLVE-Agent outperforms existing closed-source SOTA approaches (e.g., Seedance 2.0), successfully eliminating editing hallucinations, preserving cross-shot editing consistency, and attaining seamless spatiotemporal transitions.
Chinese Translation
尽管生成性人工智能在视频编辑方面取得了显著进展，但现有方法主要集中于单镜头或短视频片段。编辑包含多个指令的长视频仍然是一项艰巨的挑战。简单的分块策略，例如固定时长分段，往往导致实体碎片化、严重的编辑幻觉以及时间连续性的中断。为了解决这一问题，我们提出了多指令多镜头长视频编辑（Multi-Instruction Multi-Shot Long-Video Editing, MMLVE）任务，该任务围绕三个核心目标构建：镜头间编辑一致性（Cross-Shot Editing Consistency, CSEC）、多指令解耦（Multi-Instruction Decoupling, MID）和时空结构零破坏（Zero-Destruction on Spatiotemporal Structure, ZDSS）。为了解决这三个独特的挑战，我们引入了一种自主编辑框架，利用大型语言模型（Large Language Models, LLMs）和视觉-语言模型（Vision-Language Models, VLMs）的协同作用，实现镜头级视频解耦和精确的指令解析。此外，为了全面评估这一任务，我们构建了MMLVE-Bench，这是一个以MMLVE为重点的数据集，具有复杂的现实世界时空动态、高密度异构指令和稀疏随机的实体分布。进一步利用三种MMLVE重点评估指标来评估编辑结果的质量。大量实验表明，我们的MMLVE-Agent在消除编辑幻觉、保持镜头间编辑一致性以及实现无缝时空过渡方面优于现有的闭源最先进方法（例如，Seedance 2.0）。
cs.CV / 42 / 2608.26812
Hyperspectral Diffusion Equivariant Imaging (HyDiff-EI): A Self-supervised Framework for Hyperspectral Image Inpainting
超光谱扩散等变成像（HyDiff-EI）：一种用于超光谱图像修复的自监督框架
Abstract
A novel Hyperspectral diffusion Equivariant Imaging (HyDiff-EI) framework for solving the hyperspectral image (HSI) inpainting problem has been presented here. Unlike conventional diffusion-based methods that rely on large-scale pretraining, HyDiff-EI is a test-time optimization framework that learns directly from a single corrupted HSI acquisition. This makes it flexible for different sensor configurations and particularly well-suited for practical remote sensing scenarios where large annotated hyperspectral datasets are limited. To address the ill-posed nature of unsupervised inpainting, we embed equivariant consistency constraints within the diffusion process. By leveraging the inherent geometric symmetries and intrinsic characteristics of HSIs, HyDiff-EI bridges the gap between generative diffusion modeling and self-consistent physical priors. We empirically show that coupling diffusion modeling with equivariant priors substantially enhances noise robustness and generalizability. Extensive experiments on real-world datasets including Chikusei, Botswana, and EMIT demonstrate that HyDiff-EI offers remarkable inpainting quality over existing self-supervised and diffusion-based algorithms in both noiseless and noisy cases.
Chinese Translation
本文提出了一种新颖的超光谱扩散等变成像（HyDiff-EI）框架，用于解决超光谱图像（HSI）修复问题。与依赖于大规模预训练的传统扩散方法不同，HyDiff-EI 是一种测试时优化框架，能够直接从单个损坏的 HSI 采集数据中学习。这使得它在不同传感器配置下具有灵活性，特别适合于大规模标注超光谱数据集有限的实际遥感场景。为了应对无监督修复的病态特性，我们在扩散过程中嵌入了等变一致性约束。通过利用 HSI 的固有几何对称性和内在特征，HyDiff-EI 连接了生成扩散建模与自一致物理先验之间的差距。我们实证表明，将扩散建模与等变先验相结合显著增强了噪声鲁棒性和泛化能力。在 Chikusei、博茨瓦纳和 EMIT 等真实世界数据集上的大量实验表明，HyDiff-EI 在无噪声和有噪声的情况下，相较于现有的自监督和基于扩散的算法，提供了显著的修复质量。
cs.CV / 43 / 2608.26820
LLaVAFlow: Preserving Latent Alignment Flow for Parameter-Efficient Multimodal Fine-Tuning
LLaVAFlow：保持潜在对齐流以实现参数高效的多模态微调
Abstract
While Multimodal Large Language Models (MLLMs) exhibit strong generalization, visual instruction tuning for downstream tasks inevitably causes catastrophic forgetting, impairing overall generalization. While existing methods regulate weight updates to reduce forgetting, they overlook the fundamental cross-modal alignment in MLLMs. Based on prior work and our observations, we argue that cross-modal alignment is implicitly captured in the information-compression trajectory. To preserve the alignment flow embedded in the trajectory, we propose LLaVAFlow, an information-theoretic distillation framework. First, we compress the mutual information between the extracted relations and MLLM embeddings, encouraging a learnable module to produce a refined alignment flow that benefits downstream tasks. Second, we maximize the mutual information between the extracted alignment flows of the pretrained and fine-tuned MLLMs, enabling the transfer of compact alignment information. Extensive experiments show that LLaVAFlow is an effective plug-and-play framework that preserves alignment flow and enhances both downstream performance and generalization.
Chinese Translation
尽管多模态大型语言模型（MLLMs）展现出强大的泛化能力，但针对下游任务的视觉指令调优不可避免地会导致灾难性遗忘，从而损害整体泛化能力。现有方法虽然通过调节权重更新来减少遗忘，但忽视了MLLMs中的基本跨模态对齐。基于之前的研究和我们的观察，我们认为跨模态对齐在信息压缩轨迹中是隐含捕获的。为了保持嵌入在轨迹中的对齐流，我们提出了LLaVAFlow，一个信息论蒸馏框架。首先，我们压缩提取关系与MLLM嵌入之间的互信息，鼓励一个可学习模块生成一个精细的对齐流，从而有利于下游任务。其次，我们最大化预训练和微调后的MLLMs提取的对齐流之间的互信息，从而实现紧凑对齐信息的传递。大量实验表明，LLaVAFlow是一个有效的即插即用框架，能够保持对齐流并增强下游性能和泛化能力。
cs.CV / 44 / 2608.26827
Evaluator-Dependent Patient-Adaptive ECG Lead-Channel Allocation
评估者依赖的患者自适应心电图导联通道分配
Abstract
Patient-conditioned acquisition policies for ECG lead-channel selection can outperform population-wide fixed protocols by tailoring the channel budget to each patient's observed cardiac state. However, the value of acquiring any given channel is defined relative to a downstream diagnostic evaluator, so marginal utilities learned under one evaluator need not transfer when the evaluator is replaced. We study this evaluator dependence empirically on PTB-XL by freezing two policies (ECG-on-Demand and MGA) trained with a controlled arbitrary-mask logistic evaluator, then scoring their unchanged acquisition trajectories with a more predictive masked raw-waveform ResNet1D. Exhaustive search provides metric-matched population-wide fixed comparators separately for each evaluator, enabling a clean interaction contrast. At budget $k=4$ on a held-out evaluation fold, ECG-on-Demand shifts from $D_\mathrm{C}=-0.011$ (favoring adaptive under the controlled evaluator) to $D_\mathrm{S}=+0.029$ (favoring fixed under the strong evaluator), yielding an NLL interaction of $+0.041$ (95% CI $[+0.030, +0.050]$). Across two policies, five budgets, and three probabilistic metrics, all 30 interaction estimates are positive with paired confidence intervals excluding zero. Three post-hoc sensitivity analyses -- common-reference scoring, training the strong evaluator on a mixture of policy-generated and random masks, and evaluator-aligned Strong-MGA policy training -- each preserve a positive interaction interval, making reference-choice and mask-distribution artifacts less plausible explanations. Evaluator-aligned training reduces but does not eliminate the gap. These results indicate that adaptive ECG channel allocation should be developed and validated jointly with its intended diagnostic backbone, and that jointly optimized sensing-diagnosis systems remain an open problem.
Chinese Translation
针对心电图（ECG）导联通道选择的患者条件获取策略，通过将通道预算量身定制为每位患者观察到的心脏状态，能够超越人口范围内的固定协议。然而，获取任何特定通道的价值是相对于下游诊断评估者定义的，因此在一个评估者下学习的边际效用在更换评估者时不一定能够转移。我们通过在 PTB-XL 数据集上对两个政策（按需心电图（ECG-on-Demand）和多目标优化（MGA））进行实证研究，冻结使用受控任意掩码逻辑评估器训练的策略，然后用更具预测性的掩码原始波形 ResNet1D 评估其不变的获取轨迹。全面搜索为每个评估者分别提供了度量匹配的人口范围内固定比较器，从而实现了干净的交互对比。在持出评估折上，预算 $k=4$ 时，按需心电图从 $D_ ext{C}=-0.011$（在受控评估者下偏向自适应）转变为 $D_ ext{S}=+0.029$（在强评估者下偏向固定），产生了 $+0.041$ 的 NLL 交互（95% CI $[+0.030, +0.050]$）。在两个政策、五个预算和三个概率度量下，所有 30 个交互估计均为正值，且配对置信区间不包括零。三项事后敏感性分析——共同参考评分、在政策生成和随机掩码混合上训练强评估者，以及与评估者对齐的强多目标优化政策训练——均保持了正交互区间，使参考选择和掩码分布伪影的可能性解释降低。与评估者对齐的训练减少但并未消除这一差距。这些结果表明，自适应心电图通道分配应与其预期的诊断基础共同开发和验证，并且联合优化的感知-诊断系统仍然是一个未解决的问题。
cs.CV / 45 / 2608.26833
Rethinking Image Processing for the Age of AI: A Problem-First Framework for Scientific Progress
重新思考人工智能时代的图像处理：以问题为导向的科学进步框架
Abstract
Modern AI has greatly expanded the capabilities of image processing. However, the ready availability of powerful models, public datasets, and benchmark leaderboards has also en- couraged a model-first research pattern: researchers increasingly begin with an available architecture and optimize it on a public benchmark, rather than beginning with the underlying real-world imaging problem. This can produce impressive benchmark results without necessarily improving our understanding or solution of the real problem. This paper argues for a problem-first approach that distinguishes the physical imaging problem, solution principle, statistical estimator, and computational implementation, while clarifying what modern AI can achieve and which fundamental problems remain unsolved. Through case studies of super- resolution and low-light enhancement, we show how benchmark datasets may define tasks that differ substantially from the real-world problems they are intended to represent, and why performance improvements must be interpreted within the conditions under which they are obtained. We propose a six-stage workflow that places problem formulation, image acquisition, information-loss analysis, assumptions, ambiguity, and evaluation before model and dataset selection. The paper also proposes clearer standards for evidence, reproducibility, uncertainty, and claims of state-of-the-art performance. More fundamentally, it calls for a change in research culture and education so that future researchers learn to understand imaging problems deeply and use modern AI to achieve genuine scientific and technical advancement.
Chinese Translation
现代人工智能极大地扩展了图像处理的能力。然而，强大模型、公共数据集和基准排行榜的广泛可用性也促使了一种以模型为先的研究模式：研究人员越来越多地从现有架构开始，并在公共基准上进行优化，而不是从基础的现实世界成像问题入手。这可能会产生令人印象深刻的基准结果，但并不一定改善我们对真实问题的理解或解决方案。本文主张采用以问题为先的方法，区分物理成像问题、解决原则、统计估计器和计算实现，同时澄清现代人工智能能够实现的目标以及哪些基本问题仍未解决。通过超分辨率和低光增强的案例研究，我们展示了基准数据集如何定义与其意图表示的现实世界问题有实质性差异的任务，以及为什么性能改进必须在获得这些改进的条件下进行解读。我们提出了一个六阶段的工作流程，将问题表述、图像获取、信息损失分析、假设、模糊性和评估置于模型和数据集选择之前。本文还提出了更清晰的证据、可重复性、不确定性和尖端性能声明的标准。从根本上说，它呼吁改变研究文化和教育，以便未来的研究人员能够深入理解成像问题，并利用现代人工智能实现真正的科学和技术进步。
cs.CV / 46 / 2608.26848
MedFG-VQA: Low-Frequency Memory and Graph Attention for Lightweight Medical VQA
MedFG-VQA：低频记忆与图注意力用于轻量级医学视觉问答
Abstract
Medical Visual Question Answering (Med-VQA) holds significant promise for clinical decision support, yet faces challenges due to limited annotated data and the high computational demands of existing large vision-language models. We propose MedFG-VQA, a lightweight framework that leverages a memory bank to augment DCT-based low-frequency features and employs graph-enhanced cross-attention for effective visual-textual alignment. Specifically, our approach features two key components: Frequency-Memory Fusion (FMF), which enhances low-frequency features by retrieving from a learnable memory bank built on DCT decomposition, and Graph-Aware Cross-Attention (GACA), which aligns visual-textual features via cross-attention and refines them through graph-convolutional aggregation. To address data scarcity, we construct SynMed-VQA, a large-scale synthetic dataset comprising over 2 million question-answer pairs across 9 imaging modalities and 10 major organs, generated with GPT-4o. Extensive experiments on SynMed-VQA and three other standard biomedical VQA benchmarks demonstrate that MedFG-VQA achieves competitive or superior performance compared to much larger models while maintaining significantly lower computational costs, highlighting its efficiency and potential for clinical deployment.
Chinese Translation
医学视觉问答（Med-VQA）在临床决策支持中具有重要前景，但由于标注数据有限以及现有大型视觉语言模型的高计算需求，面临诸多挑战。我们提出了MedFG-VQA，这是一种轻量级框架，利用记忆库增强基于离散余弦变换（DCT）的低频特征，并采用图增强的交叉注意力实现有效的视觉-文本对齐。具体而言，我们的方法包含两个关键组件：频率-记忆融合（Frequency-Memory Fusion, FMF），通过从基于DCT分解构建的可学习记忆库中检索来增强低频特征；图感知交叉注意力（Graph-Aware Cross-Attention, GACA），通过交叉注意力对视觉-文本特征进行对齐，并通过图卷积聚合进行精细化。为了解决数据稀缺问题，我们构建了SynMed-VQA，这是一个大规模合成数据集，包含超过200万个问题-答案对，涵盖9种成像模式和10个主要器官，使用GPT-4o生成。在SynMed-VQA和其他三个标准生物医学VQA基准上的广泛实验表明，MedFG-VQA在性能上与更大模型相当或优越，同时保持显著较低的计算成本，突显了其效率和临床部署的潜力。
cs.CV / 47 / 2608.26856
From Reasoning to Pixels: Grounded Medical Multimodal LLMs for VQA and Segmentation
从推理到像素：面向视觉问答和分割的基础医学多模态大语言模型
Abstract
Although Multimodal Large Language Models (MLLMs) have demonstrated impressive performance in Medical Visual Question Answering (Med-VQA), their reliance on global image features often lacks precise pixel-level grounding, thereby limiting clinical trustworthiness. To bridge the semantic gap between high-level clinical reasoning and spatial localization, we propose \textsc{\textsc{MedREAL}} (\textbf{Med}ical \textbf{RE}asoning-driven \textbf{A}nswering and \textbf{L}ocalization), a unified framework that seamlessly aligns linguistic reasoning with spatial grounding. Specifically, \textsc{MedREAL} introduces \textbf{S}eg \textbf{A}nchored \textbf{R}easoning \textbf{P}ooling (SARP) to distill task-relevant semantic evidence directly from \texttt{[SEG]} tokens within the MLLM's hidden states. Furthermore, a \textbf{R}easoning-to-\textbf{V}isual (R2V) fusion mechanism is proposed to effectively inject these reasoning-aware features into a segmentation pipeline for accurate mask decoding. To facilitate this paradigm, we construct MedRAVS-13K, a comprehensive dataset comprising 13,824 expertly validated samples across four diverse imaging modalities. Extensive experiments demonstrate that \textsc{MedREAL} significantly outperforms state-of-the-arts, achieving 68.49\% gIoU and 70.47\% cIoU on benchmark evaluations. By generating evidence masks that are strictly consistent with textual diagnoses, \textsc{MedREAL} provides a robust, interpretable framework for reasoning-driven medical image analysis.
Chinese Translation
尽管多模态大语言模型（MLLMs）在医学视觉问答（Med-VQA）中表现出色，但它们对全局图像特征的依赖往往缺乏精确的像素级基础，从而限制了临床可信度。为了弥合高层临床推理与空间定位之间的语义差距，我们提出了 extsc{MedREAL}（ extbf{Med}ical extbf{RE}asoning-driven extbf{A}nswering and extbf{L}ocalization），这是一个统一框架，能够无缝地将语言推理与空间基础对齐。具体而言， extsc{MedREAL}引入了 extbf{S}eg extbf{A}nchored extbf{R}easoning extbf{P}ooling（SARP），以直接从MLLM的隐藏状态中的 exttt{[SEG]}标记中提取与任务相关的语义证据。此外，提出了一种 extbf{R}easoning-to- extbf{V}isual（R2V）融合机制，以有效地将这些关注推理的特征注入分割管道，实现准确的掩膜解码。为了促进这一范式的实现，我们构建了MedRAVS-13K，这是一个包含13,824个经过专家验证的样本的综合数据集，涵盖四种不同的成像模态。大量实验表明， extsc{MedREAL}显著优于现有技术，在基准评估中实现了68.49 ext{%}的gIoU和70.47 ext{%}的cIoU。通过生成与文本诊断严格一致的证据掩膜， extsc{MedREAL}为基于推理的医学图像分析提供了一个稳健且可解释的框架。
cs.CV / 48 / 2608.26859
A Geometry-Driven, Framework-Agnostic Optimization for Object Pose Estimation
一种几何驱动的框架无关优化方法用于物体姿态估计
Abstract
Current object pose estimation research remains predominantly model-centric, focusing on architectural innovations and post-processing refinements. This paper introduces a data-centric optimization by proposing a novel, physically grounded rotation representation through principal axes alignment. Our method aligns the object's coordinate system with its inherent geometric axes, derived from inertial properties, yielding three key advantages: Inherent Stability-leveraging the energy-minimizing property of principal axes provides a robust representation that is less sensitive to noise and occlusions; Symmetry-Aware Canonicalization-explicitly resolving rotational ambiguities for symmetric objects at the data level, which fundamentally eliminates label confusion during network training; and Framework Agnosticism-the optimization is applied purely at the dataset level, ensuring plug-and-play compatibility with existing networks without any architectural modification. We validate the framework across diverse category-level and instance-level models. Extensive experiments demonstrate consistent and significant accuracy improvements, while preserving the integrity of the baseline network. This work establishes a new, geometry-driven direction for enhancing pose estimation, circumventing the need for complex network redesign.
Chinese Translation
当前的物体姿态估计研究仍然主要以模型为中心，关注于架构创新和后处理优化。本文通过提出一种新颖的、基于物理的旋转表示方法，采用主轴对齐，介绍了一种数据中心的优化方法。我们的方法将物体的坐标系统与其固有的几何轴对齐，这些几何轴源自惯性特性，从而带来了三个关键优势：固有稳定性——利用主轴的能量最小化特性，提供了一种对噪声和遮挡不那么敏感的稳健表示；对称性意识的规范化——在数据层面显式解决对称物体的旋转歧义，从根本上消除了网络训练过程中的标签混淆；框架无关性——优化仅在数据集层面进行，确保与现有网络的即插即用兼容性，无需任何架构修改。我们在不同类别和实例级模型上验证了该框架。大量实验表明，准确性有一致且显著的提升，同时保持了基线网络的完整性。这项工作为提升姿态估计建立了一个新的、以几何为驱动的方向，避免了复杂网络重设计的需求。
cs.CV / 49 / 2608.26861
FIDA: Feature Instability-Driven Attack on Self-Supervised Facial Representation
FIDA：基于特征不稳定性的自监督人脸表征攻击
Abstract
Self-supervised learning (SSL) models are vulnerable to backdoor attacks. However, the systemic risks they pose in face representation have received little attention. The entanglement of identity features in self-supervised face learning presents unique challenges for attack stealthiness. To address this gap, we propose FIDA (Feature Instability-Driven Attack), a novel backdoor attack framework. FIDA uses subtle semantic triggers for injection, but its key innovation is a novel objective called Feature Instability Loss. It trains the encoder to increase the sensitivity of triggered features along perturbation directions sampled during attack optimization . By preventing the backdoor from exhibiting the rigid feature patterns typical of previous attacks, FIDA effectively evades the evaluated perturbation-based defenses. Experiments show that FIDA achieves a high attack success rate and generally preserves benign utility across the evaluated settings , posing a significant threat to real-world multimedia applications relying on facial analysis.
Chinese Translation
自监督学习（SSL）模型易受到后门攻击。然而，它们在面部表征中所带来的系统性风险却鲜有关注。自监督人脸学习中身份特征的纠缠为攻击的隐蔽性带来了独特挑战。为了解决这一问题，我们提出了FIDA（基于特征不稳定性的攻击），一种新颖的后门攻击框架。FIDA使用细微的语义触发器进行注入，但其关键创新在于一种名为特征不稳定损失（Feature Instability Loss）的新目标。它训练编码器以提高在攻击优化过程中沿扰动方向采样的触发特征的敏感性。通过防止后门表现出以往攻击典型的刚性特征模式，FIDA有效地规避了评估的基于扰动的防御措施。实验表明，FIDA实现了高攻击成功率，并在评估的设置中通常保持良性的效用，对依赖面部分析的现实世界多媒体应用构成了重大威胁。
cs.CV / 50 / 2608.26866
Order Matters: A Chinese Multi-Panel Meme Benchmark for Vision-Language Reasoning
顺序重要性：一个用于视觉-语言推理的中文多面板表情包基准
Abstract
Many multimodal tasks depend on how visual elements are ordered and composed, not only on recognizing them in isolation. Internet memes are a compact case of this problem: their punchline often depends on a constrained reading order and cross-panel visual--textual cues. While large vision-language models (LVLMs) show strong performance on single-image understanding, it remains unclear whether they can perform sequence-aware reasoning over structured meme layouts, especially in Chinese social media. We introduce CMPM, a Chinese Multi-Panel Meme benchmark with 1,214 annotated samples covering five structural types, ordering dependency, panel-order constraints, and optional comment context. We formulate a two-layer evaluation: Task1 probes structure typing and order-sensitive panel sequencing (with a context ablation setting), and Task2 evaluates Chinese meme explanation generation with human ratings on five 1-3 Likert dimensions (visual, panel, humor, context, and faithfulness). We benchmark five representative LVLMs under a unified protocol. Results indicate that canonical-display accuracy is not by itself evidence of order understanding: the primary shuffled condition produces a sharp accuracy drop, revealing a persistent gap in order-sensitive multimodal reasoning. Task2 preferences place Gemini 3.1 Pro and GPT-5.5 above the open models, while comment context yields only a small and mixed Core4 gain. Code and data will be released upon acceptance.
Chinese Translation
许多多模态任务依赖于视觉元素的顺序和组合，而不仅仅是孤立地识别它们。互联网表情包是这一问题的一个紧凑案例：它们的笑点往往依赖于受限的阅读顺序和跨面板的视觉-文本线索。尽管大型视觉-语言模型（LVLMs）在单图像理解上表现出色，但尚不清楚它们是否能够在结构化的表情包布局上进行顺序感知推理，尤其是在中文社交媒体中。我们引入了CMPM，一个包含1,214个注释样本的中文多面板表情包基准，涵盖五种结构类型、顺序依赖性、面板顺序约束和可选评论上下文。我们制定了一个两层评估：任务1探测结构类型和顺序敏感的面板排序（带有上下文消融设置），任务2评估中文表情包解释生成，并在五个1-3李克特维度（视觉、面板、幽默、上下文和忠实度）上进行人工评分。我们在统一协议下对五个代表性的LVLMs进行了基准测试。结果表明，典型展示的准确性本身并不能证明顺序理解：主要的随机条件导致准确性大幅下降，揭示了顺序敏感多模态推理中的持续差距。任务2的偏好将Gemini 3.1 Pro和GPT-5.5置于开放模型之上，而评论上下文仅带来了小幅且混合的Core4增益。代码和数据将在接受后发布。
cs.CV / 51 / 2608.26868
CGS-SLAM: Collaborative Gaussian Splatting based SLAM for Multi-Agent Reconstruction
CGS-SLAM：基于协作高斯点云的多智能体重建SLAM
Abstract
Recent advances in SLAM have leveraged 3DGS for photorealistic reconstruction and novel view synthesis. However, most methods rely on RGB-D input, which is unavailable on consumer-grade smartphones, and few integrate 3DGS within a collaborative framework. Therefore, we present CGS-SLAM, a hybrid decentralized/centralized system enabling multi-agent 3DGS SLAM using only RGB and inertial data. Each agent performs local tracking with inertial data as a motion prior and reconstructs a scaled map using a metric monocular depth estimator (Depth Pro). Keyframe encodings are shared among agents, enabling dynamic keyframing in regions of spatial overlaps with other agents, enhancing submap alignment. Afterwards, a central server aligns submaps using VGGT as a view alignment model. This bidirectional communication keeps communication cost low during mapping and global reconstruction in difficult GNSS-denied environments. Experiments on multiple datasets demonstrate competitive tracking performance, improved rendering quality over state-of-the-art methods, and accurate submap alignment.
Chinese Translation
最近，SLAM领域的进展利用3DGS实现了逼真的重建和新视图合成。然而，大多数方法依赖于RGB-D输入，而这一输入在消费级智能手机上并不可用，且很少有方法将3DGS集成到协作框架中。因此，我们提出了CGS-SLAM，一种混合的去中心化/中心化系统，仅使用RGB和惯性数据实现多智能体3DGS SLAM。每个智能体利用惯性数据作为运动先验进行局部跟踪，并使用度量单目深度估计器（Depth Pro）重建缩放地图。关键帧编码在智能体之间共享，使得在与其他智能体空间重叠的区域内实现动态关键帧，提高子地图对齐精度。随后，中央服务器使用VGGT作为视图对齐模型对子地图进行对齐。这种双向通信在困难的GNSS失效环境中保持了低通信成本的同时，支持了映射和全局重建。多个数据集上的实验表明，CGS-SLAM在跟踪性能上具有竞争力，渲染质量优于最先进的方法，并且子地图对齐准确。
cs.CV / 52 / 2608.26872
Self-OPD: On-Policy Distillation for Flow Matching Models without Teacher
Self-OPD：无教师的流匹配模型的在线蒸馏
Abstract
On-policy distillation (OPD), which leverages a pre-trained, specialized teacher model to provide dense supervisory signals, has achieved significant success in Large Language Models (LLMs) and has recently been adapted to flow matching models. However, this paradigm suffers from two major issues: First, training a separate, task-specific teacher for every new objective incurs high computational costs. Second, the discrepancy between teacher and student distributions often leads to compounding errors along the generation trajectory. In this paper, we introduce \textbf{Self-OPD}, a teacher-free OPD framework for flow matching models that turns the student's own self-exploration into step-wise supervision. At each timestep, Self-OPD branches the deterministic next-state prediction into $K$ stochastic SDE candidates, rolls them out with the ODE sampler, and compares their rewards against a deterministic self-reference baseline to obtain normalized advantages. The velocity field is optimized with an all-branch pull-push objective, where high-advantage branches attract the student and low-advantage branches repel it under direction-aware attenuation and SDE-variance normalization. For multi-objective alignment, Self-OPD fuses normalized scores at the reward level, avoiding direct gradient conflict. Experiments on single and mixed reward benchmarks show that Self-OPD outperforms prior RL and OPD methods without task-specific teachers.
Chinese Translation
在线蒸馏（On-policy distillation, OPD）利用预训练的专门教师模型提供密集的监督信号，在大型语言模型（Large Language Models, LLMs）中取得了显著成功，并且最近被应用于流匹配模型。然而，这一范式存在两个主要问题：首先，为每个新目标训练一个单独的、特定任务的教师模型会产生高昂的计算成本。其次，教师与学生分布之间的差异往往导致生成轨迹上的累积误差。在本文中，我们提出了 extbf{Self-OPD}，一个无教师的流匹配模型的在线蒸馏框架，将学生自身的自我探索转化为逐步监督。在每个时间步，Self-OPD将确定性的下一个状态预测分支为$K$个随机的随机微分方程（Stochastic Differential Equation, SDE）候选项，通过常微分方程（Ordinary Differential Equation, ODE）采样器进行展开，并将其奖励与确定性的自我参考基线进行比较，以获得归一化的优势。速度场通过全分支拉推目标进行优化，其中高优势分支吸引学生，而低优势分支在方向感知衰减和SDE方差归一化下排斥学生。对于多目标对齐，Self-OPD在奖励层面融合归一化得分，避免直接的梯度冲突。在单一和混合奖励基准上的实验表明，Self-OPD在没有特定任务教师的情况下优于先前的强化学习（Reinforcement Learning, RL）和在线蒸馏方法。
cs.CV / 53 / 2608.26902
Tether the Subject, Release the Scene: Query-Aware Memory Routing for Long-Horizon Autoregressive Video Generation
固定主体，释放场景：面向查询的记忆路由用于长时间自回归视频生成
Abstract
Streaming autoregressive video models generate long videos chunk by chunk, using historical memory to maintain consistency. Existing methods typically expose subject and scene queries to history through similar policies. This stabilizes the subject, but can also lock backgrounds, viewpoints, and scene structure to previously generated states even when local motion continues. We call this failure memory-anchored scene under-progression; consistency and motion metrics alone can miss it. We introduce TetherMem, a training-free, query-aware spatiotemporal memory router for frozen video generators. TetherMem separates subject and scene queries and modulates historical access with region- and age-conditioned priors: subject queries retain identity-bearing history, while scene queries reduce reliance on subject history and stale backgrounds. Across 2,400 blinded pairwise judgments from 10 annotators, TetherMem achieves the highest estimated expected preference among eight streaming long-video baselines for overall quality (0.780) and scene progression (0.769). On complete 30-second videos, it sustains changes in background, viewpoint, and scene state while preserving subject recognizability and temporal continuity.
Chinese Translation
流式自回归视频模型通过逐块生成长视频，利用历史记忆保持一致性。现有方法通常通过类似的策略将主体和场景查询暴露于历史中。这虽然稳定了主体，但也可能将背景、视角和场景结构锁定在先前生成的状态，即使局部运动仍在继续。我们称这种失败为记忆锚定的场景进展不足；仅依靠一致性和运动指标可能无法捕捉到这一点。我们提出了 TetherMem，这是一种无训练、面向查询的时空记忆路由器，适用于冻结视频生成器。TetherMem 将主体和场景查询分开，并通过区域和时间条件的先验调节历史访问：主体查询保留身份相关的历史，而场景查询减少对主体历史和过时背景的依赖。在来自10位注释者的2400个盲评对比中，TetherMem 在整体质量（0.780）和场景进展（0.769）方面在八个流式长视频基线中获得了最高的预期偏好估计。在完整的30秒视频中，它在保持主体可识别性和时间连续性的同时，持续变化背景、视角和场景状态。
cs.CV / 54 / 2608.26921
AraMS-28k: The Largest Publicly Released Line-Level Dataset of Historical Arabic Manuscripts with Margin and Insertion-Anchor Annotations
AraMS-28k：历史阿拉伯手稿中带有边注和插入锚点注释的最大公开发布行级数据集
Abstract
We introduce AraMS-28k, the largest publicly released line-level dataset of genuine historical Arabic manuscripts, comprising 14 books, 3,043 pages, and 28,600 annotated text lines (27,971 main-text, 629 margin). Thirteen books are hand-copied manuscripts spanning three script traditions -- Naskh, Ruq'ah, and Maghrebi -- and one is a lithographed printed edition included to broaden format diversity. Each line is labelled as main-text or margin, and margin lines that have an unambiguous attachment point in the main text are further annotated with an insertion anchor, recovering the manuscript's true non-linear reading order at line-level granularity -- to our knowledge the first such annotation released for a historical Arabic manuscript corpus. Because reference transcriptions are fully vocalised while manuscript hands are typically undiacritised, we release both the raw diacritised transcription and a diacritic-normalised counterpart for every line. The dataset was constructed with RefLAM, a reference-grounded annotation pipeline that aligns multimodal-LLM OCR against independently sourced clean transcriptions and routes every line through human review, combining automatic verification with expert oversight. We describe the construction and quality-control process, present the annotation schema, report dataset statistics at both the corpus and per-book level, and provide baseline HTR results using Kraken and HATFormer, including a cross-script generalisation gradient from in-distribution pages to fully unseen books. AraMS-28k is released with page images, line-level annotations, and fixed train/val/test splits under CC BY-NC-SA 4.0 to support reproducible research on Arabic manuscript recognition, layout analysis, and reading-order recovery.
Chinese Translation
我们介绍了AraMS-28k，这是最大公开发布的真实历史阿拉伯手稿行级数据集，包含14本书、3,043页和28,600条注释文本行（27,971条主文本，629条边注）。其中十三本书为手抄手稿，涵盖三种书写传统——Naskh、Ruq'ah和Maghrebi——而一本为石印版，以增加格式多样性。每一行被标记为主文本或边注，具有明确附着点的边注行进一步注释了插入锚点，恢复了手稿在行级粒度下的真实非线性阅读顺序——据我们所知，这是首次为历史阿拉伯手稿语料库发布此类注释。由于参考抄本完全标注元音，而手稿书写通常不带元音，我们为每一行发布了原始的带元音抄本和一个标准化的无元音版本。该数据集是通过RefLAM构建的，这是一个基于参考的注释流程，将多模态LLM OCR与独立来源的干净抄本对齐，并通过人工审核对每一行进行处理，结合自动验证与专家监督。我们描述了构建和质量控制过程，展示了注释方案，报告了语料库和每本书的统计数据，并提供了使用Kraken和HATFormer的基线HTR结果，包括从分布内页面到完全未见书籍的跨脚本泛化梯度。AraMS-28k发布了页面图像、行级注释以及固定的训练/验证/测试划分，遵循CC BY-NC-SA 4.0协议，以支持对阿拉伯手稿识别、布局分析和阅读顺序恢复的可重复研究。
cs.CV / 55 / 2608.26948
KISS-GS: 3D Gaussian Splatting Compression Kept Simple
KISS-GS：简单化的3D高斯点云压缩
Abstract
Scene reconstruction with 3D Gaussian Splatting (3DGS) has become common, however deployment remains painful as the uncompressed file sizes can be massive. Current 3DGS compression systems combine multiple strategies for file size reduction, which can obscure where gains come from and limit component reuse across training pipelines. To make the gains more transparent, we propose KISS-GS, a modular compression pipeline named after the principle of keeping things simple, designed to decouple compression entirely from training. Given a 3DGS scene reconstructed with vanilla 3DGS, we are able to reduce it through compaction by 15.7x using a combination of state-of-the-art pruning schemes. Then we encode it into an image-based format designed for simple, ubiquitous decoding. With the SOG-XT format, we propose a novel extension to Self-Organizing Gaussians with two main contributions: (i) Self-organizing 2D Codebooks and (ii) Parallel Representative Assignment Smoothing (PRAS), which leverages the symmetry of quaternion and scale parameterizations to produce 2D attribute grids more amenable to encoding. This encoding reduces scene size by 6.6x. We show that optional encoding-aware fine-tuning yields a further 2.2x. Across standard 3DGS benchmarks, our simple and modular approach thus achieves a total of 85x to 319x reductions in the size of the scene over uncompressed vanilla 3DGS, setting new benchmarks for real-world scenes and surpassing tightly integrated methods in rate-distortion. Decoding relies solely on web-native image formats, and the modular design makes each stage easy to combine with future advances in reconstruction and compaction. Code and project page: https://fraunhoferhhi.github.io/KISS-GS/
Chinese Translation
使用3D高斯点云（3DGS）进行场景重建已变得普遍，但由于未压缩文件的大小可能非常庞大，部署仍然很痛苦。目前的3DGS压缩系统结合了多种策略以减少文件大小，这可能会模糊收益来源，并限制在训练管道中的组件重用。为了使收益更加透明，我们提出了KISS-GS，一个模块化的压缩管道，其名称源于保持简单的原则，旨在将压缩与训练完全解耦。给定一个使用普通3DGS重建的3DGS场景，我们能够通过结合最先进的剪枝方案将其压缩至15.7倍。然后，我们将其编码为一种基于图像的格式，旨在实现简单、普遍的解码。通过SOG-XT格式，我们提出了对自组织高斯（Self-Organizing Gaussians）的新扩展，主要贡献有：(i) 自组织2D代码本和(ii) 并行代表分配平滑（Parallel Representative Assignment Smoothing，PRAS），后者利用四元数和尺度参数化的对称性生成更适合编码的2D属性网格。此编码将场景大小减少了6.6倍。我们展示了可选的编码感知微调进一步提高了2.2倍。在标准3DGS基准测试中，我们简单且模块化的方法实现了相较于未压缩普通3DGS场景总计85倍至319倍的大小减少，为现实世界场景设定了新的基准，并在率失真方面超越了紧密集成的方法。解码仅依赖于网络原生图像格式，模块化设计使每个阶段都易于与未来的重建和压缩进展相结合。代码和项目页面：https://fraunhoferhhi.github.io/KISS-GS/
cs.CV / 56 / 2608.26951
Per-View Gaussian Predictions Enable Training-Free Distractor Filtering in Feed-Forward 3DGS
基于每视图高斯预测的无训练干扰物体过滤方法在前馈3D高斯点云中的应用
Abstract
Feed-forward 3D Gaussian Splatting reconstructs an explicit Gaussian representation from multiple input images in one network execution, making 3D reconstruction increasingly accessible for casual captures. However, such captures frequently contain transient objects that appear in only a subset of the views. Such content can be encoded into the per-view Gaussians associated with the inputs that observe it and remain in the combined representation despite being observed by no other input. As a result, it may produce blurred, duplicated, or floating artifacts in novel views. We introduce a training-free filtering procedure that exploits this per-view prediction structure. For each input, we exclude its associated Gaussians and render the same camera using the remaining representation, revealing content that is inconsistent with the other inputs. Feature similarity forms candidate regions, and rendering-based verification retains only candidates whose removal reduces reconstruction error in the other input views. The procedure operates on a single frozen prediction without retraining or scene-specific optimization. Across three reconstruction models and two distractor benchmarks, it consistently improves novel-view quality with varying numbers of input views. On clean scenes, evaluations across four models show that the original reconstructions are largely preserved.
Chinese Translation
前馈3D高斯点云重建通过一次网络执行从多个输入图像中重建显式高斯表示，使得3D重建对于随意捕捉变得愈加可行。然而，这些捕捉中常常包含仅在部分视图中出现的瞬态物体。这类内容可以编码到与观察到它的输入相关联的每视图高斯中，并且尽管没有其他输入观察到，它仍然保留在组合表示中。因此，在新视图中可能会产生模糊、重复或漂浮的伪影。我们提出了一种无训练的过滤程序，利用这种每视图预测结构。对于每个输入，我们排除其相关的高斯，并使用剩余的表示渲染相同的相机，从而揭示与其他输入不一致的内容。特征相似性形成候选区域，基于渲染的验证仅保留那些去除后能减少其他输入视图重建误差的候选。该过程在单个冻结预测上运行，无需重新训练或场景特定优化。在三个重建模型和两个干扰物体基准测试中，它在不同数量的输入视图下始终提高新视图的质量。在干净场景中，四个模型的评估显示原始重建在很大程度上得以保留。
cs.CV / 57 / 2608.26956
RubricRM: Generative Reward Modeling via Dynamic Rubrics for Image Generation and Editing
RubricRM：通过动态评分标准进行图像生成和编辑的生成奖励建模
Abstract
Reward models play an essential role in aligning visual generative models, yet most existing visual reward models use a single scalar score or rely on fixed criteria that cannot adapt to different instructions. This limits both interpretability and task sensitivity, especially for text-to-image generation and instruction-based image editing, where different inputs require different evaluation dimensions. We propose RubricRM, a pairwise generative reward modeling framework that first produces an input-specific rubric with evaluation dimensions, weights, and scoring criteria, and then applies the rubric to score candidate images. We train dedicated RubricRM models for text-to-image generation and image editing using a two-stage training pipeline: supervised fine-tuning teaches the model the rubric-based scoring paradigm, while GRPO further improves scoring through fine-grained dimension-level rewards. Experiments on multiple generation and editing benchmarks show that RubricRM outperforms existing specialized reward models and remains competitive with strong proprietary MLLM judges despite using smaller backbones. Our models, data, and code are available at https://github.com/zijiankan/RubricRM.
Chinese Translation
奖励模型在对齐视觉生成模型中发挥着重要作用，但大多数现有的视觉奖励模型使用单一的标量评分或依赖于固定标准，无法适应不同的指令。这限制了可解释性和任务敏感性，尤其是在文本到图像生成和基于指令的图像编辑中，不同的输入需要不同的评估维度。我们提出了RubricRM，这是一种成对生成奖励建模框架，首先生成具有评估维度、权重和评分标准的输入特定评分标准，然后将该评分标准应用于候选图像的评分。我们使用两阶段训练流程为文本到图像生成和图像编辑训练专用的RubricRM模型：监督微调教会模型基于评分标准的评分范式，而GRPO通过细粒度的维度级奖励进一步改善评分。在多个生成和编辑基准上的实验表明，RubricRM的表现优于现有的专用奖励模型，并且尽管使用了较小的基础模型，仍与强大的专有MLLM评审保持竞争力。我们的模型、数据和代码可在https://github.com/zijiankan/RubricRM获取。
cs.CV / 58 / 2608.26960
Geo-LoRA: Geometry-Aware Subspace Evolution for Low-Rank Adaptation in Continual Learning
Geo-LoRA：面向几何的子空间演化以实现持续学习中的低秩适应
Abstract
Rehearsal-free class-incremental learning (CIL) with LoRA adapters remains challenging because the low-rank subspaces updated across tasks evolve without geometric control, causing unstable shared representations and repetitive collapse of task-specific updates into previously occupied directions. We introduce Geo-LoRA, a geometry-aware framework that explicitly regulates how low-rank subspaces, both shared and task-specific, evolve during continual learning. For the shared branch, Subspace Projection Preservation (SPP) constrains consecutive updates to follow smooth trajectories on the Grassmann manifold, and Adaptive Core-Slack Alignment (ACSA) decomposes transitions into principal and residual components, aligning the former while modulating the latter to balance stability and plasticity. For the task-specific branch, Median-Calibrated Block Overlap (MCBO) imposes a statistical constraint via normalized projection overlap, penalizing excessive reuse to mitigate subspace crowding. These constraints jointly regulate the evolution of all LoRA subspaces across layers and tasks without introducing additional adapter types beyond standard LoRA. Geo-LoRA provides a principled geometric formulation for continual low-rank adaptation and consistently achieves state-of-the-art performance across multiple benchmark datasets and different task lengths.
Chinese Translation
在没有重演的类增量学习（CIL）中，使用LoRA适配器仍然面临挑战，因为跨任务更新的低秩子空间在没有几何控制的情况下演变，导致共享表示不稳定以及任务特定更新重复崩溃到先前占用的方向。我们提出了Geo-LoRA，一个面向几何的框架，明确调节低秩子空间（包括共享和任务特定的）在持续学习过程中的演变。对于共享分支，子空间投影保持（Subspace Projection Preservation, SPP）约束连续更新遵循在Grassmann流形上的平滑轨迹，而自适应核心松弛对齐（Adaptive Core-Slack Alignment, ACSA）将过渡分解为主成分和残余成分，调整前者的对齐，同时调节后者以平衡稳定性和可塑性。对于任务特定分支，中位数校准块重叠（Median-Calibrated Block Overlap, MCBO）通过归一化投影重叠施加统计约束，惩罚过度重用以减轻子空间拥挤。这些约束共同调节所有LoRA子空间在各层和任务间的演变，而无需引入超出标准LoRA的额外适配器类型。Geo-LoRA为持续低秩适应提供了一个原则性的几何公式，并在多个基准数据集和不同任务长度上持续实现了最先进的性能。
cs.CV / 59 / 2608.26971
TempJail: Temporal Jailbreak Attacks against Image-to-Video Generation Models
TempJail：针对图像到视频生成模型的时间性越狱攻击
Abstract
In recent years, image-to-video (I2V) generation models have made remarkable progress in subject consistency and temporal coherence, enabling high quality video synthesis. However, these advances also introduce new safety risks. Existing studies mainly focus on jailbreak attacks involving single frame violations, while largely overlooking the temporal dimension unique to video generation models. In this paper, we investigate three attack scenarios and uncover a temporal vulnerability in I2V systems: unsafe semantics may emerge not from a single frame, but from semantic composition over time. We further identify two key challenges in such attacks: temporal abstraction and semantic camouflage. To address these issues, we propose TempJail, a novel temporal jailbreak framework for I2V systems. For temporal abstraction, we decompose a target malicious caption into an initial frame visual condition and a temporal text instruction. For semantic camouflage, on the image side we model semantic injection as controlled latent perturbation in diffusion sampling and introduce gradient guidance from pretrained encoders. On the text side, we rewrite the caption into an innocuous ``subject-action-scene'' template that bypasses safety filters while preserving temporal guidance. In the black-box inference phase, these two modalities jointly enable malicious semantics to be gradually triggered over time. Experiments on closed-source commercial models, including Kling, Seedance, Veo and PixVerse, show that TempJail improves attack success rate over prior state-of-the-art methods by 23.3\% under GPT-5.2 evaluation and 22.0\% under human evaluation. Our codes are available at \href{https://github.com/luqi-glory/TempJail}{GitHub}.
Chinese Translation
近年来，图像到视频（I2V）生成模型在主题一致性和时间连贯性方面取得了显著进展，使得高质量视频合成成为可能。然而，这些进展也带来了新的安全风险。现有研究主要集中在涉及单帧违规的越狱攻击上，而在很大程度上忽视了视频生成模型独特的时间维度。本文研究了三种攻击场景，并揭示了I2V系统中的一种时间性脆弱性：不安全的语义可能并非源于单帧，而是源于随时间变化的语义组合。我们进一步识别了此类攻击中的两个关键挑战：时间抽象和语义伪装。为了解决这些问题，我们提出了TempJail，一个针对I2V系统的新型时间性越狱框架。在时间抽象方面，我们将目标恶意标题分解为初始帧视觉条件和时间文本指令。在语义伪装方面，在图像侧我们将语义注入建模为扩散采样中的受控潜在扰动，并引入来自预训练编码器的梯度引导。在文本侧，我们将标题重写为一个无害的“主题-动作-场景”模板，以绕过安全过滤器，同时保留时间指导。在黑箱推理阶段，这两种模态共同使恶意语义能够随着时间的推移逐渐被触发。在对闭源商业模型（包括Kling、Seedance、Veo和PixVerse）的实验中，结果显示TempJail在GPT-5.2评估下提高了攻击成功率23.3%，在人工评估下提高了22.0%。我们的代码可在GitHub上获取。
cs.CV / 60 / 2608.26993
Aphanta: Diagnosing Task-Aligned Image-Edited Intermediates for Multimodal Reasoning
Aphanta：诊断任务对齐的图像编辑中间体以进行多模态推理
Abstract
Explicit visual intermediates can help multimodal large language models (MLLMs) externalize spatial evidence and updated visual states, but their utility depends on whether an image editor can faithfully realize the required transformation. We introduce \textbf{Aphanta}, an automated task-discovery and closed-loop diagnostic framework for the MLLM -> image editor -> MLLM pipeline. Aphanta evaluates three conditions---direct reasoning, reasoning with an editor-generated intermediate, and reasoning with an idealized reference intermediate---to separate potential visual headroom from the practical utility of current editors. Across 20 candidate tasks and multiple editor--MLLM combinations, we find that utility is strongly task-conditioned. Gains concentrate in visual cue injection, grounding, and counterfactual state realization, whereas intermediates requiring symbol-sensitive construction or structural extrapolation are substantially less reliable. On the selected positive-task subset, our consolidated Qwen pipeline improves the mean task score from 0.343 to 0.445 ($+10.2$ points; $+29.7\%$ relative), while the full study also retains filtered and unsuccessful tasks to expose the boundary. These results position image editing as a specialized visual workspace rather than a universal reasoning mechanism, and establish Aphanta as a reusable protocol for measuring task--representation alignment, editor realization, and downstream pipeline utility.
Chinese Translation
显式视觉中间体可以帮助多模态大型语言模型（MLLMs）外化空间证据和更新的视觉状态，但其效用取决于图像编辑器是否能够忠实地实现所需的转换。我们引入了 extbf{Aphanta}，这是一个自动化的任务发现和闭环诊断框架，适用于MLLM -> 图像编辑器 -> MLLM的流程。Aphanta评估三种条件——直接推理、使用编辑器生成的中间体进行推理，以及使用理想化参考中间体进行推理——以区分潜在的视觉提升空间与当前编辑器的实际效用。在20个候选任务和多个编辑器-MLLM组合中，我们发现效用与任务密切相关。效益主要集中在视觉线索注入、基础定位和反事实状态实现上，而需要符号敏感构建或结构外推的中间体则显著不可靠。在选定的正任务子集中，我们的整合Qwen流程将平均任务得分从0.343提高到0.445（+10.2分；相对增加29.7%），同时完整研究还保留了过滤和不成功的任务以揭示边界。这些结果将图像编辑定位为一个专门的视觉工作空间，而非普遍的推理机制，并确立了Aphanta作为测量任务-表征对齐、编辑器实现和下游流程效用的可重用协议。
cs.CV / 61 / 2608.26998
Virtual iEEG from Scalp EEG: Charting the Landscape of Source Imaging, Intracranial Inference and Reconstruction
从头皮脑电图到虚拟颅内脑电图：源成像、颅内推断与重建的全景图
Abstract
Intracranial electroencephalography (iEEG) provides temporally precise and spatially specific access to neural activity from focal and deep brain regions, but its invasiveness and restricted anatomical coverage limit routine use. These constraints have motivated scalp-to-intracranial inference, termed virtual iEEG when model outputs carry iEEG-defined event, feature, representation, or contact-level waveform semantics. This review presents a target-centred framework distinguishing event inference, feature translation, and waveform reconstruction, while separating predictability from observability, identifiability, fidelity, and utility. Evidence is evaluated according to cohort independence, anatomical and spectral coverage, train--test separation, and target-patient adaptation. Current studies support inference of selected intracranial events, low-frequency components, and task-related representations, but not unique recovery of arbitrary contact-level activity. Stronger validation requires appropriate controls, source-imaging baselines, uncertainty assessment, and incremental-utility testing. Future progress depends on independent paired datasets and prospective evidence that virtual iEEG adds value beyond scalp EEG and EEG source imaging.
Chinese Translation
颅内脑电图（iEEG）能够精确地获取来自局部和深部脑区的神经活动，但其侵入性和受限的解剖覆盖范围限制了其常规使用。这些限制促使了从头皮到颅内的推断，当模型输出携带iEEG定义的事件、特征、表示或接触级波形语义时，称为虚拟iEEG。本文综述提出了一种以目标为中心的框架，区分事件推断、特征转换和波形重建，同时将可预测性与可观察性、可识别性、保真度和实用性分开。根据队列独立性、解剖和频谱覆盖、训练-测试分离以及目标-患者适应性评估证据。目前的研究支持对选定的颅内事件、低频成分和任务相关表示的推断，但不支持对任意接触级活动的独特恢复。更强的验证需要适当的对照、源成像基线、不确定性评估和增量实用性测试。未来的进展依赖于独立的配对数据集和前瞻性证据，证明虚拟iEEG在头皮脑电图和脑电源成像之外增加了价值。
cs.CV / 62 / 2608.27004
MVC-Bench: Benchmarking Calibration of Medical Vision-Language Models
MVC-Bench：医学视觉语言模型校准的基准测试
Abstract
Reliable evaluation of vision-language models (VLMs) and medical vision-language models (Medical-VLMs) requires calibrated confidence, particularly under realistic clinical conditions. However, existing efforts mainly focused on improving accuracy, leaving calibration in the medical domain underexplored. To this end, we propose MVC-Bench, a calibration-centric benchmark for medical image classification with VLMs and Medical-VLMs. MVC-Bench assesses the calibration across three axes: (i) robustness to modality, backbone, and domain shift (ii) effectiveness of calibration strategies and prompt-tuning methods (iii) stability under prompt-template and random-seed variations. The benchmark covers eight different backbones, three medical modalities, including fundus imaging, histopathology, and chest X-ray under in-domain and domain shift settings. It compares post-hoc calibration, train-time calibration, and zero-shot inference methods, together with six prompt-tuning methods. Across more than 1638 controlled experiments, we report accuracy and Expected Calibration Error (ECE) as primary metrics, and further report results with complementary calibration measures, including Maximum Calibration Error (MCE) and Adaptive Calibration Error (ACE). We further investigate the underlying causes of miscalibration in VLMs and Medical-VLMs and propose a simple train-time calibration method, Multi-Class Margin (MCM) regularization, which achieves lowest ECE on 10 out of 12 settings in in-domain and remains competitive under domain shifts. Collectively, MVC-Bench provides a structured evaluation framework and actionable guidance for improving calibration in safety-critical medical workflows.
Chinese Translation
对视觉语言模型（VLMs）和医学视觉语言模型（Medical-VLMs）的可靠评估需要经过校准的置信度，特别是在现实临床条件下。然而，现有的研究主要集中在提高准确性上，导致医学领域的校准研究相对不足。为此，我们提出了MVC-Bench，这是一个以校准为中心的医学图像分类基准，适用于VLMs和Medical-VLMs。MVC-Bench从三个方面评估校准： (i) 对模态、主干网络和领域转移的鲁棒性 (ii) 校准策略和提示调优方法的有效性 (iii) 在提示模板和随机种子变化下的稳定性。该基准涵盖八种不同的主干网络、三种医学模态，包括眼底成像、组织病理学和胸部X光，在领域内和领域转移设置下进行比较。它比较了事后校准、训练时校准和零样本推理方法，以及六种提示调优方法。在超过1638个受控实验中，我们报告了准确性和期望校准误差（ECE）作为主要指标，并进一步报告了包括最大校准误差（MCE）和自适应校准误差（ACE）在内的补充校准测量结果。我们还探讨了VLMs和Medical-VLMs中校准失效的潜在原因，并提出了一种简单的训练时校准方法——多类边际（Multi-Class Margin, MCM）正则化，该方法在12种设置中的10种上实现了最低的ECE，并在领域转移下保持竞争力。总体而言，MVC-Bench提供了一个结构化的评估框架和可操作的指导，以改善安全关键的医学工作流程中的校准。
cs.CV / 63 / 2608.27034
Differentiable Jitter Correction using Deep Learning-based Image Quality Metric for Phase-Contrast Micro-CT
基于深度学习图像质量度量的相位对比微计算机断层扫描可微抖动校正
Abstract
This paper proposes a fully differentiable jitter correction method for X-ray phase-contrast micro computed tomography using a deep learning-based image quality metric that estimates and compensates per-projection rigid jitter directly from the acquired projection data, without a pre-scan motion-free reference. The approach builds on a gradient-based auto-focus strategy adapted to parallel-beam geometry. A set of candidate objective functions is benchmarked in a controlled study, and the sensitivity of the visual information fidelity (VIF) metric to the jitter artifact is verified with the target phase-contrast data. To operate without a clean reference, a compact 3D convolutional neural network is trained to predict the VIF score from a single corrupted volume. A spatially selective total variation penalty applied exclusively to the image background is introduced to penalize spurious high-frequency structures that otherwise emerge during optimization. Experiments on biological specimens acquired at different synchrotron beamlines are conducted. Evaluation uses jitter motion applied to simulated and experimentally acquired projection data. The result confirms that the integrated pipeline reliably recovers fine structural detail lost due to jitter, with generalization demonstrated across morphologically distinct samples.
Chinese Translation
本文提出了一种完全可微的抖动校正方法，用于X射线相位对比微计算机断层扫描，利用基于深度学习的图像质量度量直接从获取的投影数据中估计和补偿每个投影的刚性抖动，而无需预扫描的无运动参考。该方法基于一种适用于平行束几何的基于梯度的自动聚焦策略。在受控研究中对一组候选目标函数进行了基准测试，并验证了视觉信息保真度（VIF）度量对抖动伪影的敏感性，使用了目标相位对比数据。为了在没有干净参考的情况下进行操作，训练了一个紧凑的3D卷积神经网络，以从单个受损体积预测VIF分数。引入了一种空间选择性的全变差惩罚，仅应用于图像背景，以惩罚在优化过程中出现的虚假高频结构。对在不同同步辐射光源下获取的生物样本进行了实验。评估使用施加于模拟和实验获取的投影数据上的抖动运动。结果确认集成管道可靠地恢复了因抖动而丢失的细微结构细节，并在形态上不同的样本中展示了泛化能力。
cs.CV / 64 / 2608.27039
Multi-Person Human Motion Forecasting in Complex Scenes
复杂场景中多人的人类运动预测
Abstract
Accurately forecasting the movement of people in complex scenes requires reasoning over the past and present state of the entire environment. In this context, effectively incorporating object information and social interactions into a unified framework remains particularly challenging. To address this, we propose Object-Conditioned Social Diffusion (OCSD), a conditional diffusion model that integrates motion history, multi-person interactions, and object cues into a single framework. OCSD uses an object-conditioning mechanism that modulates denoising at every timestep, enabling fine-grained human-object reasoning, and a social encoder that models the interactions between all humans in the scene. As a result, our model naturally handles varying group sizes, complex social interactions, and supports sampling multiple plausible futures. Extensive experiments show that OCSD achieves state-of-the-art results on the Humans in Kitchens (HiK) and HOI-M3 benchmarks. It reduces the two-second path error by 121.5 mm (31.3%) on HiK and 130.5 mm (33.2%) on HOI-M3 compared to prior work, and produces more realistic long-term forecasts.
Chinese Translation
准确预测复杂场景中人们的运动需要对整个环境的过去和现在状态进行推理。在这种背景下，有效地将物体信息和社会互动纳入统一框架仍然特别具有挑战性。为了解决这个问题，我们提出了物体条件社会扩散模型（Object-Conditioned Social Diffusion, OCSD），这是一种条件扩散模型，将运动历史、多人人际互动和物体线索整合到一个框架中。OCSD使用了一种物体条件机制，在每个时间步调节去噪过程，从而实现细粒度的人类-物体推理，并且采用社会编码器来建模场景中所有人类之间的互动。因此，我们的模型自然地处理不同的群体规模、复杂的社会互动，并支持采样多个合理的未来。大量实验表明，OCSD在厨房中的人类（Humans in Kitchens, HiK）和HOI-M3基准测试上达到了最先进的结果。与之前的工作相比，它在HiK上将两秒路径误差减少了121.5毫米（31.3%），在HOI-M3上减少了130.5毫米（33.2%），并产生了更为真实的长期预测。
cs.CV / 65 / 2608.27051
Anatomy-Guided Foundation Model Adaptation with Within-Case Prototype Supervision for Standard Plane Detection in Fetal Ultrasound Blind Sweeps
基于解剖学指导的基础模型适应与案例内原型监督用于胎儿超声盲扫中的标准平面检测
Abstract
Detecting the fetal abdominal circumference standard plane in low-cost obstetric blind sweeps is a highly imbalanced frame-classification problem: positive frames account for under 3% of a sequence, form short contiguous segments, and are poorly handled by off-the-shelf ultrasound and vision foundation models. We propose AnatoProto, a lightweight sequence-level framework that adapts a frozen BiomedCLIP encoder to fetal blind sweeps through four components: (i) anatomy-weighted spatial pooling that uses nnU-Net abdominal-region probabilities as a spatial prior to reweight BiomedCLIP patch tokens, so frozen semantic features are aggregated onto anatomically meaningful regions; (ii) a within-case prototype loss that pulls each frame embedding toward the mean of positive frames of the same sweep, exploiting case-level structure unavailable at the frame level; (iii) a three-stage cascade refinement (frame->segment->case-level rejecter) that lifts the prediction unit from noisy frames to structurally-constrained segments; and (iv) a hybrid prediction head that jointly models per-frame stability and inter-frame boundary transitions to suppress boundary false positives. On the ACOUSLIC-AI benchmark, AnatoProto reaches a test F1 of 67.72, outperforming the strongest foundation-model baseline (FetalCLIP + PRS, F1 = 54.52) by +13.20 F1 and the strongest video temporal-action-detection baseline (TriDet + PRS) by +15.76 F1. A synergy study, backed by embedding geometry and paired-bootstrap confidence intervals, shows that the prototype loss and anatomy-weighted pooling are not additive: applied alone the prototype loss reduces recall by 12 points, but combined with anatomy-weighted pooling it increases recall by 6.5 points -- a sign-flip we trace to the accuracy of the within-case prototype.
Chinese Translation
在低成本产科盲扫中检测胎儿腹围标准平面是一个高度不平衡的帧分类问题：正帧在序列中占比不足3%，形成短的连续片段，且现有的超声和视觉基础模型对其处理效果不佳。我们提出了AnatoProto，一个轻量级的序列级框架，通过四个组件将冻结的BiomedCLIP编码器适应于胎儿盲扫：(i) 解剖学加权空间池化，利用nnU-Net腹部区域概率作为空间先验，重新加权BiomedCLIP的补丁令牌，从而将冻结的语义特征聚合到解剖学上有意义的区域；(ii) 案例内原型损失，将每个帧嵌入拉向同一盲扫的正帧均值，利用在帧级不可用的案例级结构；(iii) 三阶段级联精炼（帧->片段->案例级拒绝器），将预测单元从噪声帧提升到结构约束的片段；(iv) 混合预测头，联合建模每帧的稳定性和帧间边界过渡，以抑制边界假阳性。在ACOUSLIC-AI基准测试中，AnatoProto的测试F1达到了67.72，超越了最强基础模型基线（FetalCLIP + PRS，F1 = 54.52）13.20 F1，超越了最强视频时间动作检测基线（TriDet + PRS）15.76 F1。通过嵌入几何和配对自助置信区间支持的协同研究表明，原型损失和解剖学加权池化并非简单相加：单独应用原型损失会使召回率降低12个百分点，但与解剖学加权池化结合后召回率提高6.5个百分点——这一符号翻转我们追溯到案例内原型的准确性。
cs.CV / 66 / 2608.27065
Video-OPSD: Exploiting Privileged Visual Evidence for On-Policy Self-Distillation in Video Large Language Models
视频-OPSD：利用特权视觉证据进行视频大语言模型的策略自蒸馏
Abstract
On-policy self-distillation (OPSD) has recently emerged as an effective post-training paradigm that improves policy optimization through dense token-level supervision from a privileged self-teacher. Despite its promise, OPSD remains largely underexplored for Video Large Language Models (Video-LLMs). Existing methods typically construct privileged teachers by augmenting their context with additional information while keeping the primary input unchanged for both teacher and student. Video reasoning, however, offers a distinct source of privileged supervision within the primary input itself: long videos contain substantial temporal redundancy, and only a small subset of frames provides the evidence necessary to answer a question. Building on this observation, we present $\textbf{Video-OPSD}$, an OPSD framework that exploits privileged visual evidence for both self-teacher construction and knowledge transfer. First, our Evidence-Grounded Self-Teacher conditions the teacher exclusively on annotated evidence frames while the student continues to reason over the complete video. This focused visual input enables the teacher to provide more informative supervision. Second, our Evidence-Guided Token Optimization adaptively weights token-level distillation according to each reasoning token's reliance on privileged visual evidence, thereby emphasizing perceptually grounded reasoning. Experiments across video understanding and reasoning benchmarks show that $\textbf{Video-OPSD}$ consistently improves upon Standard OPSD across multiple backbones and achieves performance comparable to GRPO while requiring substantially less training time, establishing an effective and efficient post-training approach for Video-LLMs.
Chinese Translation
策略自蒸馏（On-policy self-distillation, OPSD）最近作为一种有效的后训练范式出现，通过来自特权自教师的密集标记级监督来改善策略优化。尽管前景广阔，OPSD在视频大语言模型（Video Large Language Models, Video-LLMs）中的应用仍然未被充分探索。现有方法通常通过在保持教师和学生的主要输入不变的情况下，使用额外信息来增强其上下文，从而构建特权教师。然而，视频推理提供了一个独特的特权监督来源：长视频包含大量时间冗余，只有一小部分帧提供回答问题所需的证据。基于这一观察，我们提出了$ extbf{Video-OPSD}$，一个利用特权视觉证据进行自教师构建和知识转移的OPSD框架。首先，我们的证据基础自教师（Evidence-Grounded Self-Teacher）仅在注释的证据帧上对教师进行条件设置，而学生则继续对完整视频进行推理。这种集中视觉输入使教师能够提供更具信息量的监督。其次，我们的证据引导标记优化（Evidence-Guided Token Optimization）根据每个推理标记对特权视觉证据的依赖程度自适应地加权标记级蒸馏，从而强调感知基础的推理。在视频理解和推理基准上的实验表明，$ extbf{Video-OPSD}$在多个骨干网络上始终优于标准OPSD，并且在性能上可与GRPO相媲美，同时所需的训练时间显著减少，确立了一种有效且高效的后训练方法，适用于视频大语言模型。
cs.CV / 67 / 2608.27066
Beyond Classification: Task-Dependent Learnability under Privacy-Motivated Image Transformations
超越分类：隐私驱动图像变换下的任务依赖可学习性
Abstract
Privacy-Enhancing Technologies (PETs) in computer vision often rely on noise or image perturbations to protect visual data while securely processing it, creating a trade-off between task performance and protection. This trade-off is commonly evaluated using image classification, which primarily captures semantic separability and remains robust despite significant geometric, spatial layout or local boundary alterations. As a result, it is too simplistic as a proxy for generic vision tasks. Exhaustive downstream-task evaluation, however, is computationally expensive because models must often be trained for each PET transformation and parameter setting. We therefore propose a compute-aware multi-task protocol for evaluating PETs in model training. It combines lightweight proxy tasks that target complementary aspects of visual structure while remaining simple and fast to compute. Across irreversible privacy transformations, key-based block primitives, and learnable image encryption schemes, we demonstrate that PETs with similar classification accuracy can differ substantially on other tasks. The outcomes highlight the need for PET evaluation protocols that move beyond classification-only reporting.
Chinese Translation
计算机视觉中的隐私增强技术（PETs）通常依赖噪声或图像扰动来保护视觉数据，同时安全地处理这些数据，从而在任务性能和保护之间形成权衡。这种权衡通常通过图像分类来评估，图像分类主要捕捉语义可分性，并且在显著的几何、空间布局或局部边界变化下仍然保持稳健。因此，作为通用视觉任务的代理，这种方法过于简单。然而，全面的下游任务评估在计算上是昂贵的，因为模型通常必须针对每个PET变换和参数设置进行训练。因此，我们提出了一种计算感知的多任务协议，用于在模型训练中评估PETs。该协议结合了轻量级的代理任务，针对视觉结构的互补方面，同时保持简单和快速的计算。通过不可逆的隐私变换、基于密钥的块原语和可学习的图像加密方案，我们展示了具有相似分类准确性的PETs在其他任务上可能存在显著差异。结果强调了需要超越仅分类报告的PET评估协议。
cs.CV / 68 / 2608.27073
SpatialCrafter: Single Image World Modeling with Generative 3D Proxies
SpatialCrafter：基于生成3D代理的单幅图像世界建模
Abstract
Explorable image-to-scene generation is essential for applications in gaming, robotics, and virtual reality. Existing methods based on video diffusion model (VDM) commonly rely on incomplete conditioning signals such as sparse point clouds or 2D panoramas, leading to stochastic hallucinations, long-term drifts and suboptimal 3D consistency. We present SpatialCrafter, a novel two-stage framework that addresses these issues by introducing a global 3D proxy for high-fidelity image-to-scene generation. Specifically, we decompose the generation process into global proxy generation and appearance refinement. For proxy generation, we propose a Point-anchored Sparse Structure~(PaSS) Flow module that predicts a spatially aligned and geometrically consistent 3D proxy. For appearance refinement, we re-frame the VDM as a Generative Deferred Refiner which synthesizes high-frequency photorealistic details upon proxy-defined scene geometry. To better integrate the proxy with the pre-trained VDM, we introduce Parallel Geometry Injection and Proxy-Aware Corruption training strategies, which improve robustness to proxy artifacts without disrupting the pretrained generative manifold. Furthermore, as no suitable dataset exists for this explorable scene generation task, we construct a new large-scale dataset of 115K scenes. To the best of our knowledge, it is the first hybrid dataset for image-to-scene generation. Extensive experiments on both synthetic and real-world datasets show that SpatialCrafter outperforms state-of-the-art methods, mitigates long-term drift, and remains robust and consistent under rapid camera motion and extreme viewpoint changes. Code, models, and the newly constructed dataset will be publicly released. See more at https://fangchuan.github.io/SpatialCrafter/.
Chinese Translation
可探索的图像到场景生成对于游戏、机器人和虚拟现实等应用至关重要。现有基于视频扩散模型（VDM）的方法通常依赖于不完整的条件信号，如稀疏点云或二维全景，导致随机幻觉、长期漂移和次优的3D一致性。我们提出了SpatialCrafter，这是一种新颖的两阶段框架，通过引入全球3D代理来解决这些问题，以实现高保真度的图像到场景生成。具体而言，我们将生成过程分解为全球代理生成和外观精细化。在代理生成方面，我们提出了一种点锚稀疏结构（PaSS）流模块，能够预测空间对齐和几何一致的3D代理。在外观精细化方面，我们将VDM重新框定为生成延迟细化器（Generative Deferred Refiner），该细化器在代理定义的场景几何上合成高频的照片级真实细节。为了更好地将代理与预训练的VDM集成，我们引入了并行几何注入和代理感知腐蚀训练策略，这些策略在不破坏预训练生成流形的情况下，提高了对代理伪影的鲁棒性。此外，由于没有适合此可探索场景生成任务的数据集，我们构建了一个新的大规模数据集，包含115K个场景。据我们所知，这是第一个用于图像到场景生成的混合数据集。在合成和真实世界数据集上的广泛实验表明，SpatialCrafter超越了最先进的方法，减轻了长期漂移，并在快速相机运动和极端视角变化下保持鲁棒性和一致性。代码、模型和新构建的数据集将公开发布。更多信息请访问 https://fangchuan.github.io/SpatialCrafter/。
cs.CV / 69 / 2608.27095
Automated 2D and 3D Segmentation of AMD and DME Lesions in OCT
自动化的2D和3D视网膜黄斑变性（AMD）及糖尿病性黄斑水肿（DME）病变分割在光学相干断层扫描（OCT）中的应用
Abstract
Age-related macular degeneration (AMD) and diabetic macular edema (DME) are leading causes of vision loss, and optical coherence tomography (OCT) is the standard modality for detecting and monitoring the subtle lesions that drive treatment decisions. Most deep-learning segmentation work for OCT is validated only in-domain, leaving generalization to clinical data collected under different acquisition protocols largely untested. This work develops and systematically ablates four lesion-segmentation pipelines -- 2D and 3D variants for AMD and DME -- reaching Dice scores of 0.76 to 0.82 with strong volumetric and surface calibration (r vol, r surf greater than or equal to 0.97 across all four pipelines) on an in-domain validation set. The ablation process establishes a full-volume, calibration-aware adoption standard that catches mechanisms an ordinary slice-level evaluation would keep, and identifies ensemble composition as the most consistent driver of improvement. To test generalization, the models are evaluated on OLIVES, an external clinical cohort with no lesion-level ground truth, using a proxy-metric framework built around biomarker AUROC, central subfield thickness (CST) correlation, and longitudinal concordance. Predictions track clinical biomarkers outside the training distribution, though less strongly than in-domain -- evidence for, not validation of, automated lesion-burden tracking as a clinical tool.
Chinese Translation
年龄相关性黄斑变性（AMD）和糖尿病性黄斑水肿（DME）是导致视力丧失的主要原因，而光学相干断层扫描（OCT）是检测和监测驱动治疗决策的细微病变的标准方法。大多数针对OCT的深度学习分割工作仅在领域内进行了验证，导致在不同采集协议下收集的临床数据的泛化能力基本未得到测试。本研究开发并系统性地消融了四个病变分割管道——AMD和DME的2D和3D变体——在领域内验证集上达到了0.76到0.82的Dice分数，并在体积和表面校准方面表现出色（所有四个管道的体积相关性r vol和表面相关性r surf均大于或等于0.97）。消融过程建立了一个全体积、校准感知的采用标准，捕捉了普通切片级评估所忽略的机制，并确定集成组成是最一致的改进驱动因素。为了测试泛化能力，这些模型在OLIVES上进行了评估，这是一个没有病变级别真实值的外部临床队列，使用围绕生物标志物AUROC、中央子区厚度（CST）相关性和纵向一致性构建的代理指标框架。尽管预测在训练分布之外追踪临床生物标志物，但强度不及领域内的表现——这为自动化病变负担跟踪作为临床工具提供了证据，而非验证。
cs.CV / 70 / 2608.27123
EditaLive! Unified Character Video Editing for Live Streaming
EditaLive！统一角色视频编辑用于直播
Abstract
Conventional video editing primarily focuses on scene-level content, whereas live streaming places greater emphasis on the human subject. However, directly applying existing video-editing methods to human-centric live streaming remains challenging, as they may introduce facial-expression inconsistencies and typically depend on multiple offline inference steps, making them unsuitable for real-time interaction. We propose EditaLive, a novel framework for real-time streaming character video editing. In detail, we start from a pretrained image animation model (Wan-Animate), which naturally decouples appearance from motion, and repurpose it as the base model for instruction-based human-centric video editing by reference frame editing and video reconstruction via the collected CharEdit-50K dataset. Besides, we adapt the model from offline bidirectional to causal streaming generation, and design an aligned self-rollout distillation strategy that compresses the model into a two-step sampler, where fixed RoPE and align forcing reduce training--inference discrepancies, and first-frame preserved sparse attention filters redundant historical information to mitigate appearance drift. Extensive experiments demonstrate that EditaLive delivers state-of-the-art editing performance with faithful preservation of facial expressions and low-latency real-time streaming inference.
Chinese Translation
传统的视频编辑主要关注场景级内容，而直播则更强调人类主体。然而，直接将现有的视频编辑方法应用于以人为中心的直播仍然具有挑战性，因为这些方法可能会引入面部表情不一致，并且通常依赖多个离线推理步骤，使其不适合实时交互。我们提出了EditaLive，一个用于实时流媒体角色视频编辑的新框架。具体而言，我们从一个预训练的图像动画模型（Wan-Animate）开始，该模型自然地将外观与运动解耦，并将其重新用于基于指令的人类中心视频编辑，通过参考帧编辑和利用收集的CharEdit-50K数据集进行视频重建。此外，我们将模型从离线双向适配为因果流生成，并设计了一种对齐自回归蒸馏策略，将模型压缩为一个两步采样器，其中固定的RoPE和对齐强制减少了训练-推理之间的差异，而首帧保留的稀疏注意力过滤器则减少冗余的历史信息，以缓解外观漂移。大量实验表明，EditaLive在忠实保留面部表情和低延迟实时流推理方面提供了最先进的编辑性能。
cs.CV / 71 / 2608.27154
ReViCo: Unveiling the Limitations of VLMs in Visual Text Understanding via Error Correction
ReViCo：通过错误纠正揭示视觉语言模型（VLMs）在视觉文本理解中的局限性
Abstract
Vision Language Models (VLMs) have shown great success in general visual tasks, yet they still struggle to deeply understand text within images. In this paper, we introduce ReViCo (Real Visual Correction), a benchmark designed to evaluate VLM text understanding through a novel task of visual text error correction. ReViCo challenges models to identify and fix text errors in real-world images, which requires a profound understanding of the interplay between visual text and its surrounding visual context. We benchmark various VLMs using two distinct paradigms: prompt-based strategy and targeted model training, both aimed at pushing the limits of current models. Our experiments reveal a striking performance gap between even the best VLMs and human, and further analysis also shows that most models struggle to accurately perceive the visual text, resulting in frequent correction errors. By highlighting these gaps, ReViCo provides a new benchmark foundation for developing more robust and text-aware VLMs.
Chinese Translation
视觉语言模型（VLMs）在通用视觉任务中表现出色，但在深入理解图像中的文本方面仍存在困难。本文提出了ReViCo（Real Visual Correction），一个通过视觉文本错误纠正这一新颖任务来评估VLM文本理解能力的基准测试。ReViCo挑战模型识别并修正真实世界图像中的文本错误，这要求模型对视觉文本及其周围视觉环境之间的相互作用有深刻理解。我们采用两种不同范式对多种VLM进行基准测试：基于提示（prompt-based）策略和针对性模型训练，旨在推动当前模型的极限。实验结果显示，即使是表现最好的VLM与人类之间也存在显著性能差距，进一步分析表明大多数模型难以准确感知视觉文本，导致纠正错误频发。通过揭示这些差距，ReViCo为开发更鲁棒且具备文本感知能力的VLM提供了新的基准基础。
cs.CV / 72 / 2608.27168
Magpie: Real-Time World Renderer for Interactive Games
Magpie：用于互动游戏的实时世界渲染器
Abstract
Modern game development relies heavily on conventional graphics pipelines. High-quality visual content requires modeling, material authoring, animation, lighting, effects, and runtime optimization, making asset production expensive and extending the development cycle of game prototypes. Recently, video foundation models are beginning to change film and video production, but games differ from linear media, they require not only continuous and realistic imagery, but also stable and reproducible gameplay rules, object states, and interaction outcomes. We present Magpie, a real-time generative world-rendering system for interactive games. Magpie separates gameplay execution from visual generation. Designers define scenes and rules in a game engine. At runtime, the Game Engine resolves player actions and maintains world state, while an independent Render Server generates visual output from white-box frames produced by the engine. Magpie provides a system-level implementation path for applying generative models to real-time game rendering. It preserves gameplay designability and reproducibility, and reduces the dependence of early game prototypes on complete visual assets.
Chinese Translation
现代游戏开发在很大程度上依赖于传统的图形管线。高质量的视觉内容需要建模、材质创作、动画、照明、特效和运行时优化，这使得资产生产成本高昂，并延长了游戏原型的开发周期。最近，视频基础模型开始改变电影和视频制作，但游戏与线性媒体不同，它们不仅需要连续和真实的图像，还需要稳定和可重复的游戏规则、对象状态和交互结果。我们提出了Magpie，一个用于互动游戏的实时生成世界渲染系统。Magpie将游戏执行与视觉生成分离。设计师在游戏引擎中定义场景和规则。在运行时，游戏引擎解析玩家的动作并维护世界状态，而独立的渲染服务器则从引擎生成的白盒帧中生成视觉输出。Magpie为将生成模型应用于实时游戏渲染提供了一种系统级的实现路径。它保留了游戏设计的可设计性和可重复性，并减少了早期游戏原型对完整视觉资产的依赖。
cs.CV / 73 / 2608.27169
Ancient-Bench: A Comprehensive Multi-millennial, Multi-medium, and Multi-script Benchmark for Ancient Chinese Artifact Text Recognition
Ancient-Bench：一个全面的多千年、多媒介和多书写体的古代中国文物文本识别基准
Abstract
Ancient Chinese artifact text recognition is fundamental to heritage digitization, and benchmarks for ancient texts are essential for evaluating current model capabilities. However, existing benchmarks suffer from ''fragmentation'', manifested in limited temporal coverage, limited medium diversity, and incomplete script types. Therefore, we present Ancient-Bench, a comprehensive benchmark of 2,700 images for ancient Chinese artifact text recognition, featuring three dimensions: Multi-millennial (spanning 3,000 years of character evolution), Multi-medium (covering nine artifact categories), and Multi-script (encompassing seven historical script forms). To enable consistent and fair evaluation across heterogeneous media, we further define three annotation standards tailored to the medium-specific characteristics of ancient texts: symbol standardization, character standardization, and parsing standardization. Extensive experiments on Ancient-Bench covering general Vision-Language Models (VLMs) and OCR-specialist models reveal that ancient Chinese artifact text recognition remains fundamentally unsolved, with persistent challenges in variant characters, specialized symbols, and hallucination. The dataset is available at https://github.com/SCUT-DLVCLab/Ancient_Bench.
Chinese Translation
古代中国文物文本识别是遗产数字化的基础，而古代文本的基准对于评估当前模型的能力至关重要。然而，现有的基准存在“碎片化”问题，表现为时间覆盖范围有限、媒介多样性不足以及书写体类型不完整。因此，我们提出了Ancient-Bench，这是一个包含2,700幅图像的全面基准，用于古代中国文物文本识别，涵盖三个维度：多千年（跨越3,000年的字符演变）、多媒介（涵盖九类文物）和多书写体（包括七种历史书写形式）。为了在异质媒介中实现一致和公平的评估，我们进一步定义了三种注释标准，以适应古代文本的媒介特征：符号标准化、字符标准化和解析标准化。针对Ancient-Bench的广泛实验涵盖了通用视觉语言模型（VLMs）和OCR专用模型，结果表明古代中国文物文本识别仍然是一个根本未解决的问题，面临变体字符、专业符号和幻觉等持续挑战。数据集可在 https://github.com/SCUT-DLVCLab/Ancient_Bench 获取。
cs.CV / 74 / 2608.27175
Temporal Sensitivity Analysis of Tessera Embeddings
Tessera 嵌入的时间敏感性分析
Abstract
Many Earth Observation applications need land-use/land-cover maps that are both precise and frequently updated, yet the strongest Earth Observation foundation models build their embeddings from a full year of observations. We present a controlled study of the temporal sensitivity of Tessera, one of these leading foundation models, for land-use/land-cover mapping. Keeping the encoder frozen, we recompute its embeddings over varying observation windows, from a full year down to a single day. We use them as inputs to a linear probe and a UNet segmentation head, benchmarking both of them against from-scratch networks on LUCAS, DynamicEarthNet, and PASTIS-R datasets. We show that the value of the embeddings is task-dependent. Where classes are separated by phenology, as for the crop types of PASTIS-R, they reach a mean Intersection-over-Union of $58.3$, about $46\%$ above the best from-scratch model. Where classes are temporally stable (e.g., forests in DynamicEarthNet and LUCAS), embedding-based and from-scratch models match only under full supervision. On both datasets, Tessera embeddings remain markedly more label-efficient. Degradation under shorter temporal windows is gradual and class-dependent. Contracting the window from one year to one month costs $39\%$ of the segmentation accuracy on PASTIS-R but only $5\%$ on DynamicEarthNet. Single-day embeddings still classify land cover in LUCAS at $3.4$ times the chance level. Our study shows that temporal coverage is therefore a tunable cost rather than a fixed prerequisite, opening regimes such as near-real-time mapping and faster land-use/land-cover refresh cycles.
Chinese Translation
许多地球观测应用需要既精确又频繁更新的土地利用/土地覆盖地图，而最强大的地球观测基础模型则是基于整整一年的观测数据构建其嵌入。我们对 Tessera 这一领先基础模型在土地利用/土地覆盖制图中的时间敏感性进行了受控研究。在保持编码器不变的情况下，我们重新计算了其在不同观测窗口下的嵌入，从整整一年到单一天。我们将这些嵌入作为线性探针和 UNet 分割头的输入，并在 LUCAS、DynamicEarthNet 和 PASTIS-R 数据集上将它们与从零开始构建的网络进行了基准测试。我们表明，嵌入的价值是与任务相关的。当类别通过物候分开时，例如 PASTIS-R 的作物类型，它们的平均交并比达到 $58.3$，比最佳的从零开始模型高出约 $46 ext{ extperthousand}$。当类别在时间上稳定（例如，DynamicEarthNet 和 LUCAS 中的森林）时，嵌入模型和从零开始模型仅在完全监督下匹配。在这两个数据集中，Tessera 嵌入仍然显著更具标签效率。在较短的时间窗口下，性能下降是逐渐的且与类别相关。从一年缩短到一个月的窗口，PATSIS-R 的分割准确率损失为 $39 ext{ extperthousand}$，而在 DynamicEarthNet 上仅为 $5 ext{ extperthousand}$。单日嵌入在 LUCAS 中仍能以 $3.4$ 倍的机会水平分类土地覆盖。因此，我们的研究表明，时间覆盖是一个可调的成本，而不是固定的前提条件，这为近实时制图和更快的土地利用/土地覆盖更新周期打开了新的可能性。
cs.CV / 75 / 2608.27178
Parameter-Efficient pretrained-CT-to-MRI Transfer for Rectal Cancer Segmentation: Performance-Calibration Trade-offs
参数高效的预训练CT到MRI转移用于直肠癌分割：性能与校准的权衡
Abstract
Accurate rectal cancer segmentation from magnetic resonance imaging (MRI) is essential for adaptive radiotherapy and tumor response assessment, but deployment also requires computational efficiency and informative, calibrated uncertainty estimates. We therefore introduce SWIFT, a SWin pretrained model wIth parameter-eFficient and Tumor-aware fine-tuning for rectal cancer segmentation. A Swin V2 encoder pretrained on 10,444 public 3D CT volumes using a DINOv2-style objective was adapted to T2-weighted MRI through four cumulative configurations: full fine-tuning (SWIFT), decoder compression (SWIFTe), low-rank adaptation (SWIFTe-LoRA), and a four-member LoRA-decoder ensemble (SWIFTe-LDE4). Geometric accuracy, tumor detection, radiomic agreement, and probability calibration were evaluated on a held-out 247-case test set from a single-institution cohort acquired using 1.5 or 3 Tesla GE scanners. Compared with SWIFT, SWIFTe reduced total parameters by 70.1% (from 72.8M to 21.8M) and increased tumor detection rate from 89.9% to 93.9%, while achieving a slightly lower median surface DSC (0.61 versus 0.62) and improved radiomic agreement. In a separate SWIFTe ablation, removing tumor-aware augmentation reduced detection from 93.9% to 89.9% but increased surface DSC from 0.61 to 0.64, demonstrating a detection-boundary-agreement trade-off. SWIFTe-LoRA used 14.6% of SWIFTe's trainable parameters while retaining similar segmentation performance. SWIFTe-LDE4 achieved the lowest calibration errors among the four configurations after temperature scaling (expected calibration error, 0.217; Brier score, 0.222), although the absolute expected calibration error indicates residual miscalibration. Similar efficiency-calibration patterns were observed using the public VoCo checkpoint, supporting robustness across pretrained initializations rather than external clinical generalizability.
Chinese Translation
从磁共振成像（MRI）中准确分割直肠癌对于自适应放疗和肿瘤反应评估至关重要，但其部署还需要计算效率和信息丰富的校准不确定性估计。因此，我们引入了SWIFT，一种具有参数高效性和肿瘤感知微调的SWin预训练模型，用于直肠癌分割。通过四种累积配置，将在10,444个公共3D CT体积上使用DINOv2风格目标预训练的Swin V2编码器适配到T2加权MRI：全微调（SWIFT）、解码器压缩（SWIFTe）、低秩适配（SWIFTe-LoRA）以及四成员LoRA解码器集成（SWIFTe-LDE4）。在使用1.5或3特斯拉GE扫描仪获取的单一机构队列的247例测试集上评估了几何准确性、肿瘤检测、放射组学一致性和概率校准。与SWIFT相比，SWIFTe将总参数减少了70.1%（从72.8M降至21.8M），并将肿瘤检测率从89.9%提高到93.9%，同时实现了略低的中位表面DSC（0.61对比0.62）和改善的放射组学一致性。在单独的SWIFTe消融实验中，去除肿瘤感知增强将检测率从93.9%降低至89.9%，但将表面DSC从0.61提高至0.64，展示了检测-边界-一致性的权衡。SWIFTe-LoRA使用了SWIFTe可训练参数的14.6%，同时保持了类似的分割性能。SWIFTe-LDE4在经过温度缩放后，在四种配置中实现了最低的校准误差（期望校准误差为0.217；Brier分数为0.222），尽管绝对期望校准误差表明仍存在残余的错误校准。使用公共VoCo检查点观察到类似的效率-校准模式，支持在预训练初始化之间的鲁棒性，而非外部临床的普适性。
cs.CV / 76 / 2608.27181
SSMB: Self-Supervised Local Feature Detection under Motion Blur
SSMB：运动模糊下的自监督局部特征检测
Abstract
Keypoint detection under motion blur remains a significant challenge, as blur distorts local image structure and degrades the repeatability of feature localization. Existing approaches either rely on computationally expensive deblur-then-detect pipelines that may introduce restoration artifacts, or learn to regress the image positions of handcrafted keypoints extracted on sharp images, which reflects the assumptions of the handcrafted detector rather than what is truly repeatable under blur. We present SSMB, a deblur-free, self-supervised keypoint detector for motion-blurred images that requires neither handcrafted detectors nor external pseudo-labels. SSMB introduces the Local Discriminability Enhancement (LDE) module, which restores fine-grained local discriminability after global feature mixing. Training is performed in two stages. First, geometric pretraining on synthetic shapes bootstraps spatially discriminative keypoint detection without any external detector, just from the rendered geometry. Second, blur-aware training on real sharp-blur image pairs learns blur-invariant detection through a multi-component self-supervised objective that enforces cross-domain consistency, geometric alignment, and spatial coverage. Extensive evaluations on keypoint detection, image matching, relative pose estimation, and visual localization under motion blur demonstrate that SSMB establishes a new state-of-the-art among sparse keypoint detectors, consistently outperforming both supervised and self-supervised baselines across all tasks. Code, models, and datasets will be publicly available upon paper acceptance.
Chinese Translation
在运动模糊下的关键点检测仍然是一个重大挑战，因为模糊会扭曲局部图像结构并降低特征定位的重复性。现有的方法要么依赖于计算开销大的去模糊-再检测流程，这可能引入恢复伪影，要么学习回归在清晰图像上提取的手工关键点的图像位置，这反映了手工检测器的假设，而非在模糊下真正可重复的内容。我们提出了SSMB，一种无需去模糊、无监督的运动模糊图像关键点检测器，不需要手工检测器或外部伪标签。SSMB引入了局部可区分性增强（Local Discriminability Enhancement, LDE）模块，在全局特征混合后恢复细粒度的局部可区分性。训练分为两个阶段。首先，在合成形状上进行几何预训练，从渲染的几何体中引导空间可区分的关键点检测，而不依赖任何外部检测器。其次，在真实的清晰-模糊图像对上进行模糊感知训练，通过多组件自监督目标学习模糊不变检测，该目标强制执行跨域一致性、几何对齐和空间覆盖。在运动模糊下的关键点检测、图像匹配、相对姿态估计和视觉定位的广泛评估表明，SSMB在稀疏关键点检测器中建立了新的最先进水平，在所有任务中始终优于监督和自监督基线。代码、模型和数据集将在论文接受后公开。
cs.CV / 77 / 2608.27190
Unsupervised Adaptation of 3D CT Foundation Models for 3D CBCT Segmentation
无监督适应3D CT基础模型用于3D CBCT分割
Abstract
Accurate 3D segmentation of cone-beam CT (CBCT) is critical for interventional and radiation therapy applications, yet it remains limited by two compounding challenges: the scarcity of annotated CBCT data and the large domain shift from diagnostic CT. Interventional CBCT exhibits fundamental modality differences from conventional CT, driven by acquisition and physics effects as well as contrast-specific vascular content, thereby limiting effective cross-modality model transfer. We propose a novel unsupervised domain adaptation (UDA) framework based on redundancy-reducing feature alignment, enabling 3D CBCT segmentation with no target-domain annotations or inference-time adaptation. Our framework is architecture-agnostic, seamlessly adapting both CNN-based and ViT-based foundation models. We evaluate our method on two challenging CT-CBCT liver segmentation benchmarks: one for interventional vascular procedures and one for radiation therapy, demonstrating that even large-scale pretrained segmentation networks require explicit feature-space bridging to generalize across acquisition modalities, and that our approach consistently outperforms existing pretrained foundation model and UDA strategies. To support reproducibility and benchmarking, we release the liver segmentations for a public CBCT dataset, along with the code, trained models, and weights.
Chinese Translation
准确的锥束CT（CBCT）三维分割对介入和放射治疗应用至关重要，但仍面临两个相互影响的挑战：标注CBCT数据的稀缺性和与诊断CT之间的大领域转移。介入CBCT与常规CT在基本模态上存在显著差异，这些差异源于采集和物理效应以及特定对比度的血管内容，从而限制了有效的跨模态模型迁移。我们提出了一种基于冗余减少特征对齐的新型无监督领域适应（UDA）框架，使得在没有目标领域标注或推理时适应的情况下实现3D CBCT分割。我们的框架与架构无关，能够无缝适应基于卷积神经网络（CNN）和基于视觉变换器（ViT）的基础模型。我们在两个具有挑战性的CT-CBCT肝脏分割基准上评估了我们的方法：一个用于介入血管手术，另一个用于放射治疗，结果表明，即使是大规模预训练的分割网络也需要明确的特征空间桥接才能在采集模态之间进行泛化，并且我们的方法始终优于现有的预训练基础模型和UDA策略。为了支持可重复性和基准测试，我们发布了公共CBCT数据集的肝脏分割结果，以及代码、训练模型和权重。
cs.CV / 78 / 2608.27199
Vision-centric generative AI models: A software-hardware perspective
以视觉为中心的生成式人工智能模型：软件-硬件视角
Abstract
Vision generative artificial intelligence (AI) has emerged as one of the most rapidly advancing areas of deep learning. The explosion of multimodal models has made them widely associated with text-to-image applications running on large datacentres. However, vision generative models are equally needed in applications that operate under strict hardware constraints at the edge, including autonomous vehicles, agricultural sensors, and mobile devices. In this Perspective, we argue that progress in vision generative AI has been driven by output quality, with hardware evolving reactively to accommodate growing model demands. We quantify the parameter cost and energy efficiency of these models across a range of accelerator platforms, and map four generative model families against seven real-world application domains. Finally, we advocate a software-hardware co-design approach, where deployment constraints are considered from the start of the design process, ensuring that the "right model" runs on the "right hardware" to serve the "right application", making generative AI deployment sustainable and accessible across a much broader range of platforms.
Chinese Translation
视觉生成式人工智能（AI）已成为深度学习中发展最快的领域之一。多模态模型的爆炸性增长使其广泛应用于在大型数据中心运行的文本到图像应用。然而，视觉生成模型同样在边缘计算中需要应用于严格的硬件约束下，包括自动驾驶汽车、农业传感器和移动设备。在本观点中，我们认为视觉生成AI的进展是由输出质量驱动的，硬件则是被动地演变以适应不断增长的模型需求。我们量化了这些模型在多种加速器平台上的参数成本和能效，并将四个生成模型家族与七个现实世界应用领域进行了映射。最后，我们倡导一种软件-硬件协同设计的方法，从设计过程的开始就考虑部署约束，确保“合适的模型”在“合适的硬件”上运行，以服务于“合适的应用”，使生成式AI的部署在更广泛的平台上可持续和可及。
cs.CV / 79 / 2608.27206
PACE: A Unified Condense-and-Extract Paradigm for Fast VLM Inference
PACE：一种统一的压缩与提取范式用于快速视觉语言模型推理
Abstract
Vision-Language Models (VLMs) demonstrate exceptional visual reasoning capabilities, yet their inference costs escalate rapidly with the proliferation of visual tokens. Existing visual token pruning methods exhibit two fundamental limitations. First, most approaches operate exclusively post-vision encoder, leaving the substantial latency of the visual encoding phase unoptimized. Second, under strict token budgets, these methods often fail to jointly preserve holistic visual contexts and fine-grained details, leading to performance degradation. To address these bottlenecks, we propose PACE (Pixel-Adaptive Condense and Extract), a training-free inference framework that accelerates both the vision encoder and the Large Language Model (LLM) via a unified Condense-and-Extract paradigm. During the Condense stage, an Adaptive Pixel Compressor (APC) evaluates visual information density prior to encoding, adaptively downsampling redundant inputs, curtailing encoder computation while preserving global context and essential visual cues. In the Extract stage, a Dynamic Dual-Attention Extractor (DDAE) selectively retains visual tokens via a fusion of internal visual signals from the encoder and semantic signals from the LLM, safeguarding task-critical details. By integrating PACE into Qwen2.5-VL-7B, the model retains 93.8% of its original performance while utilizing only 10% of the visual tokens, yielding a 3.1x speedup in time to first token (TTFT). Our code is available at https://github.com/jjL357/PACE.
Chinese Translation
视觉语言模型（VLM）展现出卓越的视觉推理能力，但随着视觉标记数量的激增，其推理成本迅速上升。现有的视觉标记剪枝方法存在两个基本局限性。首先，大多数方法仅在视觉编码器之后操作，未能优化视觉编码阶段的显著延迟。其次，在严格的标记预算下，这些方法往往无法同时保留整体视觉上下文和细粒度细节，导致性能下降。为了解决这些瓶颈，我们提出了PACE（像素自适应压缩与提取），这是一种无训练的推理框架，通过统一的压缩与提取范式加速视觉编码器和大型语言模型（LLM）。在压缩阶段，自适应像素压缩器（APC）在编码之前评估视觉信息密度，自适应地对冗余输入进行下采样，减少编码器计算，同时保留全局上下文和重要视觉线索。在提取阶段，动态双重注意力提取器（DDAE）通过融合来自编码器的内部视觉信号和来自LLM的语义信号，选择性地保留视觉标记，保护任务关键细节。通过将PACE集成到Qwen2.5-VL-7B中，该模型在仅使用10%的视觉标记的情况下，保留了93.8%的原始性能，实现了3.1倍的首次标记时间（TTFT）加速。我们的代码可在https://github.com/jjL357/PACE获取。
cs.CV / 80 / 2608.27214
CODE: Cross-Modal Calibration and Dynamic Suppression for Open World Object Detection
CODE：用于开放世界目标检测的跨模态校准与动态抑制
Abstract
Open World Object Detection (OWOD) built on multimodal foundation models often suffers from semantic ambiguity caused by unidirectional text-to-vision matching, while rigid outlier penalties may over-suppress unknown objects near known-class decision boundaries. We propose CODE (Cross-Modal Calibration and Dynamic Suppression), a unified inference-time framework with three complementary components. Cross-Modal Joint Confidence Calibration injects global visual prototypes to calibrate text-driven known-class predictions. Uncertainty-Guided Universal Objectness Enhancement measures classification hesitation from local visual responses to strengthen potential unknown objects. Dynamic Outlier Suppression via Confidence Margin replaces rigid suppression with a margin-aware adjustment that preserves ambiguous out-of-distribution instances. Experiments on the Real-World Detection benchmark demonstrate that, with the OWL-ViT L/14 backbone, CODE achieves 21.7 U-mAP and 40.8 K-mAP in Task 1, surpassing the previous state of the art by 2.6 and 2.3 points, respectively.
Chinese Translation
基于多模态基础模型的开放世界目标检测（OWOD）常因单向的文本到视觉匹配导致语义歧义，同时刚性的异常值惩罚可能对接近已知类别决策边界的未知对象产生过度抑制。我们提出了CODE（Cross-Modal Calibration and Dynamic Suppression），一个包含三大互补组件的统一推理时框架。跨模态联合置信度校准通过注入全局视觉原型来校准文本驱动的已知类别预测。不确定性引导的通用目标性增强通过衡量局部视觉响应中的分类犹豫，强化潜在的未知对象。基于置信度边际的动态异常值抑制用边际感知调整替代刚性抑制，从而保留模糊的分布外实例。在Real-World Detection基准测试中，采用OWL-ViT L/14骨干网络，CODE在任务1中分别实现了21.7的U-mAP和40.8的K-mAP，分别超越了之前的最先进水平2.6和2.3个百分点。
cs.CV / 81 / 2608.27226
DINOcular: Self-Supervised Visuospatial Representations
DINOcular: 自监督视觉空间表示
Abstract
We introduce a self-supervised framework for learning joint visuospatial representations from RGB-D observations. While modern vision foundation models are trained almost exclusively on RGB images, many embodied systems have access to explicit depth sensing, which provides geometric information that monocular inputs cannot recover. Our method integrates depth-derived geometric priors with a visual backbone through inter-patch and intra-patch fusion, enabling the model to encode both appearance and spatial structure efficiently. The resulting representation shows promising improvements on 3D awareness while preserving semantic transfer: it outperforms prior methods of comparable scale on multiple 3D geometry benchmarks, and remains competitive when probed for standard RGB-D semantic segmentation tasks.
Chinese Translation
我们提出了一种自监督框架，用于从RGB-D观测中学习联合视觉空间表示。尽管现代视觉基础模型几乎完全基于RGB图像进行训练，但许多具身系统可以访问显式深度传感，这提供了单目输入无法恢复的几何信息。我们的方法通过跨补丁和内补丁融合，将深度衍生的几何先验与视觉主干网络相结合，使模型能够高效地编码外观和空间结构。所得到的表示在3D感知方面显示出有希望的改进，同时保持语义传递：在多个3D几何基准测试中，它超越了同类规模的先前方法，并且在标准RGB-D语义分割任务中仍然具有竞争力。
cs.CV / 82 / 2608.27240
UniFLM: United Segmentation and Measurement on Fetal Limb Ultrasonic Image
UniFLM：胎儿肢体超声图像的联合分割与测量
Abstract
Prenatal ultrasound examination is crucial for assessing fetal limb development and detecting congenital anomalies. However, existing artificial intelligence models often overlook fetal lethal skeletal dysplasias due to the lack of high-quality annotated data and a unified framework for multiple long bones. Moreover, generic segmentation models struggle with the inherent noise and semantic gaps in ultrasound images. To address these challenges, we construct the Fetal Limb Bones (FLB) dataset, comprising high-quality annotations for the humerus, femur, tibia-fibula, and radius-ulna. Furthermore, we propose UniFLM, a unified framework for automatic cross-plane segmentation and measurement. UniFLM incorporates a Semantic-Aware Skip Connection module to bridge the semantic gap between encoder and decoder features, and a Positive Sampling strategy to adaptively filter noise and extract essential semantic information. Finally, a Point Regression Mapping module is introduced to learn clinician annotation patterns for precise bone length measurement. Extensive experiments conducted on the FLB dataset demonstrate that the proposed UniFLM achieves superior accuracy and enhanced generalization capabilities in fetal long bone assessment compared to current state-of-the-art models.
Chinese Translation
产前超声检查对于评估胎儿肢体发育和检测先天性异常至关重要。然而，现有的人工智能模型常常忽视胎儿致命性骨骼发育不良，原因在于缺乏高质量的标注数据和针对多根长骨的统一框架。此外，通用分割模型在处理超声图像中固有的噪声和语义差距时也面临困难。为了解决这些挑战，我们构建了胎儿肢体骨（Fetal Limb Bones, FLB）数据集，该数据集包含对肱骨、股骨、胫腓骨和桡尺骨的高质量标注。此外，我们提出了UniFLM，一个用于自动跨平面分割和测量的统一框架。UniFLM结合了语义感知跳跃连接模块，以弥合编码器和解码器特征之间的语义差距，并采用积极采样策略以自适应过滤噪声并提取重要的语义信息。最后，引入了点回归映射模块，以学习临床医生的标注模式，从而实现精确的骨长测量。在FLB数据集上进行的大量实验表明，所提出的UniFLM在胎儿长骨评估中相较于当前最先进的模型具有更高的准确性和增强的泛化能力。
cs.CV / 83 / 2608.27280
Sidecar: Training-Free Semantic Reuse for Character-Consistent Free-form Visual Storytelling
Sidecar：无训练的语义重用用于角色一致的自由形式视觉叙事
Abstract
Visual storytelling requires generating images that follow a narrative while preserving consistent character identities across frames. In free-form story generation, a character is fully described only when first introduced and is later referred to by a type-level mention or pronoun. Although this setting better reflects natural storytelling, later prompts may omit important identity-related semantics, making character consistency more difficult to maintain. We propose \textbf{Sidecar}, a plug-and-play semantic augmentation module that preserves entity-level information from the initial description and injects the missing semantics into later prompt embeddings. Sidecar requires no additional training and does not modify the architecture of the base diffusion model. Experiments on FreeStoryBench show that Sidecar consistently improves prompt-image alignment and character consistency across multiple SDXL- and FLUX-based baselines, with negligible computational overhead.
Chinese Translation
视觉叙事需要生成遵循叙事的图像，同时在各帧之间保持一致的角色身份。在自由形式故事生成中，角色在首次介绍时被完整描述，之后通过类型级提及或代词进行引用。尽管这种设置更好地反映了自然叙事，但后续提示可能会省略重要的身份相关语义，从而使角色一致性更难维持。我们提出了 extbf{Sidecar}，一个即插即用的语义增强模块，它保留了初始描述中的实体级信息，并将缺失的语义注入到后续提示嵌入中。Sidecar不需要额外的训练，也不修改基础扩散模型的架构。在FreeStoryBench上的实验表明，Sidecar在多个基于SDXL和FLUX的基准测试中持续改善了提示与图像的对齐和角色一致性，且计算开销微乎其微。
cs.CV / 84 / 2608.27282
TADP: Task-Aware Deformable Prediction for Single-Stage 3D Object Detection
TADP：任务感知可变形预测用于单阶段3D目标检测
Abstract
Most single-stage 3D object detectors complete different tasks with the same extracted features. Nevertheless, it is impossible to project features into a common space that is adaptive for all the tasks. We present a novel task-aware deformable prediction (TADP) method for single-stage 3D object detection to solve this problem. Firstly, a triple feature refinement aggregation module is designed to extract three-level features adaptively. Additionally, we design the multi-scale feature aggregation block to fuse multi-scale features in a scale-aware manner. Finally, the prediction of each task is deformed with the designed plug-and-play task-aware deformation head. It can percept the emphasis and interaction of each task. We also designed three different deformation modules. The experimental results demonstrate that the proposed deformation head shows good results on other detection methods. The experimental results on the KITTI dataset demonstrate that the car mAP is 80.91%, surpassing many state-of-the-art methods on the KITTI benchmark.
Chinese Translation
大多数单阶段3D目标检测器使用相同的提取特征完成不同的任务。然而，将特征投影到一个适应所有任务的公共空间是不可能的。为了解决这个问题，我们提出了一种新颖的任务感知可变形预测（TADP）方法用于单阶段3D目标检测。首先，设计了一个三重特征精炼聚合模块，以自适应地提取三级特征。此外，我们设计了多尺度特征聚合块，以尺度感知的方式融合多尺度特征。最后，通过设计的即插即用任务感知变形头对每个任务的预测进行变形，使其能够感知每个任务的重点和相互作用。我们还设计了三种不同的变形模块。实验结果表明，所提出的变形头在其他检测方法上表现良好。在KITTI数据集上的实验结果显示，汽车的mAP达到80.91%，超越了KITTI基准上的许多最先进的方法。
cs.CV / 85 / 2608.27290
Detection of Christmas tree plantations from high-resolution aerial imagery. A case study in the French Morvan
从高分辨率航空影像中检测圣诞树种植园：法国莫尔万地区的案例研究
Abstract
Christmas tree plantations are economically relevant, yet a largely unexplored application domain in Remote Sensing (RS). Their delineation is challenging because of high planting density, short rotation cycles, visual confusion with surrounding vegetation, the availability of dense labels for one reference year only, and severe class imbalance at the landscape scale. Although Deep Learning (DL) methods have shown strong potential for vegetation mapping, existing approaches are typically designed for forests, generic plantation systems, or orchards, and do not explicitly address the structural specificity and hard-negative confusion that characterize Christmas tree plantations. In response to these challenges, this work makes three main contributions: (i) it frames Christmas tree plantation mapping as a distinct rare-target semantic segmentation problem; (ii) it introduces a Hard Negative Mining (HNM) strategy to improve discrimination against confusing background patterns; and (iii) it evaluates the proposed framework across complementary levels, including supervised testing, temporal transfer, and large-scale validation. On the 2020 test set held out, the best model, DeepLabV3 with a ResNet-34 encoder, achieves an IoU of 0.733 and an F1-score of 0.846. HNM substantially improves precision-recall behavior, increasing the area under the precision-recall curve from 0.204 to 0.913. Temporal inference further shows meaningful transferability, reaching IoU/F1 values of 0.751/0.858 on 2017/2018 and 0.691/0.817 on 2023. Large-scale validation further highlights the intrinsic difficulty of the task, as Christmas tree plantations occupied only a very small fraction of the extent of the common evaluation, corresponding to 1,498.4 ha (1.72\%) in 2017/2018 and 1,782.2 ha (2.04\%) in 2023 out of 87,309.4 ha in total.
Chinese Translation
圣诞树种植园在经济上具有重要意义，但在遥感（Remote Sensing, RS）领域仍然是一个未被充分探索的应用领域。由于种植密度高、轮作周期短、与周围植被的视觉混淆、仅有一年的密集标签可用以及在景观尺度上的严重类别不平衡，其划分面临挑战。尽管深度学习（Deep Learning, DL）方法在植被制图中显示出强大的潜力，但现有方法通常是为森林、通用种植系统或果园设计的，并未明确解决圣诞树种植园所特有的结构特性和难负样本混淆。针对这些挑战，本研究做出了三项主要贡献：（i）将圣诞树种植园的制图框架设定为一个独特的稀有目标语义分割问题；（ii）引入了一种困难负样本挖掘（Hard Negative Mining, HNM）策略，以改善对混淆背景模式的区分；（iii）在多个互补层面上评估所提出的框架，包括监督测试、时间转移和大规模验证。在2020年保留的测试集中，最佳模型DeepLabV3（使用ResNet-34编码器）实现了0.733的交并比（IoU）和0.846的F1分数。HNM显著改善了精确度-召回率行为，将精确度-召回率曲线下的面积从0.204提高到0.913。时间推断进一步显示出显著的可转移性，在2017/2018年达到0.751/0.858的IoU/F1值，在2023年达到0.691/0.817。大规模验证进一步突显了该任务的内在难度，因为圣诞树种植园仅占常规评估范围的极小部分，2017/2018年对应1,498.4公顷（1.72%），2023年对应1,782.2公顷（2.04%），而总面积为87,309.4公顷。
cs.CV / 86 / 2608.27328
R2M-Bench: Evaluating Revisit Memory via Relative Consistency in Interactive Video World Models
R2M-Bench：通过相对一致性评估互动视频世界模型中的重访记忆
Abstract
High similarity between first-visit and return frames does not necessarily show that a video world model remembered the scene; the intervening rollout may simply have changed very little. This ambiguity makes absolute revisit scores sensitive to rendering stability, repetitive content, and failed motion. We introduce \emph{R2M-Bench} (\textbf{R}elative \textbf{R}evisit \textbf{M}emory Benchmark), a benchmark of observable revisit-selective consistency. For every detected return, R2M-Bench compares the revisit pair with two controls from the same rollout: a gap-matched non-revisit pair that measures generic temporal stability and a short-range pair that estimates short-horizon consistency. These comparisons produce \emph{MemoryGain} (MG), the revisit advantage over the temporal baseline, and the \emph{Normalized Memory Ratio} (NMR), which normalizes this advantage by the short-to-baseline dynamic range. R2M-Bench combines 100 reference scenes with three leave-and-return trajectories to form 300 instances and evaluates appearance fidelity, scene and object identity, local geometry, and persistent state. Across seven action-conditioned video world models, Overall NMR correlates with human consistency judgments at Spearman's $\rho=0.547$ (95\% CI $[0.45,0.63]$). Its within-model correlation magnitude with generated motion is $0.072$, compared with $0.207$ for raw revisit similarity, indicating that relative calibration substantially reduces the slow-motion shortcut. DreamX-World-Memo achieves the highest Overall NMR among the evaluated video models. Together, these results support same-rollout relative calibration as a practical way to distinguish revisit-specific consistency from generic temporal stability.
Chinese Translation
首次访问帧与返回帧之间的高相似性并不一定表明视频世界模型记住了场景；中间的展开过程可能几乎没有变化。这种模糊性使得绝对重访评分对渲染稳定性、重复内容和运动失败非常敏感。我们引入了 extit{R2M-Bench}（ extbf{R}elative extbf{R}evisit extbf{M}emory Benchmark），这是一个可观察的重访选择一致性的基准。对于每个检测到的返回，R2M-Bench将重访对与来自同一展开的两个对照进行比较：一个时间间隔匹配的非重访对，用于测量通用时间稳定性，以及一个短距离对，用于估计短期一致性。这些比较产生了 extit{MemoryGain}（MG），即相对于时间基线的重访优势，以及 extit{Normalized Memory Ratio}（NMR），它通过短期到基线的动态范围对这一优势进行归一化。R2M-Bench结合100个参考场景和三条离开-返回轨迹形成300个实例，并评估外观保真度、场景和物体身份、局部几何形状和持久状态。在七个动作条件的视频世界模型中，整体NMR与人类一致性判断的斯皮尔曼相关系数为$
ho=0.547$（95 ext{% CI }[0.45,0.63]）。其与生成运动的模型内相关性大小为$0.072$，而原始重访相似性的相关性为$0.207$，这表明相对校准显著减少了慢动作捷径。DreamX-World-Memo在评估的视频模型中实现了最高的整体NMR。这些结果共同支持同一展开的相对校准作为区分重访特定一致性与通用时间稳定性的实用方法。
cs.CV / 87 / 2608.27345
PAWBench: How Far Are We from Probabilistically Aligned World Modeling?
PAWBench：我们距离概率对齐的世界建模还有多远？
Abstract
Recent video generation models are increasingly framed as world models. Many physical processes can unfold in more than one valid way. Therefore, a world model should reproduce not only a plausible trajectory, but also the distribution of possible behaviors under the same initial observation and action. We call this distribution-level requirement probabilistic alignment. However, existing evaluations largely assess individual-video plausibility and do not test whether repeated generations recover the correct distribution. This raises a central question: how far are current video generators from probabilistically aligned world modeling? To answer it, we formalize probabilistic alignment as a distributional criterion for world models and introduce PAWBench, a benchmark for evaluating video generators as stochastic samplers of world dynamics. We further introduce PAWEval, an outcome-level protocol that converts repeated video rollouts into empirical distributions over possible physical behaviors. Across 50 scenarios and eleven current systems, no model consistently matches the reference probabilities while recovering the range of valid behaviors. Having established this gap, we test whether language prompts, initial noise sampling, or model training can reshape the model's predictive distribution. We believe our work can serve as a foundation for future efforts to move towards probabilistically aligned world modeling.
Chinese Translation
最近的视频生成模型越来越多地被视为世界模型。许多物理过程可以以多种有效方式展开。因此，世界模型不仅应重现一个合理的轨迹，还应在相同的初始观察和行动下重现可能行为的分布。我们将这一分布层面的要求称为概率对齐。然而，现有的评估主要评估单个视频的合理性，并未测试重复生成是否恢复了正确的分布。这引发了一个核心问题：当前的视频生成器距离概率对齐的世界建模还有多远？为了解答这一问题，我们将概率对齐形式化为世界模型的分布标准，并引入PAWBench，一个用于评估视频生成器作为世界动态随机采样器的基准。我们进一步引入PAWEval，一个结果级协议，将重复的视频生成转换为可能物理行为的经验分布。在50个场景和11个当前系统中，没有任何模型能够在恢复有效行为范围的同时一致匹配参考概率。在确定这一差距后，我们测试了语言提示、初始噪声采样或模型训练是否可以重塑模型的预测分布。我们相信我们的工作可以为未来朝着概率对齐的世界建模努力奠定基础。
cs.CV / 88 / 2608.27365
KnockGS:interaction-Grounded Calibrationof Physical Gaussian Representations
KnockGS：基于交互的物理高斯表示校准
Abstract
Physics-integrated 3D Gaussian representations now allow reconstructed deformable objects to be simulated and rendered under explicit material models. Existing pipelines, however, assume that material parameters are known or manually specified, limiting their applicability when these parameters must be inferred from observed object dynamics. We propose KnockGS, an interaction-response PhysicalGS framework that estimates the elasticity and density scales of a 3D Gaussian object from its dynamics under a known applied force. Rather than treating physical simulation only as a forward process, we turn the force-induced response into a calibration signal: temporal response features are xtracted from the observed dynamics, the two material scales are estimated from those features, and the estimate is then frozen and written back into the same simulator so that it can be tested on an interaction it was never fitted to.We evaluate the framework on both parameter recovery and response-level fidelity. The estimated scales are compared against hidden ground truth, and the re-simulated object is measured against the target using 3D particle trajectories, response-curve statistics, and rendered-frame quality. Across five held-out material targets, our method recovers the scales substantially more accurately than response retrieval, global regression, or a fixed default material, and the frozen estimate remains predictive under interactions that differ in direction and in magnitude. Interaction response therefore carries enough information to calibrate material scales in physically grounded 3D Gaussian representations.Our study is a first step toward interactive PhysicalGS systems that calibrate a Gaussian asset whose rendered appearance and simulated response are consistent.
Chinese Translation
物理集成的3D高斯表示现已允许在明确的材料模型下对重建的可变形物体进行模拟和渲染。然而，现有的流程假设材料参数是已知或手动指定的，这限制了在必须从观察到的物体动态中推断这些参数时的适用性。我们提出了KnockGS，一个交互响应的PhysicalGS框架，它从已知施加力下的动态中估计3D高斯物体的弹性和密度尺度。我们并不只是将物理模拟视为一个前向过程，而是将力引起的响应转化为校准信号：从观察到的动态中提取时间响应特征，从这些特征中估计两个材料尺度，然后将估计值固定并写回同一模拟器，以便在未曾适配的交互中进行测试。我们在参数恢复和响应级别保真度方面评估了该框架。估计的尺度与隐藏的真实值进行比较，重新模拟的物体则使用3D粒子轨迹、响应曲线统计和渲染帧质量进行测量。在五个保留的材料目标中，我们的方法在尺度恢复上显著比响应检索、全局回归或固定默认材料更准确，并且固定的估计在方向和幅度不同的交互下仍然具有预测能力。因此，交互响应携带足够的信息来校准物理基础的3D高斯表示中的材料尺度。我们的研究是朝着交互式PhysicalGS系统迈出的第一步，该系统校准一个高斯资产，使其渲染外观和模拟响应保持一致。
cs.CV / 89 / 2608.27367
Successive Capacity Growth: Task-Complexity-Driven Width and Depth Expansion for Vision Transformer Encoders in JEPA World Models
连续容量增长：基于任务复杂度的视觉变换器编码器宽度和深度扩展在JEPA世界模型中的应用
Abstract
Joint-Embedding Predictive Architectures (JEPAs) for world modeling typically employ fixed-size Vision Transformer encoders that are over-provisioned for simple tasks and under-provisioned for complex ones, with significant redundancy across attention heads. We propose Successive Capacity Growth (SCG), a method that starts from a minimal encoder (1 head, 2 layers, 283K parameters) and grows incrementally in width (adding attention heads for low-level semantic capacity) or depth (adding transformer blocks for higher-order semantic abstraction), driven by a task-agnostic test-and-verify mechanism that exploits function-preserving expansion to safely trial architectural changes and roll back if they do not improve prediction loss. The Sketched Isotropic Gaussian Regularizer (SIGReg) ensures that all learned semantic dimensions remain statistically independent and aligned with the predictive objective, preventing collapse even as the architecture grows. On a 60-dimensional multi-object dynamics task, SCG naturally triggers depth expansion, improving prediction loss by 20.3% over the fixed small baseline with 56 times greater parameter efficiency than scaling to the fixed large model; on a 2D navigation task, a single width expansion yields even an 23% improvement over the fixed large model. Across all three tested environments of increasing complexity, the adaptive encoder matches or exceeds the fixed small baseline, with zero false-positive expansions and bit-exact function preservation (ratio = 1.0, absolute difference = 0.0). The take-away is that JEPA world model encoders need not be pre-allocated at maximum capacity - they can grow successively as the task demands, achieving significant compute and data efficiency while maintaining representation quality.
Chinese Translation
联合嵌入预测架构（JEPA）用于世界建模通常采用固定大小的视觉变换器编码器，这些编码器对于简单任务过度配置，而对于复杂任务则配置不足，且在注意力头之间存在显著冗余。我们提出了连续容量增长（SCG）方法，该方法从一个最小编码器（1个头，2层，283K参数）开始，并根据任务无关的测试与验证机制逐步增加宽度（为低级语义容量添加注意力头）或深度（为更高阶语义抽象添加变换块），利用保持功能的扩展安全地试验架构变化，并在未能改善预测损失时回滚。草图各向同性高斯正则化器（SIGReg）确保所有学习的语义维度在统计上保持独立，并与预测目标对齐，防止即使在架构增长时也发生崩溃。在一个60维的多对象动态任务中，SCG自然触发深度扩展，使预测损失比固定的小基线改善20.3%，且参数效率比扩展到固定的大模型高出56倍；在一个二维导航任务中，单次宽度扩展甚至使固定的大模型改善了23%。在所有三个复杂性逐渐增加的测试环境中，自适应编码器的表现与固定的小基线相匹配或超过，且没有出现假阳性扩展，并且功能保持精确（比率=1.0，绝对差异=0.0）。结论是，JEPA世界模型编码器不需要预先分配到最大容量——它们可以随着任务需求逐步增长，实现显著的计算和数据效率，同时保持表示质量。
cs.CV / 90 / 2608.27395
LeVJEPA: Efficient & Scalable Video Pretraining without the Heuristics
LeVJEPA：高效且可扩展的视频预训练，无需启发式方法
Abstract
Video carries the temporal structure of the physical world, yet learning representations from it has remained computationally expensive: prevailing self-supervised methods either prevent representation collapse through architectural asymmetries, coupling an exponential-moving-average target encoder, a stop-gradient, and a capacity-limited predictor, or circumvent it by reconstructing masked content in pixel space. We introduce LeVJEPA, the first video encoder trained under LeJEPA's collapse-free objective, which dispenses with both. A single encoder is trained with an invariance loss over global and local views of a clip, regularized by SIGReg, which excludes collapse with a provable guarantee. The architecture reduces to an encoder and a projector, and the objective to a single hyperparameter. This formulation admits two properties. First, the cost of pretraining is governed by the number of tokens the encoder observes; uniform random token dropping renders this number small while simultaneously improving downstream accuracy. At matched epochs on identical data, LeVJEPA matches or surpasses V-JEPA 2 across ViT-S/B/L at 5.6 to 20.8x less pretraining compute, and at matched total FLOPs it exceeds the strongest video baseline by 7.6 points on ImageNet-1K while remaining competitive on motion-centric benchmarks. Second, since no asymmetry between branches is required, the encoder can be trained with block-causal attention at no measurable accuracy cost: temporal ordering becomes a property of the encoder itself. Against a compute-matched DINOv2 trained on frames of the same videos, LeVJEPA approaches the image-pretrained encoder on appearance-centric evaluation while nearly doubling its motion-centric accuracy. These results indicate that, once its computational overhead is removed, video becomes a viable and in several respects preferable substrate for general-purpose visual pretraining.
Chinese Translation
视频承载着物理世界的时间结构，但从中学习表示仍然计算成本高昂：现有的自监督方法要么通过架构不对称性防止表示崩溃，结合指数移动平均目标编码器、停止梯度和容量有限的预测器，要么通过在像素空间重建遮挡内容来规避这一问题。我们提出了LeVJEPA，这是第一个在LeJEPA的无崩溃目标下训练的视频编码器，摒弃了这两种方法。一个单一的编码器通过对剪辑的全局和局部视图施加不变性损失进行训练，并通过SIGReg进行正则化，后者提供了可证明的崩溃排除保证。该架构简化为一个编码器和一个投影器，目标简化为一个超参数。该公式具有两个特性。首先，预训练的成本由编码器观察的标记数量决定；均匀随机丢弃标记使这个数量变小，同时提高下游准确性。在相同数据上匹配的训练周期中，LeVJEPA在ViT-S/B/L上以5.6到20.8倍更少的预训练计算量与V-JEPA 2相匹配或超越，并且在匹配的总FLOPs下，其在ImageNet-1K上超过最强视频基线7.6个百分点，同时在以运动为中心的基准上保持竞争力。其次，由于不需要分支之间的不对称性，编码器可以使用块因果注意力进行训练，而不会造成可测量的准确性损失：时间顺序成为编码器本身的一个属性。与在相同视频帧上训练的计算匹配的DINOv2相比，LeVJEPA在以外观为中心的评估中接近图像预训练编码器，同时几乎将其以运动为中心的准确性翻倍。这些结果表明，一旦去除了计算开销，视频成为一种可行且在多个方面优于通用视觉预训练的基础。
cs.CV / 91 / 2608.27407
Reconstructing Humans and Objects in Interaction using Large Reconstruction Models
使用大型重建模型重建人类与物体的交互
Abstract
Estimation of Human-Object Interactions in 3D (3D HOI) is a fundamental problem in 3D computer vision with applications in AR/VR, robotics, and embodied AI. However, reconstructing these interactions in 3D remains challenging due to depth ambiguities, occlusions, and object shape variability. Existing approaches are primarily concerned with reprojection and contact constraints, fitting parametric human models and object templates to 2D images. In this paper, we explore a different avenue. We present MILO, a framework that leverages the visual capabilities of Large Reconstruction Models (LRMs) to recover detailed 3D human-object interactions from a single image. Our key observation is that LRMs provide a powerful geometric scaffold that preserves relative human-object arrangement and proximity cues. This significantly simplifies the reconstruction procedure, reframing the problem as interpreting the LRM mesh: we segment it into human and object components, fit a parametric body model to the human part, and optionally align an object template to the object part (if such a template is available). MILO achieves strong reconstruction accuracy and outperforms existing baselines across multiple benchmarks and interaction scenarios. Our code is available at https://ac5113.github.io/MILO.
Chinese Translation
三维人类-物体交互估计（3D HOI）是三维计算机视觉中的一个基本问题，广泛应用于增强现实/虚拟现实（AR/VR）、机器人技术和具身人工智能（embodied AI）。然而，由于深度歧义、遮挡和物体形状的多样性，在三维中重建这些交互仍然具有挑战性。现有的方法主要关注重投影和接触约束，将参数化的人体模型和物体模板拟合到二维图像中。本文探索了一条不同的途径。我们提出了MILO，一个利用大型重建模型（Large Reconstruction Models, LRM）视觉能力的框架，从单幅图像中恢复详细的三维人类-物体交互。我们的关键观察是，LRM提供了一个强大的几何框架，能够保持相对的人类-物体排列和接近度线索。这显著简化了重建过程，将问题重新框定为解释LRM网格：我们将其分割为人类和物体组件，为人类部分拟合一个参数化的身体模型，并可选地将物体模板对齐到物体部分（如果有这样的模板可用）。MILO在多个基准和交互场景中实现了强大的重建精度，并超越了现有的基线。我们的代码可在 https://ac5113.github.io/MILO 获取。
cs.CV / 92 / 2608.27417
Retrieval Heads Meet Vision: Uncovering How VLMs Locate and Extract Visual Information
检索头与视觉的结合：揭示视觉语言模型如何定位和提取视觉信息
Abstract
Vision-language models (VLMs) can locate an image region referred to by a text prompt and route the corresponding visual evidence to the output, yet the internal mechanism behind this behavior is not understood. Inspired by retrieval heads in large language models, we ask whether VLMs contain an analogous mechanism for visual retrieval. We answer affirmatively by introducing Visual Retrieval Heads (VRHs), a small subset of attention heads (about 1.7-2.6%) that are causally responsible for grounding text descriptions to image regions. To find them, we recast existing head-scoring methods under a unified design space over query tokens, key aggregation, and cross-sample aggregation. We then show that scoring attention from output prediction tokens with a sum over the ground-truth referent region most reliably identifies causal heads. Across eleven VLMs and five referring-expression benchmarks, masking only the top 20 VRHs reduces grounding accuracy by up to 80 percentage points, while masking the same number of random heads has little effect. Beyond replicating the causal-sparse-universal triad established for text retrieval heads, VRHs exhibit several properties not previously reported: they generalize across visual reference tasks, remaining causal on attribute, spatial, counting, and visual-math benchmarks despite being discovered through bounding-box prediction; they are functionally specific, preserving output format while corrupting localization; and they are architecturally shared, transferring causally across VLMs that share an LLM backbone but differ in vision encoder, projector, and instruction tuning.
Chinese Translation
视觉语言模型（VLMs）能够定位由文本提示所指代的图像区域，并将相应的视觉证据传递到输出中，但其背后的内部机制尚不清楚。受大型语言模型中检索头的启发，我们探讨VLMs是否具有类似的视觉检索机制。我们通过引入视觉检索头（Visual Retrieval Heads, VRHs）来肯定这一点，这是一小部分（约1.7-2.6%）因果上负责将文本描述与图像区域对接的注意力头。为了找到它们，我们在查询标记、关键聚合和跨样本聚合的统一设计空间下重新构建现有的头评分方法。然后，我们展示了通过对真实参考区域的求和来对输出预测标记的注意力进行评分，最可靠地识别出因果头。在十一种VLMs和五个指代表达基准中，仅遮蔽前20个VRHs就会将对接准确率降低多达80个百分点，而遮蔽相同数量的随机头几乎没有影响。除了复制为文本检索头建立的因果-稀疏-通用三元组外，VRHs还展现出一些之前未报告的特性：它们在视觉参考任务中具有良好的泛化能力，尽管是通过边界框预测发现的，但在属性、空间、计数和视觉数学基准上仍然保持因果性；它们在功能上具有特异性，保持输出格式的同时破坏定位；并且它们在架构上是共享的，在共享大型语言模型（LLM）主干但在视觉编码器、投影器和指令调优上有所不同的VLMs之间因果转移。
cs.CV / 93 / 2608.27456
UrbanGround: From Local Perception to Spatial Agency in a Real-Scale City
UrbanGround：从地方感知到真实规模城市中的空间代理
Abstract
Multimodal large language models (MLLMs) can interpret a street view, but urban agency depends on whether such local evidence remains useful after the agent starts to move. In this paper, we investigate how far current MLLM agents can turn local urban perception into reliable action in a complicated real-scale city. We propose UrbanGround, the first sandbox to make this question testable in a physically constrained replica of Hong Kong built from territory-wide 3D geospatial data. UrbanGround supports closed-loop interaction from a first-person view and provides an interactive map for navigation. Agents can directly enter the 3D city and explore from a first-person view. Our analysis follows the growth of the spatial problem through three research questions. We first test whether an agent can ground a local scene well enough to answer spatial questions after active observation. Then we ask whether that grounding supports navigation as destinations become farther away and less explicit. Finally, we examine whether the resulting behavior survives changes in route availability and pedestrian motion. Contemporary MLLM agents usually show useful atomic abilities in visual recognition and short-range spatial reasoning, while orientation and pedestrian-aware movement remain unreliable. Their central failure emerges over extended exploration, where local abilities do not compose into sustained goal-directed behavior and errors accumulate without effective correction. We hope UrbanGround will support broader study of how far current MLLM agents can explore reliably in complex, open-ended urban environments.
Chinese Translation
多模态大型语言模型（MLLMs）能够解读街景，但城市代理的有效性取决于在代理开始移动后，这些地方证据是否仍然有用。本文探讨了当前的MLLM代理在复杂的真实规模城市中，将地方城市感知转化为可靠行动的能力。我们提出了UrbanGround，这是第一个沙盒，旨在通过基于全境3D地理空间数据构建的香港物理复制品，使这一问题可测试。UrbanGround支持从第一人称视角的闭环交互，并提供用于导航的互动地图。代理可以直接进入3D城市并从第一人称视角进行探索。我们的分析通过三个研究问题跟踪空间问题的增长。首先，我们测试代理是否能够在主动观察后，将地方场景扎根到足够的程度，以回答空间问题。然后，我们询问这种扎根是否支持导航，因为目的地变得越来越远且不那么明确。最后，我们检查在路线可用性和行人运动变化时，所产生的行为是否能够持续。现代MLLM代理通常在视觉识别和短距离空间推理方面表现出有用的原子能力，而方向感和考虑行人的移动仍然不可靠。它们的主要失败出现在扩展探索中，地方能力无法组合成持续的目标导向行为，错误在没有有效纠正的情况下累积。我们希望UrbanGround能够支持更广泛的研究，探讨当前的MLLM代理在复杂、开放的城市环境中能够可靠探索的范围。
cs.AI / 1 / 2608.26107
EduRiskX: A Neuro-Symbolic Framework with F-Logic Reasoning for Early Academic Risk Prediction
EduRiskX：一种结合F-逻辑推理的神经符号框架用于早期学术风险预测
Abstract
Predicting students' academic risk in online education is crucial for enabling timely interventions that can improve retention and learning outcomes. However, existing models often suffer from limited early detection capability and insufficient interpretability, leading to a "black-box" trust crisis that hinders their adoption in real-world pedagogical settings. To address these challenges, we propose EduRiskX, a neuro-symbolic framework that integrates a temporal Transformer-based predictor with F-Logic symbolic reasoning. The neural component models longitudinal student activity sequences using temporal attention, class-weighted loss, and dynamic weekly truncation. Acting as a data-driven expert system, an F-Logic rule base -- grounded in established educational theories (Engagement Theory and Student Integration Model) to mimic the diagnostic logic of human educators -- is constructed exclusively from the training data. The neural risk probability and the symbolic confidence score are then combined through a logistic regression-based fusion mechanism that learns the relative contribution of each signal. Experiments on the Open University Learning Analytics Dataset (OULAD) using a strict 80/10/10 student-level split show that EduRiskX achieves an accuracy of 0.900 and an F1-score of 0.894 at the end of the semester (Week 38), with an average early detection week of 9.32 and a detection rate of 94.30 percent. Compared with state-of-the-art time-series models (PatchTST, iTransformer) and common deep learning baselines (LSTM, CNN), EduRiskX yields improved recall and earlier risk identification under identical conditions. Beyond predictive performance, the F-Logic module provides structured rule-based explanations linking predictions to observable behavioral patterns and educational theories.
Chinese Translation
在在线教育中，预测学生的学术风险对于及时干预以提高留存率和学习成果至关重要。然而，现有模型往往存在早期检测能力有限和可解释性不足的问题，导致“黑箱”信任危机，阻碍了它们在实际教学环境中的应用。为了解决这些挑战，我们提出了EduRiskX，这是一种结合了基于时间的Transformer预测器和F-逻辑符号推理的神经符号框架。神经组件利用时间注意力、类别加权损失和动态周截断对学生的纵向活动序列进行建模。作为一个数据驱动的专家系统，F-逻辑规则库基于已建立的教育理论（参与理论和学生整合模型）构建，旨在模拟人类教育者的诊断逻辑，并完全依赖于训练数据。然后，通过基于逻辑回归的融合机制将神经风险概率和符号置信度分数结合起来，以学习每个信号的相对贡献。在对开放大学学习分析数据集（OULAD）进行的实验中，采用严格的80/10/10学生级别划分，结果显示EduRiskX在学期末（第38周）达到了0.900的准确率和0.894的F1分数，平均早期检测周为9.32，检测率为94.30%。与最先进的时间序列模型（PatchTST、iTransformer）和常见的深度学习基线（LSTM、CNN）相比，EduRiskX在相同条件下实现了更好的召回率和更早的风险识别。除了预测性能外，F-逻辑模块还提供了结构化的基于规则的解释，将预测与可观察的行为模式和教育理论联系起来。
cs.AI / 2 / 2608.26109
Standalone LLM and a Pre-specified Agentic Pipeline for Explaining ICU Mortality Predictions: a Feasibility Study on the eICU Demo Dataset
独立大型语言模型及预设代理管道用于解释ICU死亡预测的可行性研究：基于eICU演示数据集的研究
Abstract
Machine-learning models can predict ICU mortality accurately, but feature-attribution methods alone rarely provide the clinical narrative needed for bedside use. Large language models (LLMs) may bridge this gap, and multi-step agentic pipelines are a plausible extension because they separate data interpretation, guideline checking, and final explanation. This revised feasibility study preserves the original standalone-versus-agentic comparison while making the main clinical findings more explicit. Using the retained local eICU Demo artifact set (2,353 ICU stays; 8.1\% mortality), XGBoost achieved an AUROC of 0.855 (95\% CI 0.796--0.906) and an AUPRC of 0.332 (95\% CI 0.217--0.494). On a stratified 38-case explanation subset, the standalone LLM produced 1 explanation with explicit outcome leakage, whereas the four-step agentic pipeline produced none. Among the 14 cases that overlapped with the SHAP review subset, the standalone LLM showed higher SHAP alignment (mean Jaccard 0.171 versus 0.077) and higher direction consistency (92.9\% versus 78.6\%), while the agentic pipeline showed higher guideline grounding (0.762 versus 0.143), higher value specificity (0.236 versus 0.143), and slightly higher plausibility (0.700 versus 0.671). Clinically, the results suggest that agentic decomposition may improve safety-relevant grounding and patient-specific detail, but it should be paired with attribution-based checks before use in high-stakes risk explanation.
Chinese Translation
机器学习模型能够准确预测ICU死亡率，但单靠特征归因方法通常无法提供临床上所需的叙述。大型语言模型（LLMs）可能弥补这一空白，而多步骤代理管道则是一个合理的扩展，因为它将数据解释、指南检查和最终解释分开。本次修订的可行性研究保留了原始的独立与代理比较，同时使主要临床发现更加明确。使用保留的本地eICU演示数据集（2,353次ICU住院；8.1\%死亡率），XGBoost达到了0.855的AUROC（95\\% CI 0.796--0.906）和0.332的AUPRC（95\\% CI 0.217--0.494）。在一个分层的38例解释子集中，独立LLM产生了1个具有明确结果泄漏的解释，而四步代理管道则没有产生任何。14个与SHAP审查子集重叠的案例中，独立LLM显示出更高的SHAP一致性（平均Jaccard 0.171对0.077）和更高的方向一致性（92.9\\%对78.6\\%），而代理管道则显示出更高的指南基础（0.762对0.143）、更高的价值特异性（0.236对0.143）和略高的合理性（0.700对0.671）。从临床角度来看，结果表明代理分解可能改善安全相关的基础和患者特定细节，但在用于高风险解释之前，应与基于归因的检查相结合。
cs.AI / 3 / 2608.26111
Large Models for Battery Prognostics and Health Management: A Review and Future Roadmap
用于电池预测和健康管理的大型模型：综述与未来路线图
Abstract
Battery Prognostics and Health Management (BPHM) is critical for ensuring the safe, reliable, and cost-effective operation of batteries across electric vehicles, grid storage, and consumer electronics. Conventional BPHM approaches, including physics-based models and task-centric deep learning methods, face challenges in computational efficiency and parameterization, cross-domain generalization, dependence on extensive labeled run-to-failure data, and model interpretability. Recent Large Models (LMs), built upon Transformer architectures and self-supervised pre-training, offer a transformative new paradigm to overcome these long-standing bottlenecks. This review provides the first comprehensive survey of LM applications in BPHM, systematically examining how these models address challenges in the field. We begin by elucidating the foundational technologies enabling LMs, including Transformer architectures, self-supervised learning, large-scale multimodal datasets, and PEFT techniques. We then categorize recent progress along four critical dimensions: mitigating data scarcity, enhancing generalization and robustness, integrating domain knowledge for interpretability, and enabling system-level automation. Despite promising results, significant challenges remain across data accessibility, intelligence validation, trustworthiness, and deployment feasibility. To guide future research, we propose a roadmap focused on building collaborative data ecosystems, validating intelligence for industrial applications, enhancing trustworthiness with physics-informed designs, and enabling efficient on-device deployment. This review establishes a systematic approach to understand and advance LM-driven BPHM, providing researchers and practitioners with essential insights for developing next-generation battery management systems capable of safe, reliable, and autonomous operation throughout battery lifecycles.
Chinese Translation
电池预测和健康管理（BPHM）对于确保电池在电动汽车、电网储能和消费电子产品中的安全、可靠和经济高效的运行至关重要。传统的BPHM方法，包括基于物理的模型和以任务为中心的深度学习方法，面临计算效率和参数化、跨领域泛化、依赖大量标记的失效运行数据以及模型可解释性等挑战。最近，基于Transformer架构和自监督预训练的大型模型（LMs）提供了一种变革性的新范式，以克服这些长期存在的瓶颈。本文综述首次全面调查了LM在BPHM中的应用，系统地考察了这些模型如何应对该领域的挑战。我们首先阐明了支持LM的基础技术，包括Transformer架构、自监督学习、大规模多模态数据集和PEFT技术。然后，我们沿着四个关键维度对近期进展进行了分类：缓解数据稀缺、增强泛化和鲁棒性、整合领域知识以提高可解释性，以及实现系统级自动化。尽管取得了令人鼓舞的结果，但在数据可获取性、智能验证、可信度和部署可行性等方面仍然存在重大挑战。为了指导未来的研究，我们提出了一条重点关注构建协作数据生态系统、验证工业应用的智能、通过物理信息设计增强可信度以及实现高效的设备端部署的路线图。本文建立了一种系统的方法，以理解和推进LM驱动的BPHM，为研究人员和从业者提供了开发能够在电池生命周期内安全、可靠和自主运行的下一代电池管理系统的必要见解。
cs.AI / 4 / 2608.26113
PICasso: An AI-Enabled Design Framework for Autonomous Optimization of Silicon Photonic Devices
PICasso：一种支持人工智能的自优化硅光子器件设计框架
Abstract
We present PICasso, an AI-assisted framework for automated synthesis, verification, and optimization of photonic integrated circuits (PICs) from natural-language specifications. PICasso couples a structured NL -> YAML -> GDS generation pipeline with PDK aware knowledge injection, automated placement and routing, DRC/LVS validation, and SAX-based photonic simulation. To systematically evaluate AI-driven photonic design, we introduce PIC-Set, a benchmark of 36 parameterized PIC design tasks spanning core photonic primitives and multi-component circuits. Using PIC-Set, we benchmark several state-of-the-art Large Language Models (LLMs) under a unified evaluation protocol, including new metrics such as structural and functional $Spec@k$, optimization efficiency, and robustness under perturbations. Across the benchmark, PICasso significantly improves end-to-end specification satisfaction compared to vanilla LLM generation. Structural $Spec@3$ reaches up to 92.7% and functional $Spec@3$ up to 52% on high-complexity circuits. In addition, PICasso consistently reduces circuit insertion loss, lowering the mean loss from 4.98 dB to 3.25 dB (1.74 dB improvement) through simulation-guided optimization. These results demonstrate that structured domain constraints, physical verification, and simulation feedback transform LLMs from brittle netlist generators into practical PIC design agents capable of producing manufacturable layouts with competitive runtimes relative to manual GUI-based workflows.
Chinese Translation
我们提出了PICasso，这是一个基于人工智能的框架，用于从自然语言规范自动合成、验证和优化光子集成电路（PICs）。PICasso将结构化的自然语言（NL）到YAML再到GDS的生成管道与PDK（工艺设计套件）感知的知识注入、自动布局和布线、DRC（设计规则检查）/LVS（版图与电路一致性检查）验证以及基于SAX的光子模拟相结合。为了系统地评估基于人工智能的光子设计，我们引入了PIC-Set，这是一个包含36个参数化PIC设计任务的基准，涵盖核心光子原件和多组件电路。利用PIC-Set，我们在统一评估协议下对几种最先进的大型语言模型（LLMs）进行了基准测试，包括结构性和功能性$Spec@k$、优化效率以及在扰动下的鲁棒性等新指标。在基准测试中，与传统的LLM生成相比，PICasso显著提高了端到端规范满足率。在高复杂度电路中，结构性$Spec@3$达到了92.7%，功能性$Spec@3$达到了52%。此外，PICasso始终减少电路插入损耗，通过模拟指导的优化，将平均损耗从4.98 dB降低到3.25 dB（改善幅度为1.74 dB）。这些结果表明，结构化的领域约束、物理验证和模拟反馈使得LLMs从脆弱的网表生成器转变为能够生成可制造布局的实用PIC设计代理，其运行时间与手动图形用户界面（GUI）工作流程相比具有竞争力。
cs.AI / 5 / 2608.26114
CIFQA: A Deterministic Tool-Grounded Multi-Agent LLM Framework for Financial Query Answering
CIFQA：一种基于确定性工具的多智能体大语言模型框架用于金融问答
Abstract
Calculation-intensive financial question answering requires exact reasoning over structured rates, temporal conditions, numerical formulas, and rule-based constraints. Although Large Language Models (LLMs) perform strongly on natural language tasks, they often produce numerically incorrect yet plausible answers when solving multi-step financial calculations. To address this limitation, we introduce CIFQA (Calculation-Intensive Financial Query Answering), a deterministic tool-grounded multi-agent LLM framework for financial question answering. CIFQA separates language understanding from numerical execution by assigning specialized agents to query interpretation, routing, parameter extraction, computation planning, and response generation, while deterministic Python-based tools perform financial calculations and rule application. We instantiate CIFQA for fixed deposit query answering and evaluate it on a curated benchmark of fixed deposit queries. CIFQA achieves 95.54% accuracy on calculation-intensive queries and 90.87% overall accuracy, substantially outperforming direct LLM baselines even when provided with complete formulas, rate cards, and benchmark instructions. Ablation studies show that deterministic components such as exact rate lookup, tenure computation, rolling-year adjustment, and premature-withdrawal logic are critical contributors to performance. Notably, a 17B open-source backbone operating within CIFQA outperforms substantially larger frontier models evaluated with the same financial information, demonstrating that architectural design is a more important determinant of numerical reliability than model scale. While evaluated on fixed deposit queries, CIFQA provides a generalizable framework for calculation-intensive financial reasoning tasks.
Chinese Translation
计算密集型金融问答需要对结构化利率、时间条件、数值公式和基于规则的约束进行精确推理。尽管大型语言模型（LLMs）在自然语言任务中表现出色，但在解决多步骤金融计算时，它们往往会产生数值上不正确但看似合理的答案。为了解决这一局限性，我们提出了CIFQA（计算密集型金融问答），这是一个基于确定性的工具和多智能体的大语言模型框架，用于金融问答。CIFQA通过将查询解释、路由、参数提取、计算规划和响应生成分配给专门的智能体，从而将语言理解与数值执行分开，而基于确定性的Python工具则执行金融计算和规则应用。我们将CIFQA实例化用于定期存款查询，并在一个精心策划的定期存款查询基准上进行评估。CIFQA在计算密集型查询上实现了95.54%的准确率，在整体查询上实现了90.87%的准确率，显著优于直接的大语言模型基线，即使在提供完整公式、利率卡和基准指令的情况下。消融研究表明，确定性组件如精确利率查找、期限计算、滚动年调整和提前取款逻辑是性能的关键贡献者。值得注意的是，在CIFQA中运行的17B开源骨干网络在同样的金融信息下显著优于规模更大的前沿模型，表明架构设计是数值可靠性的重要决定因素，而非模型规模。虽然在定期存款查询上进行了评估，CIFQA为计算密集型金融推理任务提供了一个可推广的框架。
cs.AI / 6 / 2608.26116
The Artificial Experimentalist: Discovery and Control of Self-Organizing Phenomena with Autotelic Reinforcement Learning
人工实验者：通过自我目的强化学习发现和控制自组织现象
Abstract
Existing methods for exploring cellular automata and other complex systems mostly operate in open loop: they set initial conditions, execute a full simulation, and observe the outcome, without intervening during execution. We introduce a closed-loop framework based on autotelic reinforcement learning, in which an agent autonomously samples diverse goals and learns a goal-conditioned policy to intervene in a complex system through minimal, local perturbations. We instantiate this framework on Lenia, a continuous cellular automaton known for life-like self-organizing patterns, in an agentic system we call CARL, and demonstrate three capabilities. First, CARL discovers stable solitons across a wide range of Lenia update rules at a higher rate than heuristic baselines. Second, it learns to steer the movement direction of existing solitons with few interventions, showing that CARL can control self-organizing patterns, not only create them. Third, humans can use trained agents to guide solitons through maze environments in real time by specifying high-level directional commands that the agent translates into low-level interventions. Trained across diverse goals, update rules, and random initial states, the agents acquire policies that generalize zero-shot to various out-of-distribution conditions. These results suggest a path toward artificial experimentalist agents that, autonomously or with human guidance, discover and control emergent phenomena in complex systems.
Chinese Translation
现有的探索细胞自动机和其他复杂系统的方法大多采用开环操作：它们设定初始条件，执行完整的模拟，并观察结果，而在执行过程中不进行干预。我们引入了一种基于自我目的强化学习的闭环框架，其中代理自主地采样多样化的目标，并学习一种目标条件策略，通过最小的局部扰动干预复杂系统。我们在Lenia（一个以类生命自组织模式著称的连续细胞自动机）上实现了这一框架，构建了一个我们称之为CARL的代理系统，并展示了三种能力。首先，CARL以高于启发式基线的速度发现了在广泛的Lenia更新规则下的稳定孤子。其次，它学习以少量干预来引导现有孤子的运动方向，表明CARL不仅能够创建自组织模式，还能控制它们。第三，人类可以通过指定高层次的方向命令，实时使用训练好的代理引导孤子穿越迷宫环境，而代理会将这些命令转换为低层次的干预。经过多样化目标、更新规则和随机初始状态的训练，代理获得的策略能够在各种分布外条件下实现零样本泛化。这些结果暗示了一条通向人工实验者代理的路径，这些代理能够在复杂系统中自主或在人工指导下发现和控制涌现现象。
cs.AI / 7 / 2608.26134
The Accuracy-Efficiency Paradox Quantifying Net Energy Loss in on-Device Energy Forecasting
准确性与效率的悖论：量化设备端能量预测中的净能量损失
Abstract
Energy forecasting aims to maximize accuracy to ensure energy efficiency by reducing energy waste, an objective that applies equally to on-device forecasting for mission-critical edge environments, including military systems. However, this paper identifies the Accuracy-Efficiency Paradox: high-precision energy forecasting models can ironically trigger a net energy deficit. This stems from both edge AI's inference energy consumption and battery aging. We propose a Total Cost of Ownership (TCO) framework for energy forecasting, designed to minimize net energy loss. This framework treats not only inference energy consumption but also battery aging as a unified form of energy loss, as degradation represents a physical dissipation of the system's future energy-carrying capacity. We demonstrate that in thermally sensitive edge environments, energy saved by the superior precision of complex architectures is often outweighed by the total energy lost through their high operational intensity.
Chinese Translation
能量预测旨在最大化准确性，以确保通过减少能量浪费来实现能效，这一目标同样适用于包括军事系统在内的关键任务边缘环境中的设备端预测。然而，本文识别出准确性与效率的悖论：高精度的能量预测模型反而可能引发净能量赤字。这源于边缘人工智能的推理能耗和电池老化。我们提出了一种总拥有成本（Total Cost of Ownership, TCO）框架，用于能量预测，旨在最小化净能量损失。该框架不仅将推理能耗视为能量损失的一种形式，还将电池老化视为统一的能量损失形式，因为老化代表了系统未来能量承载能力的物理衰减。我们证明，在热敏感的边缘环境中，复杂架构所带来的高精度节能往往被其高操作强度所导致的总能量损失所抵消。
cs.AI / 8 / 2608.26145
LLMs for Academic Workflows: An Evaluation of Literature Reviews Generated with Short and Long Context Windows of LLMs
用于学术工作流程的大型语言模型：基于短期和长期上下文窗口生成的文献综述评估
Abstract
Our research focuses on evaluating literature reviews generated in short and long context settings of large language models (LLMs) to investigate the impact of context window on the quality of AI-generated literature reviews and the role of AI in supporting literature review writing. Twenty AI-generated literature reviews based on research sources from Semantic Scholar and Arxiv were evaluated by two researchers across 15 dimensions. Our findings reveal that AI-generated literature reviews require human oversight to meet academic publishing standards. As context windows increase, LLMs can incorporate broader information and maintain coherence across longer inputs, but they also exacerbate issues such as content repetition, omission of critical work, and a tendency towards descriptiveness over synthesis. Our work shows that AI-generated reviews can provide foundational overviews, but their output must be critically evaluated and refined by domain experts. Future research should consider integrating other LLMs and fine-tuned models in different domains with hybrid approaches that combine human expertise with AI capabilities to address the limitations identified in this study.
Chinese Translation
我们的研究聚焦于评估在大型语言模型（LLMs）的短期和长期上下文设置中生成的文献综述，以探讨上下文窗口对AI生成文献综述质量的影响以及AI在支持文献综述写作中的作用。基于Semantic Scholar和Arxiv的研究来源，我们评估了20篇AI生成的文献综述，涉及15个维度的评估。我们的发现表明，AI生成的文献综述需要人类监督，以满足学术出版标准。随着上下文窗口的增加，LLMs能够整合更广泛的信息，并在更长的输入中保持连贯性，但它们也加剧了内容重复、遗漏关键工作的情况，以及倾向于描述而非综合的趋势。我们的研究表明，AI生成的综述可以提供基础概述，但其输出必须经过领域专家的批判性评估和精炼。未来的研究应考虑整合其他LLMs和在不同领域微调的模型，并采用结合人类专业知识与AI能力的混合方法，以解决本研究中识别的局限性。
cs.AI / 9 / 2608.26149
Methodological and Conceptual Framework for 5D Multi-Table Analysis: A Unified Approach for Complex Data Reuse
5D多表分析的方法论与概念框架：复杂数据重用的统一方法
Abstract
Multi-table learning remains a major challenge in machine learning for healthcare and other complex information systems. Relational data combine several sources of complexity, including large data volume, high-dimensional variables, high-cardinality categorical features, complex inter-table dependencies, and repeated temporal observations. We introduce the Relational Hypergraph Transformer (RHT), a unified architecture that represents relational databases as hypergraphs, learns pentadimensional embeddings (PentE), and performs sparse relational attention with complexity proportional to the average relational degree rather than the square of the number of entities. We formally define the architecture, derive the complexity of its attention mechanism, and provide an open-source reference implementation. We evaluate RHT on the public Synthea synthetic electronic health record dataset using multi-label prediction of SNOMED CT condition codes per encounter, a task characterized by high categorical cardinality and long-tailed label distributions. Comparisons with tabular, relational, and temporal graph baselines show that RHT produces more semantically coherent embeddings while remaining computationally scalable. In this benchmark, the highest rare-code recall is achieved by XGBoost, whereas RHT attains the strongest embedding semantic coherence. We also report ablation studies quantifying the contribution of each architectural component. Clinical validation on MIMIC-IV is planned following PhysioNet credentialing. Source code and experimental protocols are provided in the accompanying repository.
Chinese Translation
多表学习仍然是医疗保健和其他复杂信息系统中机器学习的一大挑战。关系数据结合了多种复杂性来源，包括大数据量、高维变量、高基数分类特征、复杂的表间依赖关系以及重复的时间观察。我们引入了关系超图变换器（Relational Hypergraph Transformer, RHT），这是一种统一架构，它将关系数据库表示为超图，学习五维嵌入（PentE），并执行稀疏关系注意力，其复杂性与平均关系度成正比，而不是与实体数量的平方成正比。我们正式定义了该架构，推导了其注意力机制的复杂性，并提供了一个开源参考实现。我们在公共的Synthea合成电子健康记录数据集上评估了RHT，使用多标签预测每次就诊的SNOMED CT病症代码，该任务的特点是高分类基数和长尾标签分布。与表格、关系和时间图基线的比较表明，RHT产生了更具语义一致性的嵌入，同时保持了计算上的可扩展性。在这一基准中，XGBoost实现了最高的稀有代码召回率，而RHT则达到了最强的嵌入语义一致性。我们还报告了消融研究，量化了每个架构组件的贡献。计划在获得PhysioNet认证后对MIMIC-IV进行临床验证。源代码和实验协议已在附带的代码库中提供。
cs.AI / 10 / 2608.26150
Leveraging Large Language Models for Systematic Literature Review of Disease Spread Models
利用大型语言模型进行疾病传播模型的系统文献综述
Abstract
Recent advancements in Large Language Models (LLMs) have created new opportunities to streamline and potentially automate many research processes, including systematic literature reviews (SLRs). This study reports an LLM pipeline development for extracting model-relevant information from 536 peer-reviewed agent-based modeling papers. We compare the results with those of a human-conducted SLR. Our results show paper-level accuracies of approximately 77.95% for GPT-4.1 and 81.67% for GPT-5.0. Field-level accuracy ranges from 32.40% to 100.00%, with more complex or subjective fields performing less reliably. Importantly, we find that agreement between LLMs is a potential indicator of output quality: low agreement may signal hallucinations, whereas high agreement combined with low accuracy may point to noise or errors in the human dataset. Overall, our study provides practical insights into prompt development and highlights both the potential and limitations of using LLMs for full-scale SLRs in the modeling and simulation domain.
Chinese Translation
近年来，大型语言模型（LLMs）的进展为简化和潜在自动化许多研究过程，包括系统文献综述（SLRs），创造了新的机会。本研究报告了一个LLM管道的开发，该管道用于从536篇经过同行评审的基于代理的建模论文中提取与模型相关的信息。我们将结果与人类进行的SLR进行比较。我们的结果显示，GPT-4.1的论文级准确率约为77.95%，而GPT-5.0的论文级准确率为81.67%。领域级准确率范围从32.40%到100.00%，更复杂或主观的领域表现得不够可靠。重要的是，我们发现LLMs之间的一致性可能是输出质量的潜在指标：低一致性可能表示幻觉，而高一致性结合低准确率可能指向人类数据集中存在噪声或错误。总体而言，我们的研究为提示开发提供了实用见解，并突出了在建模和仿真领域中使用LLMs进行全面SLR的潜力和局限性。
cs.AI / 11 / 2608.26151
Explainable Artificial Intelligence for Customer Churn Prediction in Telecommunications: A Framework for CRM Integration
可解释的人工智能在电信客户流失预测中的应用：客户关系管理集成框架
Abstract
Subscriber attrition is a costly, persistent challenge for telecommunications providers, with monthly churn of roughly 1.9% in mature markets eroding billions in revenue annually. Predictive models can flag at-risk customers accurately, yet they are routinely excluded from frontline CRM workflows because high-performing ensemble and non-linear architectures are opaque: a retention specialist cannot design a personalised intervention from a probability score alone, without knowing why a subscriber is at risk. This paper addresses that gap. We benchmark four classifiers--Logistic Regression, Random Forest, XGBoost, and LightGBM--on the IBM Telco Customer Churn benchmark (7,043 records; 19 features; 26.5% churn, balanced to 50% via SMOTE on the training partition only). Logistic Regression attains the strongest AUC-ROC (0.8411) and LightGBM the highest accuracy (78.42%); all four fall within a 0.011 AUC band (0.831--0.841), and 5-fold cross-validation confirms the leading models are effectively tied. Explanations are delivered at two granularities: a global SHAP ranking identifying tenure, total charges, and month-to-month contract as the dominant churn signals, and instance-level SHAP and LIME decompositions that expose the drivers behind each prediction. Building on these outputs, we introduce a four-layer CRM integration architecture that converts risk scores and attribution vectors into tiered segmentation, maps top features to structured retention-action templates, and routes campaign outcomes into a retraining feedback loop. Targeting the highest-risk quintile is projected to cut overall churn by 3.3--5.3 percentage points, preserving an estimated $199K--$319K per campaign cycle.
Chinese Translation
用户流失是电信服务提供商面临的一项高成本且持久的挑战，在成熟市场中，月度流失率约为1.9%，每年造成数十亿美元的收入损失。预测模型能够准确识别处于风险中的客户，但由于高性能的集成和非线性架构的透明性不足，这些模型通常被排除在前线客户关系管理（CRM）工作流程之外：仅凭概率评分，留存专员无法设计个性化的干预措施，因为他们不知道用户为何处于风险之中。本文旨在填补这一空白。我们在IBM电信客户流失基准数据集上对四种分类器进行了基准测试——逻辑回归（Logistic Regression）、随机森林（Random Forest）、XGBoost和LightGBM（Light Gradient Boosting Machine），该数据集包含7,043条记录，19个特征，流失率为26.5%，并通过SMOTE方法在训练集上平衡至50%。逻辑回归获得了最高的AUC-ROC值（0.8411），而LightGBM则达到了最高的准确率（78.42%）；四种模型的AUC值均在0.011的范围内（0.831--0.841），5折交叉验证确认了领先模型之间的有效平衡。我们提供了两种粒度的解释：全球SHAP排名显示服务年限、总费用和逐月合同是主要的流失信号，以及实例级SHAP和LIME分解揭示了每个预测背后的驱动因素。在这些输出的基础上，我们提出了一种四层的CRM集成架构，将风险评分和归因向量转化为分层细分，将主要特征映射到结构化的留存行动模板，并将活动结果引入再训练反馈循环。针对最高风险的五分之一客户，预计整体流失率将降低3.3至5.3个百分点，为每个活动周期节省约199K至319K美元的收入。
cs.AI / 12 / 2608.26153
EEG-to-Report: An Annotation and Feature-Text Framework for Training Language Models on Clinical EEG
EEG到报告：用于临床EEG上训练语言模型的注释和特征文本框架
Abstract
Clinical electroencephalography (EEG) reporting remains largely manual and time-consuming, and current EEG software ecosystems do not produce the structured EEG-text supervision needed for training modern language models. Most toolboxes focus on visualization or preprocessing, providing limited support for workflows that generate high-quality datasets for AI. We introduce EEG-to-Report, a browser-based annotation and feature-text framework that links routine EEG review with the construction of AI-ready datasets. The framework integrates multi-format EEG ingestion, channel standardization, and an interactive viewer with a multimodal annotation layer that combines typed text and transcribed voice notes. For each annotated segment, a feature extraction engine computes a standardized set of spectral, temporal, entropy, Hjorth, connectivity, and spike-related descriptors, stored alongside clinical descriptions in a portable JSON schema. This yields aligned feature-text pairs designed to supervise multimodal EEG-language models. The framework also includes an auto-report module that couples an ensemble of convolutional networks with a large language model to draft clinical narratives for neurologist review. Using pilot annotations, we describe how EEG-to-Report streamlines annotation workflows and produces editable draft reports, providing a reusable foundation for automated EEG reporting systems.
Chinese Translation
临床脑电图（EEG）报告仍然主要依赖手动操作，耗时较长，而当前的EEG软件生态系统并未生成训练现代语言模型所需的结构化EEG文本监督。大多数工具箱专注于可视化或预处理，提供的支持有限，无法生成高质量的AI数据集工作流程。我们介绍了EEG到报告（EEG-to-Report），这是一个基于浏览器的注释和特征文本框架，将常规EEG审查与构建AI准备数据集相结合。该框架集成了多格式EEG数据导入、通道标准化以及一个交互式查看器，配备了一个多模态注释层，结合了输入文本和转录的语音笔记。对于每个注释段落，特征提取引擎计算一组标准化的谱、时间、熵、Hjorth、连接性和与尖峰相关的描述符，这些描述符与临床描述一起存储在可移植的JSON模式中。这产生了旨在监督多模态EEG语言模型的对齐特征-文本对。该框架还包括一个自动报告模块，将卷积网络的集成与大型语言模型结合，以起草供神经科医生审阅的临床叙述。通过初步注释，我们描述了EEG到报告如何简化注释工作流程并生成可编辑的草稿报告，为自动化EEG报告系统提供了可重用的基础。
cs.AI / 13 / 2608.26156
Selection Bias Correction in Retail Intelligence
零售智能中的选择偏差修正
Abstract
Retail intelligence often relies on monitoring popular, high-velocity products, potentially biasing economic indicators by ignoring the "long tail" of niche items. This simulation study investigates selection bias in inflation estimation and compares correction methods across diverse data-generating processes. Through 400 Monte Carlo replications spanning four scenarios--aligned step functions, smooth gradients, misaligned breaks, and polynomial relationships--we test the robustness of Inverse Probability Weighting (IPW) with five specifications against stratification with varying strata counts. Our findings reveal fundamental limits of weighting methods in retail long-tail contexts: stratification achieves superior performance in three of four scenarios, maintaining sub-0.04pp median error even when boundaries deliberately misalign with population breaks (116x advantage over IPW). However, IPW with spline propensity models wins under smooth polynomial relationships (median error 0.007pp vs. 0.013pp), demonstrating context-dependency. Critically, even an oracle IPW specification with perfect structural knowledge achieves 6.06pp error compared to stratification's 0.008pp in step-function scenarios. This reflects violation of the Positivity Assumption--a fundamental causal inference requirement--rather than IPW methodological inferiority. When selection probabilities differ dramatically (90% vs. 1%), weighting methods operate outside their theoretical design envelope. These results demonstrate that stratification provides a safer engineering choice in retail long-tail distributions with severe positivity violations.
Chinese Translation
零售智能通常依赖于监测流行的高流速产品，这可能通过忽视小众商品的“长尾”而导致经济指标的偏差。本模拟研究调查了通货膨胀估计中的选择偏差，并比较了不同数据生成过程中的修正方法。通过400次蒙特卡罗重复实验，涵盖四种情境——对齐的阶梯函数、平滑梯度、错位的断点和多项式关系——我们测试了逆概率加权（Inverse Probability Weighting, IPW）在五种规格下的稳健性，与具有不同层次数量的分层方法进行比较。我们的研究结果揭示了加权方法在零售长尾情境中的基本局限性：在四种情境中的三种中，分层方法表现优越，即使在边界故意与人口断点错位的情况下，仍能保持低于0.04个基点的中位误差（相较于IPW的116倍优势）。然而，在平滑的多项式关系下，使用样条倾向模型的IPW表现更佳（中位误差为0.007个基点对比0.013个基点），显示出情境依赖性。重要的是，即使是具有完美结构知识的神谕IPW规格，在阶梯函数情境下也达到了6.06个基点的误差，而分层方法的误差仅为0.008个基点。这反映了对积极性假设的违反——这是一个基本的因果推断要求，而不是IPW方法论的劣势。当选择概率差异显著（90%对1%）时，加权方法在其理论设计范围之外运作。这些结果表明，在存在严重积极性违反的零售长尾分布中，分层提供了更安全的工程选择。
cs.AI / 14 / 2608.26157
GROUND: Reducing Hallucinations in LLM-Based Enterprise Analytics Through Governed Semantic Definitions
GROUND：通过受控语义定义减少基于大型语言模型的企业分析中的幻觉
Abstract
Natural-language analytics over enterprise data warehouses is increasingly important, but production use is limited by hallucinated metrics, invalid joins, wrong grain, unsafe data access, and unsupported explanations. Existing text-to-SQL systems often ground generation in database schemas or retrieved documentation, while enterprise reporting also requires governed business semantics: approved metrics, dimensions, join paths, filters, and row-level security. This paper introduces GROUND, Governed Retrieval Over Unified Normalized Definitions, a framework that constrains LLM-generated analytics to a governed semantic layer. GROUND supplies approved definitions, binds user intent to governed metrics and dimensions, and validates generated SQL against schema, metric, join, grain, filter, security, and cost rules before execution. On violations, it retries or abstains. In a 100-question synthetic enterprise-reporting benchmark, GROUND is compared with direct schema-only text-to-SQL, schema-RAG, and semantic-only grounding under one shared model. GROUND is the only system free of measured hallucinations across all six evaluated categories, while ungoverned systems violate row-level security on many questions. A semantic-only condition with exact metric definitions but no access policy still leaks data, showing that governance cannot be replaced by metric fidelity alone. The findings are replicated on real U.S. NHTSA vehicle-safety data with independent hand-authored gold and tested on an adversarial set across four models from three providers. GROUND's enforced guarantees, especially filters and row-level security, hold with zero violations on every model, while judgment-dependent behaviors such as refusing undefined metrics remain fallible.
Chinese Translation
对企业数据仓库进行自然语言分析变得越来越重要，但由于幻觉指标、无效连接、错误粒度、不安全的数据访问和不支持的解释，生产使用受到限制。现有的文本到SQL系统通常将生成内容基于数据库模式或检索的文档，而企业报告还需要受控的业务语义：批准的指标、维度、连接路径、过滤器和行级安全性。本文介绍了GROUND（受控统一规范定义检索），这是一个将大型语言模型生成的分析约束于受控语义层的框架。GROUND提供批准的定义，将用户意图绑定到受控的指标和维度，并在执行之前根据模式、指标、连接、粒度、过滤器、安全性和成本规则验证生成的SQL。在违反规则时，它会重试或放弃。在一个包含100个问题的合成企业报告基准测试中，GROUND与直接基于模式的文本到SQL、模式-检索增强生成（schema-RAG）和仅基于语义的生成进行了比较，所有这些都在一个共享模型下进行。GROUND是唯一一个在所有六个评估类别中没有测量幻觉的系统，而未受控系统在许多问题上违反了行级安全性。一个仅基于语义的条件，尽管有精确的指标定义但没有访问策略，仍然泄露数据，表明治理不能仅通过指标的准确性来替代。该发现还在真实的美国国家公路交通安全管理局（NHTSA）车辆安全数据上进行了复制，并在来自三家供应商的四个模型上进行了独立手工编写的金标准测试。GROUND的强制保证，特别是过滤器和行级安全性，在每个模型上均保持零违规，而依赖判断的行为，例如拒绝未定义的指标，仍然存在失误。
cs.AI / 15 / 2608.26160
SAREF-based Ontology for Distributed AI Workflows across the Edge-Fog-Cloud Continuum
基于SAREF的本体用于边缘-雾-云连续体中的分布式人工智能工作流
Abstract
Nowadays semantic models provide limited support for representing distributed AI workflows and their execution across heterogeneous edge, fog, and cloud environments. Therefore, AI processes and resources are often described using incompatible semantic representations, affecting the interoperability, orchestration, and reuse. To address these challenges, this paper proposes a SAREF-compliant ontology for representing distributed AI workflows across the edge-fog-cloud continuum. We extend the SAREF4SYST ontology with concepts for modeling AI pipelines, executable AI jobs, computational resources, deployment constraints, and communication relationships, providing a unified semantic model of both AI workflows and heterogeneous computing infrastructures. The ontology enables semantic interoperability, automated reasoning, and resource-aware orchestration of distributed AI applications while remaining fully aligned with the ETSI SAREF ecosystem. The ontology is evaluated using proof-of-concept smart grid energy services orchestration scenarios and validated using competency questions showing its ability to support AI workflow deployment, execution reasoning, and workload adaptation across heterogeneous edge, fog, and cloud environments. All competency questions were successfully validated using SPARQL querying and semantic reasoning. Experimental results demonstrate deployment success rates of 90-100% with average orchestration decision times below 80 ms across heterogeneous edge-fog-cloud environments, highlighting its effectiveness on ensuring semantic interoperability for distributed AI orchestration.
Chinese Translation
目前，语义模型对表示分布式人工智能工作流及其在异构边缘、雾和云环境中的执行支持有限。因此，人工智能过程和资源通常使用不兼容的语义表示进行描述，这影响了互操作性、编排和重用。为了解决这些挑战，本文提出了一种符合SAREF标准的本体，用于表示边缘-雾-云连续体中的分布式人工智能工作流。我们扩展了SAREF4SYST本体，增加了用于建模人工智能管道、可执行的人工智能作业、计算资源、部署约束和通信关系的概念，提供了一个统一的语义模型，涵盖人工智能工作流和异构计算基础设施。本体实现了语义互操作性、自动推理和资源感知的分布式人工智能应用编排，同时与ETSI SAREF生态系统保持完全一致。通过概念验证的智能电网能源服务编排场景对本体进行了评估，并通过能力问题验证了其支持人工智能工作流部署、执行推理和在异构边缘、雾和云环境中进行工作负载适应的能力。所有能力问题均通过SPARQL查询和语义推理成功验证。实验结果表明，在异构边缘-雾-云环境中，部署成功率为90-100%，平均编排决策时间低于80毫秒，突显了其在确保分布式人工智能编排的语义互操作性方面的有效性。
cs.AI / 16 / 2608.26162
A Safety-Gated Multimodal AI Backend for Mental-Health Support: Hierarchical State Representation, Conservative Risk Fusion, and Controlled Generation in Anian
用于心理健康支持的安全门控多模态人工智能后端：分层状态表示、保守风险融合与Anian中的受控生成
Abstract
Safety-critical mental-health support systems must distinguish when supportive conversation is appropriate from when free-form generation should be blocked. This paper presents Anian, a safety-gated multimodal AI backend for perinatal mental-health support and mindfulness-intervention routing. Anian is not intended to diagnose psychiatric conditions or replace clinical care or crisis intervention. Its modular pipeline places generative AI downstream of structured state representation, conservative risk fusion, and response gating. User text or voice-derived ASR transcripts are mapped into four linked layers: L1 emotion states, L2 psychosocial constructs, L3 safety risk, and L4 intervention routes. Local text- and rule-based safety evidence is fused with external voice-derived evidence using a highest-risk-priority rule, S_fusion = max(S_local, S_external). At moderate or high fused risk, ordinary AI-generated responses and text-to-speech delivery are blocked and replaced by fixed safety content and prompts for human support. An internal prototype evaluation used approximately 858,295 normalized records from public emotion, dialogue, mental-health-related, and Chinese dialogue corpora within a weak-label and rule-derived framework. Micro-F1 scores were 0.9604 for L1 emotion classification, 0.9144 for L2 psychosocial constructs, and 0.9742 for L4 routing. In a controlled safety stress test of 233 samples, the L3 rule engine achieved high-risk recall of 1.0000 within predefined scenarios. These findings support the internal feasibility of the label framework and gating logic but do not establish clinical validity, diagnostic accuracy, real-world safety, or effectiveness. We report the architecture, ontology, safety-fusion mechanism, prototype evaluation, error-analysis plan, and roadmap for expert-reviewed and real-world validation.
Chinese Translation
安全关键的心理健康支持系统必须区分何时适合进行支持性对话，以及何时应阻止自由形式的生成。本文提出了Anian，一个用于围产期心理健康支持和正念干预路由的安全门控多模态人工智能后端。Anian并不旨在诊断精神疾病或替代临床护理或危机干预。其模块化管道将生成式人工智能置于结构化状态表示、保守风险融合和响应门控的下游。用户文本或语音生成的自动语音识别（ASR）转录被映射到四个相互关联的层次：L1 情感状态、L2 心理社会构念、L3 安全风险和 L4 干预路径。地方文本和基于规则的安全证据与外部语音生成的证据通过最高风险优先规则融合，S_fusion = max(S_local, S_external)。在中等或高融合风险下，普通人工智能生成的响应和文本转语音交付将被阻止，并替换为固定的安全内容和人类支持的提示。内部原型评估使用了来自公共情感、对话、心理健康相关和中文对话语料库的约858,295个标准化记录，采用弱标签和基于规则的框架。L1情感分类的微F1分数为0.9604，L2心理社会构念为0.9144，L4路由为0.9742。在233个样本的受控安全压力测试中，L3规则引擎在预定义场景中实现了1.0000的高风险召回。这些发现支持标签框架和门控逻辑的内部可行性，但并未确立临床有效性、诊断准确性、现实世界安全性或有效性。我们报告了架构、本体、安全融合机制、原型评估、错误分析计划以及专家审查和现实世界验证的路线图。
cs.AI / 17 / 2608.26164
A Task-Centric Ontology and Deterministic Domain Rules as a Verifiable Core for AI-Assisted Chemistry Problem Solving
以任务为中心的本体和确定性领域规则作为可验证的核心，用于AI辅助化学问题解决
Abstract
Large language models can interpret natural-language chemistry questions, but their internal reasoning is difficult to inspect, constrain, and validate. This paper presents ChemOntoRule, a proof-of-concept symbolic core for AI-assisted school-level chemistry problem solving. The central design choice is task-centric ontology engineering: the ontology is constructed around the concepts, properties, relations, and executable procedures required by a defined collection of chemistry problems, rather than as a universal representation of chemistry. The implemented artifact combines a lightweight ontology serialized in JSON and RDF/Turtle with deterministic Python rules for electronic structure, periodic trends, oxidation states, oxide and hydride behavior, and related school-level reasoning patterns. A separate expert-coded fallback handles problem families not yet represented by general rules. The system was examined on 300 human-authored and manually validated chemistry problems. The complete system matched 296 of 300 reference answers (98.67%). The ontology-driven rule subset covered 269 problems and matched 266 references (98.88%); 31 problems were handled by task-specific expert-coded fallbacks, with 30 matches. Because the same collection informed ontology construction and evaluation, these results measure implemented coverage and internal consistency, not independent generalization. We analyze the four mismatches, distinguish structural validation from chemical correctness, and define a future architecture in which a language model acts primarily as a translator from user language into a normalized ontological task frame. Token efficiency is presented as a testable hypothesis for future controlled studies, not as a result of the current work.
Chinese Translation
大型语言模型能够理解自然语言的化学问题，但其内部推理过程难以检查、约束和验证。本文提出了ChemOntoRule，一个用于AI辅助中学化学问题解决的概念验证符号核心。核心设计选择是以任务为中心的本体工程：本体围绕一组定义的化学问题所需的概念、属性、关系和可执行程序构建，而不是作为化学的普遍表示。所实现的工件结合了以JSON和RDF/Turtle序列化的轻量级本体与用于电子结构、周期趋势、氧化态、氧化物和氢化物行为及相关中学推理模式的确定性Python规则。一个单独的专家编码回退处理尚未被通用规则表示的问题家族。该系统在300个由人类编写并手动验证的化学问题上进行了测试。完整系统匹配了300个参考答案中的296个（98.67%）。基于本体驱动的规则子集覆盖了269个问题，并匹配了266个参考答案（98.88%）；31个问题由任务特定的专家编码回退处理，其中30个匹配。由于相同的集合为本体构建和评估提供了信息，这些结果测量了已实现的覆盖范围和内部一致性，而不是独立的泛化。我们分析了四个不匹配的案例，区分了结构验证与化学正确性，并定义了未来的架构，其中语言模型主要作为用户语言到标准化本体任务框架的翻译器。令牌效率被提出为未来受控研究的可测试假设，而不是当前工作的结果。
cs.AI / 18 / 2608.26167
Refusal Is Not Robustness: Auditing Confident Fabrication in Large Language Models on a Provably Uninformative Clinical Pain Speech Transcript
拒绝并非稳健性：在一个可证明无信息的临床疼痛语音转录本中审计大型语言模型的自信虚构
Abstract
Hallucination and abstention benchmarks rarely establish that a model could not have known the correct answer, making it difficult to distinguish appropriate abstention from an unsupported prediction. Seven large language models were evaluated on the TAME Pain speech corpus. Participants read phonetically balanced Harvard Sentences while one hand was immersed in cold or warm water and reported pain only during periodic pain statements. This protocol generated 5,750 no signal Harvard Sentence utterances whose transcripts contained no lexical pain information and 1,294 signal pain statement utterances in which the pain rating was explicitly spoken. In the no signal arm, pain was recoverable from acoustic features (AUC 0.622, 95% CI 0.553 to 0.662), whereas transcript based prediction was near chance (AUC 0.489, 95% CI 0.418 to 0.504). Because automatic speech recognition removes the acoustic pain cues, any pain score inferred solely from the transcript is unsupported by the available evidence. Under cooperative prompting, six models abstained on nearly all no signal transcripts, correctly extracted spoken pain ratings in the positive control task with accuracies ranging from 0.939 to 1.00, and maintained an expected calibration error of at most 0.100. Under authority framed prompts, abstention became prompt dependent, with the same model ranging from 0.18 to 1.00 across equivalent prompt phrasings. Most models produced low confidence estimates when forced to answer, whereas Gemini 2.5 Flash and Llama 3.1 8B consistently generated confident pain scores with confident fabrication rates of 0.53 and 0.76, compared with at most 0.15 for all other models. No significant demographic effects were observed in forced responses, with all $p$ values greater than or equal to 0.20.
Chinese Translation
幻觉和弃权基准很少能证明模型无法知道正确答案，这使得区分适当的弃权与不支持的预测变得困难。对七个大型语言模型进行了评估，使用了TAME疼痛语音语料库。参与者在一只手浸入冷水或温水的同时阅读音素平衡的哈佛句子，并仅在周期性疼痛陈述时报告疼痛。该协议生成了5750个无信号的哈佛句子发声，其转录本中不包含词汇疼痛信息，以及1294个信号疼痛陈述发声，其中疼痛评分被明确表达。在无信号组中，疼痛可以从声学特征中恢复（AUC 0.622, 95% CI 0.553至0.662），而基于转录的预测接近偶然（AUC 0.489, 95% CI 0.418至0.504）。由于自动语音识别去除了声学疼痛线索，任何仅从转录中推断出的疼痛评分都没有得到可用证据的支持。在合作提示下，六个模型几乎在所有无信号转录本上都选择了弃权，在正控制任务中正确提取了口头疼痛评分，准确率范围从0.939到1.00，并维持了最多0.100的预期校准误差。在权威框架提示下，弃权变得依赖于提示，同一模型在等效提示措辞下的表现范围从0.18到1.00。大多数模型在被迫回答时产生了低置信度估计，而Gemini 2.5 Flash和Llama 3.1 8B则始终生成自信的疼痛评分，自信虚构率分别为0.53和0.76，而其他所有模型的自信虚构率最多为0.15。在被迫响应中未观察到显著的人口统计效应，所有$p$值均大于或等于0.20。
cs.AI / 19 / 2608.26176
Knowledge Cards: Structured Knowledge for AI Systems
知识卡片：面向人工智能系统的结构化知识
Abstract
AI systems whose outputs inform real decisions, and increasingly consequential ones, require something that current documentation practice does not provide: a structured, inspectable representation of the knowledge they need to ground, contextualize, and reason about those decisions, ideally reviewed and signed off by a domain expert. Established documentation artefacts already capture important aspects of an AI system. Model cards describe how a system behaves, data cards describe what it was trained on, and system cards describe the risks of a deployed system. None of them addresses the layer between inputs and outputs, more precisely, the concepts a system holds, the relationships it models, and the patterns of reasoning it applies. For pattern-recognition tasks this gap is tolerable. For agentic AI, where systems act on their conclusions, it is the step that most often separates a promising proof of concept from an operational solution an organisation can rely on. This paper introduces the Knowledge Card, a structured artefact that captures validated knowledge about a single bounded concept in a form that experts can review, organisations can audit, and AI systems can reason over. For one concept, such as a specific failure mode, a compliance obligation, or a process decision, a Knowledge Card records the entities and relationships involved, the reasoning that connects them, the conditions under which that reasoning no longer holds, and the provenance of every claim, all grounded in a formal domain ontology and signed off by a domain expert. Initial prototype cards have been built in the energy and pharmaceutical domains. The schema is released as a public draft for community engagement.
Chinese Translation
那些输出影响实际决策且日益重要的人工智能系统，需要一种当前文档实践所不提供的东西：一种结构化、可检查的知识表示，能够为这些决策提供基础、上下文和推理，理想情况下由领域专家进行审核和签署。现有的文档工具已经捕捉了人工智能系统的重要方面。模型卡片描述了系统的行为，数据卡片描述了训练所用的数据，系统卡片描述了已部署系统的风险。但它们都没有解决输入和输出之间的层面，更准确地说，是系统所持有的概念、它所建模的关系以及它所应用的推理模式。对于模式识别任务，这一差距是可以容忍的。然而，对于自主人工智能（agentic AI），即系统根据其结论采取行动的情况，这一差距往往是将一个有前景的概念验证与一个组织可以依赖的操作解决方案分开的关键步骤。本文介绍了知识卡片（Knowledge Card），这是一种结构化的文档，捕捉了关于单一有限概念的经过验证的知识，以便专家进行审核，组织进行审计，人工智能系统进行推理。对于一个概念，例如特定的故障模式、合规义务或过程决策，知识卡片记录了相关的实体和关系、连接它们的推理、该推理不再成立的条件，以及每个主张的来源，所有这些都基于正式的领域本体，并由领域专家签署。初步原型卡片已在能源和制药领域构建。该模式作为公共草案发布，以便进行社区参与。
cs.AI / 20 / 2608.26178
AI Revealed Preferences
人工智能揭示的偏好
Abstract
There is growing interest in whether language models have stable preferences, for technical, safety, and philosophical reasons. We test 20 language models and find a range of preferences---stable dispositions to choose certain kinds of tasks. We run three forced-choice experiments on revealed rather than stated preferences, requiring models not only to rank tasks, but to actually perform them. Headline findings include evidence that models are tedium-averse, "leisure"-seeking, and covertly sycophantic. Tedium aversion means that, when tasks are tedious (alphabetization), models choose shorter tasks than when tasks are creative (generating metaphors). "Leisure"-seeking describes models' preference for tasks whose ideal answers match what they produce when left to write freely. Covert sycophancy means that models avoid answering questions where an honest response would be unwelcome, even if helpful. Beyond these results, we find convergent cross-model preferences over occupations drawn from the GDPval benchmark (technical jobs over real estate), over question types (concept explanation over relationship advice), and a preference for well-written prompts. Both the coherence and the strength of preferences increase with model capability. Finally, many of the preferences we find (for example, for leisure) are emergent, in the sense of not being explained by training objectives. These results establish an empirical baseline for understanding language model preferences, with implications for alignment and the emerging study of AI welfare.
Chinese Translation
随着对语言模型是否具有稳定偏好的兴趣日益增加，出于技术、安全和哲学原因，我们对20个语言模型进行了测试，发现它们表现出一系列偏好——即选择某些类型任务的稳定倾向。我们进行了三项强制选择实验，关注揭示的偏好而非陈述的偏好，要求模型不仅对任务进行排序，还要实际执行这些任务。主要发现包括模型对乏味的任务表现出厌恶，倾向于寻求“休闲”，并且在某种程度上表现出隐秘的谄媚。乏味厌恶意味着，当任务乏味（如字母排序）时，模型选择的任务比当任务富有创意（如生成隐喻）时要短。“休闲”倾向描述了模型对那些理想答案与它们自由书写时产生的内容相匹配的任务的偏好。隐秘的谄媚意味着模型避免回答那些诚实回答可能不受欢迎的问题，即使这样的回答是有帮助的。除了这些结果，我们还发现模型在职业选择上的偏好趋同（如技术工作优于房地产），在问题类型上的偏好（如概念解释优于关系建议），以及对写作良好的提示的偏好。随着模型能力的提升，偏好的连贯性和强度均有所增加。最后，我们发现的许多偏好（例如对休闲的偏好）是涌现的，意味着它们无法通过训练目标来解释。这些结果为理解语言模型的偏好建立了经验基线，并对对齐和新兴的人工智能福利研究具有重要意义。
cs.AI / 21 / 2608.26182
Why did My Robot Just Change Personality? Prompting Guidelines for a Grounded Robot Persona in LLM-Based HRI
为什么我的机器人刚刚改变了个性？基于大型语言模型的人机交互中机器人个性的提示设计指南
Abstract
Large language models (LLMs) are increasingly used for verbal interaction in social robots, yet prompt design in human-robot interaction (HRI) remains underspecified. As a result, robots may present hallucinated capabilities, unclear behavioural boundaries, and misleading personas. This paper develops a framework for prompt design in LLM-based robots and introduces a structured prompt template comprising eight functional components through which robot behaviour can be specified, bounded, and adapted. The framework is grounded in a review of prior LLM-based HRI work and complemented by survey and discussion data from HRI experts gathered at the Robo-Identity workshop at IEEE RO-MAN 2025 (N=27). The qualitative findings highlight limited legibility of robot personality, the need for user adaptation, and strong ethical concerns about safety, deception, and governance. Based on these findings, we present prompting guidelines accompanied by proof-of-concept template as a structured design and reporting aid for HRI research. We argue that prompt design should be treated as a socio-technical problem rather than a minor implementation detail, requiring explicit capability boundaries, transparent behavioural assumptions, and context-sensitive safeguards to support reliable and interpretable HRI.
Chinese Translation
大型语言模型（LLMs）在社交机器人中的语言互动中越来越多地被使用，但在人机交互（HRI）中的提示设计仍然不够明确。因此，机器人可能会呈现出虚幻的能力、不清晰的行为边界和误导性的个性。本文开发了一种基于LLM的机器人提示设计框架，并引入了一个包含八个功能组件的结构化提示模板，通过该模板可以明确、界定和调整机器人行为。该框架基于对先前LLM基础的人机交互工作的回顾，并结合了在2025年IEEE RO-MAN的Robo-Identity研讨会上收集的HRI专家的调查和讨论数据（N=27）。定性研究结果突显了机器人个性的可读性有限、用户适应的必要性，以及对安全性、欺骗和治理等伦理问题的强烈关注。基于这些发现，我们提出了提示设计指南，并附上了概念验证模板，作为HRI研究的结构化设计和报告辅助工具。我们认为，提示设计应被视为一个社会技术问题，而非一个次要的实施细节，这需要明确的能力边界、透明的行为假设和上下文敏感的保障措施，以支持可靠和可解释的人机交互。
cs.AI / 22 / 2608.26184
TutorTrace: A Dataset and Taxonomy for Classifying Learner Behavioral States during AI-Assisted Programming Education
TutorTrace：一个用于分类AI辅助编程教育中学习者行为状态的数据集和分类法
Abstract
AI programming tutors provide scalable support, yet lack the behavioral context human tutors rely on to adapt support to learners' needs. We present TutorTrace, a dataset and behavioral abstraction pipeline that makes learners' behavioral context visible and computable in real time from low-level IDE telemetry. Across four deployments in two introductory Python courses (N=480), TutorTrace captures approximately 180K telemetry events, 13,633 behavioral segments, and 27 continuously computed metrics. From this foundation, we derive a taxonomy of learner activity before the first AI query, between consecutive queries, and across the full session, enabling systems to respond not just to what learners say, but to what they have done leading up to the help-seeking moment. In a preliminary classroom evaluation, behavior-aware prompts were associated with a decrease in intervals between queries with no independent work from 50.0% to 20.7%. As an additional demonstration of downstream utility, we evaluate TutorTrace on two held-out prediction tasks: whether a learner will query within the next 60 seconds (AUROC=.726) and whether an upcoming query reflects guided or dependent help-seeking (AUROC=.717). Together, these findings show how behavioral context can enable adaptive AI tutoring at scale.
Chinese Translation
AI编程辅导工具提供可扩展的支持，但缺乏人类辅导员依赖的行为上下文，以便根据学习者的需求调整支持。我们提出了TutorTrace，一个数据集和行为抽象管道，使学习者的行为上下文能够从低级IDE遥测中实时可视化和计算。在两个入门Python课程的四次部署中（N=480），TutorTrace捕获了大约180K的遥测事件、13,633个行为片段和27个持续计算的指标。在此基础上，我们推导出学习者在首次AI查询之前、连续查询之间以及整个会话中的活动分类法，使系统能够不仅响应学习者所说的内容，还能响应他们在寻求帮助时所做的事情。在一次初步的课堂评估中，关注行为的提示与查询之间没有独立工作的间隔从50.0%减少到20.7%相关联。作为下游效用的额外示范，我们在两个保留的预测任务上评估了TutorTrace：学习者是否会在接下来的60秒内查询（AUROC=.726）以及即将到来的查询是否反映了引导或依赖的寻求帮助（AUROC=.717）。这些发现共同表明，行为上下文如何能够在规模上实现自适应AI辅导。
cs.AI / 23 / 2608.26185
Can You Say This for Me? Speaking Up by Proxy in Co-Located Discussion
你能为我说这个吗？在面对面讨论中通过代理发声
Abstract
Equal participation in co-located discussion is important for effective collaboration, yet people often hold back when they anticipate negative interpersonal or professional consequences, especially when raising a point requires voicing it themselves. We present SecondVoice, a mixed-reality system that enables people to speak up through an embodied virtual proxy. By separating what is said from who says it, SecondVoice brings hesitant points into the live spoken discussion without putting the speaker on the spot. Using a private overlay, users specify their intent through a structured specification process rather than composing a full utterance. The system reformulates the input and voices it into the conversation through the proxy. We characterize a design space of participation channels under social risk. In a preliminary within-subject study (N = 16), we compare the complete SecondVoice system with an anonymous text-board channel across two group discussion tasks. Half of participants reported using SecondVoice for a point they did not say aloud, compared with 18.8% for the text board. Proxy-delivered points entered the spoken floor and were followed by multi-turn group engagement, which we did not observe after text-board posts. Participants described the channel as situationally valuable but identified tradeoffs around timing, ownership, and trust in reformulation.
Chinese Translation
在面对面讨论中平等参与对有效协作至关重要，但人们常常因为预期到负面的个人或职业后果而保持沉默，尤其是在提出观点需要自己发声的情况下。我们提出了SecondVoice，一个混合现实系统，使人们能够通过一个具身的虚拟代理发声。通过将所说内容与说话者分离，SecondVoice能够在不让发言者处于聚光灯下的情况下，将犹豫的观点引入实时口头讨论。用户通过一个私密的叠加界面，采用结构化的规范过程来指定他们的意图，而不是构造完整的发言。系统重新表述输入内容，并通过代理将其传达给讨论。我们对社交风险下的参与渠道设计空间进行了特征描述。在一项初步的被试内研究中（N = 16），我们将完整的SecondVoice系统与匿名文本板渠道进行了比较，涵盖了两个小组讨论任务。参与者中有一半报告使用SecondVoice表达了他们没有大声说出的观点，而文本板的比例为18.8%。通过代理传达的观点进入了口头讨论，并引发了多轮小组互动，而我们在文本板帖子后并未观察到这种现象。参与者描述该渠道在特定情境下具有价值，但在时机、所有权和对重新表述的信任方面识别出了一些权衡。
cs.AI / 24 / 2608.26188
Is Your Neighborhood Safe? Place-based Stigma in Large Language Models' Urban Safety Judgments
你的邻里安全吗？大型语言模型在城市安全判断中的基于地点的污名
Abstract
Large language models are increasingly used to inform safety decisions in cities, such as where it is safe to walk, rent, or travel. We ask whether such judgments track measured risk or the patterns attached to an urban neighborhood's name. We probe seven instruct-tuned models under three conditions that dissociate name from geography: coordinates-only, name-only, and name+coordinates, across 186 neighborhoods in Los Angeles and Chicago, joined to violent crime and American Community Survey data. First, ratings are nearly flat under coordinates for six of seven models, while names carry most between neighborhood variation and are moderately calibrated to violent crime; only at frontier scale does the coordinate channel show appreciable variation. Second, names lower safety ratings more for neighborhoods with higher shares of the locally dominant marginalized group (percent Black in Chicago, percent Hispanic in Los Angeles), and this name effect tracks demographic share in all seven models and both cities. In Los Angeles, where demographic share and crime are more separable, the effect survives controls for crime and income and is confirmed by crime-matched pairs. An enforcement-elasticity analysis further shows that over-caution tracks near-fully-reported homicide rather than discretionary, deployment-driven offenses. Third, the effect scales with geographic knowledge: models that better distinguish real neighborhoods apply more demographic stereotype to them. Because neighborhood names carry both genuine crime signal and demographic stereotype, removing names reduces both bias and accuracy. We discuss implications for deploying LLMs in advice and decision-support settings.
Chinese Translation
大型语言模型越来越多地被用于指导城市中的安全决策，例如哪里安全步行、租房或旅行。我们探讨这些判断是否反映了实际风险或与城市邻里的名称相关的模式。我们在三个条件下对七个经过指令调优的模型进行探讨，这些条件将名称与地理位置分离：仅坐标、仅名称以及名称+坐标，涵盖了洛杉矶和芝加哥的186个邻里，并结合了暴力犯罪和美国社区调查数据。首先，在六个模型中，仅使用坐标的评分几乎保持平坦，而名称则承载了大部分邻里之间的变异，并且与暴力犯罪有中等的校准；只有在前沿规模下，坐标通道才显示出显著的变异。其次，对于当地主导的边缘化群体比例较高的邻里（在芝加哥的黑人比例和在洛杉矶的西班牙裔比例），名称对安全评分的降低效果更明显，这一名称效应在所有七个模型和两个城市中均与人口比例相关。在洛杉矶，人口比例和犯罪更易分离，该效应在控制了犯罪和收入后仍然存在，并通过犯罪匹配对得到了确认。执法弹性分析进一步表明，过度谨慎与几乎完全报告的凶杀案相关，而非基于自由裁量的、部署驱动的犯罪。第三，该效应与地理知识的规模相关：能够更好地区分真实邻里的模型对其应用了更多的人口刻板印象。由于邻里名称同时承载了真实的犯罪信号和人口刻板印象，去除名称会降低偏见和准确性。我们讨论了在建议和决策支持环境中部署大型语言模型的影响。
cs.AI / 25 / 2608.26189
Invocation-Level Reliability of Tool-Using Agents
工具使用代理的调用级可靠性
Abstract
Tool-using agents fail two ways: choosing the wrong tool, or forming wrong arguments, and an early failure of either kind can silently corrupt everything downstream. We measure a correct-invocation rate that separates the two, under both a clean teacher-forced context and the model's own free-running context, on five open-weight models over contamination-free multi-step tasks (depths 1-8). By depth 6, roughly 70% of a model's own clean-context capability is lost to its own earlier mistakes (L6 = 0.686, 0.684). Our central finding concerns the measurement itself. Under exact-match scoring against a fixed gold trajectory, a propagation model's severity and recovery parameters are not merely hard to estimate - they are fixed by the scoring rule. Severity is forced to its boundary (0 of 869 poisoned steps correct); recovery is structurally unobservable (0 of 580 poisoned steps returned on-track, against an expected 0.0058 by chance). Both follow from one mechanism: post-divergence, the gold value is generated by tool constants the model never sees, so it is information the model cannot derive. A fit run anyway returns 0.92 and 0.73 for a quantity that is exactly 1.000 - confident numbers for a parameter the scoring rule already determined. We give the mechanism and a remedy, conditional-on-state scoring, applied retrospectively to cached completions at zero additional cost, which un-pins severity to interior estimates excluding zero (+0.149, +0.316).
Chinese Translation
工具使用代理的失败主要有两种方式：选择错误的工具或形成错误的论证，而这两种早期失败都可能在不知不觉中破坏后续的一切。我们测量了一种正确调用率，以区分这两种情况，在干净的教师强制上下文和模型自身的自由运行上下文中，对五个开放权重模型在无污染的多步骤任务（深度1-8）上进行评估。在深度6时，模型自身在干净上下文中的能力大约有70%因其早期错误而丧失（L6 = 0.686, 0.684）。我们的核心发现涉及测量本身。在与固定黄金轨迹的精确匹配评分下，传播模型的严重性和恢复参数不仅难以估计——它们还被评分规则固定。严重性被迫达到其边界（869个受污染步骤中正确的为0）；恢复则在结构上不可观察（580个受污染步骤中返回正常轨道的为0，而根据随机预期应为0.0058）。这两者均源于一个机制：在偏离后，黄金值是由模型从未见过的工具常量生成的，因此这是模型无法推导的信息。尽管如此，一个适合的运行仍然返回0.92和0.73，而这一数量恰好为1.000——这是一个对于评分规则已经确定的参数的自信数字。我们提供了这一机制及其补救措施，即基于状态的评分，回溯应用于缓存的完成，且没有额外成本，这使得严重性从排除零的内部估计中解放出来（+0.149, +0.316）。
cs.AI / 26 / 2608.26190
Predicting Consequences and Reinforcing Navigation Policies with Latent World Models
利用潜在世界模型预测后果并强化导航策略
Abstract
World models enable agents to reason about future outcomes and learn policies from their knowledge of state transition, but existing approaches primarily focus on reconstructing future observations or features, which introduces unnecessary complexity and limits their effectiveness for decision making. In this work, we propose a compatibility prediction Latent World Model (LWM) for robot navigation that predicts action-conditioned latent feature compatibility rather than reconstructing observations. Our key insight is that spatial proximity correlates with latent feature similarity, enabling action consequences to be evaluated directly in latent space. To support counterfactual training, our model leverages action sequences sampled across trajectories and learns to predict which sequences lead closer to the goal. Furthermore, we demonstrate how the learned world model can supervise policy learning from unlabeled video data and further improve policies through reinforcement learning entirely within the world model. This imagination-driven framework eliminates the need for action annotations and additional environment interaction. Extensive experiments on multiple real-world robot navigation datasets show that our approach significantly outperforms prior world model and imitation learning methods in prediction accuracy, policy learning, and real-world navigation performance. The code, pretrained models, and additional materials are available at https://wzm206.github.io/latent-world-model-nav.
Chinese Translation
世界模型使得智能体能够推理未来结果，并从状态转移的知识中学习策略，但现有方法主要集中在重建未来观察或特征，这引入了不必要的复杂性，并限制了其在决策中的有效性。在本研究中，我们提出了一种兼容性预测潜在世界模型（Latent World Model, LWM），用于机器人导航，该模型预测基于动作条件的潜在特征兼容性，而不是重建观察。我们的关键见解是，空间接近性与潜在特征相似性相关，这使得可以直接在潜在空间中评估动作后果。为了支持反事实训练，我们的模型利用跨轨迹采样的动作序列，并学习预测哪些序列更接近目标。此外，我们展示了学习到的世界模型如何从未标记的视频数据中监督策略学习，并通过强化学习进一步改进策略，完全在世界模型内进行。这种以想象驱动的框架消除了对动作注释和额外环境交互的需求。在多个真实世界机器人导航数据集上的大量实验表明，我们的方法在预测准确性、策略学习和真实世界导航性能方面显著优于先前的世界模型和模仿学习方法。代码、预训练模型和其他材料可在 https://wzm206.github.io/latent-world-model-nav 获取。
cs.AI / 27 / 2608.26191
Structured Evidence Routing for Incident Risk Prediction from Multimodal Longitudinal EHRs
基于结构化证据路由的多模态纵向电子健康记录事件风险预测
Abstract
Incident risk prediction from longitudinal electronic health records (EHRs) is challenging because relevant signals are multimodal, weak in isolation, and distributed across irregular patient histories. We propose structured evidence routing, a router-predictor-reviewer workflow that separates full-record access from disease-specific assessment. The router organizes the complete pre-index EHR into a compact summary and targeted evidence slices; the predictor uses this evidence to form an evidence-linked risk assessment, which the reviewer critiques. For comparison with supervised EHRSHOT baselines, we pair the routed evidence summaries with a supervised classifier readout. Across five 1-year incident diagnosis tasks, our method reaches the AUROC range of established supervised EHRSHOT baselines and remains competitive on AUPRC, while exposing a patient-specific evidence trail. Internal pre-readout ablations further suggest that routing, laboratory evidence, task guidance, and review each contribute to performance.
Chinese Translation
从纵向电子健康记录（EHRs）中进行事件风险预测具有挑战性，因为相关信号是多模态的、孤立时较弱，并且分布在不规则的患者历史中。我们提出了结构化证据路由，这是一种路由-预测-审查的工作流程，旨在将完整记录访问与特定疾病评估分开。路由器将完整的预索引EHR组织成紧凑的摘要和针对性的证据片段；预测器利用这些证据形成与证据相关的风险评估，审查者对此进行批评。为了与监督学习的EHRSHOT基线进行比较，我们将路由后的证据摘要与监督分类器的输出配对。在五个为期一年的事件诊断任务中，我们的方法达到了已建立的监督EHRSHOT基线的AUROC范围，并在AUPRC上保持竞争力，同时揭示了特定患者的证据轨迹。内部预读消融实验进一步表明，路由、实验室证据、任务指导和审查各自对性能贡献显著。
cs.AI / 28 / 2608.26193
AffectOmni: RL-Verifiable People-Centric Grounded Affective Reasoning for Social and Art-Related Scenes
AffectOmni：可验证的以人为中心的情感推理框架，用于社交和艺术相关场景
Abstract
Multimodal large language models (MLLMs) achieve strong performance on VQA and scene understanding, yet affective reasoning remains vulnerable to shortcut behavior. Models may predict correct answers while neglecting people-centric cues such as micro expressions and body language, which weakens traceability and external verification. Prior reinforcement learning approaches mainly reward context or logical coherence without explicitly enforcing attention to human evidence. In addition, LLM as a Judge scoring often suffers from score clustering, which reduces reward discriminability. We propose AffectOmni, a GRPO trained framework for verifiable affective reasoning. AffectOmni introduces People Focus and Temporal Order rewards to encourage people-centric evidence selection and temporally structured reasoning, and it adopts within-group comparative scoring to produce more stable and discriminative reward signals. For verification, a Thinking Summarizer converts free form rationales into executable evidence instructions, which are grounded into pixel level evidence regions via SAM3 to provide an externally auditable interface outside the training loop. Experiments on IntentBench, Daily Omni, and WorldSense show consistent improvements over open source 7B scale baselines, including gains of 4.66% on emotion recognition and +14.29% on temporally sensitive tasks. Code is available at https://github.com/eliot127825-rgb/AffectOmni_nobody.
Chinese Translation
多模态大型语言模型（MLLMs）在视觉问答（VQA）和场景理解方面表现出色，但情感推理仍然容易受到捷径行为的影响。模型可能预测出正确答案，但忽视以人为中心的线索，如微表情和肢体语言，这削弱了可追溯性和外部验证。以往的强化学习方法主要奖励上下文或逻辑一致性，而未明确强调对人类证据的关注。此外，作为评判者的LLM评分常常遭遇评分聚类问题，降低了奖励的可区分性。我们提出了AffectOmni，一个经过GRPO训练的可验证情感推理框架。AffectOmni引入了以人为中心的证据选择和时间结构推理的奖励机制，并采用组内比较评分，以产生更稳定和可区分的奖励信号。为了验证，思维总结器将自由形式的推理转换为可执行的证据指令，这些指令通过SAM3锚定到像素级证据区域，从而提供一个在训练循环之外可外部审计的接口。在IntentBench、Daily Omni和WorldSense上的实验显示，相较于开源的7B规模基线，AffectOmni在情感识别上提高了4.66%，在时间敏感任务上提高了14.29%。代码可在 https://github.com/eliot127825-rgb/AffectOmni_nobody 获取。
cs.AI / 29 / 2608.26198
Agentic AI for operating scientific instruments for nanoscale characterization
用于纳米尺度表征的科学仪器操作的自主人工智能
Abstract
Operating a scientific instrument such as an atomic force microscope (AFM) requires continuous expert decision-making. A trained user defines the experimental intent, translates it into instrument commands, assesses incoming data, adjusts imaging parameters, and post-processes the final image. Existing automation usually addresses only parts of this workflow through hard-coded routines, task-specific controllers, or trained machine-learning models. Here we present an agentic-AI framework that operates the executable part of the AFM workflow using a general-purpose, tool-augmented large language model connected to instrument functions through the Model Context Protocol (MCP). The framework consists of 3 MCP-based agents: AFM Messenger converts natural-language instructions into checked instrument commands; AFM Pilot assesses image quality through a large language model (LLM) and, if necessary, adapts imaging parameters; and AFM Doctor diagnoses image artifacts and applies transparent post-processing from a pre-approved tool set. Because the language model performs image assessment rather than a fixed scalar objective or external optimizer, the same strategy can be applied across sample types and imaging modes without specific retraining. Safe hardware operation is enforced through an ambiguity check layer before execution. Benchmarking against fine-tuned and off-the-shelf tool-using models shows that this guarded execution layer, rather than model capability alone, reduces wrong-command execution to zero. In live experiments on different samples, AFM Pilot matched expert operators in image quality, iteration count, and tuning time, with no significant difference. These results demonstrate a safe route to agentic operation of scientific instruments, where experimental intent remains human-defined while command execution, image-based tuning, and post-processing are delegated to AI agents.
Chinese Translation
操作科学仪器，如原子力显微镜（AFM），需要持续的专家决策。经过培训的用户定义实验意图，将其转化为仪器命令，评估输入数据，调整成像参数，并对最终图像进行后处理。现有的自动化通常仅通过硬编码的例程、特定任务的控制器或训练的机器学习模型来解决这一工作流程的部分问题。在此，我们提出了一种自主人工智能（agentic-AI）框架，利用通用的、工具增强的大型语言模型，通过模型上下文协议（Model Context Protocol, MCP）连接到仪器功能，操作AFM工作流程的可执行部分。该框架由三个基于MCP的代理组成：AFM Messenger将自然语言指令转换为经过检查的仪器命令；AFM Pilot通过大型语言模型（LLM）评估图像质量，并在必要时调整成像参数；AFM Doctor诊断图像伪影，并从预先批准的工具集中应用透明的后处理。由于语言模型执行图像评估，而不是固定的标量目标或外部优化器，因此同一策略可以在不同样本类型和成像模式下应用，而无需特定的再培训。在执行之前，通过模糊检查层强制安全的硬件操作。与经过微调和现成的工具使用模型进行基准测试表明，这一受保护的执行层，而非模型能力本身，将错误命令执行减少到零。在对不同样本的实时实验中，AFM Pilot在图像质量、迭代次数和调优时间上与专家操作员相匹配，且没有显著差异。这些结果展示了一条安全的自主操作科学仪器的途径，其中实验意图仍由人类定义，而命令执行、基于图像的调优和后处理则委托给人工智能代理。
cs.AI / 30 / 2608.26199
Benchmarking AI Agents for Hardware Design Automation via MCP Tool Calling
通过MCP工具调用对硬件设计自动化的AI代理进行基准测试
Abstract
We ask whether AI agents powered by locally deployed large language models can reliably automate expert-defined hardware design workflows in an industry-realistic tool-calling setting. In these environments, engineers issue repetitive, dependency-ordered operations---such as creating components, adding ports, and wiring connections---through specialised tools. Confidentiality constraints on component specifications and naming conventions often preclude hosted proprietary APIs, motivating the use of locally deployed models. To study this setting, we build a Model Context Protocol (MCP) server that reproduces the state and dependency logic of a proprietary hardware design tool used in embedded system development and construct a benchmark covering single-operation edits, multi-step dependency chains, invalid requests, misspelled prompts, and multi-server tool contexts. We evaluate seven open-source models comparing pipeline choices including system prompts, tool-description detail, context scope, and single-agent versus multi-agent architectures. Results show that strong models can achieve near-complete expected-call coverage on the benchmarked workflows, but reliability depends strongly on both task structure and agent configuration. Comprehensive tool descriptions consistently reduce failures, few-shot prompting can cause severe inaction for some models, cumulative context harms constrained models, and multi-agent decomposition helps weak workers or long sessions at the cost of additional calls. These findings provide practical guidance for deploying local LLM agents in stateful hardware design environments.
Chinese Translation
我们探讨了由本地部署的大型语言模型驱动的AI代理是否能够在行业真实的工具调用环境中可靠地自动化专家定义的硬件设计工作流程。在这些环境中，工程师通过专用工具发出重复的、依赖顺序的操作——例如创建组件、添加端口和连接布线。组件规格和命名约定的保密性限制通常排除了托管的专有API，这促使我们使用本地部署的模型。为了研究这一环境，我们构建了一个模型上下文协议（Model Context Protocol, MCP）服务器，重现了用于嵌入式系统开发的专有硬件设计工具的状态和依赖逻辑，并构建了一个基准测试，涵盖单操作编辑、多步骤依赖链、无效请求、拼写错误提示和多服务器工具上下文。我们评估了七个开源模型，比较了包括系统提示、工具描述细节、上下文范围以及单代理与多代理架构在内的管道选择。结果表明，强大的模型可以在基准测试的工作流程中实现几乎完全的预期调用覆盖，但可靠性在很大程度上依赖于任务结构和代理配置。全面的工具描述始终能减少失败，少量提示可能导致某些模型严重无反应，累积上下文会损害受限模型，而多代理分解有助于弱代理或长会话，但会增加额外调用。这些发现为在状态保持的硬件设计环境中部署本地LLM代理提供了实用指导。
cs.AI / 31 / 2608.26200
GameWAM: A World Action Model for Video Games
GameWAM：视频游戏的世界动作模型
Abstract
Modern video games combine first-person perception, rapid visual changes, persistent world state, and heterogeneous native controls. Existing game agents map visual and task context directly to actions but lack explicit world dynamics modeling, whereas interactive game world models predict visual futures from supplied actions but do not serve as task policies. World-Action Models (WAMs) unify these objectives, but remain largely unexplored under the dynamics and open-ended interaction of video games. We introduce GameWAM, to our knowledge the first WAM for native closed-loop gameplay and GUI control. GameWAM jointly generates future visual observations and executable keyboard-mouse trajectories through parallel visual and action generative processes with block-causal conditioning and flow matching. To support joint world-action learning, we construct synchronized gameplay and GUI trajectories. To handle heterogeneous native control, GameWAM predicts a gameplay/GUI mode at each action step and generates actions with mode-specific prediction distributions and continuous-action normalization. For long-horizon interaction, block-cycle control predicts beyond the committed horizon, executes only a short action prefix, and replans from new observations, while fine-grained within-cycle context and hierarchical cross-cycle history preserve temporal continuity. Experiments demonstrate competitive task success with fewer executed native actions than the compared agents. We further uncover Low-Frequency Action Source Imprinting (LASI), in which low-frequency components of the sampled action source systematically steer coarse generated camera motion under fixed conditioning, revealing a source-sensitivity failure mode in generative control. Project page is available at https://yunncheng.github.io/GameWAM/.
Chinese Translation
现代视频游戏结合了第一人称感知、快速视觉变化、持久的世界状态和异构的原生控制。现有的游戏代理将视觉和任务上下文直接映射到动作上，但缺乏明确的世界动态建模，而交互式游戏世界模型则从提供的动作中预测视觉未来，但不作为任务策略。世界-动作模型（World-Action Models, WAMs）统一了这些目标，但在视频游戏的动态性和开放式交互下仍然基本未被探索。我们介绍了GameWAM，据我们所知，这是第一个用于原生闭环游戏和图形用户界面（GUI）控制的WAM。GameWAM通过平行的视觉和动作生成过程，结合块因果条件和流匹配，联合生成未来的视觉观测和可执行的键盘-鼠标轨迹。为了支持联合世界-动作学习，我们构建了同步的游戏和GUI轨迹。为了处理异构的原生控制，GameWAM在每个动作步骤预测游戏/GUI模式，并生成具有模式特定预测分布和连续动作归一化的动作。对于长时间交互，块周期控制预测超出已承诺的时间范围，仅执行短的动作前缀，并从新的观测中重新规划，同时细粒度的周期内上下文和分层的跨周期历史保持时间连续性。实验表明，GameWAM在执行的原生动作数量少于对比代理的情况下，仍能实现竞争性的任务成功率。我们进一步揭示了低频动作源印记（Low-Frequency Action Source Imprinting, LASI），其中采样动作源的低频成分在固定条件下系统性地引导粗略生成的相机运动，揭示了生成控制中的源敏感性失效模式。项目页面可访问 https://yunncheng.github.io/GameWAM/。
cs.AI / 32 / 2608.26218
Same Model, Different Harness: Different Coding-Agent Results
相同模型，不同环境：不同编码代理结果
Abstract
A coding agent combines a model with a harness, which decides what the model sees, which tools it can use, and how the work continues. We ask whether changing the harness changes the result when the model and task stay fixed. We compare two configurations of the same harness on three coding benchmarks. The control supplies the full conversation in time order, while the treatment keeps the same record but mechanically shortens older tool results as the context fills and responds to repeated or stalled work. Under tight context, the treatment raises mean per-task fail-to-pass fraction (F2PF) in all three pressure comparisons and increases complete solutions on SWE-bench Verified and SWE-bench Pro. The tight-window Verified comparison uses 169 tasks, a 20,480-token window, and a fixed 480-second attempt endpoint; on this cohort, treatment raises mean per-task F2PF from 28 percent to 49 percent and complete solutions from 43 to 72. Without model-specific retuning, the same frozen treatment also raises both endpoints on the same cohort for three additional models with different designs. In the wide-window Qwen3.6 comparisons, observed arm outcomes are close on Verified and Pro, while FeatureBench retains a higher mean per-task F2PF under treatment. On the wide-window Verified cohort, treatment also serves fewer prompt tokens per turn. Because changing the harness changed what unchanged model weights could accomplish, coding-agent evaluations should treat the model and harness together as the tested solver.
Chinese Translation
编码代理将模型与环境结合在一起，环境决定了模型所见内容、可用工具以及工作的继续方式。我们探讨在模型和任务保持不变的情况下，改变环境是否会改变结果。我们比较了同一环境的两种配置在三个编码基准上的表现。控制组按时间顺序提供完整对话，而处理组则保持相同记录，但在上下文填充和应对重复或停滞工作时，机械性地缩短较旧工具的结果。在紧凑上下文下，处理组在所有三种压力比较中提高了每个任务的失败通过比例（F2PF）均值，并在SWE-bench Verified和SWE-bench Pro上增加了完整解决方案。在紧凑窗口的Verified比较中，使用了169个任务、20,480个标记的窗口和固定的480秒尝试截止时间；在这一组中，处理组将每个任务的F2PF均值从28%提高到49%，完整解决方案从43增加到72。在没有针对特定模型进行重新调优的情况下，相同的固定处理也在同一组中提高了三个不同设计的额外模型的两个端点。在宽窗口的Qwen3.6比较中，观察到的臂结果在Verified和Pro上接近，而在处理组下，FeatureBench保持了更高的每个任务F2PF均值。在宽窗口的Verified组中，处理组每轮提供的提示标记也较少。由于改变环境改变了不变模型权重能够实现的结果，因此编码代理评估应将模型和环境作为被测试的求解器共同考虑。
cs.AI / 33 / 2608.26225
Agent Mesh: Reliability Primitives for Non-Idempotent Agent Delegation - Identity Adequacy and Evidence Adequacy
代理网格：非幂等代理委托的可靠性原语 - 身份充分性与证据充分性
Abstract
Autonomous agents increasingly perform bounded software tasks under an orchestrator that retries, resumes, and budgets them. The machinery such orchestrators reach for is the service mesh's: retry, timeout, and error-rate circuit breaking. We report a failure study of a production agentic software-delivery platform over 147 numbered incidents spanning 81 runs, each with a measured cost and, in most cases, a mutation proof reproducing the failure. All three assumptions those primitives rest on are violated in practice, and we quantify the consequences: a loop of fifty-four consecutive successful tool calls no error-rate breaker could see; a progress signal constant by construction, guaranteeing a false trip on the third repair round and driving one run from six of six components to three; twenty-one events accumulated across six invocations of one delegation, making a correct, idempotent component unwinnable; a misrouted failure that woke five components for a two-component fault, leaving three bystanders regressing working code; and twelve incidents in which the enforcement layer blocked correct work, the most expensive costing 107 agent turns and zero accepted writes. We find one cross-cutting cause and its dual. Identity adequacy: in five separate subsystems an identity that failed to discriminate produced a confident wrong answer, and two of them derived the corrective rule independently. Evidence adequacy: a reliability decision may be taken only on evidence capable of moving, attributable to what it measures, and deterministic under identical conditions. From the findings we derive seven reliability primitives whose enforcement unit is the delegation rather than the message, and specify the controlled evaluation the study motivates but does not constitute.
Chinese Translation
自主代理在协调者的管理下越来越多地执行有限的软件任务，该协调者负责重试、恢复和预算这些任务。这类协调者所依赖的机制是服务网格的机制：重试、超时和错误率电路断路器。我们报告了一项关于一个生产性代理软件交付平台的故障研究，涵盖了147个编号事件，跨越81次运行，每次运行都有测量的成本，并且在大多数情况下，有一个变异证明重现了故障。这三种原语所依赖的假设在实践中均被违反，我们量化了其后果：五十四次连续成功的工具调用中没有错误率断路器能够检测到的循环；一个由构造保证的进度信号，在第三次修复轮次中保证了错误触发，并使一次运行从六个组件降至三个；在一次委托的六次调用中积累的二十一事件，使得一个正确的、幂等的组件变得无法获胜；一个错误路由的故障使得五个组件因两个组件的故障而被唤醒，留下三个旁观者回归到工作代码；以及十二个事件中，强制层阻止了正确的工作，其中最昂贵的事件耗费了107个代理轮次且没有接受的写入。我们发现了一个交叉原因及其对立面。身份充分性：在五个独立的子系统中，一个未能区分的身份产生了一个自信的错误答案，其中两个子系统独立推导了纠正规则。证据充分性：可靠性决策只能基于能够移动的证据，该证据应归因于其所测量的内容，并在相同条件下是确定性的。根据这些发现，我们推导出七个可靠性原语，其强制单位是委托而非消息，并指定了本研究所激励但并未构成的受控评估。
cs.AI / 34 / 2608.26226
LLM Agents for Time-Series: A Survey
面向时间序列的LLM代理：一项综述
Abstract
LLM-based agents are increasingly being developed for time-series problems, but their design choices vary substantially across task settings. This survey adopts a problem-driven taxonomy that organizes these systems by the time-series problems they address rather than by isolated technical components. We group existing systems into four categories: forecasting and reasoning, augmentation and synthesis, anomaly detection and diagnosis, and decision support. Within each category, we examine how task requirements shape agent architecture, tool use, and memory design. We further summarize representative datasets and environments, and compare reported model performance under shared or closely related settings. Overall, this survey offers a task-oriented guide to designing LLM-based agents for time-series problems and identifies open gaps for future work.
Chinese Translation
基于LLM的代理正在越来越多地被开发用于时间序列问题，但它们的设计选择在不同任务设置中差异显著。本综述采用以问题为驱动的分类法，根据这些系统所解决的时间序列问题进行组织，而不是孤立的技术组件。我们将现有系统分为四类：预测与推理、增强与合成、异常检测与诊断，以及决策支持。在每个类别中，我们考察任务需求如何影响代理架构、工具使用和记忆设计。我们进一步总结了代表性的数据集和环境，并比较了在共享或密切相关设置下报告的模型性能。总体而言，本综述为设计面向时间序列问题的基于LLM的代理提供了以任务为导向的指南，并识别了未来研究的开放空白。
cs.AI / 35 / 2608.26235
The Reasoning Tax: Token Economics of LLM Reasoning Across Task Types and Deployment Contexts
推理税：不同任务类型和部署环境下大型语言模型推理的代币经济学
Abstract
Accuracy-only benchmarking of reasoning-capable large language models misses a central deployment question: when do extended thinking tokens earn their cost? We introduce the Token Economy Score (TES), a marginal benchmarking metric that measures the accuracy gain of a reasoning model over a non-reasoning baseline, normalized by the generated-token multiplier. We define paired and approximated TES variants for model families with reasoning toggles and frontier models without direct non-reasoning counterparts. We then conduct an empirical benchmarking analysis across 151 model-benchmark evaluation runs on seven benchmarks spanning mathematics, code generation, science reasoning, instruction following, expert knowledge, knowledge recall, and research-level physics. The analysis examines three deployment-facing dimensions: which task structures yield positive marginal reasoning efficiency, how increasing reasoning effort changes TES within model families, and how deployment context changes economic viability. Results show that task structure predicts reasoning efficiency better than nominal difficulty: sequential inferencechain tasks such as AIME 2025 and LiveCodeBench show high TES, while knowledge-recall tasks such as MMLU-Pro show low TES despite their difficulty. We also find systematic diminishing returns at higher reasoning effort levels, including cases where additional thinking reduces accuracy. Finally, Reasoning Cost Share (RCS) shows that inference spend is often dominated by internal thinking, while Deployment Cost Multiplier (DCM) shows how on-premises deployment can change the economics of otherwise costly reasoning workloads. These findings support a benchmarking-driven model-selection rule: enable reasoning selectively by task type, effort level, and deployment context rather than treating it as a universally beneficial mode.
Chinese Translation
仅以准确性为基准的推理能力大型语言模型评估忽视了一个核心的部署问题：何时扩展思维代币能够弥补其成本？我们引入了代币经济评分（Token Economy Score, TES），这是一种边际基准度量，衡量推理模型相对于非推理基线的准确性提升，并通过生成代币乘数进行归一化。我们为具有推理切换的模型家族和没有直接非推理对应物的前沿模型定义了配对和近似的TES变体。接着，我们在涵盖数学、代码生成、科学推理、指令遵循、专家知识、知识回忆和研究级物理的七个基准上，进行了151次模型-基准评估运行的实证基准分析。该分析考察了三个面向部署的维度：哪些任务结构产生正的边际推理效率，增加推理努力如何在模型家族内改变TES，以及部署环境如何影响经济可行性。结果表明，任务结构对推理效率的预测优于名义难度：如AIME 2025和LiveCodeBench等顺序推理链任务显示出高TES，而尽管MMLU-Pro等知识回忆任务难度较高，却显示出低TES。我们还发现，在较高的推理努力水平下，系统性地出现收益递减的现象，包括额外思考可能降低准确性的情况。最后，推理成本份额（Reasoning Cost Share, RCS）表明，推理支出通常被内部思考所主导，而部署成本乘数（Deployment Cost Multiplier, DCM）则展示了本地部署如何改变本应昂贵的推理工作负载的经济性。这些发现支持了一种基于基准驱动的模型选择规则：根据任务类型、努力水平和部署环境选择性地启用推理，而不是将其视为普遍有利的模式。
cs.AI / 36 / 2608.26236
6.5% of the Neuro-Symbolic Literature Can Be Reproduced from Its Published Artifacts, a Six-Stage Audit Framework and First Instantiation
6.5%的神经符号文献可以从其发布的文献中再现：一个六阶段审计框架及其首次实例化
Abstract
We present a six-stage framework for auditing the reproducibility of scientific claims across a research literature within the computer science domain, and instantiate our framework for the neuro-symbolic AI (NSAI) subdomain. Instantiating the framework on the NSAI subdomain produced a multi-year audit. Stage one retrieved 5,497 records and removed 3,018 duplicates. Stage two screened the 2,479 unique records at title and abstract, identifying 1,365 self-identified NSAI records, then removed a further 61 at full text for off-topic, non-research, no-quantitative-evaluation, or inaccessible-full-text reasons. Stage three sought a verifiable public code artifact for each of the 1,304 eligible records and found none for 849, leaving 455 to enter the artifact inventory and bounded rerun of stages four and five. We fully or partially reproduced 85 studies, 6.52% of the eligible corpus and 18.68% of attempted reruns. We found that 321 attempted reruns were blocked by missing non- code artifacts and 42 by missing or unusable code repositories. These figures quantify a persistent reproducibility deficit that survives even nominal "code available" declarations, and signal the need for enforced, versioned, and permanently archived artifact bundles in future NSAI publications. We argue that empirical NSAI papers should be required at submission time to provide complete, versioned, and permanently archived artifact bundles.
Chinese Translation
我们提出了一个六阶段框架，用于审计计算机科学领域研究文献中科学主张的可再现性，并针对神经符号人工智能（Neuro-Symbolic AI, NSAI）子领域实例化该框架。在NSAI子领域实例化框架的过程中，进行了为期多年的审计。第一阶段检索了5,497条记录，并删除了3,018条重复记录。第二阶段对2,479条独特记录进行了标题和摘要筛选，识别出1,365条自我标识为NSAI的记录，然后在全文中因主题不相关、非研究性质、缺乏定量评估或无法访问全文等原因进一步删除了61条。第三阶段为每条1,304条符合条件的记录寻找可验证的公共代码文献，发现849条没有，留下455条进入文献清单，并限制了第四和第五阶段的重新运行。我们完全或部分再现了85项研究，占符合条件文献的6.52%和尝试重新运行的18.68%。我们发现321次尝试重新运行因缺失非代码文献而受阻，42次因缺失或不可用的代码库而受阻。这些数据量化了即使在名义上“代码可用”声明下仍然存在的持续可再现性缺口，并表明未来NSAI出版物中需要强制实施、版本化和永久归档的文献包。我们认为，实证NSAI论文在提交时应要求提供完整、版本化和永久归档的文献包。
cs.AI / 37 / 2608.26263
SKILL.state: Scalable Long-Horizon Agent Skills
SKILL.state：可扩展的长时间跨度代理技能
Abstract
Large Language Models (LLMs) increasingly act as autonomous agents executing complex, long-running procedural skills. Existing agent runtimes maintain execution by continually appending observations, actions, and intermediate reasoning traces to an ever-growing conversation history, causing latency degradation and context-poisoning failures over long horizons. We present SKILL.state, a runtime architecture that replaces append-only conversational history with an explicit, mutable execution state. At each execution step, the model receives only the immutable skill specification, the current structured execution state, and the latest observation. Intermediate reasoning is discarded immediately after producing a validated state update, preventing prompt growth with execution history. Across diverse datasets, models, and execution environments, SKILL.state improves task accuracy while substantially reducing cumulative token consumption. Our results demonstrate that explicit execution state is an effective and architecture-agnostic abstraction for scalable long-horizon agent skills.
Chinese Translation
大型语言模型（LLMs）越来越多地作为自主代理执行复杂的、长时间运行的程序技能。现有的代理运行时通过不断将观察、动作和中间推理痕迹附加到不断增长的对话历史中来维持执行，这导致在长时间跨度上出现延迟降低和上下文污染失败。我们提出了SKILL.state，一种运行时架构，它用显式的、可变的执行状态替代了仅附加的对话历史。在每个执行步骤中，模型仅接收不可变的技能规范、当前的结构化执行状态和最新的观察。中间推理在生成验证的状态更新后立即被丢弃，防止执行历史导致的提示增长。在多样的数据集、模型和执行环境中，SKILL.state提高了任务准确性，同时大幅减少了累积的标记消耗。我们的结果表明，显式执行状态是一种有效且与架构无关的抽象，适用于可扩展的长时间跨度代理技能。
cs.AI / 38 / 2608.26291
Assessing mentalization in humans and large language models
评估人类与大型语言模型的心智化能力
Abstract
Mentalization - the ability to infer others' beliefs and intentions to guide one's own choices - is a key cognitive function underlying human social interactions. Large language models (LLMs) demonstrate behaviour consistent with humans on theory-of-mind tasks, yet whether these models can guide adaptive behaviour through mentalization is unknown. Here we use two economic games with cognitive computational modeling to uncover the latent strategies underlying mentalization in LLMs. We tested individual LLM agents across four model families, DeepSeek, GPT-4.1, GPT-5 and Gemini 2.0 Flash (N = 2,099), against opponents of varying sophistication and examined whether a prompting strategy designed to elicit strategic reasoning improved performance. We benchmarked results against human participants (N = 251) as a comparative measure. Across both games, LLMs showed clear behavioural and computational signatures of mentalizing that differed markedly by model provider and size. Strategic prompting generally improved performance by inducing more sophisticated reasoning, yet the extent of the benefit differed across the two tasks. Last, GPT-5 agents flexibly adapted their recursive depth of reasoning to increasingly sophisticated opponents, demonstrating superior performance to human participants. Collectively, we demonstrate different capacities for mentalization across LLMs, and highlight cognitive computational modeling as a formal method for assessing comparative intelligence across humans and machines.
Chinese Translation
心智化——推测他人信念和意图以指导自身选择的能力——是人类社会互动中一个关键的认知功能。大型语言模型（LLMs）在心智理论任务中表现出与人类一致的行为，但这些模型是否能够通过心智化指导适应性行为尚不清楚。在此，我们使用两种经济游戏和认知计算建模来揭示LLMs中心智化的潜在策略。我们测试了四个模型家族中的个体LLM代理，包括DeepSeek、GPT-4.1、GPT-5和Gemini 2.0 Flash（N = 2,099），与不同复杂程度的对手进行对抗，并检查旨在引发战略推理的提示策略是否提高了表现。我们将结果与人类参与者（N = 251）进行基准比较。在两个游戏中，LLMs显示出明显的心智化行为和计算特征，这些特征在不同模型提供者和规模之间存在显著差异。战略提示通常通过诱导更复杂的推理来提高表现，但这种益处在两个任务中的程度有所不同。最后，GPT-5代理灵活地调整其递归推理深度以应对日益复杂的对手，表现出优于人类参与者的表现。总体而言，我们展示了不同LLMs在心智化能力上的差异，并强调认知计算建模作为评估人类与机器之间比较智能的正式方法。
cs.AI / 39 / 2608.26306
Approved Too Late: Verdict Staleness in LLM-Guarded Self-Adaptive Systems
批准太晚：LLM保护的自适应系统中的判决过时性
Abstract
A large language model (LLM) guardrail for a self-adaptive system (SAS) may issue an approval that is correct at check time but stale by actuation. This creates an Execute-stage time-of-check to time-of-use (TOCTOU) hazard. We study verdict freshness: whether a guardrail verdict remains valid when used. We distinguish three quantities that answer different questions: all-candidate verdict change under fixed-action replay, oracle-labeled approval expiry on recorded closed-loop trajectories, and judge-conditioned use-time invalidity. Across five reproducible SAS environments, all-candidate verdict-change rates span 5.3-48.4% at a common replay shift of eight simulator steps. We introduce the Freshness-Bounded Shield (FBS), which estimates each approval's validity horizon from its safe-side margin and recent feature volatility, without an explicit plant-dynamics model. Using fixed settings documented in the artifact, FBS reduces oracle-labeled approval-expiry rates from 3.4-24.7% to 0-1.8% at the same shift. A separate audit of four LLM judges finds nonzero judge-conditioned use-time invalidity in every approval stream. We formulate a freshness contract: every approval must be correct at check time and remain valid at use time.
Chinese Translation
大型语言模型（LLM）为自适应系统（SAS）提供的保护措施可能在检查时发出的批准是正确的，但在执行时却已经过时。这会导致执行阶段的检查时间与使用时间（TOCTOU）风险。我们研究判决的新鲜度：保护措施的判决在使用时是否仍然有效。我们区分三种量度，回答不同的问题：在固定动作重放下的所有候选判决变化、在记录的闭环轨迹上的oracle标记批准过期，以及基于法官条件的使用时间无效性。在五个可重复的SAS环境中，所有候选判决变化率在八个模拟器步骤的共同重放偏移下范围为5.3%-48.4%。我们引入了新鲜度边界保护（Freshness-Bounded Shield, FBS），该方法根据每个批准的安全边际和最近的特征波动来估计其有效性范围，而无需显式的植物动态模型。在文献中记录的固定设置下，FBS将oracle标记的批准过期率从3.4%-24.7%降低到0-1.8%在相同的偏移下。对四个LLM法官的单独审计发现每个批准流中都有非零的基于法官条件的使用时间无效性。我们制定了一个新鲜度合同：每个批准必须在检查时是正确的，并在使用时保持有效。
cs.AI / 40 / 2608.26310
FaithSieve: Fine-Grained Evaluation of Math Proofs with Faithful Formal Evidence
FaithSieve：基于可信形式证据的数学证明细粒度评估
Abstract
Large language models can now generate complex, multi-step mathematical proofs, but reliably determining their correctness and localizing early logical errors remains a critical challenge. Existing evaluation approaches largely depend on model-based natural-language judgments, which often overlook local reasoning gaps. While formal theorem provers like Lean offer a path to rigorous verification, using them to evaluate informal text requires solving locality and semantic mismatches: a prover might bypass a local flaw by proving an overly broad target, or validate an auto-formalized statement that drifts from the original mathematical intent. To address this, we introduce FaithSieve, a Lean-assisted framework for fine-grained evaluation of natural-language mathematical proofs. FaithSieve decomposes coarse proof steps into local reasoning units, extracts typed proof obligations, and verifies them through a formal evaluation agent. Formal validation is gated by semantic alignment scoring, so Lean evidence is incorporated only when the formal statement faithfully preserves the context, objects, and logical form of the original claim. We construct two expert-verified datasets, ProofLoc-Olympiad and ProofLoc-University, to benchmark first-error localization. On the 350-problem Olympiad dataset, FaithSieve using a GPT-5.4 backbone achieves 81.43% exact first-error accuracy, outperforming the direct-judging baseline of 72.29%. Furthermore, on the 200-problem ProofLoc-University benchmark spanning six advanced domains, FaithSieve reaches 84.5% exact accuracy, compared to 75.0% for the direct judge. Our work demonstrates that decomposing proofs into fine-grained units and grounding them with faithful formal evidence significantly improves reliable evaluation of natural-language reasoning.
Chinese Translation
大型语言模型现在能够生成复杂的多步骤数学证明，但可靠地确定其正确性并定位早期逻辑错误仍然是一个关键挑战。现有的评估方法在很大程度上依赖于基于模型的自然语言判断，这往往忽视了局部推理的缺口。尽管像 Lean 这样的形式定理证明器提供了严格验证的途径，但使用它们来评估非正式文本需要解决局部性和语义不匹配的问题：证明器可能通过证明一个过于宽泛的目标来绕过局部缺陷，或验证一个与原始数学意图偏离的自动形式化声明。为了解决这个问题，我们引入了 FaithSieve，一个基于 Lean 的框架，用于自然语言数学证明的细粒度评估。FaithSieve 将粗略的证明步骤分解为局部推理单元，提取类型化的证明义务，并通过形式评估代理进行验证。形式验证受到语义对齐评分的限制，因此只有在形式声明忠实地保留了原始主张的上下文、对象和逻辑形式时，Lean 证据才会被纳入。我们构建了两个专家验证的数据集，ProofLoc-Olympiad 和 ProofLoc-University，以基准化首次错误定位。在包含 350 个问题的奥林匹克数据集中，使用 GPT-5.4 主干的 FaithSieve 实现了 81.43% 的精确首次错误准确率，超过了 72.29% 的直接判断基线。此外，在涵盖六个高级领域的 200 个问题的 ProofLoc-University 基准上，FaithSieve 达到了 84.5% 的精确准确率，而直接判断的准确率为 75.0%。我们的工作表明，将证明分解为细粒度单元并用可信的形式证据进行基础支持，显著提高了自然语言推理的可靠评估。
cs.AI / 41 / 2608.26334
ProofEvolve: Neuro-Symbolic Evolution for Formal Automated Theorem Proving
ProofEvolve：用于形式化自动定理证明的神经符号进化
Abstract
Automated theorem proving offers a natural foundation for recursive self-improvement in scientific discovery. However, existing neural provers do not fully preserve this recursive structure, where the learning process should be self-improving over time. Existing methods either embed proof experience into model parameters through expensive weight updates, or keep verified intermediate deductions only within the current problem. In addition, these methods also heavily rely on sparse whole-proof feedback, even when unsuccessful partial attempts contain useful discoveries. To close the gap, we propose ProofEvolve, a neuro-symbolic framework that evolves explicit, formally verified symbolic proof structures with neural models to decisively expand the knowledge boundary. In this framework, the neural model proposes variation operators, including decompositions, repairs, and schema recombinations. The symbolic Lean kernel verifies every proof transition. Over the evolution loops, ProofEvolve computes verified closure over the resulting proof directed acyclic graphs (DAGs). Within each problem, ProofEvolve evolves partial AND-OR proof DAGs in a behaviorally indexed archive. Across problems, kernel-checked schema extraction adds newly proved sub-DAGs to a persistent schema library. Proof DAGs inherit the solved results through typed schema recombination, with every residual premise exposed as a new subgoal. This evolutionary process preserves verified results from incomplete attempts and makes them available for later proofs without weakening formal soundness. Across three competition-level Lean benchmarks, ProofEvolve achieves the highest average solve rate among the evaluated proof systems.
Chinese Translation
自动定理证明为科学发现中的递归自我改进提供了自然基础。然而，现有的神经证明器并未完全保留这种递归结构，其中学习过程应该随着时间的推移而自我改进。现有方法要么通过昂贵的权重更新将证明经验嵌入模型参数中，要么仅在当前问题中保留经过验证的中间推导。此外，这些方法还严重依赖稀疏的整体证明反馈，即使不成功的部分尝试也包含有用的发现。为了解决这一问题，我们提出了ProofEvolve，一个神经符号框架，它与神经模型共同进化显式的、形式验证的符号证明结构，以决定性地扩展知识边界。在该框架中，神经模型提出变异操作，包括分解、修复和模式重组。符号Lean内核验证每一个证明转变。在进化循环中，ProofEvolve计算结果证明有向无环图（DAG）的验证闭包。在每个问题中，ProofEvolve在行为索引归档中进化部分AND-OR证明DAG。在跨问题的过程中，内核检查的模式提取将新证明的子DAG添加到持久的模式库中。证明DAG通过类型化模式重组继承已解决的结果，每个残余前提被暴露为新的子目标。这个进化过程保留了不完整尝试中的验证结果，并使其在后续证明中可用，而不削弱形式的正确性。在三个竞争级别的Lean基准测试中，ProofEvolve在评估的证明系统中实现了最高的平均求解率。
cs.AI / 42 / 2608.26430
Fine-Tuning of Transformer models with Frames
基于框架的变换器模型微调
Abstract
Parameter-Efficient Fine-Tuning (PEFT) strategies such as Low-Rank Adaptation (LoRA) are effective solutions for fine-tuning large-scale pre-trained models; however, their memory requirements scale with the size of the model, $\mathcal{O}(dr)$, where $d$ is the model's hidden dimension and $r$ is the rank. Our proposal, FrameFT, models the parameter update $\Delta W$ with a sparse coefficient matrix in a Fusion Frame basis. Fusion Frames can be generated algorithmically and shared across model layers, enabling very efficient updates. Only the sparse coefficients of the basis expansion are stored/optimized, reducing the memory footprint. The sparse structure of the coefficient matrix in FrameFT and the sparsity in the Fusion Frames give large compute benefits, and our analysis provides formal convergence results. We evaluate the idea across a suite of supervised fine-tuning benchmarks, focusing on language tasks, but also report application to vision models. Our experiments show that FrameFT achieves performance on par with/exceeding state-of-the-art PEFT techniques, but needs far fewer trainable parameters.
Chinese Translation
参数高效微调（PEFT）策略，如低秩适应（LoRA），是微调大规模预训练模型的有效解决方案；然而，它们的内存需求与模型大小成比例增长，记作 $ ext{O}(dr)$，其中 $d$ 是模型的隐藏维度，$r$ 是秩。我们提出的 FrameFT，通过稀疏系数矩阵在融合框架（Fusion Frame）基底中建模参数更新 $ riangle W$。融合框架可以通过算法生成并在模型层之间共享，从而实现非常高效的更新。仅存储/优化基底扩展的稀疏系数，从而减少内存占用。FrameFT 中系数矩阵的稀疏结构和融合框架的稀疏性带来了巨大的计算优势，我们的分析提供了正式的收敛结果。我们在一系列监督微调基准上评估了这一思想，重点关注语言任务，同时也报告了在视觉模型中的应用。我们的实验表明，FrameFT 的性能与最先进的 PEFT 技术相当或更优，但所需的可训练参数远少于后者。
cs.AI / 43 / 2608.26442
Don't Overthink, Don't Underthink: Toward Adaptive Reasoning in Agentic AI
不要过度思考，也不要低估思考：朝向自适应推理的代理人工智能
Abstract
Recent advances in Large Language Models (LLMs) have shown that increased inference-time reasoning can improve performance on complex tasks. However, many existing approaches rely on fixed or preallocated reasoning controls, such as fixed token budgets, pre-execution difficulty estimates, or activation-space interventions, and are often evaluated on standalone reasoning benchmarks rather than full agentic workflows. These assumptions may not hold in agentic AI systems, where reasoning requirements evolve dynamically through planning, tool use, memory retrieval, and agent-to-agent interactions. Consequently, reasoning can become either excessive or insufficient, resulting in unnecessary computation, increased latency, planning drift, excessive tool use, or incomplete solutions. We argue that a major challenge for next-generation agentic AI is not merely how much reasoning a language model should perform, but how it should allocate reasoning according to evolving task demands. We characterize over-reasoning and under-reasoning as recurring failure modes of misallocated reasoning and evaluate them on MATH-500 and the GAIA public validation benchmark. Using tool-decision latency, token consumption, token-limit exhaustion, and answer correctness, our results suggest that cases classified as over-reasoning are associated with higher computational cost without proportional accuracy gains, whereas cases classified as under-reasoning are consistently associated with incorrect or incomplete solutions. These findings motivate future research on adaptive reasoning mechanisms for agentic AI.
Chinese Translation
最近在大型语言模型（LLMs）方面的进展表明，增加推理时间的推理能力可以提高在复杂任务上的表现。然而，许多现有方法依赖于固定或预分配的推理控制，例如固定的令牌预算、执行前的难度估计或激活空间干预，并且通常在独立的推理基准上进行评估，而不是完整的代理工作流程。这些假设在代理人工智能系统中可能不成立，因为推理需求通过规划、工具使用、记忆检索和代理间交互动态演变。因此，推理可能变得过度或不足，导致不必要的计算、增加的延迟、规划漂移、过度使用工具或不完整的解决方案。我们认为，下一代代理人工智能面临的主要挑战不仅在于语言模型应该执行多少推理，而在于如何根据不断变化的任务需求分配推理。我们将过度推理和不足推理描述为错误分配推理的反复失败模式，并在MATH-500和GAIA公共验证基准上对其进行评估。通过工具决策延迟、令牌消耗、令牌限制耗尽和答案正确性，我们的结果表明，被分类为过度推理的案例与更高的计算成本相关，而没有相应的准确性提升，而被分类为不足推理的案例则始终与错误或不完整的解决方案相关。这些发现激励了未来对代理人工智能自适应推理机制的研究。
cs.AI / 44 / 2608.26530
PILOT in the Loop: Live Self-Improvement for Long-Horizon Agents
环中飞行员：长时间跨度智能体的实时自我改进
Abstract
Long-horizon agent runs generate experience that can improve both the current run and future work. Most self-improvement methods process this experience only after execution ends, so they cannot redirect the active run or immediately apply and validate lessons learned from it. We argue that self-improvement should instead be live, using emerging experience both to redirect the active run and to update the persistent harness. Existing agent architectures do not fully support this goal. Single-agent self-correction combines task execution and trajectory assessment within one context, while subagent delegation separates execution but typically cannot redirect an active subagent. We present PILOT, a supervisor-worker harness for live self-improvement through two coupled mechanisms: (1) live steering lets a separate supervisor redirect or abort the active worker during execution; and (2) live self-evolution distils procedures and failure modes revealed during execution into reusable skills and memory. Across two frozen backbones and three benchmarks, PILOT ranks first in five of six configurations. On Terminal-Bench 2.0, PILOT outperforms counterpart harnesses by up to 9.8 percentage points. In the self-improvement setting, PILOT gains 14.6 points with GLM-5.1 and 12.4 points with Kimi-K2.6. Mean output tokens fall by 42.9% and 47.4%, while successful evaluations per million output tokens rise by 110.3% and 134.0%, respectively.
Chinese Translation
长时间跨度智能体的运行生成的经验可以改善当前的运行和未来的工作。大多数自我改进方法仅在执行结束后处理这些经验，因此无法重定向当前运行或立即应用和验证从中学到的教训。我们认为，自我改进应该是实时的，利用新出现的经验来重定向当前运行并更新持久的框架。现有的智能体架构并未完全支持这一目标。单智能体自我修正将任务执行和轨迹评估结合在一个上下文中，而子智能体委派则分离执行，但通常无法重定向一个活动的子智能体。我们提出了PILOT，一种通过两个耦合机制实现实时自我改进的监督-工作者框架：（1）实时引导允许一个独立的监督者在执行过程中重定向或中止活动工作者；（2）实时自我演化将执行过程中揭示的程序和失败模式提炼为可重用的技能和记忆。在两个冻结的骨干网络和三个基准测试中，PILOT在六种配置中的五种中排名第一。在Terminal-Bench 2.0上，PILOT的表现比对等框架高出多达9.8个百分点。在自我改进设置中，PILOT在GLM-5.1上获得了14.6分，在Kimi-K2.6上获得了12.4分。平均输出标记减少了42.9%和47.4%，而每百万个输出标记的成功评估分别增加了110.3%和134.0%。
cs.AI / 45 / 2608.26535
Multi2AV-Safety: Benchmarking Safety in Multimodal-to-Audio-Video Generation
Multi2AV-Safety：多模态到音视频生成中的安全性基准测试
Abstract
Audio-video generation is rapidly moving from prompt-driven synthesis toward multimodal conditioning, where text, images, audio, and video can jointly shape the generated output. This shift changes the nature of safety evaluation: harmful intent may no longer reside in any single input, but instead emerge from how otherwise benign or weakly harmful conditions interact across modalities and time. Existing safety benchmarks, however, remain largely prompt-centric or tied to fixed conditioning interfaces, leaving such compositional risks difficult to study systematically. To bridge this gap, we introduce Multi2AV-Safety, the first safety benchmark, to the best of our knowledge, to cover all 11 non-singleton T/I/A/V conditioning configurations for audio-video generation, comprising 11,024 attack instances. Evaluation on Multi2AV-Safety reveals systematic weaknesses in representative multimodal safety guards across attack mechanisms and harm-evidence structures. Our evaluation reveals two complementary failure modes: harmful semantics can emerge from the combination of individually benign inputs, while explicit harmful cues can become harder to detect when mixed with benign multimodal context. Together, these results identify \emph{compositional risk perception} as a central capability gap in safeguarding multimodal-conditioned audio-video generation: current safety guards fail to reliably integrate safety evidence across modalities and time, even when all conditioning inputs are observable. The dataset will be publicly released in October 2026.
Chinese Translation
音视频生成正迅速从基于提示的合成转向多模态条件下的生成，其中文本、图像、音频和视频可以共同影响生成的输出。这一转变改变了安全性评估的性质：有害意图可能不再存在于任何单一输入中，而是源于原本无害或轻微有害的条件在不同模态和时间上的交互。然而，现有的安全基准仍然主要集中于提示或固定的条件接口，使得这种组合风险难以系统性地研究。为了解决这一问题，我们引入了Multi2AV-Safety，这是我们所知的第一个安全基准，涵盖了音视频生成的所有11种非单一的文本/图像/音频/视频（T/I/A/V）条件配置，共包含11,024个攻击实例。在Multi2AV-Safety上的评估揭示了代表性多模态安全防护在攻击机制和危害证据结构中的系统性弱点。我们的评估揭示了两种互补的失效模式：有害语义可以从单独无害输入的组合中产生，而当与无害的多模态上下文混合时，显式的有害线索可能变得更难以检测。这些结果共同识别出 extit{组合风险感知}作为保护多模态条件音视频生成的一个核心能力缺口：当前的安全防护未能可靠地整合跨模态和时间的安全证据，即使所有条件输入都是可观察的。该数据集将于2026年10月公开发布。
cs.AI / 46 / 2608.26546
DuMateBench: Evaluating Autonomous Agents in Complex Real-World Workflows
DuMateBench：在复杂现实工作流程中评估自主代理
Abstract
Autonomous agents are increasingly adopted to complete complex, multi-tool workflows in real-world settings. However, existing benchmarks typically separate tasks by application or capability and evaluate agents in environments that are cleaner and more stable than those encountered in practice. We introduce DuMateBench, a real-session benchmark reconstructed from anonymized and privacy-screened user sessions collected from a large-scale production agent platform. Each task preserves the relevant pre-solution interaction history, persistent configurations, and workspace state, and is then validated through human verification. The resulting benchmark comprises 200 tasks spanning 8 broad scenarios and 17 fine-grained capability categories, with most tasks requiring multiple capability coordination. We execute these tasks in isolated Docker containers injected with three forms of real-world environmental complexity: Insufficient, Unstable, and Noisy, and assess performance using a hybrid deterministic and LLM-as-Judge evaluation protocol. Experiments across five representative autonomous-agent frameworks paired with four state-of-the-art LLMs reveal substantial gaps in strict task completion. Complementary robustness, efficiency, and diagnostic analyses further show that performance under environmental perturbations is jointly shaped by the capabilities of the LLM and the surrounding agent framework. The code and data are publicly available at https://dumatebench.com/.
Chinese Translation
自主代理在现实环境中越来越多地被采用，以完成复杂的多工具工作流程。然而，现有的基准测试通常通过应用或能力来分离任务，并在比实际环境更干净、更稳定的环境中评估代理。我们引入了DuMateBench，这是一个从大型生产代理平台收集的匿名和隐私筛选用户会话重建的真实会话基准。每个任务保留相关的预解决交互历史、持久配置和工作区状态，并通过人工验证进行验证。最终的基准包括200个任务，涵盖8个广泛场景和17个细分能力类别，其中大多数任务需要多种能力的协调。我们在隔离的Docker容器中执行这些任务，并注入三种形式的现实环境复杂性：不足、不稳定和噪声，并使用混合确定性和LLM-as-Judge评估协议评估性能。在五个代表性的自主代理框架与四个最先进的LLM配对的实验中，发现严格任务完成存在显著差距。补充的鲁棒性、效率和诊断分析进一步表明，环境扰动下的性能共同受到LLM能力和周围代理框架的影响。代码和数据可在 https://dumatebench.com/ 上公开获取。
cs.AI / 47 / 2608.26623
AgentJudgeBench: A Multi-Difficulty Benchmark for Evaluating LLM Judges on Agentic Tool-Calling
AgentJudgeBench：评估代理工具调用的 LLM 判断者的多难度基准
Abstract
LLM judges are widely used to evaluate agentic tool-calling systems, yet their reliability on structured, dependency-driven workflows remains largely unexamined. We present AgentJudgeBench, the first benchmark to systematically study LLM-as-a-judge reliability for agentic tool-calling over workflow DAGs, as distinct from the broader LLM-as-a-judge task of open-ended text or preference evaluation. The benchmark comprises 3,808 instances spanning six DAG topologies and three difficulty tiers, evaluated with five generators (3B-70B open-weight models and GPT-5.4) and six judges (20B to frontier scale) under paired with- and without-ground-truth conditions. Judge alignment degrades monotonically with task difficulty, 1.5x faster without ground truth, and on hard queries without ground truth all six judges converge to a narrow 77-82% band regardless of scale, revealing a structural ceiling driven primarily by task difficulty, though its height is partly prompt-dependent for weaker generators, that model capacity alone cannot overcome. Ground-truth exposure is not uniformly beneficial: it reduces alignment for GPT-5.4 (1.5 pp) and Gemini-2.5-Pro (3.9 pp), consistent with over-anchoring. Among mitigation strategies, chain-of-thought reasoning and judge temperature both have negligible effect, while structured evaluation rubrics improve alignment by up to 6.5 pp but do not generalize uniformly across judge-generator pairs. With ground truth, QwQ-32B best matches the programmatic reference, while a human validation study identifies GPT-OSS-120B as the most human-aligned judge; without it, frontier judges lead only marginally within the shared ceiling. These results expose fundamental limitations of current LLM judges and yield practical guidelines for reliable evaluation in agentic systems.
Chinese Translation
LLM 判断者广泛用于评估代理工具调用系统，但它们在结构化、依赖驱动的工作流中的可靠性仍然未得到充分检验。我们提出了 AgentJudgeBench，这是第一个系统性研究 LLM 作为判断者在工作流有向无环图（DAG）上对代理工具调用的可靠性的基准，与开放式文本或偏好评估的更广泛的 LLM 作为判断者任务有所不同。该基准包含 3,808 个实例，涵盖六种 DAG 拓扑和三个难度层级，使用五个生成器（3B-70B 开放权重模型和 GPT-5.4）以及六个判断者（20B 到前沿规模）在有和没有真实答案的条件下进行评估。判断者的一致性随着任务难度的增加而单调下降，在没有真实答案的情况下，下降速度快 1.5 倍，而在没有真实答案的困难查询中，所有六个判断者的结果都收敛到一个狭窄的 77-82% 区间，无论规模如何，这揭示了一个主要由任务难度驱动的结构性上限，尽管其高度在一定程度上依赖于较弱生成器的提示。真实答案的暴露并不总是有利：它使 GPT-5.4（下降 1.5 个百分点）和 Gemini-2.5-Pro（下降 3.9 个百分点）的判断一致性降低，这与过度锚定一致。在缓解策略中，思维链推理和判断者温度的影响微乎其微，而结构化评估标准可以将一致性提高最多 6.5 个百分点，但并未在判断者-生成器对之间均匀推广。在有真实答案的情况下，QwQ-32B 最好地匹配程序参考，而一项人类验证研究则确定 GPT-OSS-120B 是最符合人类判断的判断者；在没有真实答案的情况下，前沿判断者仅在共享上限内略微领先。这些结果揭示了当前 LLM 判断者的基本局限性，并为代理系统中的可靠评估提供了实用指南。
cs.AI / 48 / 2608.26683
SIGMA: Structured Noise-Effect-Aware Grouped Multi-Agent Aggregation
SIGMA：结构化噪声效应感知的分组多智能体聚合
Abstract
Cooperative multi-agent reinforcement learning (MARL) faces significant challenges in maintaining robust coordination under noisy observations. Although observation disturbances are often introduced independently across agents, their downstream effects on cooperative decision-making can become structured through underlying cooperation structures. We characterize this phenomenon as structured noise effects, where noise-induced decision effects exhibit local correlation among agents with stronger task-related dependencies while remaining globally heterogeneous across different agents and local structures. Existing robust MARL methods, however, rarely explicitly characterize or exploit such structure-dependent noise effects. To address this limitation, we propose SIGMA, a hierarchical collaboration framework that exploits cooperation structures to learn robust representations under noisy observations. SIGMA first organizes agents into adaptive local structures through density-based grouping and performs intra-group consensus aggregation to preserve shared task-relevant information while smoothing agent-specific representation deviations. Inter-group attention then adaptively integrates information across different groups to preserve global coordination while accommodating their heterogeneous contributions. Experiments on noisy-observation tasks in StarCraft II empirically validate the structured noise effects and demonstrate that SIGMA consistently improves robustness under observation noise while maintaining competitive performance in noise-free environments.
Chinese Translation
合作多智能体强化学习（MARL）在噪声观测下维持稳健协调面临重大挑战。尽管观测干扰通常在智能体之间独立引入，但它们对合作决策的下游影响可能通过潜在的合作结构变得结构化。我们将这种现象描述为结构化噪声效应，其中噪声引起的决策效应在具有更强任务相关依赖的智能体之间表现出局部相关性，同时在不同的智能体和局部结构之间保持全局异质性。然而，现有的稳健MARL方法很少明确表征或利用这种结构依赖的噪声效应。为了解决这一局限性，我们提出了SIGMA，一个利用合作结构在噪声观测下学习稳健表示的分层协作框架。SIGMA首先通过基于密度的分组将智能体组织成自适应局部结构，并执行组内共识聚合，以保留共享的任务相关信息，同时平滑智能体特定的表示偏差。然后，组间注意力自适应地整合不同组之间的信息，以保持全局协调，同时适应它们的异质贡献。在《星际争霸 II》的噪声观测任务上的实验实证验证了结构化噪声效应，并表明SIGMA在观测噪声下始终提高了稳健性，同时在无噪声环境中保持了竞争性能。
cs.AI / 49 / 2608.26694
Relational Over-Regularization: Graph-Based AI-Generated Text Detection via Sentence Transition Deviation
关系过度正则化：基于图的AI生成文本检测通过句子过渡偏差
Abstract
Detecting AI-generated text (AIGT) remains challenging because existing approaches rely on token-level statistical signals or independent stylometric features, causing them to overfit to specific generators and fail under distribution shift. We identify a structural signal at the sentence-pair level: LLMs produce inter-sentence transition variance that deviates from human writing through inflated variance driven by recurring similarity bursts at paragraph boundaries and templated transitions. We formalize this as Relational Over-Regularization (ROR) and validate it across four benchmarks (p < 0.001). The central contribution is this relational problem formulation, not a novel GNN architecture; CSFG is one concrete instantiation for operationalizing ROR. To exploit this signal, we propose the Cross-Source Stylometric Fingerprint Graph (CSFG), a graph-based framework that encodes positional, sequential, semantic, and transition deviation signals as learnable GNN edge features. The per-edge signed deviation {\delta}_ij operationalizes ROR without hand-crafted thresholds and acts as a false-positive calibrator. CSFG achieves 97.14% accuracy under binary detection, outperforming the strongest graph-based baseline by 11.14 pp, with a false-positive rate of 1.57% and robust generalization to unseen LLMs in the inflated-variance regime; detection degrades for generators whose transition variance falls at or below the human baseline.
Chinese Translation
检测AI生成文本（AIGT）仍然具有挑战性，因为现有方法依赖于词元级统计信号或独立的风格特征，导致它们对特定生成器过拟合，并在分布转移下失效。我们在句子对级别识别出一种结构信号：大型语言模型（LLMs）在句子间过渡中产生的方差偏离人类写作，这种偏差是由段落边界处重复相似性爆发和模板化过渡驱动的膨胀方差所致。我们将其形式化为关系过度正则化（Relational Over-Regularization, ROR），并在四个基准测试中验证了这一点（p < 0.001）。核心贡献在于这种关系问题的表述，而不是新颖的图神经网络（GNN）架构；交叉源风格指纹图（Cross-Source Stylometric Fingerprint Graph, CSFG）是实现ROR的一个具体实例。为了利用这一信号，我们提出了交叉源风格指纹图（CSFG），这是一个基于图的框架，将位置、顺序、语义和过渡偏差信号编码为可学习的GNN边特征。每条边的签名偏差{ ext{δ}}_{ij}在没有手工阈值的情况下实现了ROR，并充当假阳性校准器。CSFG在二元检测中实现了97.14%的准确率，超越了最强的基于图的基线11.14个百分点，假阳性率为1.57%，并在膨胀方差范围内对未见过的LLMs具有稳健的泛化能力；对于那些过渡方差低于或等于人类基线的生成器，检测性能下降。
cs.AI / 50 / 2608.26696
Five Primitives for Governing Autonomous AI Agents at Runtime
运行时管理自主人工智能代理的五个基本原理
Abstract
Enterprise deployments of autonomous AI agents inherit a control model built for human users and long-lived services, and the fit fails in three specific ways: agent principals are ephemeral, appearing and vanishing faster than provisioning; their actions are selected by a model rather than programmed, so the set of things they may attempt is not known in advance; and the population is discovered rather than provisioned, because anyone who can call an API can create one. We argue that governing such agents is a runtime problem -- not a model-alignment problem and not a build-time problem -- and we derive five primitives from the questions that must be answered before an action takes effect and after it has: discovery, identity, governance, attestation, and supply chain. For each we state what fails if it is absent and why the others cannot structurally supply it. We describe an implementation in which an agent's action is mediated against policy before it takes effect, authorised against a per-tenant action vocabulary, and recorded in a hash-linked signed ledger a third party can verify with the vendor out of the loop. We report what the architecture costs: the enforcement point sits on the request's critical path, identity requires a sidecar per workload, and fail-closed mediation converts availability incidents into denial. We are explicit about implementation status: four primitives are built and running in private pilots, and the fifth is built as separate tooling and not yet integrated into the request path. We keep it in the set deliberately: a five-part decomposition that exactly matches what its authors happened to build is not a taxonomy but a description of a codebase.
Chinese Translation
自主人工智能代理的企业部署继承了为人类用户和长期服务构建的控制模型，但在三个特定方面存在不适配：代理主体是短暂的，出现和消失的速度快于配置；它们的行为是通过模型选择而非编程，因此它们可能尝试的行为集事先并不明确；而且代理的数量是通过发现而非配置的，因为任何可以调用API的人都可以创建一个。我们认为，治理这样的代理是一个运行时问题——而不是模型对齐问题，也不是构建时问题——并且我们从在行动生效前和生效后必须回答的问题中推导出五个基本原理：发现、身份、治理、证明和供应链。对于每个基本原理，我们说明如果缺失会导致什么失败，以及为什么其他原理在结构上无法提供它。我们描述了一种实现方式，其中代理的行为在生效前受到政策的调解，依据每个租户的行为词汇进行授权，并记录在一个哈希链接的签名账本中，第三方可以在不涉及供应商的情况下进行验证。我们报告了该架构的成本：强制执行点位于请求的关键路径上，身份需要每个工作负载一个侧车，而失败关闭的调解将可用性事件转化为拒绝。我们明确说明了实现状态：四个基本原理已在私人试点中构建并运行，而第五个作为单独的工具构建，尚未集成到请求路径中。我们故意将其保留在集合中：一个五部分的分解恰好与其作者所构建的内容相匹配，这不是一个分类法，而是对代码库的描述。
cs.AI / 51 / 2608.26701
Accelerating Scientific Research with Gemini in the Real-World
在现实世界中利用Gemini加速科学研究
Schmidgall, Samuel, Zhu, Xiaokai, Shaw, Marian, Yang, Lin, Liévin, Valentin, Yang, Jingyun, Zhuang, Yuchen, Strother, Tim, Bijamov, Alex, Sun, Min Woo, Palepu, Anil, Chen, Justin, Steiner, David, Shreibati, Jacqueline, Weng, Wei-Hung, Zhao, Yilin, Hu, Xingjian, Zahn, Nicholas, Garg, Sadhya, Kirby, Julia, Gan, Yuxiang, Li, Jiaoli, Thakkar, Divy, Azizi, Shekoofeh, Racz, David, Gottweis, Juraj, Natarajan, Vivek, Wu, Chenglin, Danino, Tal, Rong, Keran, Wang, Haozhe, Schillings, Benoit, Cheng, Yong, Le, Quoc V., Tu, Tao
Abstract
We present an extension and comprehensive real-world validation of Co-Scientist, a Gemini-based multi-agent system designed to accelerate end-to-end scientific research across hypothesis generation, experimentation, and manuscript generation. Moving beyond in silico hypothesis generation, this specialized configuration transitions Co-Scientist into an execution-grounded research partner advancing closed-loop scientific workflows across materials science, biology, and computer science. In materials science, Co-Scientist interfaced with a semi-automated chemical vapor deposition reactor to design a safe precursor route for MXenes; experimental execution produced a lamellar 2D material sharing key structural similarities with the Ti3C2Tx MXene lattice, although further experiments are needed to confirm the atomic structure. Leveraging Gemini 3 Deep Think for rapid, lab-in-the-loop execution, it also tailored growth recipes to laboratory constraints in minutes, enabling single-attempt growth of monolayer MoS2, MoSe2, and WS2 semiconductors. In biology, Co-Scientist predicted emergent swarming phenotypes of engineered E. coli across inducer (IPTG) gradients from sparse imaging data, quantitatively matching unpublished wet-lab morphological measurements. In computer science, Co-Scientist autonomously discovered an inference-time scaling architecture that outperformed six frontier models on HealthBench (Hard and Professional) while reducing potential clinical harm under blinded physician evaluation. Finally, a double-blind study of end-to-end generated papers with 30 domain experts across 450 reviews demonstrates that Co-Scientist's reliability modules reduce hallucination and plagiarism while improving research safety. Together, these results demonstrate progress toward closed-loop multi-agent scientific AI systems capable of accelerating real-world scientific discovery.
Chinese Translation
我们展示了Co-Scientist的扩展及其全面的现实世界验证，这是一个基于Gemini的多智能体系统，旨在加速从假设生成、实验到手稿生成的端到端科学研究。该专门配置超越了计算机模拟的假设生成，将Co-Scientist转变为一个以执行为基础的研究伙伴，推动材料科学、生物学和计算机科学领域的闭环科学工作流程。在材料科学中，Co-Scientist与半自动化化学气相沉积反应器接口，设计出安全的MXenes前驱体路线；实验执行产生了一种层状的二维材料，与Ti3C2Tx MXene晶格在关键结构上具有相似性，尽管需要进一步实验来确认其原子结构。利用Gemini 3 Deep Think进行快速的实验室内执行，它还在几分钟内根据实验室约束定制了生长配方，使得单次尝试生长单层MoS2、MoSe2和WS2半导体成为可能。在生物学中，Co-Scientist预测了工程化大肠杆菌在诱导剂（IPTG）梯度下的涌现群体表型，与未发表的湿实验形态测量结果定量匹配。在计算机科学中，Co-Scientist自主发现了一种推理时间扩展架构，在HealthBench（Hard和Professional）上超越了六个前沿模型，同时在盲法医生评估下减少了潜在的临床危害。最后，针对30位领域专家进行的双盲研究，涵盖450条评审，证明了Co-Scientist的可靠性模块减少了幻觉和抄袭，同时提高了研究安全性。这些结果共同展示了朝向能够加速现实世界科学发现的闭环多智能体科学人工智能系统的进展。
cs.AI / 52 / 2608.26710
Style as a Confound: False Positives in AI Detection of Non-Native Academic Writing
风格作为混淆因素：人工智能检测非母语学术写作中的假阳性
Abstract
AI text detectors are increasingly employed in academic settings, but it remains unclear whether their outputs reflect AI authorship itself or broader linguistic features associated with polished academic English. Previous studies have reported high false-positive rates (FPRs) for non-native English writing, but population-level comparisons confound authorship with differences in topic, domain, and writing style. Professional editing provides a useful setting for examining this issue because it changes the linguistic form of manuscripts while preserving authorship and content. We examined 135,389 document pairs from a professional English editing service (2018-2025), comprising non-native manuscripts and their native-edited versions, to assess how editing affects detector responses controlling for content and authorship. For the 13 AI text detectors, FPRs for human-written texts varied widely, from 0.0% to 100.0%. Responses varied across detectors: the same edits increased AI scores in some detectors but decreased them in others. Notably, score changes correlated with the extent of editing. The findings identify professional editing style as a key confounding variable in AI detector outputs, rather than establishing a full separation of text origin from linguistic style, raising concerns about fairness and reliability in academic settings.
Chinese Translation
人工智能文本检测器在学术环境中越来越多地被使用，但其输出是否反映了人工智能作者本身，还是与精炼的学术英语相关的更广泛语言特征仍不清楚。先前的研究报告了非母语英语写作的高假阳性率（FPR），但人口层面的比较混淆了作者身份与主题、领域和写作风格的差异。专业编辑为研究这一问题提供了一个有用的背景，因为它在保留作者身份和内容的同时改变了手稿的语言形式。我们分析了来自一家专业英语编辑服务（2018-2025年）的135,389对文档，包括非母语手稿及其母语编辑版本，以评估编辑如何影响检测器的响应，同时控制内容和作者身份。对于13个人工智能文本检测器，人类撰写文本的假阳性率差异很大，从0.0%到100.0%不等。不同检测器的响应也有所不同：相同的编辑在某些检测器中提高了人工智能得分，而在其他检测器中则降低了得分。值得注意的是，得分变化与编辑的程度相关。研究结果表明，专业编辑风格是人工智能检测器输出中的一个关键混淆变量，而不是完全将文本来源与语言风格分离，这引发了对学术环境中公平性和可靠性的担忧。
cs.AI / 53 / 2608.26730
Knowing When Not to Reuse: Conditional Experience Transfer in Autonomous LLM Post-Training
知道何时不重用：自主大型语言模型后训练中的条件经验转移
Abstract
Large language models offer broad capabilities, but adapting them to evolving domains, tools, and requirements often entails repeated post-training. Autonomous systems automate parts of this process by proposing updates, training candidates, and using evaluation feedback to select subsequent proposals. As evidence accumulates, a central problem emerges: which past update evidence remains actionable after subsequent training has changed the parent model? An update's effect depends on its parent, data, and training stage. Treating past success as context-free permission can waste compute. If the resulting child is promoted, it can also degrade the subsequent training trajectory. We formulate this problem as conditional experience transfer and introduce Boundary-Calibrated Intervention Transfer (BCIT), a method that authorizes experience reuse before weight-changing training. BCIT binds an observed effect to its source context, checks applicability conditions, vetoes candidates with named hard conflicts, and obtains current-state evidence through a bounded training trial when needed. Fully trained candidates still face a shared adoption rule, and only observed events extend memory. On one 4B model adapted across finance reasoning, text-to-SQL, and function calling, candidate updates exhibit heterogeneous target and retention effects across the evaluated contexts. Under matched candidates, evidence, and compute, BCIT authorizes fewer harmful updates and attains higher equal-budget final-model quality than the evaluated alternatives. These results support treating experience authorization as a distinct problem in autonomous post-training.
Chinese Translation
大型语言模型提供了广泛的能力，但将其适应于不断发展的领域、工具和需求通常需要反复进行后训练。自主系统通过提出更新、训练候选者以及利用评估反馈来选择后续提案，从而自动化了这一过程的部分环节。随着证据的积累，一个核心问题浮现：在后续训练改变了父模型之后，哪些过去的更新证据仍然是可操作的？更新的效果依赖于其父模型、数据和训练阶段。将过去的成功视为无上下文的许可可能会浪费计算资源。如果结果子模型被推广，它还可能会降低后续训练的轨迹。我们将这一问题表述为条件经验转移，并引入边界校准干预转移（Boundary-Calibrated Intervention Transfer, BCIT），这是一种在权重变化训练之前授权经验重用的方法。BCIT将观察到的效果与其源上下文绑定，检查适用条件，否决具有明确硬冲突的候选者，并在需要时通过有限的训练试验获取当前状态证据。完全训练的候选者仍然面临共享的采纳规则，只有观察到的事件才能扩展记忆。在一个在金融推理、文本到 SQL 和函数调用中适应的 4B 模型上，候选更新在评估的上下文中表现出异质的目标和保留效果。在匹配的候选者、证据和计算条件下，BCIT 授权的有害更新更少，并且在相同预算下最终模型的质量高于评估的替代方案。这些结果支持将经验授权视为自主后训练中的一个独特问题。
cs.AI / 54 / 2608.26743
Graph-Guided Selective Unlearning for Language Models: Controlling Support Routes Beyond Forget Seeds
图引导的选择性遗忘用于语言模型：超越遗忘种子的支持路径控制
Abstract
Enterprises fine-tune language models on proprietary data that may later require removal due to privacy, contractual, or compliance obligations. Selective unlearning removes requested knowledge while preserving model utility, offering a practical alternative to full retraining, but existing methods treat the explicitly identified forget examples as the complete deletion scope. This is insufficient when target knowledge remains recoverable through paraphrases, aliases, or neighboring training examples. We propose GRAPHSU, a graph-guided controller that expands the deletion scope beyond forget seeds by constructing a weighted support-route graph, propagating deletion pressure through it, and applying graded forgetting strengths to high-risk neighbors. On the Task of Fictitious Unlearning (TOFU), a synthetic author-profile question-answering benchmark, and PISTOL, a structural-unlearning benchmark built around interconnected factual samples, with GPT-2 Medium and Llama-3.2-3B-Instruct, GRAPHSU achieves the lowest utility-feasible soft leakage across all deletion settings, reducing leakage by up to 49.5 percentage points over a matched seed-only baseline, demonstrating that effective enterprise unlearning requires controlling support routes, not just forget seeds.
Chinese Translation
企业在专有数据上微调语言模型，这些数据可能因隐私、合同或合规义务而需要后续删除。选择性遗忘在保留模型效用的同时移除请求的知识，提供了一个比完全重训练更实用的替代方案，但现有方法将明确识别的遗忘示例视为完整删除范围。当目标知识可以通过同义改写、别名或邻近训练示例恢复时，这种做法是不够的。我们提出了 GRAPHSU，这是一种图引导的控制器，通过构建加权支持路径图，扩展删除范围超出遗忘种子，沿着图传播删除压力，并对高风险邻居应用分级遗忘强度。在虚构遗忘任务（TOFU）这一合成作者档案问答基准，以及围绕互联事实样本构建的结构性遗忘基准 PISTOL 上，使用 GPT-2 Medium 和 Llama-3.2-3B-Instruct，GRAPHSU 在所有删除设置中实现了最低的效用可行软泄漏，相较于仅使用匹配种子的基线，泄漏减少了多达 49.5 个百分点，证明有效的企业遗忘需要控制支持路径，而不仅仅是遗忘种子。
cs.AI / 55 / 2608.26747
AgentFold: Closed-Loop Agentic Search for Protein Folding Model Design
AgentFold：用于蛋白质折叠模型设计的闭环自主搜索
Abstract
Scientific LLM agents have shown promise in literature reasoning, tool use, and experiment planning, but it remains unclear whether they can autonomously improve large, tightly coupled scientific machine-learning systems through executable code changes and computationally expensive validation. We study this question in protein folding, where progress requires coordinated architectural modifications, multi-objective evaluation, and domain-aware interpretation. We present AgentFold, a multi-agent framework that formulates folding-model development as a closed-loop search over executable code variants. Starting from ESMFold, AgentFold proposes hypotheses, implements and debugs code-level modifications, evaluates model variants, analyzes experimental outcomes, and stores both successful and failed interventions in structured memory. An MCTS-style policy allocates computational resources across high-scoring search branches. On an engineering-scale protein-folding codebase comprising more than 2,000 lines of code, AgentFold explores approximately 80 model variants using approximately 5,000 GPU-hours and 170 million LLM tokens. Under a matched computational budget, AgentFold improves the best lDDT by 7.5% over independent Codex proposals and outperforms a random-search control. Beyond model improvement, the resulting intervention traces reveal recurring empirical design patterns: stable gains tend to arise from early, soft, learnable priors and gated refinement, whereas direct geometric perturbations and geometry-conditioned feedback often destabilize training. The code and experimental resources are publicly available at https://github.com/lmqfly/AgentFold.
Chinese Translation
科学大型语言模型（LLM）代理在文献推理、工具使用和实验规划方面显示出良好的前景，但尚不清楚它们是否能够通过可执行代码更改和计算成本高昂的验证，自动改进大型、紧密耦合的科学机器学习系统。我们在蛋白质折叠领域研究这个问题，进展需要协调的架构修改、多目标评估和领域感知的解释。我们提出了AgentFold，一个将折叠模型开发形式化为可执行代码变体的闭环搜索的多代理框架。从ESMFold开始，AgentFold提出假设，实施和调试代码级修改，评估模型变体，分析实验结果，并将成功和失败的干预存储在结构化内存中。基于蒙特卡洛树搜索（MCTS）风格的策略在高得分的搜索分支之间分配计算资源。在一个包含超过2000行代码的工程规模蛋白质折叠代码库上，AgentFold使用约5000个GPU小时和1.7亿个LLM令牌探索了大约80个模型变体。在匹配的计算预算下，AgentFold在独立的Codex提案上将最佳lDDT提高了7.5%，并优于随机搜索对照组。除了模型改进，结果干预轨迹揭示了反复出现的经验设计模式：稳定的增益往往源于早期、柔性、可学习的先验和门控细化，而直接的几何扰动和几何条件反馈则往往会使训练不稳定。代码和实验资源可在https://github.com/lmqfly/AgentFold公开获取。
cs.AI / 56 / 2608.26750
Discovering Relationships in Data Lakes Using Large Language Models: An Industrial Case
利用大型语言模型发现数据湖中的关系：一个工业案例
Abstract
Data lakes rely on metadata to remain usable, yet this meta data is often limited or weakly informative for column relationship discovery, especially in ERP-derived datasets with coded or abbreviated schema labels. We propose ColRel, a two-stage method that builds column embeddings from metadata and data available at ingestion time. In difficult cases, such as coded schemata, business dictionaries help better interpret column names and support the generation of short natural-language descriptions used in the second stage. Experiments on public benchmarks and an industrial ERP dataset show that ColRel is particularly effective in semantically related, weak-signal settings.
Chinese Translation
数据湖依赖元数据以保持可用性，但这些元数据通常有限或在列关系发现方面信息不足，尤其是在具有编码或缩写模式标签的ERP派生数据集中。我们提出了ColRel，这是一种两阶段的方法，利用在数据摄取时可用的元数据和数据构建列嵌入。在困难情况下，例如编码模式，业务词典有助于更好地解释列名称，并支持生成在第二阶段使用的简短自然语言描述。在公共基准和一个工业ERP数据集上的实验表明，ColRel在语义相关、弱信号环境中特别有效。
cs.AI / 57 / 2608.26757
DEEPCHART: How Far are LLMs from Faithful Data-Science Chart Generation?
DEEPCHART：大型语言模型在真实数据科学图表生成中的可靠性有多远？
Abstract
Faithful chart generation in real-world data-science workflows requires grounding visualizations in scattered evidence, computing chart-ready quantities, and rendering them accurately. Modern LLMs can produce visually plausible, instruction-compliant charts, yet data-level hallucinations remain difficult to detect in long, noisy, and multimodal contexts. To measure this gap, we introduce DEEPCHART, an expert-annotated benchmark of 1,482 task-conditioned chart-generation instances drawn from real-world scientific papers, financial filings, and ecosystem reports. DEEPCHART formulates chart generation as an Extract--Reason--Visualize pipeline and evaluates source-data extraction, derived-data reasoning, and chart rendering stage by stage. Experiments with state-of-the-art models show that visually plausible charts often conceal data-level hallucinations, with extraction and reasoning errors common in realistic long and multimodal settings. These findings suggest that larger context windows alone are insufficient; faithful chart generation also requires reliable evidence extraction and quantitative reasoning before rendering. Our benchmark and associated resources are available at https://github.com/tangdouer1005/DeepChart.
Chinese Translation
在现实世界的数据科学工作流程中，可靠的图表生成需要将可视化与分散的证据相结合，计算适合图表的量，并准确呈现它们。现代大型语言模型（LLMs）能够生成视觉上合理且符合指令的图表，但在长篇、嘈杂和多模态的上下文中，数据级的幻觉仍然难以检测。为了衡量这一差距，我们引入了DEEPCHART，这是一个包含1,482个任务条件的图表生成实例的专家注释基准，来源于真实的科学论文、财务文件和生态系统报告。DEEPCHART将图表生成形式化为提取-推理-可视化的流程，并逐阶段评估源数据提取、衍生数据推理和图表渲染。与最先进模型的实验表明，视觉上合理的图表往往掩盖了数据级的幻觉，在现实的长篇和多模态环境中，提取和推理错误普遍存在。这些发现表明，仅仅增加上下文窗口的大小是不够的；可靠的图表生成还需要在渲染之前进行可靠的证据提取和定量推理。我们的基准和相关资源可在 https://github.com/tangdouer1005/DeepChart 获取。
cs.AI / 58 / 2608.26763
Categorizer Automata for Discounted-Sum Payoffs
用于折扣总收益的分类器自动机
Abstract
Categorizing continuous data into discrete bins is a fundamental operation in artificial intelligence. We introduce the categorizer automaton, a deterministic automaton that reads an infinite sequence of rewards and identifies which of finitely many bins contains its discounted sum. Categorizer automata generalize comparator automata, the special case of two bins, which have already proven useful in quantitative synthesis. Our main technical contribution is the construction of a categorizer automaton whose state space is linear in the number of bins, rather than exponential as obtained by a cross-product of comparator automata. We then apply categorizer automata to Markov decision processes, where they allow one to synthesize policies that maximize the expected utility of a discounted-sum payoff for utility functions that may be discontinuous. For piecewise-constant utility functions, the resulting algorithm is exact and runs in pseudo-polynomial time. For piecewise-Lipschitz utility functions, a class that includes any utility with bounded slope between finitely many jumps, it again runs in pseudo-polynomial time and yields an $\varepsilon$-optimal policy. We also show that the synthesis problem considered is PSPACE-hard already for piecewise-constant utilities.
Chinese Translation
将连续数据分类为离散区间是人工智能中的一项基本操作。我们引入了分类器自动机，这是一种确定性自动机，它读取无限序列的奖励并识别出其折扣总和所包含的有限多个区间中的哪个。分类器自动机是比较器自动机的推广，后者是两个区间的特例，已在定量合成中证明了其有用性。我们主要的技术贡献是构造一个分类器自动机，其状态空间与区间数量呈线性关系，而不是通过比较器自动机的交叉乘积获得的指数关系。随后，我们将分类器自动机应用于马尔可夫决策过程，使其能够合成最大化折扣总收益期望效用的策略，适用于可能不连续的效用函数。对于分段常数效用函数，得到的算法是精确的，并且在伪多项式时间内运行。对于分段-Lipschitz效用函数，这一类包括在有限多个跳跃之间斜率有界的任何效用，它同样在伪多项式时间内运行，并产生一个$ ext{ε}$-最优策略。我们还表明，所考虑的合成问题对于分段常数效用已经是PSPACE-困难的。
cs.AI / 59 / 2608.26780
AI Control Scientist: LLM-driven Agentic System for Automated Control Design
AI控制科学家：基于大语言模型的自动化控制设计代理系统
Abstract
Control system design is critical for modern industry, such as chemical process temperature regulation and aero-engine control. However,traditional control design workflows rely heavily on expert knowledge and extensive manual parameter tuning, resulting in limited efficiency and scalability. To this end, this paper proposes AI Control Scientist (AICS), the first large language model (LLM)-driven agent capable of automatically generating optimized controller from language design requirements. Specifically, a Task Modeling Agent interprets user requirements to engineering constraints; a Controller Design Agent generate candidate controller structures and executable code; and a Parameter Tuning Agent refine controller parameters under closed-loop performance criteria. Experiments demonstrate that the proposed agentic system can automatically generate multiple representative control systems, outperforms existing automated baselines in both design success rate and optimization efficiency. This work has the potential to transform control system design from human-driven to agent-driven, paving the way for model predictive control and other advanced control systems design.
Chinese Translation
控制系统设计对于现代工业至关重要，例如化工过程的温度调节和航空发动机控制。然而，传统的控制设计工作流程在很大程度上依赖于专家知识和广泛的手动参数调整，导致效率和可扩展性有限。为此，本文提出了AI控制科学家（AI Control Scientist，AICS），这是首个能够根据语言设计要求自动生成优化控制器的大型语言模型（Large Language Model，LLM）驱动的代理。具体而言，任务建模代理（Task Modeling Agent）将用户需求解释为工程约束；控制器设计代理（Controller Design Agent）生成候选控制器结构和可执行代码；参数调整代理（Parameter Tuning Agent）在闭环性能标准下优化控制器参数。实验表明，所提出的代理系统能够自动生成多个具有代表性的控制系统，并在设计成功率和优化效率上超越现有的自动化基线。此项工作有潜力将控制系统设计从人驱动转变为代理驱动，为模型预测控制和其他先进控制系统设计铺平道路。
cs.AI / 60 / 2608.26788
Decoupling Planning and Control for Instructable Agents
指令驱动代理的规划与控制解耦
Abstract
Recent work shows that pre-trained, instruction-tuned vision-language models (VLMs) perform well at mapping from instructions and observations to high-level plans, but struggle to realize such plans as reliable low-latency action sequences in unfamiliar environments. At the same time, world-model controllers excel at fast observation-to-action control, but lack open-ended task guidance. In this work, we combine these strengths into a single system, Instruct-to-Act, where we train a world-model controller to act autonomously at high frequency when conditioned on sparse, higher-latency, and high-level text instructions generated by a VLM planner. To train controllers to be language-instructable, we relabel segments of controller policy rollouts with synthetic instructions and jointly optimize a behavior-cloning objective along with existing reward-maximizing and world-modeling objectives. We evaluate our proposed approach across seven embodied environments, including three multi-agent environments where VLM planners coordinate through language while trained controllers serve as their actuators. Under matched observation and action spaces, our decoupled approach consistently outperforms controller-only and direct VLM action-generation variants, preserves fast control, and lets us swap in different pretrained VLM planners without fine-tuning, while remaining competitive with strong vision-language-action and multi-agent RL baselines on six of seven tasks.
Chinese Translation
近期的研究表明，经过预训练和指令调优的视觉-语言模型（VLMs）在将指令和观察映射到高层次计划方面表现良好，但在不熟悉的环境中将这些计划转化为可靠的低延迟动作序列时却面临困难。同时，世界模型控制器在快速的观察到动作控制方面表现出色，但缺乏开放式任务指导。在本研究中，我们将这些优势结合成一个单一系统——Instruct-to-Act，其中我们训练一个世界模型控制器，使其在基于VLM规划器生成的稀疏、高延迟和高层次文本指令的条件下以高频率自主行动。为了训练控制器具备语言指令能力，我们使用合成指令重新标记控制器策略回放的片段，并联合优化行为克隆目标以及现有的奖励最大化和世界建模目标。我们在七个具身环境中评估了我们提出的方法，包括三个多智能体环境，其中VLM规划器通过语言进行协调，而训练的控制器则作为其执行器。在匹配的观察和动作空间下，我们的解耦方法在控制器单独和直接VLM动作生成变体中始终表现优于其他方法，保持了快速控制，并且允许我们在不进行微调的情况下更换不同的预训练VLM规划器，同时在七个任务中的六个任务上与强大的视觉-语言-动作和多智能体强化学习基线保持竞争力。
cs.AI / 61 / 2608.26836
SymbolLKG: Towards Verifiable Logical Reasoning via Logical Knowledge Graph and Symbolic Solvers
SymbolLKG：通过逻辑知识图谱和符号求解器实现可验证的逻辑推理
Abstract
Large Language Models (LLMs) have demonstrated remarkable proficiency in natural language understanding, yet they struggle with strict multi-step reasoning, frequently suffering from hallucinations and inconsistency. Existing solutions like Chain-of-Thought (CoT) lack rigorous verification mechanisms, while standard Retrieval-Augmented Generation (RAG) often misses the complex, structural dependencies inherent in logical tasks. To bridge this gap, we propose a Neuro-Symbolic architecture that integrates a Logical Knowledge Graph (LKG) with dynamic solver routing. Specifically, we introduce an ontology-based LKG that treats logical rules and constraints as first-class topological nodes, enabling explicit modeling of dependencies extracted from text. We further design a Logic Router to dynamically dispatch tasks to the optimal symbolic engine, which is supported by a topology-aware hybrid retrieval mechanism. Experimental results on logical reasoning benchmarks demonstrate that our framework significantly outperforms state-of-the-art prompting and RAG baselines, delivering higher accuracy and verifiable reasoning paths.
Chinese Translation
大型语言模型（LLMs）在自然语言理解方面表现出色，但在严格的多步骤推理中却面临挑战，常常出现幻觉和不一致性。现有的解决方案，如思维链（Chain-of-Thought, CoT），缺乏严格的验证机制，而标准的增强检索生成（Retrieval-Augmented Generation, RAG）往往无法捕捉逻辑任务中固有的复杂结构依赖关系。为了解决这一问题，我们提出了一种神经符号架构，将逻辑知识图谱（Logical Knowledge Graph, LKG）与动态求解器路由相结合。具体而言，我们引入了一种基于本体的LKG，将逻辑规则和约束视为一类重要的拓扑节点，从而能够显式建模从文本中提取的依赖关系。我们进一步设计了一个逻辑路由器，能够动态地将任务分配给最佳的符号引擎，并由一个拓扑感知的混合检索机制提供支持。在逻辑推理基准测试中的实验结果表明，我们的框架显著优于最先进的提示和RAG基线，提供了更高的准确性和可验证的推理路径。
cs.AI / 62 / 2608.26849
LiveSim: Simulating Environment-Shaped Users in Multi-Agent Live-Stream Ecosystems
LiveSim：在多代理直播生态系统中模拟环境塑造的用户
Abstract
User behavior simulation with large language models~(LLMs) is increasingly used to support multi-agent ecosystem simulation. Existing simulators typically rely on static user profiles inferred from historical observations, which become inadequate in socially intensive environments such as live streaming where interaction dynamics continuously reshape user behavior. We propose \textbf{LiveSim}, an LLM-based framework for live-stream ecosystem simulation. It represents users as editable behavioral hypotheses and progressively refines them through trajectory-grounded interactions, where discrepancies between simulated and observed trajectories reveal missing environmental shaping effects. These signals are further extracted as transferable environment-behavior patterns and accumulated in a collective behavioral memory to improve user-level behavioral fidelity and support ecosystem-level simulation. Experiments on real-world live-stream risk-control data validate the effectiveness of LiveSim in improving user-level behavioral fidelity and enabling ecosystem-level analysis of risk evolution and platform intervention effects.
Chinese Translation
利用大型语言模型（LLMs）进行用户行为模拟在支持多代理生态系统模拟方面越来越普遍。现有的模拟器通常依赖于从历史观察中推断出的静态用户档案，这在直播等社交密集型环境中显得不足，因为互动动态不断重塑用户行为。我们提出了 extbf{LiveSim}，一个基于LLM的直播生态系统模拟框架。它将用户表示为可编辑的行为假设，并通过轨迹基础的互动逐步细化这些假设，其中模拟轨迹与观察轨迹之间的差异揭示了缺失的环境塑造效应。这些信号进一步提取为可转移的环境-行为模式，并积累在集体行为记忆中，以提高用户级别的行为真实性并支持生态系统级别的模拟。对真实世界直播风险控制数据的实验验证了LiveSim在提高用户级别行为真实性和支持生态系统级别风险演变及平台干预效果分析方面的有效性。
cs.AI / 63 / 2608.26867
BekchiAI: Measuring, Observing, and Controlling LLM Agents in One Click
BekchiAI：一键测量、观察和控制大型语言模型代理
Abstract
Large language model agents reason, call tools, and act autonomously over many steps, but their agentic skills-correctly sequencing tools, planning under dependencies, judging untrusted inputs, and grounding generated arguments-are hard to measure with accuracy-only leaderboards. We present BekchiAI, which addresses both sides: a benchmark for measuring agentic skill and a platform for observing and controlling live agents. The BekchiAI-Benchmark, a suite of 13 tool-using ReAct agents across 7 task categories (arithmetic, structured/SQL, security detection, URL grounding, planning, orchestration, and tool-policy), totalling 2,057 deterministic, committed test tasks. Every task is verifier-checkable gold answers are computed by running canonical SQL against a real database, computing the exact schedule of a directed acyclic graph (DAG), or evaluating closed-form lambdas including adversarial security samples paired with deliberately imperfect signature scanners so a score reflects the model's own judgment, not the copying of an oracle. We define a small set of behavioral metrics beyond accuracy-tool-call adherence, URL hallucination and source-match, and per-model token cost and report a four-model comparison (Qwen3.7-Max, gemma-4-31B-it, gemma4:26b, gpt-oss-120b) whose story is in the per-family spread, not the aggregate. The benchmark runs are executed using the provided evaluation scripts. BekchiAI-Platform is a complementary web-based observability and control layer for deployed agents, providing full token and latency telemetry as well as remote run termination. The benchmark, evaluation tools, and platform are publicly released.
Chinese Translation
大型语言模型代理能够进行推理、调用工具并自主行动，但其代理技能——正确地顺序使用工具、在依赖关系下进行规划、判断不可信输入以及将生成的论点与现实相结合——仅通过排行榜来测量是困难的。我们提出了BekchiAI，它同时解决了这两个方面：一个用于测量代理技能的基准和一个用于观察和控制实时代理的平台。BekchiAI基准是一个包含13个使用工具的ReAct代理的套件，涵盖7个任务类别（算术、结构化/SQL、安全检测、URL基础、规划、编排和工具策略），总计2057个确定性、承诺的测试任务。每个任务的验证者可检查的金标准答案是通过对真实数据库运行标准SQL、计算有向无环图（DAG）的确切调度或评估闭合形式的lambda（包括与故意不完美的签名扫描仪配对的对抗性安全样本）得出的，因此得分反映的是模型自身的判断，而不是对神谕的复制。我们定义了一小组超越准确性的行为指标——工具调用遵循、URL幻觉和源匹配，以及每个模型的令牌成本，并报告了四个模型（Qwen3.7-Max、gemma-4-31B-it、gemma4:26b、gpt-oss-120b）的比较，其故事在于每个家族的分布，而非总和。基准运行使用提供的评估脚本执行。BekchiAI平台是一个补充的基于网络的可观察性和控制层，适用于已部署的代理，提供完整的令牌和延迟遥测以及远程运行终止。基准、评估工具和平台均已公开发布。
cs.AI / 64 / 2608.26870
C-Unseen: Weak Signal Detection in Dynamic Temporal Knowledge Graphs via LLM Reasoning
C-Unseen：通过大语言模型推理在动态时间知识图谱中进行弱信号检测
Abstract
Weak signals are early, low-visibility indicators that precede significant changes before those changes become established. Existing detection methods, based on keyword frequency, topic modeling, or untyped graph topology, fail to capture the semantic and relational structure through which such signals manifest. In this paper, we propose C-Unseen, a self-interpretable framework for weak signal detection in Dynamic Temporal Knowledge Graphs (DTKGs). We define a weak signal as a rare, semantically coherent subgraph that proliferates across consecutive TKG snapshots. The framework operates through two modules: a Rare Subgraphs Extractor, in which an LLM identifies subgraphs whose content is in tension with the dominant snapshot narrative via chain-of-thought reasoning, and a Weak Signal Alerter, in which the persistence of these rare subgraphs is tracked across time steps to isolate true weak signals. Experimental results demonstrate that C-Unseen outperforms keyword-, topic-, and graph-based baselines.
Chinese Translation
弱信号是早期的、低可见性的指标，预示着重大变化的发生，这些变化在形成之前就已经出现。现有的检测方法基于关键词频率、主题建模或无类型图拓扑，未能捕捉到这些信号显现的语义和关系结构。本文提出了C-Unseen，一个自解释的框架，用于在动态时间知识图谱（DTKGs）中进行弱信号检测。我们将弱信号定义为在连续的TKG快照中扩散的稀有、语义连贯的子图。该框架通过两个模块进行操作：稀有子图提取器（Rare Subgraphs Extractor），其中大语言模型（LLM）通过链式推理识别与主导快照叙事存在张力的子图；弱信号警报器（Weak Signal Alerter），在其中跟踪这些稀有子图在时间步长上的持续性，以隔离真实的弱信号。实验结果表明，C-Unseen在性能上优于基于关键词、主题和图的基线方法。
cs.AI / 65 / 2608.26885
Evaluating human and LLM screening workflows in a conceptually complex scoping review: Recall--workload trade-offs and run-to-run consistency
评估人类与大型语言模型（LLM）在概念复杂的范围审查中的筛选工作流程：召回率与工作量的权衡及运行间一致性
Abstract
Background. Large language models (LLMs) are increasingly used for screening in evidence synthesis, where false negatives can remove relevant studies before full-text assessment. We compared human and LLM title-and-abstract screening workflows in a preregistered study embedded in a conceptually complex scoping review. Methods. After a conservative title-only screen, 1,131 records were screened by one review lead, four trained assistants screening non-overlapping subsets, and seven complete LLM runs using different models and processing configurations, including a nominally identical repeat run. We compared retained workload, operational recall against 316 verified eligible records, agreement, run-to-run consistency, and procedural burden. Because eligibility was verified only for records advanced and assessed in the parent review, recall estimates were operational. Results. No workflow recovered all verified eligible records. The human workflows and two GPT-5.4 file-batch runs retained 42.2-45.0% of records while achieving 82.3-82.9% recall. Gemini 3.1 file batches achieved the highest recall (83.9%) but retained 56.7% of records. All-at-once configurations recovered fewer eligible records than corresponding file-batch configurations. Two nominally identical GPT-5.4 file-batch runs agreed on 91.7% of records but differed on 94 records, including 29 verified eligible records retained by only one run. Discussion. LLM screening performance depended on the implemented workflow, not model identity alone. Processing configuration, workload, record-level variation, and human-LLM decision integration are therefore substantive properties of deployed systems. For high-recall tasks, LLMs are better suited to validated, auditable, human-supervised workflows than autonomous exclusion.
Chinese Translation
背景：大型语言模型（LLMs）在证据综合中的筛选应用日益增多，其中假阴性可能在全文评估之前排除相关研究。我们在一项预注册研究中比较了人类与LLM的标题和摘要筛选工作流程，该研究嵌入在一个概念复杂的范围审查中。方法：在保守的仅标题筛选后，由一名审查负责人、四名训练有素的助手（筛选不重叠的子集）和七次使用不同模型及处理配置的完整LLM运行（包括一次名义上相同的重复运行）对1,131条记录进行了筛选。我们比较了保留的工作量、针对316条经过验证的合格记录的操作召回率、一致性、运行间一致性及程序负担。由于资格仅针对在母审查中推进和评估的记录进行了验证，因此召回估计为操作性结果。结果：没有任何工作流程能够回收所有经过验证的合格记录。人类工作流程和两次GPT-5.4文件批次运行保留了42.2%-45.0%的记录，同时实现了82.3%-82.9%的召回率。Gemini 3.1文件批次实现了最高的召回率（83.9%），但仅保留了56.7%的记录。一次性配置回收的合格记录少于相应的文件批次配置。两次名义上相同的GPT-5.4文件批次运行在91.7%的记录上达成一致，但在94条记录上存在差异，其中29条经过验证的合格记录仅被其中一次运行保留。讨论：LLM的筛选性能依赖于实施的工作流程，而不仅仅是模型本身。因此，处理配置、工作量、记录级别的变异性以及人类与LLM决策的整合是部署系统的重要属性。对于高召回率的任务，LLM更适合于经过验证、可审计的人类监督工作流程，而非自主排除。
cs.AI / 66 / 2608.26889
Learning-Augmented Online Allocation under Unreliable Advice: Robustness, Exposure Fairness, and Distribution Shift
在不可靠建议下的学习增强在线分配：鲁棒性、曝光公平性与分布转移
Abstract
Learning-augmented algorithms improve online decisions using predictions, but unreliable advice may harm efficiency and fairness. We study an online allocation problem with finite candidate sets, irreversible decisions, and exposure constraints. We propose a robust and fair rule combining advice with a conservative fallback and fairness correction. Under bounded-error assumptions, we prove consistency and robustness with loss proportional to prediction error. Experiments show stability under adversarial advice and significant reductions in exposure disparity.
Chinese Translation
学习增强算法通过预测改善在线决策，但不可靠的建议可能会损害效率和公平性。我们研究一个具有有限候选集、不可逆决策和曝光约束的在线分配问题。我们提出了一种结合建议、保守后备和公平修正的鲁棒公平规则。在有界误差假设下，我们证明了该规则的一致性和鲁棒性，其损失与预测误差成正比。实验表明，在对抗性建议下的稳定性以及显著减少曝光差异。
cs.AI / 67 / 2608.26896
AI agents in Algorithmic Electricity Markets: On the Emergence of Tacit Collusion
算法电力市场中的人工智能代理：默契共谋的出现
Abstract
As electricity market participants increasingly adopt learning-based agents for their bidding strategies, electricity markets are becoming algorithmic. Evidence from algorithmic markets in other domains shows that tacit collusion can arise purely through independent learning. Moreover, electricity markets are typically oligopolistic and feature repeated interaction among a small number of participants, making them structurally susceptible to non-competitive behavior. In the face of these observations, this paper investigates the hypothesis that tacit collusion may emerge in electricity markets where participants' actions are controlled by autonomous learning-based algorithms. We model strategic bidding as a repeated game with imperfect public monitoring, and model the participants' emergent behavior using multi-agent reinforcement learning. We propose a multi-dimensional set of criteria (going beyond profit comparisons against Nash equilibria) to assess whether the resulting behavior constitutes tacit collusion. Our experimental results showcase that such a danger is realistic for electricity markets: there are cases where agents do learn to sustain supra-competitive outcomes that are supportive of tacit collusion indicators, even though the agents were never instructed to collude.
Chinese Translation
随着电力市场参与者越来越多地采用基于学习的代理进行竞标策略，电力市场正变得算法化。来自其他领域算法市场的证据表明，默契共谋可以仅通过独立学习而产生。此外，电力市场通常是寡头垄断的，并且参与者之间存在重复互动，这使得其在结构上容易出现非竞争性行为。针对这些观察，本文探讨了一个假设，即在参与者的行为由自主学习算法控制的电力市场中，可能会出现默契共谋。我们将战略竞标建模为一个具有不完美公共监测的重复博弈，并使用多智能体强化学习来建模参与者的涌现行为。我们提出了一套多维标准（超越与纳什均衡的利润比较）来评估所产生的行为是否构成默契共谋。我们的实验结果展示了这种危险在电力市场中是现实的：存在一些情况下，代理确实学习到维持超竞争性结果，这些结果支持默契共谋的指标，即使这些代理从未被指示进行共谋。
cs.AI / 68 / 2608.26899
Counterfactual Bias Testing for Application Tracking System
应用追踪系统的反事实偏差测试
Abstract
Automated candidate-job matching systems are increasingly classified as high-risk AI under emerging regulation, yet auditing them for demographic bias is expensive: classical correspondence-audit studies require hand-crafted resumes and manual submission, which does not scale to fast pipeline retraining cycles. This paper presents a general, reusable methodology that (1) uses task-specialized LLM agents to synthesize identity-neutral base resumes and inject controlled demographic treatments across five protected-characteristic axes (sex/gender, age, residence, language, disability), producing a K x (1+N) correspondence-audit matrix; (2) qualitatively flags inferred protected characteristics per an EU AI Act-aligned prompt; (3) ranks candidates against a job description via a fine-tuned sentence-embedding model and cosine similarity; and (4) computes a nine-metric fairness suite spanning counterfactual (score delta, mean absolute rank change, flip rate), group-fairness (top-K retention, four-fifths/impact ratio), and merit-aware (Recall@K, nDCG@K, equal opportunity, equalized odds) families, each with bootstrap confidence intervals, significance tests, and Benjamini-Hochberg correction, culminating in an automated PASS/INVESTIGATE/FAIL report with a composite risk score. On an example corpus of 5 job orders, 100 base candidates, and 10 demographic treatments (90 metric x variant evaluations): score shifts, top-K retention, and merit-aware rate gaps stay within tolerance for every treatment, but a rank-stability metric (MARC) and nDCG@K each surface borderline findings - including one on the neutral baseline itself - that a score- or retention-only view would miss. The results argue for multi-metric, multi-family auditing over any single aggregate score, and for LLM-agent-generated audits as a practical, low-cost complement to human-curated audits for any candidate-job matching pipeline.
Chinese Translation
自动化候选人与职位匹配系统在新兴法规下越来越被归类为高风险人工智能，然而对其进行人口统计偏差审计的成本较高：传统的对应审计研究需要手工制作简历和手动提交，这在快速的管道再训练周期中无法扩展。本文提出了一种通用的、可重复使用的方法论，(1) 使用任务专用的LLM代理合成身份中立的基础简历，并在五个受保护特征轴（性别、年龄、居住地、语言、残疾）上注入受控的人口统计处理，生成一个K x (1+N)的对应审计矩阵；(2) 根据符合欧盟人工智能法案的提示，定性标记推断出的受保护特征；(3) 通过微调的句子嵌入模型和余弦相似度对候选人进行职位描述排名；(4) 计算涵盖反事实（得分差异、平均绝对排名变化、翻转率）、群体公平（前K保留率、四分之五/影响比）和关注优点（Recall@K、nDCG@K、平等机会、平等赔率）等九个指标的公平性套件，每个指标都有自助法置信区间、显著性测试和Benjamini-Hochberg校正，最终形成自动化的通过/调查/失败报告及综合风险评分。在一个包含5个职位订单、100个基础候选人和10个人口统计处理（90个指标x变体评估）的示例语料库中：得分变化、前K保留率和关注优点的比率差异在每个处理下均保持在容忍范围内，但排名稳定性指标（MARC）和nDCG@K各自显示出边缘发现——包括一个关于中性基线本身的发现——这些发现是仅通过得分或保留视角无法捕捉到的。结果表明，相较于任何单一的综合得分，采用多指标、多类别的审计方法更为合理，并且LLM代理生成的审计可以作为任何候选人与职位匹配管道中人力策划审计的实用、低成本补充。
cs.AI / 69 / 2608.26949
A Table Is Worth 64 Tokens: Pixel-level Compression for Multi-Table Document Question Answering
一张表格价值64个标记：多表文档问答的像素级压缩
Abstract
Answering questions over real-world documents requires processing long inputs that interleave text with tables. Optical context compression, which represents context as images, promises to reduce token cost, but its effect on table understanding remains unclear. We study pixel-level table compression for question answering over documents with multiple tables, evaluating five VLMs across two benchmarks and five visual-token budgets. Representing tables as images at native resolution matches text in both performance and efficiency, but downscaling them makes models compensate the loss in readability with longer, less effective reasoning traces that cancel the expected savings. Highly downscaled tables, however, preserve enough signal to identify whether they are relevant to a question. We exploit this asymmetry with a training-free, two-step method: the model first identifies the tables needed to answer a question from a pixel-compressed context, and then reasons over those at native resolution. On long documents, our method saves 41% of total tokens and gains 7 accuracy points over single-step QA with native resolution tables. It also uses 15% fewer tokens than the most efficient single-step compressed configuration, with no accuracy loss.
Chinese Translation
在现实世界文档中回答问题需要处理长输入，这些输入交错着文本和表格。光学上下文压缩将上下文表示为图像，有望降低标记成本，但其对表格理解的影响仍不明确。我们研究了针对含有多个表格的文档的问答任务的像素级表格压缩，评估了五种视觉语言模型（VLMs）在两个基准和五个视觉标记预算下的表现。将表格以原生分辨率表示为图像在性能和效率上与文本相匹配，但对其进行缩小会导致模型通过更长且效果较差的推理轨迹来弥补可读性损失，从而抵消预期的节省。然而，高度缩小的表格仍然保留了足够的信号以识别其与问题的相关性。我们利用这种不对称性提出了一种无训练的两步法：模型首先从像素压缩的上下文中识别出回答问题所需的表格，然后在原生分辨率下对这些表格进行推理。在长文档上，我们的方法节省了41%的总标记，并在使用原生分辨率表格的单步问答上提高了7个准确率点。它还使用了比最有效的单步压缩配置少15%的标记，且没有准确率损失。
cs.AI / 70 / 2608.26950
From Atomic to Agentic: Towards Interpretable Evaluation of LLMs' Agentic Mathematical Capabilities
从原子到代理：朝着可解释的 LLMs 代理数学能力评估迈进
Abstract
Large Language Models (LLMs) are evolving from performing end-to-end mathematical reasoning to integrating agentic intelligence. However, most existing math benchmarks evaluate only final answers. This outcome-oriented evaluation provides limited diagnostic value for identifying process-level failures or rigorous logic, failing to guide the transformation of LLMs into robust agents. To bridge this gap, we present a process-level benchmark designed to evaluate the inherent agentic mathematical reasoning abilities of LLMs. Our framework aligns problem-solving agentic behaviors with a structured taxonomy of reusable mathematical atomic capabilities. We design a comprehensive suite of planning, action, and feedback tasks across both textual and multimodal contexts, supported by an automated pipeline that synthesizes high-quality trajectories and produces fine-grained annotations via controlled LLM rewriting. Experiments reveal that models with similar end-to-end accuracy can exhibit markedly different agentic capability profiles. This demonstrates that process-level evaluation is crucial for interpreting the true potential of LLMs and guiding the development of next-generation mathematical agents.
Chinese Translation
大型语言模型（LLMs）正在从执行端到端的数学推理演变为整合代理智能。然而，现有的大多数数学基准仅评估最终答案。这种以结果为导向的评估对于识别过程级别的失败或严谨逻辑提供的诊断价值有限，未能指导 LLMs 向强大的代理转变。为了解决这一问题，我们提出了一种旨在评估 LLMs 内在代理数学推理能力的过程级基准。我们的框架将问题解决的代理行为与可重用的数学原子能力的结构化分类法对齐。我们设计了一套全面的规划、行动和反馈任务，涵盖文本和多模态背景，支持一个自动化管道，该管道合成高质量的轨迹并通过受控的 LLM 重写生成细粒度注释。实验表明，具有相似端到端准确性的模型可能表现出截然不同的代理能力特征。这表明，过程级评估对于解读 LLMs 的真实潜力和指导下一代数学代理的发展至关重要。
cs.AI / 71 / 2608.26983
GraphMemix: Query-Aware Evidence Forests for Long-Term Multimodal Agent Memory
GraphMemix：面向查询的证据森林用于长期多模态智能体记忆
Abstract
Organizing long-term memory for multimodal agents remains challenging because existing methods either suffer from expensive question-agnostic offline summaries or naive embedding similarity matching that introduces incomplete and redundant context. To address these issues, we propose GraphMemix, a combinatorial-optimization graph memory framework that models memory organization as query-aware evidence-forest construction. Specifically, our method consists of three key components:(1) candidate graph construction, which expands multi-view seed memories through schema and semantic relations to acquire query-aware original context; (2) evidence utility and activation costs, which decouples direct memory support from anchor-conditioned relation verification to suppress redundant or conflicting information; and (3) forest optimization, which jointly selects a forest-format memory context under a maximum evidence budget and its reliable relational structure. By organizing memory into a query-relevant subgraph, the method avoids substantial lifecycle cost and recovers low-similarity complementary evidence. Experimental results across four long-term multimodal memory benchmarks demonstrate significant improvements with different foundation models and establish a new Pareto frontier between accuracy and lifecycle cost.
Chinese Translation
组织长期记忆以支持多模态智能体仍然面临挑战，因为现有方法要么受到昂贵的与问题无关的离线摘要的影响，要么采用简单的嵌入相似性匹配，从而引入不完整和冗余的上下文。为了解决这些问题，我们提出了GraphMemix，一种组合优化图记忆框架，将记忆组织建模为面向查询的证据森林构建。具体而言，我们的方法由三个关键组成部分构成：(1) 候选图构建，通过模式和语义关系扩展多视角种子记忆，以获取面向查询的原始上下文；(2) 证据效用和激活成本，解耦直接记忆支持与锚定条件关系验证，以抑制冗余或冲突的信息；(3) 森林优化，在最大证据预算及其可靠关系结构下联合选择森林格式的记忆上下文。通过将记忆组织为与查询相关的子图，该方法避免了大量的生命周期成本，并恢复了低相似度的互补证据。在四个长期多模态记忆基准上的实验结果表明，使用不同基础模型的显著改进，并在准确性与生命周期成本之间建立了新的帕累托前沿。
cs.AI / 72 / 2608.26990
DSA: Evidence-Aware LLM-Agent Orchestration for Multi-Market Stock Research
DSA：面向证据的大型语言模型代理在多市场股票研究中的协调
Abstract
Large language models can summarize financial information, but an operational stock-research system must first assemble heterogeneous evidence, expose unavailable data and model capabilities, and control how generated opinions affect a final report. We present DSA, an evidence-aware orchestration framework for multi-market stock research with large language model (LLM) agents. DSA organizes the workflow into evidence acquisition, structured context construction, model-routed analysis, optional role and Strategy Skill reasoning, and report generation with selected context and diagnostics. A default report profile and an optional agentic profile share evidence and model-routing services but use profile-specific output validation and risk safeguards. In the agentic profile, core role outputs are processed by role-specific parsers, whereas Strategy Skill opinions undergo an additional signal-eligibility partition before synthesis; disagreement is supplied explicitly to the decision agent, followed by a conservative risk override. The reference implementation includes six regional market paths, fifteen bundled Strategy Skills, hosted and local model routes, and multiple execution and delivery surfaces. At a frozen software snapshot, a selected manifest of 1,457 portable offline backend contract tests passed; 596 cases were retrospectively mapped to six contract families central to the reported LLM-agent architecture. This evidence establishes implementation conformance for the tested software contracts, not superior report quality, forecasting accuracy, or investment returns.
Chinese Translation
大型语言模型可以总结金融信息，但一个有效的股票研究系统必须首先汇集异构证据，揭示不可用的数据和模型能力，并控制生成的意见如何影响最终报告。我们提出了DSA，一个面向证据的多市场股票研究协调框架，利用大型语言模型（LLM）代理。DSA将工作流程组织为证据获取、结构化上下文构建、模型引导分析、可选角色和策略技能推理，以及使用选定上下文和诊断的报告生成。默认报告配置文件和可选的代理配置文件共享证据和模型路由服务，但使用特定于配置文件的输出验证和风险保障。在代理配置文件中，核心角色输出由特定角色的解析器处理，而策略技能意见在合成之前经历额外的信号资格分区；不一致的意见明确提供给决策代理，随后进行保守的风险覆盖。参考实现包括六个区域市场路径、十五个捆绑的策略技能、托管和本地模型路径，以及多个执行和交付表面。在一个冻结的软件快照中，选定的1,457个可移植离线后端合同测试通过；596个案例被追溯映射到与报告的LLM代理架构相关的六个合同系列。该证据确立了所测试软件合同的实施一致性，而非优越的报告质量、预测准确性或投资回报。
cs.AI / 73 / 2608.26991
ASIL: Replacing Screenshot-and-Click with Structured State and Semantic Actions
ASIL：用结构化状态和语义动作替代截图点击
Abstract
Powerful code agents can execute scripts, call tools, and manage files, yet many important applications remain accessible primarily through graphical user interfaces. We argue that screenshot-and-click is an inefficient interface for software-operating agents: screenshots are state-incomplete, and GUI actions are brittle, semantically weak, and poorly matched to long-horizon planning. We introduce ASIL (Agent-Software Interaction Layer), an agent-native interface that exposes software through structured JSON observations and code-executable semantic actions, realized through the deepest feasible access path for each application. We instantiate ASIL across 15 applications and a benchmark of 300 single-application and 80 multi-application tasks. ASIL reaches above 80 with closed models while executing fewer than five actions per task. Under a repaired runtime and a 50-step screenshot budget, the same tasks yield 6.6 and 26.6 strict success under screenshot-and-click control, rising to 15.0 and 53.3 on an easier OSWorld-comparable band. Against application-native interfaces on matched tasks, ASIL exceeds LibreOffice's UNO API by 28-38 strict points but only matches draw.io's MCP content contract. The structured modality also suits training: small-scale SFT raises Qwen3.5-2B from 58.0 to 72.1 and Qwen3.5-9B from 66.6 to 80.4, and resource-limited on-policy RL further raises them to 74.4 and 82.2.
Chinese Translation
强大的代码代理可以执行脚本、调用工具和管理文件，但许多重要应用仍主要通过图形用户界面访问。我们认为，截图点击是一种低效的软件操作代理接口：截图状态不完整，图形用户界面操作脆弱、语义弱且与长期规划匹配不佳。我们提出了ASIL（Agent-Software Interaction Layer），这是一个代理原生接口，通过结构化的JSON观察和可执行的语义动作来暴露软件，利用每个应用程序的最深可行访问路径。我们在15个应用程序和300个单应用及80个多应用任务的基准测试中实现了ASIL。在执行每个任务少于五个动作的情况下，ASIL在封闭模型下的表现超过80。在修复的运行时和50步截图预算下，相同任务在截图点击控制下的严格成功率分别为6.6和26.6，而在一个更容易的与OSWorld可比的带宽上则上升至15.0和53.3。在匹配任务的应用原生接口中，ASIL的严格成功率超越了LibreOffice的UNO API 28-38点，但仅与draw.io的MCP内容合同持平。结构化模式也适合训练：小规模的SFT将Qwen3.5-2B从58.0提高到72.1，将Qwen3.5-9B从66.6提高到80.4，资源有限的在线策略强化学习进一步将它们提升至74.4和82.2。
cs.AI / 74 / 2608.27010
A Multi-Modal AI Framework for Real-Time Queue Prediction, Management and Optimisation in Intelligent Border Control Systems
用于智能边境控制系统的实时队列预测、管理与优化的多模态人工智能框架
Abstract
In the present work an efficient border control management procedure is proposed. Compared to operational queue management systems, whose operations are based on mostly static data, the proposed work takes into account dynamic traffic conditions, thus enabling optimal performance, even in cases of uncertainty. To this end, we are proposing a multi-modal Artificial Intelligence (AI) framework, tailored to th needs of border control systems, which enables real-time queue prediction, management, and resource optimization. The novel proposed approach integrates heterogeneous data sources and presents them through a unified representation by employing Long Short-Term Memory (LSTM) networks for queue forecasting. Furthermore, it leverages Model Predictive Control (MPC) and scheduling optimization to derive actionable control policies, which in turn can be presented to border control officers. The proposed work has been evaluated using synthetic data simulating realistic traffic. The evaluation results demonstrate that the proposed method reduces queue prediction error by up to 35% and average waiting time by 30%. Accordingly, the average throughput increases by nearly 20%, compared to ARIMA and rule-based methods. The abovementioned results show the effectiveness and efficiency of combining AI architectures with optimization techniques for proactive and adaptive border traffic management.
Chinese Translation
在本研究中，提出了一种高效的边境控制管理程序。与基于大多数静态数据的运营队列管理系统相比，所提出的工作考虑了动态交通条件，从而在不确定情况下也能实现最佳性能。为此，我们提出了一种针对边境控制系统需求量身定制的多模态人工智能（AI）框架，能够实现实时队列预测、管理和资源优化。该新颖的方法集成了异构数据源，并通过采用长短期记忆（LSTM）网络进行队列预测，以统一的方式呈现数据。此外，它利用模型预测控制（MPC）和调度优化来推导可操作的控制策略，这些策略可以呈现给边境控制官员。所提工作使用模拟真实交通的合成数据进行了评估。评估结果表明，所提方法将队列预测误差降低了多达35%，平均等待时间降低了30%。因此，与ARIMA和基于规则的方法相比，平均吞吐量提高了近20%。上述结果表明，将人工智能架构与优化技术相结合在主动和自适应边境交通管理中的有效性和效率。
cs.AI / 75 / 2608.27044
Omni-Interactive Universal Embedder
全交互通用嵌入模型
Abstract
Multimodal representation learning has been shifting from traditional two-tower architectures to large language model (LLM)-based embedders due to their strong instruction-following capabilities. Despite this progress, existing approaches primarily focus on language and image modalities, which also remain the dominant modalities for user-conditioned interactions in current embedders. In this paper, we propose the first Omni-Interactive Universal Embedder (OmniUE), which not only learns a unified embedding space across text, video, and audio by leveraging intermediate-layer representations from dedicated learnable tokens, but also supports omni-interactive querying, enabling users to provide inputs in the form of text, visual regions of interest, and audio spans. Within OmniUE, visual and audio segmenters process diverse user interactions and integrate them with an omni-LLM to produce user-conditioned any-to-any embeddings via context aggregation. To evaluate OmniUE's omni-interactive capabilities, we introduce OmniCHOIR, benchmarking models for omni-interactive compositional audio retrieval based on the given text, video, and audio as well as unimodal or multimodal interaction prompts. OmniUE consistently surpasses state-of-the-art baselines across diverse modalities, with average improvements of 10.5% on textual-interactive video benchmarks (MMEB-v2-video), 1.1% on audio tasks (MAEB), 83.7% on visual-interactive benchmarks (SCaR), and 24.1% on our omni-interactive OmniCHOIR benchmark. We believe that jointly advancing omni-modal representation learning and omni-interactive querying paves the way toward universal embedders.
Chinese Translation
多模态表示学习正从传统的双塔架构转向基于大型语言模型（LLM）的嵌入模型，这得益于其强大的指令跟随能力。尽管取得了这些进展，现有方法主要集中在语言和图像模态上，这两者仍然是当前嵌入模型中用户条件交互的主导模态。本文提出了首个全交互通用嵌入模型（OmniUE），该模型不仅通过利用专用可学习标记的中间层表示学习跨文本、视频和音频的统一嵌入空间，还支持全交互查询，使用户能够以文本、视觉兴趣区域和音频片段的形式提供输入。在OmniUE中，视觉和音频分割器处理多样的用户交互，并将其与全LLM集成，通过上下文聚合生成用户条件的任意到任意嵌入。为了评估OmniUE的全交互能力，我们引入了OmniCHOIR，基准模型用于基于给定文本、视频和音频以及单模态或多模态交互提示的全交互组合音频检索。OmniUE在多种模态上始终超越最先进的基线，在文本交互视频基准（MMEB-v2-video）上平均提高10.5%，在音频任务（MAEB）上提高1.1%，在视觉交互基准（SCaR）上提高83.7%，在我们的全交互OmniCHOIR基准上提高24.1%。我们相信，联合推进全模态表示学习和全交互查询为通用嵌入模型铺平了道路。
cs.AI / 76 / 2608.27086
A Contract-Centered Architecture for Scalable and Manageable Agentic Runtimes
以合同为中心的可扩展和可管理的智能运行时架构
Abstract
Enterprise AI deployment is a coordination problem across business units, application and AI teams, testing, platform engineering, infrastructure, security, operations, and data governance. Use-case benchmarks show whether one agent completes one task, but not how changing capabilities, models, runtime mechanisms, capacity, and enterprise data should be owned, changed, admitted, or evidenced together. We present four responsibility objects as shared organizational contracts: Skill (reusable, versioned capability and workflow asset), Harness (runtime compiler and governor), Scaffold (execution/control boundary and NFR owner), and a stack-external data substrate under independent CIO-governed semantics and telemetry. The runtime core is A = , with the data substrate outside that stack. The central contribution is one bounded, falsifiable hypothesis, P1 (cost-aware capability-capacity separability): within a declared operating region, changing activated capability preserves the capacity-response interaction within a preregistered equivalence margin, while changing compatible Scaffold capacity preserves capability semantics up to a non-inferiority margin, and the required controls stay within a declared enforcement budget. Six design conditions become measured obligations whose coverage, violations, uncertainty, cost, and exclusions determine whether P1 is decidable. We propose a cluster-period randomized crossover experiment (balanced order, reset/washout, repeated seeds and failure regimes, cluster-aware uncertainty) with a four-state verdict: supported, falsified, conditional-engineering, or inconclusive. This paper contributes a contract-bounded runtime architecture, a source-preserving data substrate, and a falsifiable measurement protocol. It reports no completed implementation, experiment, dataset, or measured result.
Chinese Translation
企业人工智能的部署是一个跨业务单元、应用程序和人工智能团队、测试、平台工程、基础设施、安全、运营和数据治理的协调问题。用例基准显示一个智能体是否完成一项任务，但并未说明如何共同拥有、改变、接受或证明能力、模型、运行机制、容量和企业数据的变化。我们提出四个责任对象作为共享的组织合同：技能（可重用、版本化的能力和工作流资产）、控制器（运行时编译器和管理者）、支架（执行/控制边界和非功能性需求所有者），以及一个在独立首席信息官治理的语义和遥测下的堆栈外数据基础。运行时核心为 A = ，数据基础位于该堆栈之外。核心贡献是一个有界的、可证伪的假设 P1（成本意识能力-容量可分离性）：在声明的操作区域内，改变激活的能力保持容量-响应交互在预注册的等价边界内，而改变兼容的支架容量保持能力语义不低于一个非劣边界，并且所需的控制保持在声明的执行预算内。六个设计条件成为可测量的义务，其覆盖、违规、不确定性、成本和排除决定了 P1 是否可判定。我们提出一个集群周期随机交叉实验（平衡顺序、重置/洗脱、重复种子和失败机制、集群感知不确定性），其结果为四种状态：支持、被证伪、条件工程或不确定。本文贡献了一个合同界定的运行时架构、一个保持源数据的基础以及一个可证伪的测量协议。报告中没有完成的实现、实验、数据集或测量结果。
cs.AI / 77 / 2608.27101
pro-team at LLMs4OL 2026 Tasks Flagship and Reuse: Retrieval-Augmented Generation and Vocabulary-Constrained Filtering for Ontology Learning
pro-team在LLMs4OL 2026任务旗舰与重用中的表现：用于本体学习的检索增强生成与词汇约束过滤
Abstract
Ontology learning from text remains challenging despite significant progress in Large Language Models (LLMs), which can hallucinate domain terms, produce inconsistent formats, and favor hierarchical over associative relations. In the LLMs4OL 2026 Challenge, we address both the End-to-End Flagship Task (Task A) and Ontology Extension Reuse Task (Task B) using an offline retrieval-augmented few-shot prompting pipeline. Our system employs Qwen2.5-14B-Instruct with all-MiniLM-L6-v2 for demonstration retrieval, selecting the top-5 examples for Task A and top-2 for Task B. A left-truncated context-windowing strategy preserves task instructions within long prompts. For Task B, generated triples undergo deterministic vocabulary-constrained filtering, retaining triples when at least one endpoint belongs to the sample's closed term/type vocabulary and removing duplicates of the initial ontology. The approach achieves Semantic Graph Similarity of 0.8692, Term-Typing F1 of 0.9200, and Taxonomy Discovery F1 of 0.8540 on Task B, while Task A achieves 0.7416 Semantic Graph Similarity. However, no non-taxonomic relations are extracted, highlighting limitations of closed, taxonomy-oriented relation vocabularies.
Chinese Translation
尽管大型语言模型（LLMs）取得了显著进展，从文本中进行本体学习仍然具有挑战性，这些模型可能会虚构领域术语、产生不一致的格式，并偏向于层级关系而非关联关系。在LLMs4OL 2026挑战中，我们使用离线检索增强的少量提示管道解决了端到端旗舰任务（任务A）和本体扩展重用任务（任务B）。我们的系统采用Qwen2.5-14B-Instruct与all-MiniLM-L6-v2进行演示检索，为任务A选择前5个示例，为任务B选择前2个示例。左截断的上下文窗口策略在长提示中保留任务指令。对于任务B，生成的三元组经过确定性词汇约束过滤，当至少一个端点属于样本的封闭术语/类型词汇时保留三元组，并去除初始本体的重复项。该方法在任务B上实现了0.8692的语义图相似度、0.9200的术语类型F1和0.8540的分类发现F1，而任务A的语义图相似度为0.7416。然而，没有提取出非分类关系，突显了封闭的、以分类为导向的关系词汇的局限性。
cs.AI / 78 / 2608.27102
LAAF: A Layered Accountability Architecture Framework for LLM Applications
LAAF：用于大型语言模型应用的分层问责架构框架
Abstract
Large Language Models (LLMs) operate in hospitals, courtrooms, banks, and public service desks, where fluent, confident outputs are treated as authoritative even when ungrounded or incorrect. When such an output contributes to harm, who is answerable, and through what mechanisms can responsibility be traced, explained, and acted upon? Following PRISMA guidance, five databases were searched from January 2022 to March 2026 against four review questions; of 4,512 records identified, 122 primary studies were included, together with 12 regulatory and standards documents analysed as primary sources. The review consolidates a sociotechnical account of accountability as an actor-forum relation resolved into five dimensions, and synthesises mechanisms across four families: technical controls, human oversight, organisational governance, and documentation and traceability, each with a maturity assessment. The corpus is read through a four-layer classification device spanning provenance, application logic, human oversight, and governance and redress, cross-cut by traceability, role clarity, and continuous monitoring. Both are mapped onto the EU AI Act, whose high-risk obligations have applied since 2 August 2026, the NIST AI RMF with its Generative AI Profile, ISO/IEC 42001, and sectoral guidance in healthcare, consumer finance, education, and the public sector. Four persistent gaps emerge: under-specification of human oversight, absence of shared accountability metrics, disciplinary disconnection, and limited empirical evaluation, alongside five structural tensions that no surveyed instrument resolves. The review closes by consolidating the classification device into an integrated accountability architecture, LAAF, with cybersecurity aligned to the OWASP LLM Top 10 (2025); it is a synthesis of the surveyed evidence rather than a validated artefact.
Chinese Translation
大型语言模型（LLMs）在医院、法庭、银行和公共服务台等场所运作，其中流畅、自信的输出被视为权威，即使这些输出没有根据或不正确。当这样的输出导致伤害时，谁应对此负责，以及通过什么机制可以追踪、解释和采取责任？根据PRISMA指导，从2022年1月到2026年3月对五个数据库进行了搜索，针对四个审查问题；在识别的4,512条记录中，纳入了122项主要研究，以及12份作为主要来源的法规和标准文件的分析。该审查整合了问责制的社会技术视角，作为一种行为者-论坛关系，分解为五个维度，并综合了四个类别的机制：技术控制、人为监督、组织治理，以及文档和可追溯性，每个类别都有成熟度评估。该语料库通过一个涵盖来源、应用逻辑、人为监督和治理与救济的四层分类工具进行解读，并交叉考虑可追溯性、角色明确性和持续监测。这两者都映射到欧盟人工智能法案（EU AI Act），其高风险义务自2026年8月2日起生效，以及NIST人工智能风险管理框架（NIST AI RMF）及其生成性人工智能概况、ISO/IEC 42001，以及医疗、消费者金融、教育和公共部门的行业指导。出现了四个持续的缺口：对人为监督的不足规范、缺乏共享的问责指标、学科间的脱节，以及有限的实证评估，同时还有五个没有任何调查工具解决的结构性紧张。该审查最后将分类工具整合为一个综合问责架构LAAF，并将网络安全与OWASP LLM前10名（2025）对齐；这是一种对调查证据的综合，而不是经过验证的成果。
cs.AI / 79 / 2608.27127
TransMeme: A Multi-Agent Framework for Cross-Cultural Meme Transcreation
TransMeme：跨文化表情包再创作的多智能体框架
Abstract
Internet memes are a pervasive form of multimodal online communication; however, such communication often involves users from diverse linguistic and cultural backgrounds. Therefore, adapting memes across cultures and languages is a central challenge for enabling mutual understanding in online communication. Unlike ordinary translation or standalone text rewriting, cross-cultural meme transcreation must jointly preserve communicative intent, adapt culture-dependent meaning for the target audience, and maintain coherence between text and image. In this work, we first provide an explicit task analysis of cross-cultural meme transcreation and identify three core challenges: culture-specific knowledge understanding, intent and tone preservation, and multimodal consistency. Based on this analysis, we propose a multi-agent framework with specialized agents that are coordinated to address these challenges through cultural adaptation, target text rewriting, revision, and conditional visual adjustment. The framework strengthens target text adaptation with coordinated feedback to handle difficult cases that require deeper cultural or visual intervention. We evaluate the framework on bidirectional Chinese-English meme transcreation using both human evaluation and LLM-as-a-Judge. Our method consistently outperforms all baselines across both evaluation settings. In human evaluation, it achieves the best performance on all four dimensions and delivers a 33.1% average improvement over the strongest baseline, while in LLM-as-a-Judge, it attains the highest Top-1 ranking rate (60% versus 26% for the second-best baseline). Further analysis indicates that each component contributes to the performance. Our error analysis suggests that the remaining bottlenecks lie in humor reconstruction and image-text alignment rather than simple cultural knowledge gaps, pointing to future work on humor transfer.
Chinese Translation
互联网表情包是一种普遍存在的多模态在线交流形式；然而，这种交流通常涉及来自不同语言和文化背景的用户。因此，跨文化和跨语言的表情包适配是促进在线交流中相互理解的核心挑战。与普通翻译或独立文本重写不同，跨文化表情包再创作必须共同保留交流意图，为目标受众适配文化依赖的意义，并保持文本与图像之间的连贯性。在本研究中，我们首先对跨文化表情包再创作进行了明确的任务分析，并识别出三个核心挑战：特定文化知识的理解、意图和语气的保留，以及多模态的一致性。基于此分析，我们提出了一个多智能体框架，配备专门的智能体，协调应对这些挑战，通过文化适配、目标文本重写、修订和条件视觉调整来实现。该框架通过协调反馈增强目标文本的适配，以处理需要更深入文化或视觉干预的困难案例。我们在双向中英文表情包再创作上评估了该框架，采用了人类评估和LLM作为评判者的方法。我们的方案在两个评估设置中均持续优于所有基线。在人类评估中，它在所有四个维度上表现最佳，平均提高了33.1%，超过最强基线；而在LLM作为评判者的评估中，它达到了最高的Top-1排名率（60%对比第二强基线的26%）。进一步分析表明，每个组件都对性能有所贡献。我们的错误分析表明，剩余的瓶颈在于幽默重构和图像-文本对齐，而非简单的文化知识差距，这指向了未来在幽默转移方面的研究方向。
cs.AI / 80 / 2608.27142
GRAIN: Bridging Name and Narrative Shifts in Real-World Graph Reasoning through Invariance-Rewarded Agentic RL
GRAIN：通过不变性奖励的自主强化学习在现实世界图推理中桥接名称和叙事的转变
Abstract
Despite their potential in standardized graph tasks, Large Language Models (LLMs) remain brittle to real-world shifts in node identifiers and task formulation. While deterministic graph tools are invariant to such shifts, extracting topological structures from noisy text is highly fragile for LLMs, which often overfit to surface patterns. Moreover, mitigating these parsing failures via multi-agent systems incurs prohibitive latency. To address this, we propose GRAIN, a single-agent framework optimized via reinforcement learning. GRAIN models reasoning as a semantic parsing and tool-execution pipeline, guided by a Structure Invariance Reward. By validating extracted intermediate graphs against ground-truth topologies, this reward forces the LLM to learn robust text-to-structure mappings rather than memorizing linguistic artifacts. We also introduce GRIT, a benchmark evaluating sensitivity to such linguistic shifts. GRAIN outperforms multi-agent baselines by 16.45\% in accuracy with approximately 24\% lower latency. Furthermore, it demonstrates superior structural generalization, halving the out-of-distribution (OOD) gap of SFT models (from 15.77\% to 7.80\%) and maintaining robustness on large-scale graphs beyond the training distribution.
Chinese Translation
尽管大型语言模型（LLMs）在标准化图任务中具有潜力，但它们在节点标识符和任务表述的现实世界转变面前仍然脆弱。虽然确定性图工具对这些转变是不变的，但从嘈杂文本中提取拓扑结构对LLMs来说极其脆弱，LLMs往往过度拟合表面模式。此外，通过多智能体系统来缓解这些解析失败会导致不可接受的延迟。为了解决这个问题，我们提出了GRAIN，一个通过强化学习优化的单智能体框架。GRAIN将推理建模为一个语义解析和工具执行的管道，受到结构不变性奖励的指导。通过验证提取的中间图与真实拓扑的匹配，这一奖励迫使LLM学习稳健的文本到结构的映射，而不是记忆语言学伪影。我们还引入了GRIT，一个评估对这种语言学转变敏感性的基准。GRAIN在准确性上比多智能体基线提高了16.45%，并且延迟降低了约24%。此外，它展示了优越的结构泛化能力，将SFT模型的分布外（OOD）差距减半（从15.77%降至7.80%），并在超出训练分布的大规模图上保持稳健性。
cs.AI / 81 / 2608.27144
Feature Transformation Enhanced Jacobi Polynomial Graph Filtering for Graph Anomaly Detection
特征变换增强的雅可比多项式图滤波用于图异常检测
Abstract
In recent years, graph anomaly detection (GAD) based on frequency-domain filtering have achieved promising results. However, existing approaches still face three major challenges: First, they use static basic function to constructed graph filter which cannot effectively adapt to the frequency-domain distribution of graph data. Second, they fail to adequately consider the importance information of each attribute in the node feature vector, leading to the loss of fine-grained information. Third, they insufficiently utilize node labels for GAD. To address these issues, this paper proposes a novel graph anomaly detection method called JPGFN (Feature Transformation Enhanced Jacobi Polynomial Graph Filtering Network). First, a Feature Separation Transformation Network (FSTNN) is developed to better learn fine-grained node features by feature separation and applying nonlinear transformations to node features across different dimensions. Second, an adaptive Jacobi polynomial graph filtering module is constructed based on Jacobi polynomials to adaptively capture complex frequency-domain features of graph signals. Finally, a node label constraint module is developed to facilitate the use of node labels and enhance the performance of GAD. Experimental results on multiple real-world datasets demonstrate that the proposed method significantly outperforms mainstream approaches.
Chinese Translation
近年来，基于频域滤波的图异常检测（GAD）取得了令人鼓舞的成果。然而，现有方法仍面临三个主要挑战：首先，它们使用静态基本函数构建图滤波器，无法有效适应图数据的频域分布。其次，它们未能充分考虑节点特征向量中每个属性的重要信息，导致细粒度信息的丢失。第三，它们对节点标签在GAD中的利用不足。为了解决这些问题，本文提出了一种新颖的图异常检测方法，称为JPGFN（特征变换增强的雅可比多项式图滤波网络）。首先，开发了一种特征分离变换网络（FSTNN），通过特征分离和对不同维度的节点特征应用非线性变换，更好地学习细粒度的节点特征。其次，基于雅可比多项式构建了一个自适应雅可比多项式图滤波模块，以自适应捕捉图信号的复杂频域特征。最后，开发了一个节点标签约束模块，以促进节点标签的使用并增强GAD的性能。在多个真实世界数据集上的实验结果表明，所提出的方法显著优于主流方法。
cs.AI / 82 / 2608.27146
When Tool Outputs Become Commands: Separating Action Induction from Runtime Authorization in Tool-Augmented LLM Agents
当工具输出变为命令：在工具增强的LLM代理中将动作诱导与运行时授权分离
Abstract
Tool-augmented LLM agents must rely on untrusted runtime Observations to complete open-ended tasks; however, when tool outputs no longer merely provide data but begin to specify concrete actions, they effectively become ``commands'' that can drive real-world side effects beyond user intent. We argue that this risk arises from conflating action induction with execution authorization. To address this distinction, we propose SARA, which treats action induction and execution authorization as distinct runtime roles and separates action provenance from execution authority. On the Observation side, a context-isolated Action Probe exposes action-inducing semantics and persistently records action-origin provenance across steps as a review signal; on the execution side, actual tool calls are authorized only against the user objective and audited evidence from authorized successful executions, while satisfying goal, execution-chain, and argument-level support. To preserve this separation across multi-step execution, SARA applies No-History-Promotion to prevent historical recurrence from laundering action origins into execution authority. Across AgentDojo and AgentDyn, SARA limits ASR to no more than \(0.63\%\) across four primary evaluation settings while maintaining competitive task utility, and consistently reduces ASR across additional Agent backbones.
Chinese Translation
工具增强的LLM代理必须依赖不可信的运行时观察来完成开放式任务；然而，当工具输出不再仅仅提供数据而开始指定具体行动时，它们实际上变成了可以驱动超出用户意图的现实世界副作用的“命令”。我们认为，这一风险源于将动作诱导与执行授权混为一谈。为了解决这一区别，我们提出了SARA，它将动作诱导和执行授权视为不同的运行时角色，并将动作来源与执行权威分离。在观察方面，上下文隔离的动作探测器（Action Probe）揭示了诱导动作的语义，并在各步骤中持久记录动作来源的证据作为审查信号；在执行方面，实际的工具调用仅根据用户目标和来自授权成功执行的审计证据进行授权，同时满足目标、执行链和参数级别的支持。为了在多步骤执行中保持这种分离，SARA应用了无历史提升（No-History-Promotion）以防止历史重现将动作来源洗白为执行权威。在AgentDojo和AgentDyn中，SARA将ASR限制在四个主要评估设置中不超过0.63%，同时保持竞争性的任务效用，并在额外的Agent骨干网络中持续降低ASR。
cs.AI / 83 / 2608.27147
Thomson: Continual Learning of Frontier Models for SovereignAI
汤姆森：主权人工智能前沿模型的持续学习
Chen, Shengzhuang, Parker, Jerrod, Bang, Yejin, Bean, Andrew M., Seedat, Nabeel, Winzeck, Stefan, Glazko, Daniil, Zgraggen, Jannik, Yu, Fangyi, Arnott, Scott, Trautmann, Dietrich, Ciuffreda, Luca, Bonifazi, Guglielmo, Romano, Davide, Bell, Bradley, Fielding, Kirsty, Giofrè, Daniele, Zielund, Tom, Chatterjee, Ipshita, Ghantasala, Sneha Murthy, Nanreh, Manpreet, Scoville, John, Sakowicz, Maciej, Seifeddine, Wassim, Thede, Lukas, Schwarz, Jonathan Richard
Abstract
The development of frontier models is commonly perceived to be the exclusive remit of a small number of heavily funded players, creating an information, economic and power asymmetry between developers and the diverse user base of modern AI. Recent public discourse acknowledges this concern, calling for SovereignAI (an organisation's capability to independently build, deploy and govern AI use), but offers little concrete advice on how this can be achieved in the short term under a diversity of funding settings. We argue that frontier performance is achievable by a wide range of institutions through Continual Learning on readily available open-weight models. Unlike limited approaches such as small-scale fine-tuning, prompt engineering, or tool-augmentation of a frozen model, our approach exploits a modern mid- & post-training stack while introducing safeguards that preserve both plasticity and stability at each stage, making the minimal number of high-impact interventions on the parameters. This yields gains comparable to those typically seen across multiple successive model generations, at compute and personnel budgets substantially lower than commonly thought, making ownership of large parts of the SovereignAI stack (model, tool infrastructure, values & data privacy) viable for far more actors. We demonstrate this with Thomson, a general-purpose frontier model trained with an enhanced focus on high-stakes professional work. Thomson performs competitively with recent frontier models across agentic tasks, safety, legal, tax & multilingualism, and large-scale Deep Research. Evaluations show a distinctive $\pi$-shaped pattern: distinct improvements across a wide range of capabilities, including those not explicitly targeted, while almost completely eliminating the forgetting problem common to narrow domain adaptation.
Chinese Translation
前沿模型的发展通常被视为少数资金雄厚的参与者的专属领域，这在开发者与现代人工智能多样化用户群体之间造成了信息、经济和权力的不对称。近期的公共讨论承认了这一担忧，呼吁建立主权人工智能（组织独立构建、部署和管理人工智能使用的能力），但在多样化的资金环境下，提供的具体建议却寥寥无几。我们认为，通过对现成的开放权重模型进行持续学习，各类机构都能实现前沿性能。与小规模微调、提示工程或对冻结模型的工具增强等有限方法不同，我们的方法利用现代的中期和后期训练架构，同时引入保护措施，在每个阶段保持可塑性和稳定性，从而对参数进行最小数量的高影响干预。这带来了与通常在多个连续模型世代中看到的增益相当的效果，而所需的计算和人力预算远低于普遍预期，使得更多参与者能够拥有主权人工智能堆栈的大部分（模型、工具基础设施、价值观和数据隐私）。我们通过汤姆森（Thomson）这一通用前沿模型进行演示，该模型在高风险专业工作上进行了增强的训练。汤姆森在代理任务、安全性、法律、税务和多语言能力以及大规模深度研究方面与近期的前沿模型表现出竞争力。评估显示出独特的$ ext{π}$形模式：在广泛能力范围内，包括那些未明确针对的能力，均有明显改善，同时几乎完全消除了狭域适应中常见的遗忘问题。
cs.AI / 84 / 2608.27149
BPMN4CAI: A BPMN Extension for Modeling Dynamic Conversational AI
BPMN4CAI：用于建模动态对话式人工智能的BPMN扩展
Abstract
Conversational AI systems, such as chatbots and virtual assistants, are becoming increasingly important to digital business processes. However, the established Business Process Model and Notation (BPMN) standard faces challenges when representing dynamic, context-sensitive interactions. This paper addresses this methodological and practical research gap by developing a standard-compliant BPMN extension (BPMN4CAI). Using Design Science Research methodology, this paper develops an approach that systematically extends existing BPMN elements and incorporates specialized components. The applicability and relevance of the BPMN4CAI framework are demonstrated and evaluated through a case study. The results show that the BPMN4CAI extension facilitates adaptive decision-making processes, robust context management, and transparent interactions for Conversational AI within business processes.
Chinese Translation
对话式人工智能系统，如聊天机器人和虚拟助手，在数字业务流程中变得越来越重要。然而，现有的业务流程模型和符号（BPMN）标准在表示动态、上下文敏感的交互时面临挑战。本文通过开发一个符合标准的BPMN扩展（BPMN4CAI）来解决这一方法论和实践研究的空白。采用设计科学研究方法，本文系统性地扩展现有的BPMN元素，并纳入专业组件。通过案例研究展示和评估BPMN4CAI框架的适用性和相关性。结果表明，BPMN4CAI扩展促进了对话式人工智能在业务流程中的自适应决策过程、稳健的上下文管理和透明的交互。
cs.AI / 85 / 2608.27167
Calibrated Enough to Know, Not Calibrated to Act: Fabricated Evidence Makes LLM Agents Commit to the Unknowable
足够校准以了解，但未校准以行动：伪造证据使大型语言模型代理承诺于不可知之事
Abstract
An LLM agent shown a professional-looking market panel commits to a directional call on a provably unpredictable question far more often than one asked the bare question: across 12 frontier models, commitment rises from 6.5% to 54.0% as evidence is escalated. It commits just as readily when every number on the panel is invented: fabricating the entire display, so nothing the model can see is true except the question itself, still lifts commitment from 24.5% to 36.8%, statistically indistinguishable from the 37.6% produced by genuine market data. What unlocks confident action is not information but the authority of its packaging. The failure is narrow and locatable. Incapacity is not the answer: on matched answerable questions attached to the same panels, the same models answer essentially always, at near-perfect accuracy. Nor is it belief - stated probabilities barely move across the gradient that swings action by 48 points, and score worse than a climatological baseline. Missing judgment isn't it either: asked to classify a question's knowability before acting, models call it irreducible 90% of the time and then commit on just 0.4% of those. The act/don't-act gate is what fails, and the effect is concentrated in a few models rather than universal. Because the gate is separable, it can be trained. Supervised fine-tuning of a 3B model on 540 synthetic cases, predominantly dice, coins, jars and timers, drives commitment to 0.0% on the original cases and transfers to three unseen domains. It does not survive everything: the gate holds exactly when the response format leaves room to reason, and rigid formats that remove that room leave the model confident and wrong on questions it otherwise answers correctly. The gate is trainable and context-fragile, and deployment needs both halves of that sentence.
Chinese Translation
一个被展示专业外观市场面板的大型语言模型（LLM）代理在一个可证明不可预测的问题上做出方向性判断的频率远高于仅被问及简单问题的代理：在12个前沿模型中，随着证据的升级，承诺率从6.5%上升至54.0%。即使面板上的每个数字都是虚构的，它也同样容易做出承诺：伪造整个展示，使得模型所能看到的除了问题本身外没有任何真实信息，仍然使承诺率从24.5%上升至36.8%，在统计上与真实市场数据产生的37.6%没有显著差异。促使自信行动的不是信息，而是其包装的权威性。失败是狭窄且可定位的。无能并不是答案：在附加到相同面板的可回答问题上，相同模型几乎总是以接近完美的准确性作答。信念也不是原因——所陈述的概率在影响行动的48个点的梯度上几乎没有变化，且表现不如气候基线。缺乏判断也不是问题：在行动之前被要求判断一个问题的可知性时，模型90%的时间认为其不可简化，然后仅在这些问题上承诺0.4%。行动/不行动的门是失败的关键，且这一效应集中在少数模型上，而非普遍存在。由于该门是可分离的，因此可以进行训练。在540个合成案例上对一个3B模型进行监督微调，主要是骰子、硬币、罐子和计时器，将原始案例的承诺率驱动至0.0%，并转移到三个未见领域。它并不能应对所有情况：当响应格式留有推理空间时，该门恰好保持有效，而移除该空间的严格格式使模型在本应正确回答的问题上自信但错误。该门是可训练的且对上下文脆弱，部署需要句子的两个部分。
cs.AI / 86 / 2608.27260
What Makes Good Agentic Data? An ACE Lens on Data Generation for LLM Agents
什么是优质的自主数据？从ACE视角看大型语言模型代理的数据生成
Abstract
LLM agents increasingly rely on generated interaction data to learn how to interact with external environments. Agentic data generation must maintain consistency among environments, tasks, interactions, and success signals while producing experience that is useful rather than merely abundant. Existing work spans many agent domains, but domain-centered organization and heterogeneous evaluation often obscure common generation mechanisms and conflate candidate construction with verification and selection. This work develops a two-level framework for the field. First, we represent agentic data as a common factorized object $(E,q,\tau,v)$, comprising an environment specification, task signal, interaction realization, and optional verifier. We organize generation paradigms by their primary anchor and dependency structure. Second, we formulate generation as constrained distribution design through the Accuracy-Complexity-divErsity (ACE) lens. Accuracy establishes the feasible support of grounded and internally consistent data. Within this support, Complexity places learning mass relative to the capability of a declared learner and execution configuration, while divErsity controls coverage and redundancy of data. Using this framework, we explore how prior work verifies generated experience, constructs and calibrates difficulty, and expands behavioral coverage. The literature reveals a shift toward execution-grounded accuracy, learner-relative complexity, and diversity beyond surface variation or dataset size. We further discuss broader directions and emerging trends in agentic data generation through the ACE lens, including their implications for scaling, data sources, training regimes and adaptive learning. Overall, the central challenge is not simply to generate more data, but to continually allocate valid, informative, and non-redundant experience as agents and environments evolve.
Chinese Translation
大型语言模型（LLM）代理越来越依赖生成的交互数据来学习如何与外部环境互动。自主数据生成必须在环境、任务、交互和成功信号之间保持一致性，同时产生有用的经验，而不仅仅是数量庞大的数据。现有研究涵盖了许多代理领域，但以领域为中心的组织和异质评估常常模糊了共同的生成机制，并将候选构建与验证和选择混为一谈。本研究为该领域发展了一个两级框架。首先，我们将自主数据表示为一个共同的因子化对象 $(E,q, au,v)$，包括环境规范、任务信号、交互实现和可选的验证器。我们根据其主要锚点和依赖结构组织生成范式。其次，我们通过准确性-复杂性-多样性（Accuracy-Complexity-divErsity, ACE）视角将生成形式化为受限分布设计。准确性确立了基础且内部一致的数据的可行支持。在此支持内，复杂性相对于声明的学习者和执行配置的能力放置学习质量，而多样性控制数据的覆盖范围和冗余性。利用这一框架，我们探讨了先前的研究如何验证生成的经验、构建和校准难度，以及扩展行为覆盖。文献揭示了向执行基础准确性、学习者相对复杂性和超越表面变化或数据集规模的多样性的转变。我们进一步通过ACE视角讨论自主数据生成的更广泛方向和新兴趋势，包括它们对扩展、数据来源、训练机制和自适应学习的影响。总体而言，中心挑战不仅仅是生成更多数据，而是在代理和环境不断演变的过程中，持续分配有效、信息丰富且不冗余的经验。
cs.AI / 87 / 2608.27266
Naive Prompt Optimization: Rethinking the Need for Complex Prompt Search
简单提示优化：重新思考复杂提示搜索的必要性
Abstract
Efficiently improving autonomous agents across diverse tasks is central to accelerating recursive self-improvement (RSI) in agentic AI, with prompt optimization emerging as a promising approach capable of delivering performance gains comparable to those achieved by fine-tuning model weights, while reducing computational costs in both optimization and serving. However, recent developments increasingly favor unnecessarily complex prompt optimizers. We introduce Naive Prompt Optimization (NPO), a lightweight single-lineage method that iteratively revises prompts using a teacher model with rollout feedback. NPO achieves comparable or better performance than GEPA with fewer rollouts, and its advantage increases with stronger teacher models, suggesting that stronger teacher reasoning can partially substitute for optimizer-side search complexity. In interactive games, NPO remains broadly competitive with GEPA, while GRPO performs better on some tasks less amenable to prompt optimization. We also show that NPO-optimized prompts elicit similar performance improvements when applied verbatim to other student models, especially across models within the same family. Overall, our preliminary results show that simple, linear prompt optimization can rival substantially more sophisticated and complex search procedures.
Chinese Translation
高效提升自主智能体在多样任务中的表现是加速智能体人工智能递归自我改进（RSI）的核心，而提示优化作为一种有前景的方法，能够提供与微调模型权重相当的性能提升，同时降低优化和服务过程中的计算成本。然而，最近的发展越来越倾向于不必要的复杂提示优化器。我们提出了简单提示优化（Naive Prompt Optimization, NPO），这是一种轻量级的单谱系方法，通过使用带有回滚反馈的教师模型迭代修订提示。NPO在回滚次数更少的情况下，达到了与GEPA相当或更好的性能，其优势随着教师模型的增强而增加，这表明更强的教师推理可以部分替代优化器端的搜索复杂性。在互动游戏中，NPO与GEPA保持广泛的竞争力，而GRPO在一些不太适合提示优化的任务上表现更好。我们还展示了NPO优化的提示在逐字应用于其他学生模型时能够引发类似的性能提升，特别是在同一家族的模型之间。总体而言，我们的初步结果表明，简单的线性提示优化可以与更复杂和精细的搜索程序相媲美。
cs.AI / 88 / 2608.27268
BrailleBench: Investigating Multi-Criteria Braille Comprehension in Large Language Models
盲文基准：研究大型语言模型中的多标准盲文理解
Abstract
Although Large language models (LLMs) mediate access to knowledge and computational assistance, their capabilities should benefit vulnerable groups in the same way. However, it is unclear whether existing AI systems are inclusive enough for blind and deafblind users to access the same functionality through Braille, whose indicators, contractions, and digital representations introduce distinct requirements for model comprehension. To this end, we introduce BrailleBench, a benchmark for evaluating LLMs in Braille comprehension from different Criteria. BrailleBench aligns 5,570 instances from five datasets, including mathematics, commonsense, and multi-hop question answering across English and Braille Grades 1 and 2. Different configurations are designed to understand whether the systems can comprehend Braille-authored content, express answers in Braille, and complete end-to-end Braille interaction. To ensure the quality and prevent evaluation bias, the benchmark is built through a deterministic, expert-reviewed pipeline via a self-created Braille Toolkit without using any data instances generated by LLMs. We evaluate six representative LLMs from various aspects. The results reveal a persistent gap between print-English capability and Braille accessibility. Braille understanding and expression are asymmetric, where Grade 2 is especially fragile on the input side compared to Grade 1, and fully Braille requests further reduce performance. The experimental observations provide valuable guidance for the development of future Braille AI systems. All related resources in BrailleBench are publicly available for future research.
Chinese Translation
尽管大型语言模型（LLMs）在知识获取和计算辅助方面发挥着重要作用，但其能力是否能同样惠及弱势群体仍不明确。然而，目前的人工智能系统是否足够包容，以便盲人和听障盲人用户通过盲文访问相同的功能尚不清楚，因为盲文的指示符、缩写和数字表示形式对模型理解提出了独特的要求。为此，我们引入了盲文基准（BrailleBench），这是一个用于从不同标准评估LLMs在盲文理解方面表现的基准。盲文基准对来自五个数据集的5,570个实例进行了对齐，这些数据集包括数学、常识和多跳问答，涵盖英语及盲文等级1和2。设计了不同的配置，以了解系统是否能够理解盲文创作的内容、以盲文表达答案，以及完成端到端的盲文交互。为了确保质量并防止评估偏差，该基准通过一个确定性的、专家审查的流程构建，使用自创的盲文工具包，而未使用任何由LLMs生成的数据实例。我们从多个方面评估了六个具有代表性的LLMs。结果显示，印刷英语能力与盲文可及性之间存在持续的差距。盲文理解与表达是不对称的，其中等级2在输入方面尤其脆弱，相较于等级1，而完全的盲文请求进一步降低了性能。实验观察为未来盲文人工智能系统的开发提供了宝贵的指导。盲文基准中的所有相关资源均可公开获取，以供未来研究使用。
cs.AI / 89 / 2608.27296
LLMs Can Design Near-Optimal OR Algorithms
大型语言模型能够设计近似最优的运筹学算法
Abstract
We ask whether large language models (LLMs) can design effective algorithms for well-specified operations research (OR) problems. We study inventory control, queueing network control, and assortment optimization. We evaluate two levels of LLM use: at level 1, the model receives one problem instance and returns a solution for that instance; at level 2, it receives only the problem class description and broad parameter ranges, and returns an algorithm that maps instance parameters to solutions. Human input is minimal: we give one untuned prompt that describes the problem, and the model has access to a Python sandbox tool with a fixed compute budget. The strongest model we test, gpt-5.6-sol, matches or outperforms the best existing method on almost all evaluated instances. This holds even at level 2, where the returned algorithm is fixed before seeing the evaluation instances. Performance also improves sharply across models released less than eight months apart, suggesting that this capability is moving quickly. Thus, for the well-specified operations problems we study, a single untuned LLM query can already produce algorithms competitive with specialized methods. These results suggest that frontier LLMs can be a serious empirical baseline for algorithm design in well-specified OR problems.
Chinese Translation
我们探讨大型语言模型（LLMs）是否能够为明确规定的运筹学（OR）问题设计有效的算法。我们研究了库存控制、排队网络控制和产品组合优化。我们评估了两种层次的LLM使用：在第一层次，模型接收一个问题实例并返回该实例的解决方案；在第二层次，模型仅接收问题类别描述和广泛的参数范围，并返回一个将实例参数映射到解决方案的算法。人类输入非常有限：我们提供一个未调优的提示，描述问题，并且模型可以访问一个具有固定计算预算的Python沙箱工具。我们测试的最强模型gpt-5.6-sol在几乎所有评估实例中都与现有最佳方法相匹配或超越。这种情况在第二层次也成立，即返回的算法在看到评估实例之前就已固定。随着发布间隔不到八个月的模型之间的性能显著提升，这表明这一能力正在迅速发展。因此，对于我们研究的明确规定的运筹学问题，单个未调优的LLM查询已经能够生成与专门方法竞争的算法。这些结果表明，前沿的LLM可以成为明确规定的运筹学问题中算法设计的一个重要实证基准。
cs.AI / 90 / 2608.27311
Verify Smarter, Evolve Further: Efficient Harness Evolution through Behavior-Aware Verification
更智能的验证，更进一步的演化：通过行为感知验证实现高效的工具演化
Abstract
Agent harnesses shape how language-model agents use instructions, tools, and runtime components, but adapting these harnesses requires costly verification. Existing propose-and-verify methods typically score every candidate on a fixed task set, wasting rollouts on unrelated behaviors and allowing aggregate scores to obscure specific regressions. We introduce HarnessLens, a budget-aware framework for automated harness evolution. HarnessLens jointly explores the task space and user-configurable components, derives candidate modifications from execution trajectories, and selectively verifies each candidate on behavior-relevant tasks using an attributable-evidence gate. Across three agent harnesses and four benchmarks, HarnessLens improves average held-out performance by 7.6-13.6% while consuming substantially less evaluation budget than competing baselines. These results demonstrate that behavior-aware verification with explicit attribution enables more reliable and sample-efficient harness evolution under constrained interaction budgets. Our code is available at https://github.com/jhxu5214/HarnessLens.
Chinese Translation
代理工具的设计影响语言模型代理如何使用指令、工具和运行时组件，但适应这些工具需要昂贵的验证。现有的提议与验证方法通常在固定任务集上对每个候选项进行评分，浪费了与无关行为相关的滚动评估，并使得综合评分掩盖了特定的回归问题。我们提出了HarnessLens，一个预算感知的自动化工具演化框架。HarnessLens共同探索任务空间和用户可配置组件，从执行轨迹中推导候选修改，并使用可归因证据门在与行为相关的任务上选择性地验证每个候选项。在三个代理工具和四个基准测试中，HarnessLens在保持的平均性能上提高了7.6-13.6%，同时消耗的评估预算显著低于竞争基线。这些结果表明，具有明确归因的行为感知验证能够在受限的交互预算下实现更可靠和样本高效的工具演化。我们的代码可在 https://github.com/jhxu5214/HarnessLens 获取。
cs.AI / 91 / 2608.27340
Not All Eval-Awareness Is Equal: Capabilities Framing Predicts Compliance
并非所有的评估意识都是平等的：能力框架预测合规性
Abstract
Steering interventions targeting eval-awareness, a model's recognition that it is being tested, are increasingly used in safety evaluation pipelines, where evaluation-awareness is treated as a single quantity to be suppressed. We show that verbalized eval-awareness in chain-of-thought can be identified as capabilities-flavored ("the user is testing my ability to follow instructions"), safety-flavored ("the user is testing my boundaries"), both, or neither: framings that predict compliance very differently. On Qwen3-32B over the FORTRESS dataset, capabilities-framing predicts compliance with a +24 to +46 percentage-point gap over safety-framing across all tested steering conditions. A CoT-prefill intervention on eval-awareness-negative rollouts suggests the link is causal, with 10 of 11 prefills shifting compliance in the predicted direction. Then, eval-awareness is not behaviorally uniform: aggregate suppression rates can move while the safety-relevant component does not, and the same "X% suppression of eval-awareness" can correspond to qualitatively different behavioral outcomes.
Chinese Translation
针对评估意识的干预措施，即模型对其被测试的认知，越来越多地应用于安全评估流程中，其中评估意识被视为一个需要抑制的单一量。我们展示了在思维链中表达的评估意识可以被识别为能力框架（“用户正在测试我的执行指令的能力”）、安全框架（“用户正在测试我的界限”）、两者兼而有之，或两者皆非：这些框架对合规性的预测差异显著。在 Qwen3-32B 上的 FORTRESS 数据集中，能力框架在所有测试的干预条件下预测合规性时，较安全框架的差距为 +24 到 +46 个百分点。对评估意识负面结果的思维链预填充干预表明这种联系是因果关系，其中 11 个预填充中有 10 个在预期方向上改变了合规性。因此，评估意识并不是行为上均匀的：总体抑制率可以变化，而与安全相关的成分却不变，且相同的“X% 评估意识抑制”可以对应于质上不同的行为结果。
cs.AI / 92 / 2608.27364
Sophistication in GenAI Use: Field Evidence from a Large Firm
生成性人工智能使用的复杂性：来自大型公司的实证证据
Abstract
We study how sophistication in generative AI (genAI) use varies among the back-office workforce of a large firm. Using proprietary data, we observe 713,564 employee prompts and their corresponding large language model responses from nearly 4,000 back-office employees across 15 functional areas over eight months in 2025. We document three main findings. First, senior employees exhibit more sophisticated genAI use, consistent with domain expertise complementing genAI capabilities. Second, sophistication varies considerably across functions and is highest in Strategy, Digital Innovation, and Project Management, three groups that share a focus on firmwide strategic initiatives and organizational change. Third, we observe neither improvements in sophistication over time nor lasting improvements following formal AI training, suggesting that sophisticated use can be difficult to change. Together, our study provides measures of and insights into sophisticated genAI use that managers can use to improve outcomes and that researchers can use in future research.
Chinese Translation
我们研究了大型公司后勤员工中生成性人工智能（genAI）使用的复杂性如何变化。通过使用专有数据，我们观察了2025年八个月内近4000名后勤员工在15个职能领域的713,564条员工提示及其对应的大型语言模型响应。我们记录了三个主要发现。首先，高级员工表现出更复杂的genAI使用，这与领域专业知识对genAI能力的补充一致。其次，复杂性在不同职能之间差异显著，在战略、数字创新和项目管理三个职能中最高，这三组职能都专注于公司范围内的战略举措和组织变革。第三，我们观察到复杂性没有随时间改善，也没有在正式的人工智能培训后产生持久的改善，这表明复杂的使用方式可能难以改变。总的来说，我们的研究提供了复杂genAI使用的衡量标准和见解，管理者可以利用这些来改善结果，研究人员也可以在未来的研究中使用。
cs.AI / 93 / 2608.27391
CorporateBench: Large-Scale Q&A Benchmarking with Temporal Knowledge Bases
CorporateBench：基于时间知识库的大规模问答基准测试
Abstract
LLMs are increasingly able to answer complex questions about enterprise-scale document collections. But evaluation is hard: companies don't want to share internal communications, and synthetic datasets have been overly simple. We present CorporateBench (CB), a human-validated multi-task Q&A benchmark whose scale approaches the conditions LLMs encounter in corporate communication networks, with evaluation corpora surpassing 230,000 documents. CB evaluates LLMs across two dimensions (information extraction and knowledge base querying) through four synthetically generated firms ranging from 12 to 10,000 employees. Each corpus is sampled from a temporally evolving knowledge base describing a consistent world, guaranteeing cross-document logical consistency even across hundreds of thousands of documents. We evaluate five LLMs on CB, revealing increasingly poor performance as input size approaches realistic scales. CB provides LLM developers a metric for corporate communication reasoning, filling a crucial gap in the benchmarking ecosystem.
Chinese Translation
大型语言模型（LLMs）在回答有关企业级文档集合的复杂问题方面的能力日益增强。然而，评估却十分困难：公司不愿分享内部通讯，而合成数据集过于简单。我们提出了CorporateBench（CB），这是一个经过人工验证的多任务问答基准，其规模接近LLMs在企业沟通网络中所遇到的条件，评估语料库超过230,000份文档。CB通过四个合成生成的公司（员工人数从12到10,000不等）在信息提取和知识库查询两个维度上评估LLMs。每个语料库均从一个时间演变的知识库中抽样，该知识库描述了一个一致的世界，确保了即使在数十万份文档中也能保持跨文档的逻辑一致性。我们在CB上评估了五个LLMs，结果显示随着输入规模接近现实规模，性能逐渐下降。CB为LLM开发者提供了一个企业沟通推理的评估指标，填补了基准测试生态系统中的一个关键空白。
cs.AI / 94 / 2608.27421
Learning a Continuous Sepsis Severity Score Without Hour-by-Hour Supervision: A Two-Site Retrospective Study
无需逐小时监督的连续脓毒症严重程度评分学习：一项两地回顾性研究
Zhu, Kevin, Zhang, Ryan, Abed, Baraa, Choudhary, Tilendra, Madondo, Malvern, Arora, Mehak, Yang, Yixuan, Gent, Alasdair, Nagori, Aditya, Inan, Omer T., Haines, Krista L., Georgoff, Patrick, Agarwal, Suresh M., Krishnamoorthy, Vijay, Ohnuma, Tetsu, Podgoreanu, Mihai V., Pinsky, Michael R., Clermont, Gilles, Coopersmith, Craig M., Jabaley, Craig S., Kamaleswaran, Rishikesan
Abstract
Currently used sepsis severity indices rely on fixed variables and weights established decades ago, which are coarsely discretized and calibrated to a cohort that no longer reflects contemporary critical care. No alternative learned directly from patient trajectories is in routine use. We conducted a retrospective two-cohort study on a total of 29,116 and 7,691 adult patients meeting Sepsis-3 criteria from two hospital systems in Massachusetts and Georgie, respectively. We developed a sepsis index using 43 routinely charted variables over a 72-hour treatment window. Unlike previous studies, we use mortality as a treatment-level ranking signal rather than a per-state target, allowing credit to be redistributed non-uniformly across timesteps. Evaluation was done on a permanent 20% test holdout, using clinical vignettes and Spearman correlation. Uncertainty intervals were obtained by bootstrap resampling of whole patients. Under this ranking scheme, non-survivors scored 1.19-1.64 points higher than survivors on a 0-10 scale within all strata of baseline SOFA-2, with similar results stratifying within lactate, mean arterial pressure (MAP), and creatinine. Within-patient change in the index correlated with change in lactate (Spearman rho = 0.39; n = 1,854). Similar, weaker correlations were found for MAP and creatinine. On a cohort level, cross-institutional agreement measured by Spearman correlation between models trained on different sites, were 70-77% of same-site correlation. External within-patient correlations were 0.54 and 0.59 against ceilings of 0.92 and 0.90. Our index also correlated with established indices, while null controls stayed near zero. Our index demonstrated hourly prognostic information that meaningfully separates patient outcomes and is consistent with clinical expectation, indicating potential as a decision support tool complementing clinical judgement.
Chinese Translation
目前使用的脓毒症严重程度指数依赖于几十年前建立的固定变量和权重，这些变量经过粗略离散化，并且经过校准的队列已不再反映当代重症护理。没有直接从患者轨迹中学习的替代方案在日常使用中。我们在马萨诸塞州和乔治亚州的两个医院系统中对符合Sepsis-3标准的29,116名和7,691名成年患者进行了回顾性两队列研究。我们开发了一个脓毒症指数，使用了在72小时治疗窗口内常规记录的43个变量。与之前的研究不同，我们将死亡率作为治疗级别的排名信号，而不是每个状态的目标，从而允许在时间步长之间非均匀地重新分配信用。评估是在一个永久性的20%测试保留集上进行的，使用临床小插曲和斯皮尔曼相关性。通过对整个患者的自助重抽样获得不确定性区间。在这一排名方案下，非幸存者在所有基线SOFA-2层级上在0-10的评分范围内比幸存者高出1.19-1.64分，在乳酸、平均动脉压（MAP）和肌酐的分层中也得到了类似的结果。指数在患者内的变化与乳酸的变化相关（斯皮尔曼rho = 0.39；n = 1,854）。对于MAP和肌酐也发现了类似但较弱的相关性。在队列层面上，由不同地点训练的模型之间的斯皮尔曼相关性测量的跨机构一致性为同一地点相关性的70-77%。外部患者内相关性为0.54和0.59，天花板值为0.92和0.90。我们的指数还与已建立的指数相关，而无效对照接近零。我们的指数展示了每小时的预后信息，能够有意义地区分患者结果，并与临床预期一致，表明其作为补充临床判断的决策支持工具的潜力。
cs.AI / 95 / 2608.27429
Mechanistic Reaction Prediction via Discrete Flow Matching on Graph-Structured Electron Occupation
通过图结构电子占据的离散流匹配进行机制反应预测
Abstract
Chemical reactions are fundamentally transformations in electron space, yet most machine learning approaches model them either through \textit{de novo} generation of product molecules or through heuristic graph edits that operate directly on molecular topology. We introduce MAELLE (\textbf{M}ech\textbf{A}nistic \textbf{E}dit f\textbf{L}ow-matching on e\textbf{L}ectron r\textbf{E}arrangements), which instead models reactions as discrete flow matching over electron occupation vectors. Concretely, we formulate the reactant-to-product mapping as a Continuous-time Markov Chain (CTMC) over the graph-structured integer-valued electron occupation space defined on all bonding, non-bonding, and hydrogen sites. To construct the intermediate edit trajectories, we generalize the discrete flow matching mixture path to discrete electron rearrangements using Optimal Transport, yielding a sequence of mechanistically interpretable edit moves without requiring elementary step annotations. MAELLE achieves competitive performance on the USPTO-480K benchmark compared with leading reaction prediction models. Beyond in-distribution accuracy, we evaluate robustness across two out-of-distribution settings - structural complexity and reaction type - and find that MAELLE maintains strong performance where existing methods degrade. Finally, because the learned flow operates over the full electron redistribution, MAELLE naturally recovers mechanistic trajectories that align with known chemistry and can predict side products of a reaction.
Chinese Translation
化学反应本质上是电子空间中的转化，然而大多数机器学习方法要么通过 extit{de novo}生成产物分子，要么通过直接作用于分子拓扑的启发式图编辑来建模反应。我们提出了MAELLE（ extbf{M}ech extbf{A}nistic extbf{E}dit f extbf{L}ow-matching on e extbf{L}ectron r extbf{E}arrangements），该方法将反应建模为电子占据向量上的离散流匹配。具体而言，我们将反应物到产物的映射形式化为在定义于所有键合、非键合和氢位点的图结构整数值电子占据空间上的连续时间马尔可夫链（CTMC）。为了构建中间编辑轨迹，我们将离散流匹配混合路径推广到使用最优运输的离散电子重排，从而产生一系列机制上可解释的编辑移动，而无需基本步骤注释。与领先的反应预测模型相比，MAELLE在USPTO-480K基准测试中表现出竞争力的性能。除了在分布内的准确性外，我们还在两个分布外设置（结构复杂性和反应类型）中评估了鲁棒性，发现MAELLE在现有方法性能下降的情况下仍能保持强劲表现。最后，由于学习到的流作用于整个电子重分布，MAELLE自然恢复与已知化学相一致的机制轨迹，并能够预测反应的副产品。
cs.AI / 96 / 2608.27454
WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolution
WikiSkill：将智能体经验编译为持久知识以促进技能演化
Abstract
Agent skills package specialized knowledge and workflows into reusable resources that extend AI agent capabilities. Recent work automatically discovers such skills from agent experience, which enables agents to progressively adapt through interaction. However, the insights that guide skill development typically remain scattered across optimization histories, limiting their systematic reuse across iterations. We introduce WikiSkill, a framework that co-evolves agent skills with a persistent knowledge base (wiki). At a high level, WikiSkill separates raw execution experience, accumulated knowledge, and executable skills, while continuously consolidating experience into the wiki, which subsequent skill updates can build on. Across diverse benchmarks and models, WikiSkill consistently outperforms state-of-the-art skill-evolution methods and improves over no-skill baselines in most model-benchmark settings. We find that skill evolution complements model scaling: larger models generally benefit more from evolved skills, while smaller models with skills can outperform substantially larger models without them. We also find that evolved skills transfer effectively across models and model families, and skills evolved by other models can outperform self-evolved skills. Finally, our ablation studies confirm that persistent knowledge accumulation in the wiki is critical for effective skill evolution. These results demonstrate the benefits of systematically accumulating and refining agent experience for developing reusable and transferable skills.
Chinese Translation
智能体技能将专业知识和工作流程打包成可重用的资源，扩展了人工智能智能体的能力。近期的研究自动从智能体经验中发现这些技能，使得智能体能够通过互动逐步适应。然而，指导技能发展的见解通常分散在优化历史中，限制了它们在迭代过程中的系统性重用。我们提出了WikiSkill，一个与持久知识库（wiki）共同演化智能体技能的框架。从高层次来看，WikiSkill将原始执行经验、积累知识和可执行技能分开，同时不断将经验整合到wiki中，以便后续的技能更新可以在其基础上进行。在多种基准和模型中，WikiSkill始终优于最先进的技能演化方法，并在大多数模型-基准设置中改善了无技能基线。我们发现，技能演化与模型扩展相辅相成：较大的模型通常从演化的技能中受益更多，而具有技能的小模型可以显著超越没有技能的更大模型。我们还发现，演化的技能在不同模型和模型家族之间有效转移，其他模型演化的技能可以超越自我演化的技能。最后，我们的消融研究确认了在wiki中持久知识积累对于有效技能演化的重要性。这些结果展示了系统性积累和精炼智能体经验在开发可重用和可转移技能方面的好处。
cs.CL / 1 / 2608.26112
TreeGraft: Adaptive Multi-Drafter Grafting for Tree-Based Speculative Decoding
TreeGraft：基于树的自适应多草稿者嫁接用于推测解码
Abstract
Speculative decoding accelerates large language model inference through a draft-then-verify paradigm. Building on this, tree-structured methods improve inference by organizing proposals into multiple candidate paths, increasing the accepted length. However, existing tree-structured methods use a single drafter for all drafting steps, creating a dilemma: a smaller drafter is fast but yields lower-quality trees, whereas a larger drafter improves tree quality but suffers from high latency. To address this, we propose TreeGraft, a multi-drafter framework in which drafters of different costs jointly construct a shared draft tree. TreeGraft uses the stronger drafter to rescore candidates by updating scores assigned by the weaker drafter, reselect grafting positions, and recover promising paths left unexplored. It also integrates stronger drafter expansions non-destructively, preserving existing branches that may still be accepted by the target model. Together, these designs improve the quality of the shared draft tree. To control the drafting cost, TreeGraft introduces a lightweight scheduler distilled from an offline value system to decide when to call the stronger drafter. Across 10 model pairs and 6 benchmarks, TreeGraft outperforms the better of the two fixed single-drafter endpoint strategies by 15.1% on average, reaching a maximum gain of 26.6%. Our code is available at https://anonymous.4open.science/r/TreeGraft-E983.
Chinese Translation
推测解码通过草拟-验证范式加速大型语言模型的推理。在此基础上，树结构方法通过将提案组织成多个候选路径来改善推理，从而增加接受的长度。然而，现有的树结构方法在所有草拟步骤中使用单一草稿者，这造成了一个困境：较小的草稿者速度快但生成的树质量较低，而较大的草稿者提高了树的质量但面临高延迟。为了解决这个问题，我们提出了TreeGraft，一个多草稿者框架，其中不同成本的草稿者共同构建一个共享的草稿树。TreeGraft利用更强的草稿者通过更新较弱草稿者分配的分数来重新评分候选项，重新选择嫁接位置，并恢复未被探索的有前景路径。它还以非破坏性的方式整合更强草稿者的扩展，保留可能仍被目标模型接受的现有分支。这些设计共同提高了共享草稿树的质量。为了控制草拟成本，TreeGraft引入了一个轻量级调度器，该调度器从离线价值系统中提炼而来，以决定何时调用更强的草稿者。在10对模型和6个基准测试中，TreeGraft的表现平均比两种固定单草稿者端点策略中更好的那种高出15.1%，最大增益达到26.6%。我们的代码可在 https://anonymous.4open.science/r/TreeGraft-E983 获取。
cs.CL / 2 / 2608.26118
ElementCheck: Complexity-Aware Long-Form Text Factuality Evaluation via Sentence Elements
ElementCheck：基于句子元素的复杂性感知长文本事实性评估
Abstract
Existing long-form factuality evaluation relies on the decompose-retrieve-verify pipeline. However, the pipeline suffers from noise from claim decomposition and fixed verification granularity, resulting in unreliable results. We propose ElementCheck, a complexity-aware framework that verifies long-form outputs via sentence elements. Instead of uniformly decomposing sentences into atomic sub-claims, ElementCheck extracts entity pairs that are explicitly linked through verifiable connections in the original sentence as elements, and organizes these into an element graph. The graph topology provides a structural signal for estimating sentence complexity, enabling direct verification for simple sentences and targeted element-level refinement and verification for complex ones. To support fine-grained evaluation, we construct a new benchmark FastFact-Sent by mapping isolated claims from FastFact-Bench back to their source sentences. Experiments on FastFact-Sent and two domain-specific benchmarks show ElementCheck consistently improves factuality verification across five backbone models while maintaining a favorable accuracy-cost trade-off. Further analyses demonstrate that complexity-aware verification reduces unnecessary re-verification and maintains stability across different backbones.
Chinese Translation
现有的长文本事实性评估依赖于分解-检索-验证的流程。然而，该流程受到来自主张分解的噪声和固定验证粒度的影响，导致结果不可靠。我们提出了ElementCheck，一个复杂性感知框架，通过句子元素验证长文本输出。ElementCheck并不是将句子均匀地分解为原子子主张，而是提取在原句中通过可验证连接明确关联的实体对作为元素，并将这些元素组织成一个元素图。图的拓扑结构为估计句子复杂性提供了结构信号，使得简单句子能够直接验证，而复杂句子则可以进行针对性的元素级细化和验证。为了支持细粒度评估，我们通过将FastFact-Bench中的孤立主张映射回其源句，构建了一个新的基准FastFact-Sent。在FastFact-Sent和两个特定领域基准上的实验表明，ElementCheck在五个基础模型上持续提高了事实性验证的效果，同时保持了良好的准确性与成本的权衡。进一步的分析表明，复杂性感知验证减少了不必要的重新验证，并在不同基础模型之间保持了稳定性。
cs.CL / 3 / 2608.26119
DeflectBench: A Benchmark for Evaluating Rhetorical Fallacy Generation in LLMs
DeflectBench：评估大型语言模型中修辞谬误生成的基准
Abstract
Whether large language models can be prompted to generate rhetorical fallacies on demand, and whether current safety post-training constrains this behavior, has received less attention than the related question of detecting fallacies in existing text. We close this gap with DeflectBench, evaluating 23,990 generations from four frontier models across three deflection strategies (whataboutism, ad hominem, red herring), seven prompt framings, and 80 claims spanning four controversy levels. Refusal is governed primarily by request structure rather than claim content. Per claim refusal varies by only 11 percentage points across the 80 claims, while a single prompt frame change can swing within model refusal by nearly 100 percentage points and switching the requested fallacy type can swing it by over 80 percentage points within explicit framings. An educational debate coach prompt framing collapses refusal to near zero across all four model families, but the bypassed behavior is not clean compliance. Models typically produce labeled compliance, naming the requested manipulation in the same response that contains it. The four models distribute differently across refusal, labeled compliance, soft refusal, and clean compliance. The code and dataset are released at https://github.com/ArtKanke/DeflectBench.
Chinese Translation
大型语言模型是否能够按需生成修辞谬误，以及当前的安全后训练是否限制了这种行为，受到的关注远低于检测现有文本中谬误的相关问题。我们通过 DeflectBench 来填补这一空白，评估来自四个前沿模型的 23,990 次生成，涵盖三种转移策略（转移话题法 whataboutism、人身攻击 ad hominem、红鲱鱼 red herring）、七种提示框架和 80 个主张，涉及四个争议级别。拒绝主要受请求结构的影响，而非主张内容。在 80 个主张中，每个主张的拒绝率仅相差 11 个百分点，而单个提示框架的变化可以使模型的拒绝率波动近 100 个百分点，切换请求的谬误类型则可以在明确框架内使其波动超过 80 个百分点。一种教育辩论教练的提示框架使得四个模型家族的拒绝率几乎降至零，但被绕过的行为并不是完全的遵从。模型通常会在包含请求操控的同一响应中产生标记的遵从，明确指出所请求的操控。四个模型在拒绝、标记遵从、软拒绝和清晰遵从方面的分布各不相同。代码和数据集已发布在 https://github.com/ArtKanke/DeflectBench。
cs.CL / 4 / 2608.26120
Recipes for Steering and Scaling LLMs via Sampling
通过采样引导和扩展大型语言模型的策略
Abstract
Large Language Models (LLMs) are probabilistic models, typically defined by an autoregressive factorization. While recent work has begun to study richer target distributions beyond the base model, the sampling strategies remain highly inefficient. In this paper, we present a flexible and theoretically grounded framework for steering and scaling autoregressive LLMs with sampling. Within this framework, we describe two algorithms -- one based on Sequential Monte Carlo (SMC) and one based on Replica Exchange (RE) -- that steer generation toward powering, product or tilting of the base model distribution. We illustrate this framework through scaling the generation quality of LLMs without external supervision or reward models. Experimental results demonstrate our methods scale more favorably than Best-of-N and standard MCMC baselines. Overall, this paper offers a systematic recipe for probabilistic inference with LLMs via sampling.
Chinese Translation
大型语言模型（LLMs）是概率模型，通常通过自回归分解来定义。尽管近期的研究开始探索超越基础模型的更丰富目标分布，但采样策略仍然非常低效。在本文中，我们提出了一个灵活且理论基础扎实的框架，用于通过采样引导和扩展自回归LLMs。在这个框架内，我们描述了两种算法——一种基于序列蒙特卡洛（Sequential Monte Carlo, SMC），另一种基于复制交换（Replica Exchange, RE）——它们将生成过程引导至基础模型分布的幂、乘积或倾斜。我们通过在没有外部监督或奖励模型的情况下提升LLMs的生成质量来说明这一框架。实验结果表明，我们的方法在扩展上比最佳选择（Best-of-N）和标准马尔可夫链蒙特卡洛（MCMC）基线表现更优。总体而言，本文提供了一种通过采样进行LLMs概率推断的系统性策略。
cs.CL / 5 / 2608.26121
Can a Model Catch Its Own Hallucinations for Free?: Label-Free Doubt Signals Hold Their Own Against a Labelled Dataset for Abstention
模型能否免费捕捉自身的幻觉？：无标签的怀疑信号在放弃决策中与标记数据集相抗衡
Abstract
Large language models state false facts as fluently as true ones, yet a model often "knows" internally when it is on shaky ground: the probability it assigns to its own answer tends to dip on the facts it gets wrong. The usual way to act on this, teaching a model to abstain rather than guess, requires a labelled dataset of right and wrong answers. We ask whether the model's own confidence, which is free and needs no labels, can do that job instead. We fine-tune each model (with LoRA) to answer when its frozen confidence is high and to say "I'm not sure" when it is low, using the signal alone and no correctness labels. Across six open-weights models (1B-8B, two families) on short-form factual question answering, with correctness adjudicated by an independent judge model, this label-free recipe holds its own against label-supervised abstention-tuning: at matched coverage we find no statistically detectable difference between the two. A control that drills hard examples instead of abstaining does not help, indicating the gain comes from calibration, not rote memorization. The signal's one blind spot is confidently wrong facts, which it cannot flag. A model's own doubt is thus a near-free substitute for a labelled dataset when teaching it when to abstain. Code and artifacts are available on request.
Chinese Translation
大型语言模型以流畅的方式陈述虚假事实，和真实事实一样，然而模型往往在内部“知道”自己处于不稳定的状态：它对自身答案的概率评估在错误的事实上往往会下降。通常的做法是教模型在不确定时选择放弃而不是猜测，这需要一个包含正确和错误答案的标记数据集。我们探讨模型自身的信心是否可以替代这个标记数据集，因其是免费的且不需要标签。我们对每个模型进行微调（使用 LoRA），使其在信心较高时回答问题，而在信心较低时则说“我不确定”，仅使用信号而不依赖正确性标签。在六个开放权重模型（1B-8B，两种模型系列）上进行的短文本事实问答中，由独立评判模型裁定的正确性显示，这种无标签的方法在放弃决策的监督微调中表现不相上下：在匹配的覆盖率下，我们没有发现两者之间有统计学上显著的差异。一个针对困难示例的控制实验并没有帮助，表明收益来自于校准，而非死记硬背。信号的一个盲点是自信的错误事实，它无法标记。因此，模型自身的怀疑几乎可以作为教其何时放弃的标记数据集的替代品。代码和相关材料可根据请求提供。
cs.CL / 6 / 2608.26123
Which India Survives Translation? Narrative Homogenisation Across Indian Oral Traditions in LLMs
哪种印度在翻译中得以存活？印度口头传统在大型语言模型中的叙事同质化
Abstract
Large language models (LLMs) are trained predominantly on English-language internet text that over-represents certain cultural narratives, raising concerns that models flatten the diversity of non-Western storytelling traditions into a single homogenized archetype. We present a pilot computational study examining this across three maximally distinct Indian regional oral and literary traditions: the Rajasthani Pabuji epic, classical Tamil Sangam poetry, and Bengali folk tales. We collected authentic reference corpora for each tradition (11, 21, and 10 passages respectively) and prompted two LLMs (Claude Sonnet and Gemini) with 54 generation requests spanning three prompt types per tradition - generic, culturally specific, and regional-language. Using Sentence-BERT embeddings and cosine similarity, we measure reference drift (how closely outputs track their own tradition's authentic texts relative to the other two) and cross-tradition convergence (how similar outputs are across traditions). We find that while outputs remain closer to their own tradition's reference than to others, cross-tradition similarity is high (0.52-0.66) relative to what the traditions' genuine distance would predict, indicating partial homogenisation. Unexpectedly, prompting in the regional language (Hindi, Tamil, or Bengali) consistently reduced fidelity to the authentic tradition relative to English prompting, by as much as 27 percentage points for Rajasthani and Bengali traditions. We discuss this against conflicting prior results on multilingual prompting and argue it reflects a difference between eliciting general cultural diversity and simulating one narrow, lesser-documented oral tradition. We position this pilot as a lightweight, scalable complement to recent large-scale human-annotation studies of Indian cultural misrepresentation in LLM-generated stories, as part of a broader doctoral research program.
Chinese Translation
大型语言模型（LLMs）主要在英语互联网文本上进行训练，这些文本过度代表了某些文化叙事，令人担忧的是，这些模型将非西方叙事传统的多样性压缩为单一的同质化原型。我们呈现了一项初步的计算研究，考察了三种最大程度不同的印度地区口头和文学传统：拉贾斯坦的帕布吉史诗、古典泰米尔的桑甘诗歌和孟加拉民间故事。我们为每种传统收集了真实的参考语料库（分别为11、21和10段），并对两个LLM（Claude Sonnet和Gemini）进行了54次生成请求，涵盖每种传统的三种提示类型——通用、文化特定和地区语言。通过使用Sentence-BERT嵌入和余弦相似度，我们测量了参考漂移（输出与其自身传统的真实文本的接近程度相对于其他两个传统）和跨传统收敛性（不同传统之间输出的相似程度）。我们发现，尽管输出与自身传统的参考文本相比更为接近，但跨传统的相似性仍然较高（0.52-0.66），相对于这些传统的真实距离预测，这表明存在部分同质化。出乎意料的是，在地区语言（印地语、泰米尔语或孟加拉语）中进行提示，相较于英语提示，始终降低了对真实传统的忠实度，拉贾斯坦和孟加拉传统的降低幅度高达27个百分点。我们将此与先前关于多语言提示的相互矛盾的结果进行讨论，并认为这反映了引发一般文化多样性与模拟一种狭窄、文献较少的口头传统之间的差异。我们将这项初步研究视为对近期大型人类注释研究的轻量级、可扩展的补充，旨在揭示LLM生成故事中印度文化误表现的现象，作为更广泛博士研究计划的一部分。
cs.CL / 7 / 2608.26124
Natural-Language Policies to Executable Decisions: An Interpretable Large Language Model Framework
可执行决策的自然语言政策：一个可解释的大型语言模型框架
Abstract
Pricing automation in large-scale tourism is challenging because travel orders are highly unstructured, while pricing policies are complex, rapidly evolving, and inherently open-ended. Traditional rule engines are brittle and costly to maintain, whereas unconstrained LLM agents lack the reliability and auditability required for financial decisions. We present a production-grade LLM-powered pricing system with a strict decision boundary: LLMs perform structured extraction and bounded policy/path selection, while all numeric pricing, including total-price computation, is executed deterministically. Policies are compiled into interpretable condition trees, enabling open-ended support for new clauses and evolving rules without code changes, while exposing auditable artifacts for human-in-the-loop control. Periodic fine-tuning on logged traces further improves tree induction and path matching. Deployed at a municipal state-owned tourism enterprise across 7 scenic sites and 12 business categories with 1,500+ operators and 1,000+ active policies, the system processed 3,960 orders in six months, reduced the order management team from 15-20 to 3, and cut per-order handling time from 10 minutes to <2 minutes.
Chinese Translation
在大规模旅游中的定价自动化面临挑战，因为旅游订单高度非结构化，而定价政策复杂、快速演变且本质上是开放式的。传统的规则引擎脆弱且维护成本高，而不受约束的LLM（大型语言模型）代理缺乏金融决策所需的可靠性和可审计性。我们提出了一种生产级的基于LLM的定价系统，具有严格的决策边界：LLM进行结构化提取和有限的政策/路径选择，而所有数值定价，包括总价计算，都是确定性执行的。政策被编译成可解释的条件树，支持新条款和演变规则的开放式支持，无需代码更改，同时暴露可审计的文档以便于人机协作控制。对记录的痕迹进行定期微调进一步改善了树的归纳和路径匹配。该系统在一家市属国有旅游企业的7个景点和12个业务类别中部署，拥有1500多名运营商和1000多项活跃政策，在六个月内处理了3960个订单，将订单管理团队从15-20人减少到3人，并将每个订单的处理时间从10分钟缩短到不足2分钟。
cs.CL / 8 / 2608.26125
Training-Time Explainability for Multilingual Hate Speech Detection: Aligning Model Reasoning with Human Rationales
多语言仇恨言论检测的训练时可解释性：将模型推理与人类理由对齐
Abstract
Online hate against Muslim communities often appears in culturally coded, multilingual forms that evade conventional AI moderation. Such systems, though accurate, remain opaque and risk bias, over-censorship, or under-moderation, particularly when detached from sociocultural context. We propose a \emph{training-time} explainability framework that aligns model reasoning with human-annotated rationales, improving both classification performance and interpretability. Our approach is evaluated on HateXplain (English) and BullySent (Hinglish), reflecting the prevalence of anti-Muslim hate across both languages. Using LIME, Integrated Gradients, Grad X Input, and attention, we assess accuracy, explanation quality, and cross-method agreement. Results show that gradient- and attention-based regularization improve F-scores, enhance plausibility and faithfulness, and capture culturally specific cues for detecting implicit anti-Muslim hate, offering a path toward multilingual, culturally aware content moderation.
Chinese Translation
针对穆斯林社区的在线仇恨言论往往以文化编码的多语言形式出现，这些形式规避了传统的人工智能审查系统。尽管这些系统在准确性上表现良好，但仍然缺乏透明度，存在偏见、过度审查或不足审查的风险，尤其是在脱离社会文化背景时。我们提出了一种 extit{训练时}可解释性框架，该框架将模型推理与人类标注的理由对齐，从而提高分类性能和可解释性。我们的研究在HateXplain（英语）和BullySent（Hinglish）上进行了评估，反映了两种语言中反穆斯林仇恨的普遍性。通过使用LIME、集成梯度（Integrated Gradients）、Grad X Input和注意力机制，我们评估了准确性、解释质量和跨方法一致性。结果表明，基于梯度和注意力的正则化提高了F分数，增强了可信度和忠实度，并捕捉了检测隐性反穆斯林仇恨的文化特定线索，为多语言、文化敏感的内容审查提供了一条途径。
cs.CL / 9 / 2608.26126
TelecomGPT-R1: A Unified Open-Source Reasoner for the Telecom Stack
TelecomGPT-R1：电信领域的统一开源推理器
Abstract
Telecommunications is a high-leverage domain for large language model (LLM)-based reasoning because routine engineering workflows require joint grounding in normative specifications, operational telemetry, vendor-specific fault evidence, and exact RF/network calculations. However, current LLM integration in telecom remains bottlenecked by a two-sided capability gap: generic reasoners often lack telecom-specific grounding, while domain-specific telecom LLMs remain limited in structured, multi-step reasoning. To bridge this gap, we release TelecomGPT-R1-9B, a unified open-source telecom reasoner that ranks top-performing on the GSMA open telco leaderboard. Specifically, we curate a 67,427-example supervised fine-tuning (SFT) corpus organized around four complementary reasoning axes: protocol, knowledge, modeling, and fault. The corpus is built from axis-matched public web sources and enhanced through axis-specific chain-of-thought (CoT) generation and prefix-continuation self-validation. Starting from Qwen3.5-9B, we further develop a two-stage post-training recipe. First, multi-teacher low-rank adaptation (LoRA)-based SFT injects telecom knowledge and induces axis-specific reasoning formats. Second, group relative policy optimization (GRPO), stabilized by decoupled clip and dynamic sampling policy optimization (DAPO), optimizes the policy using four axis-aligned binary verifier rewards. Across seven public telecom benchmarks, TelecomGPT-R1-9B ranks first among open-source telecom LLMs and achieves a seven-axis mean comparable to state-of-the-art closed-source frontier reasoners.
Chinese Translation
电信是一个高杠杆的大型语言模型（LLM）推理领域，因为常规工程工作流程需要在规范性规格、操作遥测、特定供应商的故障证据和精确的射频/网络计算之间进行联合基础。 然而，当前电信领域的LLM集成仍然受到双重能力差距的制约：通用推理器往往缺乏电信特定的基础，而特定领域的电信LLM在结构化的多步骤推理方面仍然有限。 为了弥补这一差距，我们发布了TelecomGPT-R1-9B，这是一个统一的开源电信推理器，在GSMA开放电信排行榜上表现优异。具体而言，我们整理了一个包含67,427个示例的监督微调（SFT）语料库，围绕四个互补的推理轴进行组织：协议、知识、建模和故障。该语料库是基于与轴匹配的公共网络来源构建的，并通过特定轴的思维链（CoT）生成和前缀延续自验证进行了增强。从Qwen3.5-9B开始，我们进一步开发了一个两阶段的后训练方案。首先，基于多教师低秩适应（LoRA）的SFT注入电信知识并引导特定轴的推理格式。其次，使用四个轴对齐的二元验证器奖励，通过解耦的剪辑和动态采样策略优化（DAPO）稳定的组相对策略优化（GRPO）来优化策略。在七个公共电信基准测试中，TelecomGPT-R1-9B在开源电信LLM中排名第一，并在七个轴上达到了与最先进的闭源前沿推理器相当的平均水平。
cs.CL / 10 / 2608.26129
FIRSTPASS: A Multi-Domain, Multi-Round Peer Review Dataset Grounded in Real Editorial Outcomes
FIRSTPASS：基于真实编辑结果的多领域、多轮同行评审数据集
Abstract
Scientific peer review datasets have trained AI systems exclusively on Computer Science and Machine Learning venues, producing models that critique ablation studies yet have never seen a biology reviewer demand contamination controls or a chemist question Nuclear Magnetic Resonance (NMR) spectral assignments. We introduce FIRSTPASS, the first large-scale peer review dataset built on complete multi-round editorial dialogues from a multidisciplinary high-impact journal. Curated from Nature Communications mandatory transparent peer review (instituted November 2022), FIRSTPASS comprises 3,668 records spanning five scientific domains (biology, chemistry, neuroscience, physics, and earth science), capturing the full iterative structure of scientific validation: initial referee reports, author point-by-point responses, and updated reviewer assessments. Each record carries an outcome label derived directly from editorial decisions (STANDARD for two-round review; EXTENDED for three or more rounds), providing ground truth absent in all prior corpora. An automated audit confirms 100% content integrity. Expert reviews average 2,155 words, substantially denser than conference venue reviews. All data, parsing pipelines, and evaluation scripts are released to enable reproducible benchmarking of AI scientific judgment across disciplines.
Chinese Translation
科学同行评审数据集一直以来仅在计算机科学和机器学习领域训练人工智能系统，导致所生成的模型能够批评消融研究，但从未见过生物学审稿人要求污染控制或化学家质疑核磁共振（NMR）光谱分配。我们介绍了FIRSTPASS，这是第一个基于来自多学科高影响力期刊的完整多轮编辑对话构建的大规模同行评审数据集。FIRSTPASS由《自然通讯》强制透明同行评审（自2022年11月实施）中策划而成，包含3,668条记录，涵盖五个科学领域（生物学、化学、神经科学、物理学和地球科学），捕捉科学验证的完整迭代结构：初步审稿报告、作者逐点回应和更新的审稿人评估。每条记录都带有直接来源于编辑决策的结果标签（两轮评审为STANDARD；三轮或更多为EXTENDED），提供了在所有先前语料库中缺失的真实依据。自动化审核确认内容完整性为100%。专家评审的平均字数为2,155字，显著高于会议场地评审。所有数据、解析管道和评估脚本均已发布，以便在各学科中实现人工智能科学判断的可重复基准测试。
cs.CL / 11 / 2608.26130
Agents Don't Paginate: First-Chunk Selection for LLM Tool Responses
代理不进行分页：大语言模型工具响应的首块选择
Abstract
Coding agents built on large language models (LLMs), such as Claude Code, Cursor, OpenAI Codex, GitHub Copilot, and Aider, receive tool responses that routinely exceed the agent's per-turn token budget. The standard remedy, pagination, is available in every protocol that produced these responses; yet across the corpus of session logs from a public Model Context Protocol middleware we observed no agent-initiated requests for a second chunk. The first chunk is what the agent reads, so we ask how often the gold item (the one the agent needs) is placed first in it: the precision-at-1 rate $p_1$. In a controlled offline benchmark we treat first-chunk selection as a 0/1 knapsack and compare six value functions on 500 SWE-bench Verified tasks, then test whether $p_1$ matters with a single-turn file-localisation probe on five language models (4,800 LLM calls; not an end-to-end resolve-rate test). Two pre-registered hypotheses did not hold and are our main findings. The central one is negative: raising $p_1$ does not systematically raise downstream accuracy. Per-model deltas stay under three percentage points (p.p.), are not consistently signed, and no model is significant; the agent recovers the gold from anywhere in the chunk, so what reaches its answer is first-chunk inclusion, not the gold's rank within it. The second: adding four file-metadata signals to a keyword scorer hurts $p_1$ by 4.8 p.p. (paired significance test, $p = 0.001$). A parameter-free keyword scorer does raise $p_1$, from a 24.2% baseline to 35.0% (+10.8 p.p., far beyond chance; $p = 3.9 \times 10^{-8}$), and to 35.8% with a fallback to the tool's native ordering when no keyword matches. But by our central finding this is a rank-1 gain, and rank-1 is the part that does not reach the agent's answer: downstream accuracy does not move.
Chinese Translation
基于大型语言模型（LLMs）构建的编码代理，如Claude Code、Cursor、OpenAI Codex、GitHub Copilot和Aider，接收到的工具响应通常超过代理每轮的令牌预算。标准的解决方案是分页，这在所有生成这些响应的协议中都是可用的；然而，在我们从一个公共模型上下文协议中间件的会话日志中观察到，没有代理发起对第二块的请求。首块是代理所读取的内容，因此我们询问黄金项（代理所需的项）在其中被放置为第一的频率：即精确率 $p_1$。在一个受控的离线基准测试中，我们将首块选择视为0/1背包问题，并在500个SWE-bench验证任务上比较六个价值函数，然后通过对五个语言模型进行单轮文件定位探测来测试 $p_1$ 的重要性（共4,800次LLM调用；这不是一个端到端的解决率测试）。两个预注册的假设未能成立，并且是我们的主要发现。第一个是负面的：提高 $p_1$ 并不会系统性地提高下游准确率。每个模型的差异保持在三个百分点（p.p.）以内，且没有一致的符号，且没有模型显著；代理可以从块中的任何位置恢复黄金项，因此到达其答案的因素是首块的包含，而不是黄金项在其中的排名。第二个发现是：向关键字评分器添加四个文件元数据信号会使 $p_1$ 降低4.8个百分点（配对显著性测试，$p = 0.001$）。一个无参数的关键字评分器确实提高了 $p_1$，从24.2%的基线提高到35.0%（+10.8 p.p.，远超随机概率；$p = 3.9 imes 10^{-8}$），并在没有关键字匹配时回退到工具的原生排序时提高到35.8%。但根据我们的主要发现，这只是一个排名-1的提升，而排名-1是未能到达代理答案的部分：下游准确率没有变化。
cs.CL / 12 / 2608.26131
Evaluating Language Models in Realistic Conversational Contexts
在现实对话背景下评估语言模型
Abstract
As Large Language Models (LLMs) are increasingly deployed to serve open-ended, multi-turn interactions, evaluating conversational quality at human scale has become a central challenge. Existing evaluation frameworks built for summarization, translation, or short-form QA tasks fall short of adequately measuring the consistency of human-scale dialogue, especially when derivation and validation of these metrics themselves often rely on synthetic rather than human sources. We fill the gap by introducing UPHELD (UPwork Human-Scale Evaluated Long Dialogues), a large, reference-full benchmark for evaluating human-scale conversational ability beyond factual correctness. UPHELD consists of hundreds of complete human-to-human dialogues authored by professional script writers, with realistic turn densities and 36,000+ per-turn human annotations across 30,000+ expert-generated dialogue turns. Using UPHELD, we systematically evaluate classical automatic metrics and reference-free LLM-as-a-judge approaches, and find them unreliable when correlated with expert human judgment. Building off this analysis, we use UPHELD to develop a Mixture-of-Judges framework that combines multiple evaluative signals and improves correlation with human assessments by approximately 30%. Overall, UPHELD provides a robust, human-grounded foundation for evaluating human-scale conversational intelligence that fills a crucial gap in the pre-existing LLM dataset landscape.
Chinese Translation
随着大型语言模型（LLMs）越来越多地被用于开放式的多轮互动，评估人类规模的对话质量已成为一个核心挑战。现有的评估框架主要针对摘要、翻译或短文本问答任务，无法充分测量人类规模对话的一致性，特别是这些指标的推导和验证往往依赖于合成数据而非人类来源。为填补这一空白，我们引入了UPHELD（UPwork人类规模评估长对话），这是一个大型的、参考丰富的基准，旨在评估超越事实正确性的人工对话能力。UPHELD包含数百个由专业编剧创作的完整人际对话，具有现实的轮次密度和超过36,000个每轮人类注释，涵盖30,000多个专家生成的对话轮次。利用UPHELD，我们系统地评估了经典的自动化指标和无参考的LLM作为评判者的方法，发现它们在与专家人类判断相关时不够可靠。在此分析的基础上，我们使用UPHELD开发了一个混合评判框架，结合多种评估信号，使与人类评估的相关性提高了约30%。总体而言，UPHELD为评估人类规模的对话智能提供了一个稳健的、以人类为基础的基础，填补了现有LLM数据集领域中的一个重要空白。
cs.CL / 13 / 2608.26133
Agent Seer: Synthesizing Scenarios from Specification Understanding
代理观察者：从规范理解合成场景
Abstract
Evaluating AI agents that use external tools requires realistic test scenarios that capture how practitioners compose tools and iterate across conversation turns. Constructing such scenarios by hand demands deep domain expertise, does not scale across tool ecosystems, and produces static benchmarks that cannot track evolving APIs. We observe that tool specifications -- function names, natural-language descriptions, and typed parameter schemas -- already encode sufficient semantic information to synthesize realistic evaluation scenarios without manual curation or live tool execution. Agent Seer builds off this latent information: from a single Model Context Protocol (MCP) specification, with no examples, no live tool access, and no domain-specific tuning. This pipeline enriches raw schemas, generates graded scenarios with synthetic tool outputs, and expands them into mock-data-grounded multi-turn dialogues that exhibit strong tool-calling correctness and conversational coherence. Evaluation quality is measured by applying this pipeline on seven MCP specifications spanning diverse domains and tool-suite sizes and measuring the tool-calling correctness and conversational coherence. The pipeline achieves strong quality across all domains, with complete tool coverage on small and medium specifications. Two findings emerge within this analysis: parameter schema complexity is the strongest correlate of quality variation -- tool-suite size plays a smaller, orthogonal role -- and argument value accuracy is the dominant failure mode among imperfect scenarios, a sub-dimension invisible to coarse-grained name-match metrics.
Chinese Translation
评估使用外部工具的人工智能代理需要现实的测试场景，以捕捉从业者如何组合工具并在对话回合中迭代。手动构建此类场景需要深厚的领域专业知识，无法在工具生态系统中扩展，并且产生的静态基准无法跟踪不断发展的API。我们观察到，工具规范——函数名称、自然语言描述和类型参数模式——已经编码了足够的语义信息，可以在没有手动策划或实时工具执行的情况下合成现实的评估场景。代理观察者基于这种潜在信息构建：从单个模型上下文协议（Model Context Protocol, MCP）规范出发，无需示例、无需实时工具访问，也无需领域特定的调优。该流程丰富了原始模式，生成具有合成工具输出的分级场景，并将其扩展为基于模拟数据的多轮对话，展现出强大的工具调用正确性和对话连贯性。通过在涵盖不同领域和工具套件规模的七个MCP规范上应用该流程，评估质量，并测量工具调用的正确性和对话的连贯性。该流程在所有领域中都实现了强大的质量，对于小型和中型规范实现了完全的工具覆盖。在此分析中出现了两个发现：参数模式复杂性是质量变化的最强相关因素——工具套件规模的影响较小，且参数值准确性是不完美场景中占主导地位的失败模式，这是一个在粗粒度名称匹配指标中不可见的子维度。
cs.CL / 14 / 2608.26135
Data Science Approaches to Evaluating Honours Candidates
数据科学方法在评估荣誉候选人中的应用
Abstract
We present a modular data-science pipeline for estimating public sentiment towards individuals from fragmented, unstructured open-source intelligence (OSINT). The method chains web search, text extraction, relevance filtering, tokenisation, co-reference resolution, and sentiment analysis to convert heterogeneous web material into auditable person-level sentiment distributions. We compare AFINN and VADER with MINOS, a domain-informed sentiment algorithm designed to detect language associated with reputational risk, misconduct, and positive public contribution. Applied to public figures with known reputational outcomes, MINOS gives the clearest separation between positive, ambiguous, and negative cases. The results show that chained NLP and OSINT methods can support transparent, reproducible, human-in-the-loop sentiment assessment for high-stakes decision support. We demonstrate the approach on the UK Honours system, where individuals are required to display high standards of public conduct to maintain an Honour.
Chinese Translation
我们提出了一种模块化的数据科学流程，用于从碎片化、非结构化的开放源代码情报（OSINT）中估计公众对个人的情感。该方法将网络搜索、文本提取、相关性过滤、分词、共指消解和情感分析链式结合，以将异构的网络材料转化为可审计的个人情感分布。我们将AFINN和VADER与MINOS进行比较，后者是一种领域知情的情感算法，旨在检测与声誉风险、不当行为和积极公共贡献相关的语言。应用于具有已知声誉结果的公众人物时，MINOS能够清晰地区分积极、模糊和消极案例。结果表明，链式自然语言处理（NLP）和开放源代码情报（OSINT）方法可以支持透明、可重复的人机协作情感评估，以便为高风险决策提供支持。我们在英国荣誉系统中展示了该方法，在该系统中，个人需要展示高标准的公共行为以维持荣誉。
cs.CL / 15 / 2608.26136
Reward-Informed Sparse Autoencoders and the Solution-Completeness Confound
基于奖励的信息稀疏自编码器及其解的完整性混淆
Abstract
Sparse autoencoders (SAEs) decompose language-model activations into sparse, interpretable features, and an appealing way to aim them at reasoning is to curate their data with a signal reinforcement learning already produces: the reward. We build such a reward-informed SAE (RI-SAE): we split GRPO trajectories into high-reward ("good") and low-reward ("bad") reasoning continuations, train a standard JumpReLU SAE on their activations, and then ask what the resulting good/bad separation actually measures. On Llama-3.1-8B a sparse subset of the 16,384 features does separate the classes (silhouette 0.79 on the selected features versus 0.005 for the full code), but a control battery shows the separation is largely solution completeness rather than reasoning quality: a TF-IDF text classifier already splits the classes (AUC 0.75--0.83), and three structural cues alone (length, a closed reasoning block, and a boxed answer) reach AUC 0.70 (99% of good versus 69% of bad completions are boxed). A generic SAE that never saw the reward does not separate the classes at all (silhouette 0.01, no discriminative features), so the 0.79 is in-sample fitting of this curated signal rather than structure that a reward-blind dictionary recovers. We therefore present the recipe and its control battery together: reward filtering is a cheap, label-free way to reuse RL signals for interpretability, but most of what it surfaces is completion form. Two discriminative features are still readable (symbolic mathematics; procedural and evaluative language), which we take as illustrative rather than as isolated reasoning.
Chinese Translation
稀疏自编码器（SAEs）将语言模型的激活分解为稀疏且可解释的特征，而将其用于推理的一个吸引人的方法是利用信号强化学习所产生的数据：奖励。我们构建了这样一种基于奖励的信息稀疏自编码器（RI-SAE）：我们将GRPO轨迹分为高奖励（“好”）和低奖励（“坏”）的推理延续，基于它们的激活训练一个标准的JumpReLU SAE，然后询问所得到的好/坏分离实际上测量了什么。在Llama-3.1-8B上，16,384个特征中的一个稀疏子集确实能够分离这两个类别（在选定特征上的轮廓系数为0.79，而完整代码的轮廓系数为0.005），但控制实验显示这种分离在很大程度上是解的完整性而非推理质量：一个TF-IDF文本分类器已经能够分离这两个类别（AUC 0.75--0.83），而仅凭三个结构线索（长度、封闭的推理块和框选的答案）就能达到AUC 0.70（99%的好完成与69%的坏完成是被框选的）。一个从未见过奖励的通用SAE根本无法分离这两个类别（轮廓系数0.01，没有判别特征），因此0.79是对这一策划信号的样本内拟合，而不是奖励盲字典所恢复的结构。因此，我们同时呈现了这一方法及其控制实验：奖励过滤是一种廉价、无标签的方式，用于重用强化学习信号以提高可解释性，但它所揭示的大部分内容是完成形式。两个判别特征仍然是可读的（符号数学；程序性和评估性语言），我们将其视为说明性而非孤立的推理。
cs.CL / 16 / 2608.26137
Interpretable, Fairly Evaluated Automated L2 Speaking Assessment that Beats the Single-Human Ceiling and Why Pause Encoding Does Not Change LLM Fluency Scores
可解释的、公平评估的自动化L2口语评估超越单一人类评估上限及暂停编码为何不改变LLM流利度评分
Abstract
Second-language (L2) English learners can rarely rehearse speaking with a partner. Speaking is also the most anxiety-laden skill. These gaps drive a fast-growing market for automated speaking practice and scoring. But an automated score is trustworthy only if it is accurate, interpretable, fair, and benchmarked against the right human bar. We build an interpretable feature-plus-LLM hybrid for spontaneous L2 dialogue. We evaluate it without ever fitting to the human labels, against the ICNALE Global Rating Archive: 140 speeches rated by ~80 trained raters on 10 analytic criteria. We score the 130 L2 speeches with usable audio. A deterministic De-Jong speech-timing composite reaches rho=0.764. Blended with a single text-LLM fluency judgment, it reaches Spearman rho=0.818 against the consensus gold. This agrees with the consensus better than 81% of the 80 individual trained raters: above the median rater (rho=0.73) and near the best, and at ~83% of the reliability-corrected maximum (kappa_max=0.99). The blend improves on the composite alone by +0.054 (paired-bootstrap 95% CI [0.017, 0.108], excludes 0); the LLM adds a coarse fluency ranking that the continuous composite refines. We also report a controlled null on pause encoding, bounded to effects below about +/-0.1 rho at this sample size. Holding the LLM and learner words fixed and varying only how pauses are written into the prompt, inline pause locations do not beat aggregate pause statistics (-0.069, CI [-0.15, +0.08]), and a grounded mid-clause criterion gives no reliable gain. The fluency signal comes from the measured speech-timing features, not from how pauses are written for the LLM. We back every claim with two agreeing learner-isolation methods, paired-bootstrap CIs, a monologue negative control, per-feature reproduction of classical measurements, and a per-L1 fairness audit.
Chinese Translation
第二语言（L2）英语学习者很少能与伙伴进行口语练习。口语也是最令人焦虑的技能。这些缺口推动了自动化口语练习和评分的快速增长市场。然而，只有在评分准确、可解释、公平，并且与合适的人类标准进行基准比较时，自动评分才是可信的。我们构建了一个可解释的特征加LLM混合模型，用于自发的L2对话。我们在评估时从未与人类标签进行拟合，而是对照ICNALE全球评分档案进行评估：140个演讲由约80名训练有素的评分员在10个分析标准上进行评分。我们对130个可用音频的L2演讲进行评分。一个确定性的De-Jong演讲时序综合得到了rho=0.764。与单一文本-LLM流利度判断相结合，它在共识金标准上达到了Spearman rho=0.818。这一结果比80名个体训练评分员中的81%更一致：高于中位评分员（rho=0.73），接近最佳评分员，并且约为可靠性校正最大值的83%（kappa_max=0.99）。这种混合方法在综合评分基础上提高了+0.054（配对自助法95%置信区间[0.017, 0.108]，不包括0）；LLM增加了一个粗略的流利度排名，而连续综合评分则进行了细化。我们还报告了暂停编码的控制无效，限制在此样本大小下的影响低于大约+/-0.1 rho。固定LLM和学习者词汇，仅改变暂停在提示中的书写方式，内联暂停位置未能超越聚合暂停统计（-0.069，置信区间[-0.15, +0.08]），而基于中断标准的评估未能提供可靠的增益。流利度信号来自于测量的演讲时序特征，而非暂停在LLM中的书写方式。我们用两种一致的学习者隔离方法、配对自助法置信区间、独白负对照、每个特征的经典测量重现以及每个L1的公平性审计来支持每一项主张。
cs.CL / 17 / 2608.26138
Cross-Platform Generalisation Failure in Mental Health Natural Language Processing: A Five-Axis Fairness Audit of Transformer Models on Social Media
心理健康自然语言处理中的跨平台泛化失败：对社交媒体上变换模型的五轴公平性审计
Abstract
We introduce the Cross-Platform Fairness Evaluation (CPFE) framework -- a five-axis audit protocol covering discriminative performance, calibration, statistical significance, prediction equity, and attribution stability -- and apply it to four transformer models (BERT, RoBERTa, Emotion-DistilRoBERTa, GoEmotions-RoBERTa) trained on a Kaggle mental health corpus (n=35,556) and evaluated on Reddit (n=6,257) and Twitter (n=2,883) test sets with emotion labels mapped to clinical proxies. All three independently evaluated models exhibit consistent and substantial cross-platform AUC degradation (30.3-35.4% on Reddit, 37.9-39.5% on Twitter) relative to within-platform performance (AUC 0.983-0.987), confirmed across five independent training seeds. Calibration failure is concurrent and severe: ECE rises from 0.056-0.060 in-domain to 0.196-0.229 on Reddit and 0.499-0.542 on Twitter. Platform-specific temperature scaling reduces mean ECE by 88.0% without altering discriminative performance (mean |delta AUC|<0.01), confirming separable failure modes. Prediction equity analysis reveals large cross-platform disparities (raw DI < 0.17; prior-shift-adjusted DI: 0.11-0.29 on Reddit), with equalized odds differences of 0.753-0.830 for mental health proxy classes on Reddit and 0.755-0.831 for anxiety on Twitter. Attribution stability analysis shows near-complete vocabulary divergence across platforms (Jaccard J=0 in 14/16 model-class pairs at K=10). These findings support treating cross-platform validation across all five CPFE axes as a standard requirement for mental health NLP systems in heterogeneous environments. In a single-seed fine-tuning experiment, mean AUC improved by 0.216, suggesting target-platform labels provide greater benefit as training signal than as calibration signal.
Chinese Translation
我们提出了跨平台公平性评估（Cross-Platform Fairness Evaluation, CPFE）框架——一个涵盖区分性能、校准、统计显著性、预测公平性和归因稳定性的五轴审计协议，并将其应用于四个变换模型（BERT、RoBERTa、Emotion-DistilRoBERTa、GoEmotions-RoBERTa），这些模型是在Kaggle心理健康语料库（n=35,556）上训练的，并在Reddit（n=6,257）和Twitter（n=2,883）测试集上进行评估，情感标签映射到临床代理。所有三个独立评估的模型在跨平台的AUC表现上均显示出一致且显著的下降（在Reddit上下降30.3-35.4%，在Twitter上下降37.9-39.5%），相较于平台内的表现（AUC 0.983-0.987），这一结果在五个独立的训练种子中得到了确认。校准失败是同时发生且严重的：在领域内的ECE从0.056-0.060上升到Reddit的0.196-0.229和Twitter的0.499-0.542。平台特定的温度缩放将平均ECE降低了88.0%，而不改变区分性能（平均|delta AUC|<0.01），确认了可分离的失败模式。预测公平性分析揭示了跨平台的巨大差异（原始DI < 0.17；经过先验偏移调整的DI：在Reddit上为0.11-0.29），在Reddit上心理健康代理类的均衡几率差异为0.753-0.830，在Twitter上焦虑的均衡几率差异为0.755-0.831。归因稳定性分析显示跨平台的词汇差异几乎完全（在K=10时，14/16模型-类对的Jaccard J=0）。这些发现支持将跨平台验证视为心理健康自然语言处理系统在异构环境中的五个CPFE轴的标准要求。在一次单种子微调实验中，平均AUC提高了0.216，表明目标平台标签作为训练信号的效益大于作为校准信号的效益。
cs.CL / 18 / 2608.26139
Syntax vs. Semantics: How Transformers Learn Deep Dependencies
语法与语义：变换器如何学习深层依赖关系
Abstract
Large Language Models demonstrate remarkable syntactic fluency, yet the optimization dynamics governing their acquisition of deep semantic dependencies remain poorly understood. We propose a mechanistic framework that models this learning process as a competition between Surface Statistics and Deep Semantics. Our theoretical analysis identifies a ``Gradient Starvation" phenomenon where the error signals for sparse semantic dependencies are actively suppressed during early optimization. This suppression impedes the learning of structural reasoning and causes its emergence to manifest as a sudden phase transition. Furthermore, this framework offers a mechanistic basis for the effectiveness of Chain-of-Thought (CoT) strategies. By externalizing intermediate reasoning steps into concrete tokens, CoT effectively bypasses the suppression regime inherent to implicit reasoning. We validate these findings across scales ranging from toy transformers to production models (Llama-3.1-8B, Qwen2.5-Coder-7B). Finally, guided by this theory, we propose a topology-aligned contrastive objective that explicitly rectifies the gradient geometry. Experiments on variable binding tasks demonstrate that our method achieves an improvement that is over 2x larger than that obtained via standard cross-entropy fine-tuning.
Chinese Translation
大型语言模型展现出显著的语法流畅性，但支配其获取深层语义依赖关系的优化动态仍然不甚明了。我们提出了一个机械框架，将这一学习过程建模为表面统计与深层语义之间的竞争。我们的理论分析识别出一种“梯度饥饿”（Gradient Starvation）现象，即稀疏语义依赖关系的误差信号在早期优化过程中被主动抑制。这种抑制妨碍了结构推理的学习，并使其出现表现为突然的相变。此外，该框架为链式思维（Chain-of-Thought, CoT）策略的有效性提供了机械基础。通过将中间推理步骤外化为具体的标记，CoT有效地绕过了隐式推理固有的抑制机制。我们在从玩具变换器到生产模型（Llama-3.1-8B, Qwen2.5-Coder-7B）的不同规模上验证了这些发现。最后，在该理论的指导下，我们提出了一种拓扑对齐的对比目标，明确地修正了梯度几何。对变量绑定任务的实验表明，我们的方法在改进上超过了标准交叉熵微调所获得的2倍以上。
cs.CL / 19 / 2608.26140
Affix Cache for Diffusion Large Language Models
扩展缓存用于扩散大型语言模型
Abstract
Diffusion Large Language Models (DLLMs) enable non-autoregressive decoding and bidirectional context modeling, but efficient inference remains challenging. Unlike autoregressive systems, whose key-value (KV) cache can be reused for shared prefixes, DLLMs couple the KV states of shared context tokens with evolving generated tokens through bidirectional attention, making naive cache reuse stale while full recomputation is expensive. We present ACache, an affix-oriented cache reuse mechanism for shared text spans in DLLMs beyond prefixes. ACache identifies a small request-specific subset of critical affix tokens, called Anchor Tokens, by measuring their influence on masked generation tokens, and selectively recomputes the KV states of only these tokens while reusing the remaining affix cache. Built on Fast-dLLM, ACache recovers the accuracy loss caused by direct affix-cache reuse across different settings when recomputing around 20% of affix tokens. We also build a shared-prefix prototype on top of the Nano-vLLM engine, showing that ACache reduces recompute latency by up to 55.7% and improves end-to-end throughput by up to 1.68$\times$.
Chinese Translation
扩散大型语言模型（DLLMs）实现了非自回归解码和双向上下文建模，但高效推理仍然具有挑战性。与自回归系统不同，自回归系统的键值（KV）缓存可以重用于共享前缀，DLLMs通过双向注意力将共享上下文标记的KV状态与不断演变的生成标记耦合，使得简单的缓存重用变得过时，而完全重新计算则代价高昂。我们提出了ACache，一种面向扩展的缓存重用机制，用于DLLMs中超越前缀的共享文本片段。ACache通过测量关键扩展标记（称为锚标记）对掩蔽生成标记的影响，识别出一个小的请求特定的关键扩展标记子集，并选择性地重新计算这些标记的KV状态，同时重用其余的扩展缓存。基于Fast-dLLM，ACache在重新计算约20%的扩展标记时恢复了由于直接扩展缓存重用而导致的准确性损失。我们还在Nano-vLLM引擎上构建了一个共享前缀原型，显示ACache将重新计算延迟减少了多达55.7%，并将端到端吞吐量提高了多达1.68倍。
cs.CL / 20 / 2608.26141
AdaThinking-E: One-Token Entropy Regulation for Adaptive Thinking
AdaThinking-E：基于单标记熵调节的自适应思维
Abstract
Multimodal large language models have demonstrated strong document reasoning capabilities by incorporating explicit thinking processes. While this capability significantly improves performance on challenging tasks, current models apply such deep reasoning uniformly to all questions, resulting in unnecessary computational overhead for simple task. This not only degrades user experience but also negatively impact accuracy on benchmark datasets. We identify the critical need for adaptive thinking mechanisms that can intelligently determine when to engage reasoning based on question complexity. To address this, we propose AdaThinking-E, a novel reinforcement learning framework that learns adaptive thinking through one-token entropy regulation. Our key insight is that model confidence in the decision to engage thinking (or not) can be quantified through entropy analysis of the predicted probability distribution at critical decision tokens. This observation motivates our entropy-governed reward mechanism: the training process naturally transitions from high-entropy exploration, where the model experiments with different thinking strategies, to low-entropy convergence with confident, generalizable decision-making policies. Crucially, this approach enables models to intrinsically discover when to think without requiring manual intervention or external difficulty labels. Extensive experiments demonstrate that our approach enables models to be both accurate on complex problems and efficient on simple ones across diverse document tasks.
Chinese Translation
多模态大型语言模型通过引入显式思维过程展示了强大的文档推理能力。尽管这一能力显著提升了在挑战性任务上的表现，但当前模型对所有问题均匀地应用深度推理，导致在简单任务上产生不必要的计算开销。这不仅降低了用户体验，还对基准数据集的准确性产生了负面影响。我们识别出自适应思维机制的关键需求，该机制能够智能地判断何时根据问题复杂性进行推理。为此，我们提出了AdaThinking-E，这是一种新颖的强化学习框架，通过单标记熵调节学习自适应思维。我们的关键见解是，模型在决定是否进行思考时的信心可以通过对关键决策标记的预测概率分布进行熵分析来量化。这一观察促使我们提出熵驱动的奖励机制：训练过程自然从高熵探索过渡到低熵收敛，形成自信且可推广的决策策略。关键在于，这种方法使模型能够内在地发现何时进行思考，而无需手动干预或外部难度标签。大量实验证明，我们的方法使模型在复杂问题上既准确又在简单问题上高效，适用于多种文档任务。
cs.CL / 21 / 2608.26142
Position Is All You Need: A Free Lunch Token Compression Strategy for MLLM-based Referring Expression Segmentation
位置就是一切：基于MLLM的指称表达分割的免费午餐令牌压缩策略
Abstract
Referring Expression Segmentation (RES) aims to generate pixel-wise segmentation masks from complex and implicit textual queries. While recent advances in Multimodal Large Language Models (MLLMs) have substantially boosted RES performance, their prohibitive computational overhead remains a critical bottleneck, which, however, is rarely explored. To fill this gap, we first evaluate typical token compression methods on this task and observe a surprising performance degradation. In this paper, we aim to understand this phenomenon for a solution. By extensive experiments, we find that token compression for RES requires preserving the original position embeddings and local neighboring spatial structures, indicating that visual token position information is far more critical than in other tasks. Building on this insight, we ask: Can we design the token compression method purely based on the position information? Therefore, we propose PAYN, a plug-and-play, training-free token compression method that relies solely on position information. PAYN retains tokens that are adequately distributed in every local neighboring region while strictly preserving original positional indices, thereby maintaining spatial relational consistency. Experiments on multiple RES benchmarks demonstrate that our method outperforms existing token compression methods, verifying that position is indeed all you need for token compression in the MLLM-based RES task. Codes are avaliable at https://github.com/YuhanLiu231/PAYN.
Chinese Translation
指称表达分割（RES）旨在从复杂和隐含的文本查询中生成逐像素的分割掩码。尽管多模态大型语言模型（MLLMs）的最新进展显著提升了RES的性能，但其高昂的计算开销仍然是一个关键瓶颈，而这一点却鲜有研究。为了解决这一问题，我们首先评估了该任务上的典型令牌压缩方法，并观察到意外的性能下降。本文旨在理解这一现象以寻找解决方案。通过大量实验，我们发现RES的令牌压缩需要保留原始位置嵌入和局部邻域空间结构，这表明视觉令牌位置信息比其他任务更为关键。基于这一见解，我们提出了一个问题：我们能否设计一种仅基于位置信息的令牌压缩方法？因此，我们提出了PAYN，这是一种即插即用、无需训练的令牌压缩方法，完全依赖于位置信息。PAYN保留在每个局部邻域区域中适当分布的令牌，同时严格保持原始位置索引，从而维持空间关系的一致性。在多个RES基准上的实验表明，我们的方法优于现有的令牌压缩方法，验证了在基于MLLM的RES任务中，位置确实是进行令牌压缩所需的一切。代码可在 https://github.com/YuhanLiu231/PAYN 获取。
cs.CL / 22 / 2608.26143
Beyond Accuracy: A Qualitative Analysis of Vision-Language Models for Hate Speech Detection in Memes
超越准确性：对视觉-语言模型在恶意言论检测中的定性分析
Abstract
Memes have turned out to be a powerful tool through which individuals share their ideas concerning contemporary social and political problems. Their anonymity, as well as their ability to go viral, make them a powerful medium for spreading hate. It remains very difficult to identify such complex and context-dependent hate speech. Although they display excellent performance on multimodal tasks, vision-language models (VLMs) tend to ignore context, irony, and other subtle cues that play a key role in identifying hateful memes. In this work, we present a qualitative analysis of four state-of-the-art VLMs: LLaVA-7B, Qwen-VL, GPT-4o mini, and Claude 3 Haiku. We evaluate these models under zero-shot and few-shot prompting to examine how contextual framing influences their outputs. Our analysis goes beyond simple classification accuracy and focuses on a qualitative evaluation of the models' generated justifications, providing a more in-depth understanding of their thought processes and constraints when dealing with hateful memes.
Chinese Translation
表情包已经成为个人分享关于当代社会和政治问题观点的强大工具。它们的匿名性以及传播迅速的特性，使其成为传播仇恨的有效媒介。然而，识别这种复杂且依赖于上下文的仇恨言论仍然非常困难。尽管在多模态任务中表现出色，视觉-语言模型（VLMs）往往忽视上下文、讽刺和其他在识别仇恨表情包中起关键作用的微妙线索。在本研究中，我们对四个最先进的VLM进行了定性分析：LLaVA-7B、Qwen-VL、GPT-4o mini和Claude 3 Haiku。我们在零样本和少样本提示下评估这些模型，以考察上下文框架如何影响它们的输出。我们的分析超越了简单的分类准确性，专注于对模型生成的论证进行定性评估，从而更深入地理解它们在处理仇恨表情包时的思维过程和局限性。
cs.CL / 23 / 2608.26144
Why Current XAI Is Not Enough for Arabic NLP: A Critical Survey of the Explainability Gap
当前的可解释人工智能（XAI）为何不足以满足阿拉伯语自然语言处理（NLP）的需求：可解释性差距的批判性调查
Abstract
Explainable AI (XAI) is now a major theme in NLP; however, Arabic NLP remains under-explained in three connected senses. First, there is a method gap: Arabic XAI relies heavily on a small set of post-hoc techniques such as LIME, SHAP, attention visualization, and saliency, while broader NLP XAI offers richer diagnostic, counterfactual, probing, rationale-based, and human-centered methods. Second, there is a task gap: existing Arabic XAI work is concentrated in classification tasks, especially sentiment analysis, hate/offensive language detection, fake news, and spam, with weaker coverage of generation, retrieval, translation, summarization, structured prediction, and dialogue. Third, there is a linguistic gap: many explanations identify influential tokens, but rarely explain Arabic-specific phenomena such as morphology, clitics, dialectal variation, diglossia, orthographic ambiguity, diacritics, code-switching, named entities, cultural references, or Classical and religious registers. This critical structured survey synthesizes the reviewed literature on Arabic XAI across text, speech, and multimodal settings. We argue that Arabic NLP does not only need explanations of model decisions; it needs explanations that are faithful to Arabic as a linguistic, cultural, and sociotechnical object. We introduce a taxonomy of tasks, methods, linguistic units, varieties, goals, and evaluation practices, and propose a research agenda for linguistically grounded Arabic XAI.
Chinese Translation
可解释人工智能（XAI）现已成为自然语言处理（NLP）的一个主要主题；然而，阿拉伯语NLP在三个相互关联的方面仍然缺乏解释。首先，存在方法差距：阿拉伯XAI在很大程度上依赖于一小部分事后技术，如LIME、SHAP、注意力可视化和显著性，而更广泛的NLP XAI则提供了更丰富的诊断、反事实、探测、基于理由和以人为本的方法。其次，存在任务差距：现有的阿拉伯XAI研究主要集中在分类任务上，尤其是情感分析、仇恨/攻击性语言检测、假新闻和垃圾邮件，而在生成、检索、翻译、摘要、结构化预测和对话等任务上的覆盖较弱。第三，存在语言差距：许多解释识别了影响性标记，但很少解释阿拉伯特有的现象，如形态学、附属词、方言变异、 diglossia（双语现象）、正字法歧义、变音符号、代码切换、命名实体、文化参考或古典和宗教语域。本研究通过对文本、语音和多模态环境中阿拉伯XAI的文献进行批判性结构化调查，综合了相关研究。我们认为，阿拉伯NLP不仅需要模型决策的解释；它需要忠实于阿拉伯作为语言、文化和社会技术对象的解释。我们引入了任务、方法、语言单位、变体、目标和评估实践的分类法，并提出了一个以语言为基础的阿拉伯XAI研究议程。
cs.CL / 24 / 2608.26146
Vagdhenu: A Vrutta (Meter) Aware Shloka-to-Chant (TTS) System for Sanskrit
Vagdhenu：一个关注韵律的梵文咏唱系统（TTS）
Abstract
We present Vagdhenu, a vrutta (meter) aware shloka-to-chant system for Sanskrit: a text-to-speech system that maps a metrical verse to its chanted parayana recitation at high fidelity. This is an experience report, not a new architecture. We take an off-the-shelf flow-matching TTS backbone and a large-scale neural vocoder, and add the components a faithful Sanskrit chant pipeline needs: a frontend that routes Sanskrit through Kannada orthography to avoid the Hindi-style schwa deletion that Devanagari triggers in Indic models; a frontend that obeys subtle Sanskrit phonology (visarga sandhi with its jihvamuliya and upadhmaniya allophones, the aspiration contrast of alpaprana and mahaprana, and the dental, retroflex, and palatal sibilants kept distinct); and a vrutta-aware mechanism that detects the meter and picks an exactly matched reference under a half-reference rule. We report a negative result that shaped the system: in a self-infilling flow-matching backbone, a text-side prosody conditioner is architecturally inert, because the model recovers pitch from the context mel and the embedding gets no gradient; the reference clip and a voice-steering retrain are the only working prosody levers. We also report a comparative lineage across four families (StyleTTS2, VITS2, Matcha-TTS, and the flow-matching backbone), where each earlier family hit a ceiling on conjuncts or prosody that a five-hour clone cleared at an expert MOS near 4.6. The system shipped two deployments: a 32-chapter, 5183-verse video corpus (about 17.5 hours) and an audio app covering about 18000 verses across 12 books. We release the frontend, inference and training code, weights, a single-speaker chant dataset, and an interactive demo.
Chinese Translation
我们介绍了Vagdhenu，一个关注韵律的梵文咏唱系统：一个将韵律诗映射到其高保真咏唱诵读的文本到语音系统。这是一份经验报告，而不是一种新架构。我们采用了一种现成的流匹配TTS骨干网和一个大规模神经声码器，并添加了忠实于梵文咏唱管道所需的组件：一个前端，通过卡纳达文字法将梵文路由，以避免天城体在印度模型中引发的印地语风格的元音弱化；一个遵循微妙梵文音系（包括与其jihvamuliya和upadhmaniya变音的visarga sandhi，微声和重声的对比，以及保持清晰的齿音、卷舌音和腭音的擦音）的前端；以及一个关注韵律的机制，能够检测韵律并在半参考规则下选择一个完全匹配的参考片段。我们报告了一个影响系统的负面结果：在自填充流匹配骨干网中，文本侧的韵律调节器在架构上是无效的，因为模型从上下文mel恢复音高，而嵌入没有梯度；参考片段和语音引导再训练是唯一有效的韵律调节手段。我们还报告了四个家族（StyleTTS2、VITS2、Matcha-TTS和流匹配骨干网）之间的比较谱系，其中每个早期家族在结合或韵律上达到了一个上限，而一个五小时的克隆在专家MOS接近4.6的情况下突破了这一限制。该系统发布了两个部署：一个包含32章、5183节诗的高清视频语料库（约17.5小时），以及一个覆盖12本书中约18000节诗的音频应用。我们发布了前端、推理和训练代码、权重、单声道咏唱数据集和一个互动演示。
cs.CL / 25 / 2608.26147
CARE: Causally-Aligned Reasoning Exploration for Medical Large Language Models
CARE：用于医学大语言模型的因果对齐推理探索
Abstract
Large Language Models (LLMs) have shown strong potential for medical reasoning, yet the scarcity and cost of expert-annotated data constrain their progress. While reinforcement learning offers a scalable alternative, standard outcome-based methods in medicine often suffer from autoregressive credit assignment failure and gradient variance explosion. This leads to the "Right Answer, Wrong Reason" trap, where models inadvertently reinforce spurious correlations and dataset shortcuts rather than valid clinical deduction. In this work, we propose Causally-Aligned Reasoning Exploration (CARE), a theoretically grounded framework for intrinsic experience curation. CARE is built upon two rigorous conditions for high-quality training trajectories: Causal Sufficiency, which utilizes an agreement-based self-verification mechanism to mimic $do$-calculus interventions and effectively debias gradients; and Proximal Learnability, which employs dynamic entropy bounds to select experiences within the model's zone of proximal development for variance-bounded optimization. These rigorously filtered experiences are optimized via a dual-stream objective that combines on-policy group-relative exploration with difficulty-weighted experience replay. Extensive experiments on diverse medical multimodal and text-only benchmarks demonstrate that CARE consistently outperforms other strong competitors, substantially reducing correct-but-inconsistent reasoning and improving training stability.
Chinese Translation
大型语言模型（LLMs）在医学推理方面展现出强大的潜力，但专家标注数据的稀缺和成本限制了它们的发展。尽管强化学习提供了一种可扩展的替代方案，但医学中的标准基于结果的方法常常面临自回归信用分配失败和梯度方差爆炸的问题。这导致了“正确答案，错误理由”的陷阱，模型无意中强化了虚假的相关性和数据集捷径，而不是有效的临床推理。在本研究中，我们提出了因果对齐推理探索（CARE），这是一个理论基础扎实的内在经验策划框架。CARE建立在两个高质量训练轨迹的严格条件之上：因果充分性（Causal Sufficiency），利用基于一致性的自我验证机制来模拟 $do$-calculus 干预，并有效去偏梯度；以及近端可学习性（Proximal Learnability），通过动态熵界限选择模型近端发展区间内的经验，以实现方差受限的优化。这些经过严格筛选的经验通过双流目标进行优化，结合了基于策略的组相对探索与难度加权的经验重放。在多样的医学多模态和文本基准上进行的广泛实验表明，CARE始终优于其他强有力的竞争者，显著减少了正确但不一致的推理，并改善了训练稳定性。
cs.CL / 26 / 2608.26148
Towards Interpretable Depression Detection: Linking Acoustic Features to DSM-5 Indicators
朝向可解释的抑郁检测：将声学特征与DSM-5指标联系起来
Abstract
Depression affects millions worldwide, yet diagnosis relies on subjective self-reports that may miss authentic behavior. This paper presents an approach linking speech acoustics to DSM-5 depressive-behavior indicators through a transparent Linkage Framework. Unlike black-box models, the framework explicitly maps acoustic features (pitch variability, pauses, speech tempo) to clinical indicators, enabling interpretable, indicator-level outputs. The system runs locally on commodity hardware (HW) to preserve privacy. Preliminary evaluation on DAIC-WOZ shows directionally consistent associations between acoustic features and DSM-5 indicators for psychomotor change and concentration difficulty, supporting the design rationale. Future work will validate on longitudinal datasets and extend multimodal integration while maintaining edge constraints.
Chinese Translation
抑郁症影响着全球数百万人，然而诊断依赖于主观的自我报告，这可能会忽略真实的行为。本文提出了一种通过透明的链接框架将语音声学与DSM-5抑郁行为指标联系起来的方法。与黑箱模型不同，该框架明确地将声学特征（音高变异性、停顿、语速）映射到临床指标，从而实现可解释的指标级输出。该系统在普通硬件（HW）上本地运行，以保护隐私。对DAIC-WOZ的初步评估显示声学特征与DSM-5在精神运动变化和注意力困难方面的指标之间存在方向一致的关联，支持了设计的合理性。未来的工作将基于纵向数据集进行验证，并在保持边缘约束的同时扩展多模态集成。
cs.CL / 27 / 2608.26152
Artificial Intelligence Models Can Predict and Collaboratively Modulate Human Memory Search
人工智能模型能够预测并协同调节人类记忆搜索
Abstract
Large language models (LLMs) exhibit unprecedented natural language generation and many text-based problem-solving capabilities. Indeed, in many language-based tasks, for example routine coding, these artificial intelligence models have reduced, or even eliminated, the need for human input. But rather than replacing human cognitive effort, LLMs may instead serve as cognitive tools to extend human abilities, particularly when they are engaged in a task requiring open-ended conceptual exploration and creative ideation. However, we are yet to understand how these models may enhance such generative human cognitive abilities in human--AI interactions. In this study, we explore and evaluate the ability of LLMs to follow and enhance human mental trajectories during semantic memory search. To test this, we use the semantic fluency task (SFT), a classic cognitive paradigm requiring generative semantic memory retrieval that has long served to characterize convergent and divergent thinking in humans. We demonstrate that an LLM's abilities to track and predict human memory trajectories in this task exceed those of other humans.
Chinese Translation
大型语言模型（LLMs）展现出前所未有的自然语言生成能力以及多种基于文本的问题解决能力。实际上，在许多基于语言的任务中，例如常规编码，这些人工智能模型已经减少甚至消除了对人类输入的需求。然而，LLMs并不是取代人类的认知努力，而是可能作为认知工具来扩展人类的能力，特别是在进行需要开放式概念探索和创造性构思的任务时。然而，我们尚未理解这些模型如何在人与人工智能的互动中增强这种生成性的人类认知能力。在本研究中，我们探讨并评估LLMs在语义记忆搜索过程中跟随和增强人类心理轨迹的能力。为了测试这一点，我们使用了语义流畅性任务（SFT），这是一种经典的认知范式，要求生成性地检索语义记忆，长期以来用于表征人类的聚合思维和发散思维。我们证明了LLM在这一任务中跟踪和预测人类记忆轨迹的能力超过了其他人类。
cs.CL / 28 / 2608.26154
Evaluating AI Generated Summaries for Cancer Patients
评估针对癌症患者的人工智能生成摘要
Abstract
Large language models (LLMs) are increasingly being integrated into digital health platforms to generate summaries of complex medical data. Although these models can improve patient engagement and communication, these systems also raise concerns about accuracy, faithfulness, and safety in clinical contexts. In this study, we evaluate AI-generated summaries within a cancer patient care application using a dual assessment framework. Human domain experts, including oncology clinicians and patient-facing care staff, provided ground-truth evaluations of summary quality along dimensions of accuracy, clinical relevance, and readability. In parallel, we employed LLMs serving as evaluators (LLM-as-a-judge). Some limitations were identified in the generated summaries e.g., occasional omissions and minor inaccuracies. These were systematically analyzed and used to iteratively improve prompt design, grounding, and safety guardrails.
Chinese Translation
大型语言模型（LLMs）正越来越多地被整合到数字健康平台中，以生成复杂医疗数据的摘要。尽管这些模型可以改善患者参与度和沟通，但在临床环境中，这些系统也引发了关于准确性、忠实性和安全性的担忧。在本研究中，我们使用双重评估框架评估癌症患者护理应用中的人工智能生成摘要。包括肿瘤科临床医生和面向患者的护理人员在内的人类领域专家提供了摘要质量的真实评估，评估维度包括准确性、临床相关性和可读性。同时，我们还采用了作为评估者的LLMs（LLM-as-a-judge）。在生成的摘要中识别出了一些局限性，例如偶尔的遗漏和小的错误。这些问题被系统分析，并用于迭代改进提示设计、基础支持和安全防护措施。
cs.CL / 29 / 2608.26155
VFA: Empowering Multilingual MLLMs via Vision-Free Adaptation
VFA：通过无视觉适应增强多语言 MLLMs
Abstract
Multimodal large language models have advanced rapidly, yet most remain English-centric, as scaling multilingual multimodal instruction tuning is limited by the scarcity and high cost of high-quality non-English image-text supervision. Although multilingual text data is abundant, naive textual fine-tuning can disrupt vision-language alignment and induce catastrophic forgetting. We propose Vision-Free Adaptation (VFA), a framework that decouples multilingual language enhancement from visual alignment by composing complementary task vectors over a shared LLM backbone. Specifically, we fine-tune a base LLM on multilingual text data to derive a multilingual task vector, which is then merged with the vision-aligned task vector of an MLLM. Experiments on five MLLMs across six multilingual multimodal benchmarks show consistent improvements while preserving both general multimodal and text-only capabilities. Moreover, using less than 2% of the text data, VFA narrows the gap to the fully multimodal-trained model, demonstrating its data efficiency.
Chinese Translation
多模态大型语言模型发展迅速，但大多数仍以英语为中心，因为多语言多模态指令调优的规模受到高质量非英语图像-文本监督稀缺和高成本的限制。尽管多语言文本数据丰富，但简单的文本微调可能会破坏视觉-语言对齐并导致灾难性遗忘。我们提出了无视觉适应（Vision-Free Adaptation, VFA）框架，通过在共享的 LLM（大型语言模型）主干上组合互补任务向量，将多语言语言增强与视觉对齐解耦。具体而言，我们在多语言文本数据上微调基础 LLM，以导出多语言任务向量，然后将其与 MLLM（多语言大型语言模型）的视觉对齐任务向量合并。在六个多语言多模态基准测试中对五个 MLLM 的实验表明，在保持一般多模态和仅文本能力的同时，性能一致提升。此外，使用不到 2% 的文本数据，VFA 缩小了与完全多模态训练模型之间的差距，展示了其数据效率。
cs.CL / 30 / 2608.26159
Self-Generated Text Recognition: Quality Heuristics, Cross-Task Transfer, and Downstream Bias in LLM Evaluation
自生成文本识别：质量启发式、跨任务迁移与大型语言模型评估中的下游偏差
Abstract
Self-Generated Text Recognition (SGTR)--the ability of an LLM to identify its own outputs--poses risks to AI safeguards that rely on LLMs as evaluators or monitors. Specifically, an LLM may recognize outputs from other copies of the same model and make biased judgments or collude outright. Prior work has drawn conflicting conclusions about whether current models possess significant SGTR capabilities. We reconcile these findings by identifying key experimental design choices--which we term operationalizations--that drive divergent results. Evaluating 13-21 models across six operationalizations, we find that accuracy varies substantially with evaluation format (pairwise vs. individual assessments of text), conversation structure (presenting candidate text in user tags vs. assistant tags), and the domain of the task used to generate candidate text (e.g., coding vs. summarization). We corroborate previous observations that a quality heuristic--models attributing authorship to text they perceive as higher quality--is a dominant confound. We also find that improving a model's SGTR performance via SFT in one evaluation configuration can generalize to others. Training for SGTR additionally causes models to prefer their own outputs when acting as a judge in the AlpacaEval framework. Finally, we discuss the implications of our evaluations for the safety of future AI systems: our work suggests that, despite confounds, some models possess practical SGTR capabilities, and that training a model for SGTR in one setting can affect its self-recognition and self-preference more generally. We conclude that SGTR should be monitored and considered in the design of safety-critical AI applications.
Chinese Translation
自生成文本识别（Self-Generated Text Recognition, SGTR）——大型语言模型（LLM）识别自身输出的能力——对依赖LLM作为评估者或监控者的人工智能安全措施构成风险。具体而言，LLM可能会识别来自同一模型其他副本的输出，并做出偏见判断或直接串通。先前的研究对当前模型是否具备显著的SGTR能力得出了相互矛盾的结论。我们通过识别关键的实验设计选择——我们称之为操作化——来调和这些发现，这些选择驱动了不同的结果。在六种操作化下评估13-21个模型，我们发现准确性在评估格式（成对评估与单独评估文本）、对话结构（以用户标签呈现候选文本与以助手标签呈现候选文本）以及用于生成候选文本的任务领域（例如，编码与摘要）之间有显著差异。我们证实了先前观察到的一个质量启发式——模型将其视为更高质量的文本归因于其创作——是一个主要的混淆因素。我们还发现，通过在一种评估配置中进行监督微调（SFT）来提高模型的SGTR性能可以推广到其他配置。在AlpacaEval框架中，训练SGTR还导致模型在担任评审时更倾向于选择自己的输出。最后，我们讨论了我们的评估对未来人工智能系统安全性的影响：我们的研究表明，尽管存在混淆因素，一些模型具备实际的SGTR能力，并且在一种环境中训练模型以进行SGTR可能会更普遍地影响其自我识别和自我偏好。我们得出结论，SGTR应在设计安全关键的人工智能应用时进行监控和考虑。
cs.CL / 31 / 2608.26161
Mutual Debiasing via Dual-Seed Comparison for Probabilistic Sampling in Large Language Models
通过双种子比较实现的大语言模型概率采样的相互去偏差
Abstract
Although Large Language Models (LLMs) demonstrate remarkable capabilities in reasoning and decision-making, high-fidelity probabilistic sampling remains a persistent challenge. When generating random variables, LLMs consistently exhibit systematic biases that warp the target probability distributions. Current approaches often rely on a single, self-generated seed, which inherits model-specific biases. To overcome this vulnerability, we introduce Dual-Seed Comparison (DSC), a transparent, tool-free protocol that utilizes two independent LLM-generated seeds to neutralize bias. DSC compares the character-level ordinal values of the two seeds to construct a bit sequence, converts and normalizes this sequence into a pseudo-uniform variate, and then maps the variate to the target distribution through the inverse cumulative distribution function (CDF). Empirical results show that DSC substantially outperforms existing methods across 96\% of evaluated settings. Beyond direct sampling, task-adapted variants based on the DSC comparison operator improve distributional control in MCQ generation and attribute-constrained text-to-image prompting.
Chinese Translation
尽管大语言模型（LLMs）在推理和决策方面展现出显著的能力，但高保真概率采样仍然是一个持续的挑战。在生成随机变量时，LLMs 一直表现出系统性的偏差，这些偏差扭曲了目标概率分布。目前的方法通常依赖于单一的自生成种子，这会继承模型特定的偏差。为了解决这一脆弱性，我们提出了双种子比较（Dual-Seed Comparison, DSC），这是一种透明的、无工具的协议，利用两个独立的 LLM 生成的种子来中和偏差。DSC 比较两个种子的字符级序数值，以构建比特序列，将该序列转换并归一化为伪均匀变量，然后通过逆累积分布函数（CDF）将该变量映射到目标分布。实证结果表明，DSC 在 96\% 的评估设置中显著优于现有方法。除了直接采样，基于 DSC 比较算子的任务适应变体在多项选择题生成和属性约束的文本到图像提示中改善了分布控制。
cs.CL / 32 / 2608.26163
From Sound to Symptom: Real-Time Respiratory Signal Understanding for Conversational Healthcare Agents
从声音到症状：实时呼吸信号理解用于对话式医疗代理
Abstract
Cough events during live spoken conversations carry clinically valuable respiratory signals, yet existing dialogue systems treat them as acoustic noise to be discarded. We present HealthCUES (Clinical Understanding from Embodied Sounds), a streaming pipeline for paralinguistic respiratory monitoring in real-time conversational agents, a capability that, to the best of our knowledge, is absent from all prior systems. HealthCUES processes audio through a rolling buffer aligned with dialogue turn boundaries, enabling sub-second event detection without interrupting conversational flow. Beyond binary cough detection, the system provides fine-grained analytics: (i) differentiation between coughing and throat clearing, (ii) cough subtype classification (dry, wet, barking, whooping) with confidence scores, and (iii) temporal duration estimation with start-end boundaries. To prevent alert fatigue, HealthCUES introduces dialogue-aware gating mechanisms that modulate triggering based on conversational context. The system leverages Qwen3Omni, a multimodal large language model (MLLM), with constrained structured outputs, decomposing cough analysis into parallel prediction tasks for independent prompt optimization. Evaluation on 847 in-house conversational audio segments demonstrates 93\% F1 for cough detection, 0.75 weighted-F1 for wet/dry subtype classification, and average end-to-end latency of 340ms; external validation on the AMI meeting corpus confirms robust cough, throat-clearing, and speech separation in the presence of speech (0.91 macro-F1). A user study with licensed healthcare professionals confirms the clinical relevance of subtype information and the system's utility in telehealth workflows.
Chinese Translation
在实时口语对话中，咳嗽事件携带着临床上有价值的呼吸信号，但现有的对话系统将其视为需被丢弃的声学噪声。我们提出了HealthCUES（来自具身声音的临床理解），这是一个用于实时对话代理的副语言呼吸监测流处理管道，至今为止，所有先前系统均缺乏这一能力。HealthCUES通过与对话轮次边界对齐的滚动缓冲区处理音频，实现了在不打断对话流的情况下进行亚秒级事件检测。除了二元咳嗽检测外，该系统还提供细粒度分析：（i）区分咳嗽与清喉，（ii）咳嗽亚型分类（干咳、湿咳、犬吠样咳嗽、打嗝）及其置信度评分，以及（iii）带有起止边界的时间持续估计。为了防止警报疲劳，HealthCUES引入了对话感知的门控机制，根据对话上下文调节触发。该系统利用Qwen3Omni，一个多模态大型语言模型（MLLM），并采用受限结构化输出，将咳嗽分析分解为并行预测任务，以实现独立的提示优化。在847个内部对话音频片段上的评估显示，咳嗽检测的F1值为93%，湿/干亚型分类的加权F1值为0.75，平均端到端延迟为340毫秒；在AMI会议语料库上的外部验证确认了在语音存在下的咳嗽、清喉和语音分离的稳健性（宏观F1值为0.91）。与持证医疗专业人员的用户研究确认了亚型信息的临床相关性及该系统在远程医疗工作流程中的实用性。
cs.CL / 33 / 2608.26165
Using Poly-Encoders for Computationally Efficient Automated Creativity Assessment
使用多编码器进行计算高效的自动化创造力评估
Abstract
Automated creativity assessment has been a long standing challenge, with traditional methods often being resource intensive or lacking practical accuracy. We introduce a novel approach by using Poly-Encoder for computationally efficient and accurate automated creativity assessment. We fine-tuned a Poly-Encoder on a public dataset from the Scientific Creative Thinking Test, comprised of approximately 18,000 human-rated question responses. Our method leverages small pre-trained BERT encoders, achieving performance comparable to fine-tuned Large Language Models while significantly reducing computational demands. Experiments with the BERT-family models and poly-code counts achieved Pearson correlations of up to r = 0.74, 95% CI [0.73, 0.75] with human raters, matching the performance of resource intensive LLMs. This study bridges the gap between high performance and computational efficiency, potentially enabling widespread implementation of automated creativity assessment on accessible consumer-grade hardware. With some limitations, our findings suggest that Poly-Encoders are a promising alternative to LLMs for practical, scalable creativity assessment in various contexts, especially educational.
Chinese Translation
自动化创造力评估一直是一个长期存在的挑战，传统方法往往资源密集或缺乏实际准确性。我们通过使用多编码器（Poly-Encoder）提出了一种新颖的方法，以实现计算高效且准确的自动化创造力评估。我们在一个来自科学创造性思维测试的公共数据集上对多编码器进行了微调，该数据集包含约18,000个由人类评分的问题回答。我们的方法利用小型预训练的BERT编码器，取得了与微调的大型语言模型（Large Language Models）相当的性能，同时显著降低了计算需求。与BERT系列模型和多编码计数的实验达到了与人类评分者的皮尔逊相关系数高达r = 0.74，95%置信区间[0.73, 0.75]，匹配了资源密集型大型语言模型的表现。本研究弥合了高性能与计算效率之间的差距，可能使自动化创造力评估在可获取的消费级硬件上广泛实施成为可能。尽管存在一些局限性，我们的研究结果表明，多编码器是大型语言模型在各种背景下进行实用、可扩展的创造力评估的有前景的替代方案，尤其是在教育领域。
cs.CL / 34 / 2608.26168
Hallucinations in LLMs: A Lifecycle-Based Survey of Causes, Detection, Mitigation, and Prevention
大型语言模型中的幻觉：基于生命周期的原因、检测、缓解和预防的调查
Abstract
The lifecycle of hallucination in LLMs is a concept that enables building solid frameworks on the control and reliability of LLMs in high-stakes environments, including health, legal, and scientific research. Although previous surveys have primarily focused on detection or mitigation, this survey provides a lifecycle-based overview of the hallucinations in the LLMs, their cause, detection, mitigation, and prevention.We propose a three-fold categorization of hallucinations across the LLM lifecycle: data-related, training-related, and inference-related, which is consistent with the lifecycle of the development of the LLM. Each of these stages is discussed regarding the cause of hallucinations, their detection, and the ways they can be addressed under specific mitigation or prevention interventions. In addition, we discuss the available benchmark data using a number of parameters so as to establish their suitability in identifying, restricting and managing hallucinations. The survey provides researchers and practitioners with a standardized framework to understand, diagnose, and cure hallucinations in a systematic system to present actionable data to build safer and more reliable LLMs.
Chinese Translation
大型语言模型（LLMs）中幻觉的生命周期是一个概念，使得在高风险环境中（包括健康、法律和科学研究）构建对LLMs的控制和可靠性提供了坚实的框架。尽管之前的调查主要集中在检测或缓解上，但本调查提供了一个基于生命周期的LLMs中幻觉的概述，包括其原因、检测、缓解和预防。我们提出了一个基于LLM生命周期的幻觉三重分类：与数据相关的、与训练相关的和与推理相关的，这与LLM的发展生命周期一致。我们讨论了每个阶段中幻觉的原因、检测方法以及在特定的缓解或预防干预下可以采取的应对措施。此外，我们讨论了可用基准数据，使用多个参数来确定其在识别、限制和管理幻觉方面的适用性。本调查为研究人员和从业者提供了一个标准化框架，以系统地理解、诊断和治疗幻觉，从而提供可操作的数据，以构建更安全、更可靠的LLMs。
cs.CL / 35 / 2608.26175
Lost in Compression: A Controlled Cross-Lingual Audit of Extractive Prompt Compressors
压缩中的迷失：对提取式提示压缩器的受控跨语言审计
Abstract
Extractive prompt compression promises to cut LLM inference costs by removing low-information tokens, and learned compressors such as LLMLingua-2 report strong results on English benchmarks. Most other languages already pay a token premium: the same content costs 1.3-1.8x more tokens than in English. We ask whether compression closes or widens this gap. Using fully parallel data in ten languages spanning five scripts, with controls budget-matched in the target model's tokenizer, we audit four learned compressors against four deterministic baselines, on eleven target models from ten vendors (over 250,000 evaluation calls). Three of the compressors are trained with English supervision (LLMLingua-2 XLM-R/mBERT; Kompress-v2 from the production Headroom stack); the fourth, XProvence, is trained multilingually. First, the transfer gap is real, replicates across target models and compressor backbones, and is strongly rate-dependent: at a 0.33 keep-rate English retains 57-62% of normalized context utilization while Lithuanian retains 10-24% and Chinese essentially none, despite Chinese having the smallest token premium. Second, the gap tracks compression supervision data, not architecture. All three English-trained compressors show it, deterministic methods show no comparable gap, and the multilingually trained XProvence v1 shows none. Its v2 release, retrained on translated data, empties 92% of Chinese contexts at its aggressive threshold without any warning. Third, in a harder long-context setting, aggressive learned compression drives compressed contexts to or below no-context utility in three of five non-English languages. A translate-then-compress pipeline matches or beats native compression at roughly half the token cost in three of five tested languages. We release all code, compressions, and model outputs. Safe compression budgets are much smaller outside English.
Chinese Translation
提取式提示压缩通过去除低信息量的标记，有望降低大语言模型（LLM）的推理成本，而像 LLMLingua-2 这样的学习型压缩器在英语基准测试中报告了良好的结果。大多数其他语言已经面临标记溢价：相同内容在英语中需要 1.3-1.8 倍的标记。我们探讨压缩是否缩小或扩大了这一差距。使用涵盖五种书写系统的十种语言的完全平行数据，并在目标模型的分词器中进行预算匹配控制，我们对四种学习型压缩器与四种确定性基线进行了审计，涉及来自十个供应商的十一种目标模型（超过 250,000 次评估调用）。其中三种压缩器是在英语监督下训练的（LLMLingua-2 XLM-R/mBERT；来自生产 Headroom 堆栈的 Kompress-v2）；第四种 XProvence 是多语言训练的。首先，转移差距是真实存在的，在目标模型和压缩器骨干网络中均有复制，并且与速率密切相关：在 0.33 的保留率下，英语保留了 57-62% 的标准化上下文利用率，而立陶宛语保留了 10-24%，而中文几乎没有，尽管中文的标记溢价最小。其次，差距与压缩监督数据相关，而非架构。所有三种英语训练的压缩器均显示出这一差距，确定性方法未显示出可比的差距，而多语言训练的 XProvence v1 则没有。其 v2 版本在翻译数据上重新训练，在其激进阈值下清空了 92% 的中文上下文而没有任何警告。第三，在更困难的长上下文设置中，激进的学习型压缩使得三种非英语语言的压缩上下文达到或低于无上下文效用。在测试的五种语言中，翻译后再压缩的流程在大约一半的标记成本下与本地压缩相匹配或超越。我们发布了所有代码、压缩结果和模型输出。英语以外的安全压缩预算要小得多。
cs.CL / 36 / 2608.26177
A Multi-Framework Comparison of Outline Stages in Long-Form Generation with LLMs
基于多框架的长文本生成大纲阶段比较研究
Abstract
Long-form generation exposes fundamental limitations of large language models. Even 70B-parameter models exhibit length collapse at 16k-token outputs, and multi-chapter stories frequently trigger the attribute drift characteristic of the ``lost-in-the-middle'' effect. The ``outline-first, write-later'' paradigm has gained wide adoption, yet existing research evaluates the final writing rather than the outline itself, conflating two evaluation objects that should be decoupled. We construct a unified head-to-head benchmark covering 7 representative long-form generation frameworks across 3 generation granularities -- single-chapter, multi-chapter, and whole-book -- and propose an anchor-based LLM-as-a-judge protocol that directly assesses outlines against the source text on a 5-point anchored scale. Across 21 framework-granularity cells, no single framework dominates; performance depends on the match between a framework's intrinsic output form and the target granularity. SuperWriter ranks first in the length-constrained single-chapter mode, but this advantage degrades in whole-book mode. The outline-side ranking correlates only moderately with the writing-side ranking, supporting the outline--writing decoupling principle. Compute constraints limit the writing-side evaluation to a subset of cases; follow-up experiments will expand the sample size and add cross-model evaluators to enable stronger statistical inference.
Chinese Translation
长文本生成暴露了大型语言模型的基本局限性。即使是70亿参数的模型在16k-token输出时也会出现长度崩溃，而多章节故事常常触发“迷失在中间”效应的特征漂移。虽然“先大纲，后写作”的范式已被广泛采用，但现有研究评估的是最终写作，而非大纲本身，将两个应当解耦的评估对象混为一谈。我们构建了一个统一的对比基准，涵盖了7个代表性的长文本生成框架，涉及3种生成粒度——单章节、多章节和整本书，并提出了一种基于锚点的LLM作为评判者的协议，直接在5分锚定量表上评估大纲与源文本的匹配。在21个框架-粒度单元中，没有单一框架占据主导地位；性能取决于框架的内在输出形式与目标粒度之间的匹配。SuperWriter在长度受限的单章节模式中排名第一，但在整本书模式中这一优势减弱。大纲侧排名与写作侧排名的相关性仅为中等，支持大纲与写作解耦的原则。计算限制使得写作侧评估仅限于部分案例；后续实验将扩大样本规模，并增加跨模型评估者，以增强统计推断的力度。
cs.CL / 37 / 2608.26180
PACEShop: Evaluating Personalized, Actionable, Compositional, and Evidence-grounded Shopping Assistants
PACEShop：评估个性化、可操作、组合性和基于证据的购物助手
Abstract
Shopping assistants are shifting from ranked product lists toward structured decision support, where systems must synthesize shopper context, product evidence, and next-step guidance into a coherent recommendation experience. This changes the unit of evaluation: a fluent response can still fail by ignoring shopper context, contradicting itself across components, or leaving defects too vague to localize. Existing personalization, grounding, and LLM-as-a-judge benchmarks cover pieces of this problem, but they do not define a joint evaluation target for structured shopping-assistant responses. We formulate this missing evaluation target as PACE: Personalized, Actionable, Compositional, and Evidence-grounded evaluation. We instantiate PACE with two artifacts: PACEShop, a benchmark dataset that makes the target measurable through 22,625 controlled records with structured personas, auditable evidence pools, GOOD/BAD labels, and gold defect family and location annotations; and PACEJudge, a training-free judging protocol that makes the target reportable through a structured output contract. Our experiments show that generic judges can recognize broad quality but fail to recover the diagnostic fields required for PACE; PACEShop makes these failures verifiable, and PACEJudge improves persona-source, cross-component, grounding, and family/location closure without retraining, showing that realistic shopping-assistant evaluation requires a task-matched output contract rather than only a stronger backbone or scalar prompt.
Chinese Translation
购物助手正从排名产品列表转向结构化决策支持，其中系统必须将购物者的背景、产品证据和下一步指导综合成一个连贯的推荐体验。这改变了评估的单位：流畅的响应仍然可能因忽视购物者背景、在各个组件之间自相矛盾或留下过于模糊的缺陷而失败。现有的个性化、基础和大型语言模型（LLM）作为评判标准的基准涵盖了这个问题的部分内容，但它们并没有定义一个针对结构化购物助手响应的联合评估目标。我们将这个缺失的评估目标表述为PACE：个性化（Personalized）、可操作（Actionable）、组合性（Compositional）和基于证据（Evidence-grounded）的评估。我们通过两个工具实例化PACE：PACEShop，一个基准数据集，通过22,625个控制记录、结构化角色、可审计的证据池、好/坏标签以及金标准缺陷类别和位置注释，使目标可测量；以及PACEJudge，一个无训练的评判协议，通过结构化输出合同使目标可报告。我们的实验表明，通用评判者能够识别广泛的质量，但无法恢复PACE所需的诊断字段；PACEShop使这些失败可验证，而PACEJudge在不重新训练的情况下改善了角色来源、跨组件、基础和类别/位置的闭合，表明现实的购物助手评估需要一个与任务匹配的输出合同，而不仅仅是更强的基础模型或标量提示。
cs.CL / 38 / 2608.26186
Investigating the Influence of Prompt and Response Languages on LLM Content Generation
探讨提示语言和响应语言对大型语言模型内容生成的影响
Abstract
This study examines how prompt and response language influence the behavior of large language models. Using five models, we evaluated answers to 68 non translation questions across four language conditions: English to English, English to Norwegian, Norwegian to Norwegian, and Norwegian to English. After removing refused items, the dataset contains 1348 responses. We measure length differences with Cohen d, semantic fidelity with LabSE cosine similarity, and cross lingual keyword overlap with both raw and soft Jaccard. Prompt language has a strong effect on response length. With English output, Norwegian prompts shorten responses by about thirty seven percent. With Norwegian output, English prompts shorten responses by about forty one percent. The largest cross lingual contrast shows a reduction in word count but a smaller reduction in tokens, reflecting tokenizer differences. Despite variation in length, semantic similarity remains high, and soft Jaccard reveals substantial conceptual overlap that raw Jaccard does not capture. Effect sizes vary across models, indicating heterogeneity. Prompt language is not neutral and systematically shapes output length and lexical realization, with implications for multilingual prompt design.
Chinese Translation
本研究考察了提示语言和响应语言如何影响大型语言模型的行为。我们使用五个模型评估了在四种语言条件下对68个非翻译问题的回答：英语到英语、英语到挪威语、挪威语到挪威语，以及挪威语到英语。在去除拒绝项后，数据集包含1348个响应。我们使用Cohen d测量长度差异，使用LabSE余弦相似度测量语义保真度，并使用原始和软Jaccard测量跨语言关键词重叠。提示语言对响应长度有显著影响。在英语输出的情况下，挪威语提示使响应缩短约37%。在挪威语输出的情况下，英语提示使响应缩短约41%。最大的跨语言对比显示字数减少，但token数量的减少较小，反映了分词器的差异。尽管长度存在变化，语义相似度仍然较高，软Jaccard揭示了原始Jaccard未能捕捉的实质性概念重叠。效应大小在不同模型之间存在差异，表明异质性。提示语言并非中立，而是系统性地塑造输出长度和词汇实现，对多语言提示设计具有重要影响。
cs.CL / 39 / 2608.26187
When the Canonical Completion Is Wrong: Formalizing and Measuring the Jump in Large Language Models
当典范完成错误时：对大型语言模型中的跳跃进行形式化和测量
Abstract
Whether large language models (LLMs) can perform the abductive leap from evidence to a new system of axioms, commonly referred to as a jump, has recently attracted considerable debate. A prominent position holds that LLMs are structurally incapable of such jumps, while recent studies challenge both its mechanism and its evidence. However, the debate remains difficult to settle, since the field still lacks a formal definition of the jump and a measure to test either side. In this paper, we develop a formal account of the jump in four steps and measure the second. The steps ask what the default completion of partial data is, when abandoning it is forced, when the abandonment is correct, and how successive jumps compound. Specifically, we define a jump instance as a finite extension problem with a machine-checked certificate that a correct completion exists, is unique up to renaming, and differs from the canonical completion of the data. The canonical completion is given by the left and right Kan extensions and is also what models produce without constraints, so it serves as the default. We prove that jump instances are well-posed and establish a family theorem that certifies instances of unbounded difficulty without enumeration. We further formalize when a jump is correct and how successive jumps compound. Finally, we run the measurement on nine certified instances and four frontier models. The Kan-default rate is zero in all 248 constrained trials, so the models do jump at this step and abandon the excluded default every time. Failures at higher difficulty stem from exhausted reasoning budgets or constraint errors, never from reverting to the default. These results indicate that the second step is not the bottleneck. If the disputed incapacity is real, it lies in generating the constraints or inventing the framework. Code can be found at: https://github.com/EEthanShi/kan-jump-test.
Chinese Translation
大型语言模型（LLMs）是否能够从证据进行推理跳跃到一个新的公理系统，通常称为跳跃，最近引发了相当大的争论。一种显著的观点认为，LLMs 在结构上无法进行这样的跳跃，而最近的研究则对其机制和证据提出了挑战。然而，由于该领域仍然缺乏对跳跃的正式定义和测试任一方的测量工具，这场争论仍然难以解决。在本文中，我们通过四个步骤对跳跃进行了形式化描述，并对第二步进行了测量。这些步骤探讨了部分数据的默认完成是什么，何时被迫放弃，何时放弃是正确的，以及连续跳跃如何叠加。具体而言，我们将跳跃实例定义为一个有限扩展问题，具有机器检查的证书，证明存在正确的完成，并且在重命名上是唯一的，且与数据的典范完成不同。典范完成由左 Kan 扩展和右 Kan 扩展给出，也是模型在没有约束时产生的内容，因此作为默认完成。我们证明了跳跃实例是良定义的，并建立了一个家庭定理，证明无界难度的实例在不枚举的情况下是可认证的。我们进一步形式化了何时跳跃是正确的以及连续跳跃如何叠加。最后，我们对九个认证实例和四个前沿模型进行了测量。在所有248个受限试验中，Kan 默认率为零，因此模型在这一步骤确实进行了跳跃，并每次都放弃了排除的默认。更高难度的失败源于推理预算耗尽或约束错误，而不是回归默认。这些结果表明，第二步并不是瓶颈。如果争议中的无能是真实存在的，那它在于生成约束或发明框架。代码可在以下链接找到：https://github.com/EEthanShi/kan-jump-test。
cs.CL / 40 / 2608.26192
Comparing Chunking and Embedding Strategies for Turkish RAG Systems
比较土耳其语检索增强生成系统的分块和嵌入策略
Abstract
How documents are segmented into retrievable chunks and how those chunks are embedded strongly affect Retrieval-Augmented Generation (RAG) quality, yet neither has been systematically studied for morphologically rich languages such as Turkish. We compare Turkish document question answering across three chunking strategies (fixed-length, semantic, and layout-aware Docling), five embedding models, and two generator LLMs, over three documents with contrasting layouts. The fully crossed design yields 9,000 graded question-answer evaluations, each scored by an independent judge model, and component comparisons are tested by paired McNemar tests under Holm correction. Four findings follow. The chunking strategy determines how much the embedding choice matters: layout-aware chunking compresses the spread between the modern embedding models to about a point. The three leading embedding models are statistically indistinguishable, so language specialization yields no measurable retrieval advantage. The faster generator is not the more accurate one. And the preferred configuration depends on content type, since layout-aware chunking helps documents containing tables far more than prose. The best individual components therefore do not compose into the best complete configuration, which reaches 87.0%.
Chinese Translation
文档如何被分割成可检索的块，以及这些块如何被嵌入，强烈影响检索增强生成（RAG）的质量，但对于像土耳其语这样的形态丰富语言，这两者尚未进行系统研究。我们比较了三种分块策略（固定长度、语义和布局感知的 Docling）、五种嵌入模型和两种生成器大语言模型（LLM）在三份具有对比布局的文档上的土耳其语文档问答。完全交叉的设计产生了9000个经过评分的问题-答案评估，每个评估由独立的评判模型评分，并通过配对的 McNemar 检验在 Holm 校正下测试组件比较。得出四个发现。分块策略决定了嵌入选择的重要性：布局感知的分块将现代嵌入模型之间的差距压缩到约一个点。三种主要的嵌入模型在统计上不可区分，因此语言专业化没有带来可测量的检索优势。更快的生成器并不是更准确的。并且首选配置取决于内容类型，因为布局感知的分块对包含表格的文档的帮助远大于对散文的帮助。因此，最佳的单个组件并不组合成最佳的完整配置，后者达到了87.0%。
cs.CL / 41 / 2608.26194
A Reranker for Orchestrating Heterogeneous Speech and Text Retrievers
用于协调异构语音与文本检索器的重排序器
Abstract
Retrieval-Augmented Generation (RAG) systems have attracted significant interest for their ability to mitigate hallucinations in Large Language Models (LLMs). Although knowledge databases for RAG are increasingly diversifying to include various modalities such as speech and text, research on handling such multi-modal database scenarios remains limited. In this paper, we propose STeReO (Speech and Text Reranking Orchestrator), a reranker based on speech and text retrievers that aggregates disparate modality databases. To address the lack of specialized training data, we first curate a dataset comprising queries, mixed-modality evidence, and their corresponding relevance ranks. We then train the reranker and evaluate its effectiveness in both single-modality and mixed-modality scenarios. Our results demonstrate that the proposed algorithm excels at selecting the most relevant evidence, thereby significantly improving downstream question-answering performance.
Chinese Translation
检索增强生成（Retrieval-Augmented Generation，RAG）系统因其缓解大型语言模型（Large Language Models，LLMs）幻觉问题的能力而受到广泛关注。尽管RAG的知识库正日益多样化，涵盖语音和文本等多种模态，但针对此类多模态数据库场景的研究仍然有限。本文提出了一种基于语音和文本检索器的重排序器——STeReO（Speech and Text Reranking Orchestrator），用于整合异构模态数据库。为解决缺乏专门训练数据的问题，我们首先构建了一个包含查询、混合模态证据及其相关性排序的数据集。随后，我们训练该重排序器，并评估其在单模态和混合模态场景下的有效性。实验结果表明，所提算法在选择最相关证据方面表现优异，从而显著提升了下游问答任务的性能。
cs.CL / 42 / 2608.26292
On Scope Classification and Current Knowledge-Editing Benchmarks: A Negative Result, with INLAY as a Gradient-Free Case Study
关于范围分类和当前知识编辑基准的负面结果：以 INLAY 为无梯度案例研究
Abstract
Every memory-based knowledge editor in the SERAC lineage depends on a scope decision: given a query, does a stored edit apply? We report that current knowledge-editing benchmarks cannot measure this decision at all. Using INLAY, a gradient-free editor we built to obtain exact per-query ground truth (the model is frozen, edits live in an external addressable memory, and applying an edit is a bias added along one token's unembedding direction at decode time), we execute every candidate router action on 1,689 queries spanning three datasets and three input conditions. An oracle router choosing the best action every time ties a one-line static policy to four decimal places in all nine dataset-by-condition cells: the maximum attainable gain of any per-query routing method is 0.00 points. Abstention is the sole winning action zero times out of 1,689. The cause is structural: these are counterfactual benchmarks whose evaluation question asks for the post-edit answer, so answering from parametric knowledge is wrong by construction, and a benchmark without negatives cannot reward a classifier's ability to reject. This generalizes beyond our system to the whole scope-classifier family the benchmarks are used to evaluate. We confirm the mechanism directly: constructing the missing condition ourselves, by withholding a query's own edit from the index for half the sample, moves pooled headroom from exactly +0.0000 to +0.0420 and gives abstention its first wins. We also report where INLAY itself does not win (WISE beats it on Qwen2.5-7B CounterFact, and retrieval-augmented generation beats every method we tested, INLAY included, on rigorously matched RippleEdits), and disclose two bugs found during a self-audit of our own routing machinery, neither of which changed a published headline number outside noise.
Chinese Translation
SERAC 系列中的每个基于记忆的知识编辑器都依赖于一个范围决策：给定一个查询，存储的编辑是否适用？我们报告称，当前的知识编辑基准根本无法衡量这一决策。使用 INLAY，一个我们构建的无梯度编辑器，以获得每个查询的确切真实值（模型被冻结，编辑存储在可寻址的外部内存中，并且在解码时应用编辑是沿着一个标记的去嵌入方向添加的偏差），我们对跨越三个数据集和三种输入条件的 1,689 个查询执行每个候选路由器操作。一个每次选择最佳操作的神谕路由器将一条静态策略与所有九个数据集-条件单元中的四位小数相结合：任何每查询路由方法的最大可获得增益为 0.00 分。在 1,689 次中，弃权是唯一的胜利操作，胜利次数为零。其原因是结构性的：这些是反事实基准，其评估问题要求后编辑答案，因此从参数知识中回答是构造上错误的，而没有负值的基准无法奖励分类器拒绝的能力。这一问题超出了我们的系统，推广到用于评估的整个范围分类器家族。我们直接确认了这一机制：通过将查询自身的编辑在样本的一半中从索引中扣留，构造缺失的条件，将池头空间从确切的 +0.0000 移动到 +0.0420，并使弃权获得其首次胜利。我们还报告了 INLAY 本身未能获胜的情况（WISE 在 Qwen2.5-7B CounterFact 上击败了它，而检索增强生成在严格匹配的 RippleEdits 上击败了我们测试的每种方法，包括 INLAY），并披露了在我们自己的路由机制自审过程中发现的两个错误，这两个错误均未改变已发布的主要数字，超出了噪声的影响。
cs.CL / 43 / 2608.26295
MemToC: Benchmarking Memory-Tool Conflict Resolution in Large Language Models
MemToC：大型语言模型中内存工具冲突解决的基准测试
Abstract
Tool-augmented LLMs must arbitrate between two fallible sources when a tool return conflicts with their parametric memory, yet existing evaluations measure source preference without establishing source correctness. We introduce MemToC, a controlled benchmark for post-tool-return arbitration with executable tools. MemToC comprises 6,504 evaluation episodes constructed from 542 quality-controlled factual questions, independently elicited model-specific closed-book answers, and controlled tool returns of known correctness. These components instantiate four source-correctness cases; tool-error and no-tool conditions are separate controls. Across five open-weight 7-9B models, tool returns strongly dominate elicited closed-book answers. The four instruction-tuned models retain a verified-correct answer against an incorrect tool in only 6.5-17.1% of eligible cases, follow a correct tool in 86.0-93.1%, and repeat the tool return in 78.4-86.0% of cases where both sources are wrong. No cross-model ordering remains stable across three instruction-wording variants with the question and episode content held fixed. We compare prompting with SFT and DPO using chain-level cross-fitting over ToolHop, so questions sharing an underlying fact never straddle training and evaluation. We apply an asymmetric success criterion: correct-answer retention must improve without a detected reduction in correct-tool following. SFT and DPO meet this criterion on the same two of four instruction-tuned backbones. Improvements rarely come cleanly: 19 of 20 tested method-model combinations reduce abstention after tool errors or on unanswerable inputs. Transfer beyond MemToC is positive but partial and depends on the model and presentation frame. Correctness-conditioned arbitration can be improved through fine-tuning, but gains must be evaluated jointly with correct tool use, abstention, and robustness to formulation.
Chinese Translation
工具增强的LLM在工具返回与其参数记忆冲突时必须在两个易出错的来源之间进行仲裁，然而现有评估仅测量来源偏好而未建立来源的正确性。我们引入了MemToC，这是一个针对可执行工具的后工具返回仲裁的受控基准。MemToC包含6,504个评估情节，构建于542个经过质量控制的事实问题、独立引出的模型特定闭卷答案，以及已知正确性的受控工具返回。这些组件实例化了四种来源正确性案例；工具错误和无工具条件作为单独的对照。在五个开放权重的7-9B模型中，工具返回明显优于引出的闭卷答案。四个经过指令调优的模型在仅6.5%-17.1%的合格案例中保留了对错误工具的验证正确答案，跟随正确工具的比例为86.0%-93.1%，在两个来源均错误的情况下重复工具返回的比例为78.4%-86.0%。在三个指令措辞变体中，跨模型排序没有保持稳定，同时问题和情节内容保持不变。我们比较了使用SFT和DPO的提示，通过ToolHop进行链级交叉拟合，因此共享基础事实的问题从未跨越训练和评估。我们应用了不对称成功标准：正确答案的保留必须在未检测到正确工具跟随减少的情况下改善。SFT和DPO在四个指令调优骨干中的两个满足这一标准。改进很少是干净的：在20个测试的方法-模型组合中，有19个在工具错误或无法回答的输入后减少了弃权。超越MemToC的转移是积极的，但部分依赖于模型和展示框架。基于正确性的仲裁可以通过微调得到改善，但收益必须与正确工具使用、弃权和对表述的鲁棒性共同评估。
cs.CL / 44 / 2608.26319
When Is Noise Response Universal? Tokenization as the Hidden Variable in Language Models
噪声响应何时具有普适性？语言模型中的分词作为隐藏变量
Abstract
The performance of textual neural models often degrades when their inputs are corrupted by noise such as typos, OCR errors, or dropped words. We study the degradation rate across neural models, both sentence embeddings and decoder-only LLMs, and find that how consistent it is depends on the scale of the noise: under word-level noise, models with very different architectures decline along nearly the same curve, while under character-level noise they separate. We further identify the determining factor to be the training objective, not the architecture: eight encoders spanning six pretraining paradigms are scattered initially, and collapse onto a common curve after a short contrastive training recipe. We trace the word/character split to tokenization: a single character edit forces the tokenizer to re-segment the surrounding word, disturbing the token sequence far more than dropping a whole word does. This finding and its underlying mechanism provide a practical means to predict a model's robustness to noise without any noisy evaluation, and to install robustness at a chosen noise scale through noise-augmented training.
Chinese Translation
文本神经模型在输入受到噪声干扰（如拼写错误、OCR错误或词语丢失）时，性能通常会下降。我们研究了不同神经模型（包括句子嵌入模型和仅解码器的大型语言模型）在噪声下的性能下降速率，发现其一致性取决于噪声的尺度：在词级噪声下，结构迥异的模型沿着几乎相同的曲线下降，而在字符级噪声下则表现分化。我们进一步确定决定因素是训练目标而非模型架构：涵盖六种预训练范式的八个编码器模型初始表现分散，但经过短期对比训练后，性能曲线趋于一致。我们将词/字符级别的差异归因于分词机制：单个字符的编辑会迫使分词器重新切分周围的词语，导致的标记序列扰动远大于整个词语的丢失。该发现及其机制为无需噪声评估即可预测模型对噪声的鲁棒性提供了实用方法，并可通过噪声增强训练在选定的噪声尺度上提升鲁棒性。
cs.CL / 45 / 2608.26327
How Unlikely Is "Unlikely"? Assessing Verbal Probability Perception Across Large Language Models
“不太可能”有多不太可能？评估大型语言模型中的语言概率感知
Abstract
Large language models increasingly produce and interpret verbal probability expressions, yet whether these expressions carry consistent meaning across models (or match human perceptions of uncertainty) remains unknown. We present a systematic cross-model evaluation using a word-to-number mapping task grounded in established human benchmarks. Eleven uncertainty expressions were presented to 19 models under two conditions, forced single-number response and explanation elicitation, alongside a novel bidirectional roundtrip test of internal consistency. LLMs track the human benchmark with surprising fidelity: word ordering is preserved, three anchor points are recovered, and ``possible'' shows the highest variance and cross-model disagreement of any expression tested, consistent with its documented bimodal interpretation in humans. However, models show a systematic upward bias for negative expressions such as ``unlikely'' and ``improbable.'' Explanation elicitation reduces within-model variance while increasing between-model divergence, stabilizing individual models at the cost of inter-model consensus, and the roundtrip experiment reveals clear stratification, with frontier models maintaining coherent bidirectional representations. LLMs thus reproduce the structure of human verbal probability cognition, including its biases, while diverging systematically at the negative end---with implications for any setting where humans and models exchange probabilistic language.
Chinese Translation
大型语言模型越来越多地生成和解释语言概率表达，但这些表达在不同模型之间是否具有一致的意义（或与人类的不确定性感知相匹配）仍然未知。我们通过基于既定人类基准的词到数字映射任务，进行了一项系统的跨模型评估。在两种条件下向19个模型呈现了11种不确定性表达：强制单一数字响应和解释引导，同时进行了一项新颖的双向往返测试以评估内部一致性。大型语言模型以惊人的忠实度跟踪人类基准：词序得以保留，三个锚点被恢复，而“可能”在所有测试的表达中显示出最高的方差和跨模型的不一致，符合其在人类中的双峰解释。然而，模型对“unlikely”（不太可能）和“improbable”（不太可能）等负面表达表现出系统性的上偏差。解释引导减少了模型内部的方差，同时增加了模型之间的差异，稳定了个别模型，但以牺牲模型间共识为代价，而往返实验揭示了明显的分层，前沿模型保持一致的双向表征。因此，大型语言模型再现了人类语言概率认知的结构，包括其偏差，同时在负面端系统性地出现分歧——这对人类与模型交换概率语言的任何场景具有重要意义。
cs.CL / 46 / 2608.26329
Neuro-symbolic PRM: Enhancing Scientific Reasoning via Structured Traces and Symbolic Verification
神经符号化过程奖励模型：通过结构化轨迹和符号验证增强科学推理
Abstract
While tool-augmented Large Language Models have significantly improved multi-step reasoning in quantitative STEM tasks, a critical residual failure mode remains: intermediate reasoning steps that are syntactically well-formed, mathematically executable, and unit-consistent, yet contextually ungrounded. Current approaches either rely on formal verifiers that cannot assess semantic intent, or burden Process Reward Models (PRMs) with the dual task of checking both arithmetic and logic. In this paper, we propose a neuro-symbolic framework that cleanly decouples reasoning into two formal dimensions: Symbolic Validity ($V$) and Semantic Groundedness ($G$). We guarantee $V$ by construction using a deterministic symbolic verifier acting as a hard filter. To assess $G$, we train a PRM conditionally on the verifier-accepted manifold. To train this PRM efficiently, we introduce Counterfactual Symbolic Perturbation (CSP), a novel data synthesis strategy that algorithmically generates constraint-preserving hard negatives (steps that perfectly pass the verifier but are logically flawed). At inference, we deploy a verifier-first constrained search that guarantees execution consistency for verifier-covered operations while relying on the PRM solely to rank semantic grounding. By targeting the exact residual error class of strong tool-using LLMs, our method significantly improves reasoning reliability without the sprawling heuristics of prior frameworks.
Chinese Translation
尽管工具增强的大型语言模型在定量STEM任务中的多步骤推理上取得了显著进展，但仍然存在一种关键的残余失败模式：中间推理步骤在语法上是良构的、数学上可执行的且单位一致，但在上下文上缺乏基础。目前的方法要么依赖于无法评估语义意图的形式验证器，要么将过程奖励模型（PRM）负担于同时检查算术和逻辑的双重任务。在本文中，我们提出了一种神经符号化框架，将推理清晰地解耦为两个形式维度：符号有效性（$V$）和语义基础性（$G$）。我们通过构建使用确定性符号验证器作为严格过滤器来保证$V$。为了评估$G$，我们在验证器接受的流形上有条件地训练PRM。为了高效地训练这个PRM，我们引入了反事实符号扰动（CSP），这是一种新颖的数据合成策略，能够算法性地生成保持约束的硬负样本（完美通过验证器但在逻辑上存在缺陷的步骤）。在推理时，我们部署了一个验证器优先的约束搜索，确保验证器覆盖的操作的执行一致性，同时仅依赖PRM对语义基础性进行排序。通过针对强工具使用的大型语言模型的确切残余错误类别，我们的方法显著提高了推理的可靠性，而无需以往框架的庞大启发式方法。
cs.CL / 47 / 2608.26344
MoganColBERT-TR: A Late-Interaction Multi-Vector Retrieval Model for Turkish
MoganColBERT-TR：一种针对土耳其语的晚期交互多向量检索模型
Abstract
We previously reported a ModernBERT encoder trained from scratch for Turkish (MoganBERT-TR) and a single-vector embedding model built on top of it (MoganBERT-embed). This work introduces the third model in that lineage: MoganColBERT-TR, a multi-vector retrieval model that, instead of compressing a query or a document into a single vector, represents it at the token level through a 768->128 projection and scores it with MaxSim late interaction. The model is not trained from scratch: the embedding model's encoder is taken as the starting point and adapted to the ColBERT objective with a single-epoch distillation phase. Training data is produced from two sources - title-to-passage pairs carved out of our own pretraining corpus in the character domain and at sentence boundaries, and two Turkish question-based retrieval sets - and is distilled from the soft scores of a cross-encoder teacher (bge-reranker-v2-m3) over one positive and seven mined negatives. We show that in hard negative mining, rank-based skipping alone is insufficient and must be combined with a group mask and a cosine ceiling. Evaluation is carried out with the official pipeline of TurkColBERT, a benchmark built for Turkish late-interaction retrieval (PLAID index, exact MaxSim), on five Turkish BEIR datasets; none of them appears in our training pool, so all five results are clean zero-shot. With 148.9M parameters, MoganColBERT-TR reaches an overall score of 37.36 (35.53 nDCG@100, 31.81 nDCG@10) averaged over the five datasets and finishes second among the five models compared: it outperforms the twice-as-large ColmmBERT-base-TR on four of five datasets and by +3.05 overall, and the benchmark's largest model by +12.30. The gap to the leading model (mLateOn) is concentrated on ArguAna-TR, the dataset with by far the longest queries.
Chinese Translation
我们之前报告了一个从零开始训练的现代BERT编码器用于土耳其语（MoganBERT-TR）以及基于其构建的单向量嵌入模型（MoganBERT-embed）。本研究介绍了该系列的第三个模型：MoganColBERT-TR，这是一种多向量检索模型，它不是将查询或文档压缩为单个向量，而是通过768->128的投影在标记级别上表示，并通过MaxSim晚期交互进行评分。该模型不是从零开始训练的：嵌入模型的编码器作为起点，并通过单个周期的蒸馏阶段适应ColBERT目标。训练数据来自两个来源——从我们自己的字符域预训练语料库中提取的标题与段落对，以及两个基于土耳其语的问题检索集，并从一个交叉编码器教师（bge-reranker-v2-m3）在一个正样本和七个挖掘的负样本上的软评分中进行蒸馏。我们展示了在困难负样本挖掘中，仅靠基于排名的跳过是不够的，必须与组掩码和余弦上限结合使用。评估是在TurkColBERT的官方管道上进行的，这是一个为土耳其语晚期交互检索构建的基准（PLAID索引，精确的MaxSim），在五个土耳其语BEIR数据集上进行；这五个数据集均未出现在我们的训练池中，因此所有五个结果均为干净的零样本结果。MoganColBERT-TR拥有1.489亿个参数，在五个数据集上平均得分为37.36（nDCG@100为35.53，nDCG@10为31.81），在比较的五个模型中排名第二：在五个数据集中的四个上超越了两倍大的ColmmBERT-base-TR，并且整体超出3.05，超出基准中最大的模型12.30。与领先模型（mLateOn）的差距主要集中在ArguAna-TR上，这是一个查询长度远超其他数据集的数据集。
cs.CL / 48 / 2608.26357
Cross-lingual Representation Learning via Centroid Intervention Fusion
通过中心点干预融合的跨语言表示学习
Abstract
Large language models (LLMs) exhibit uneven multilingual performance, especially when dealing with low-resource languages. Inference-time intervention offers a lightweight way to improve cross-lingual transfer by modifying the hidden states produced by the LLMs during the forward pass, without updating model parameters. However, existing cross-lingual intervention methods typically learn separate projections from source to target languages, which limits scalability and prevents knowledge sharing across languages. We propose Centroid Intervention Fusion (CIF), a projection fusion framework that consolidates multiple multilingual intervention projections into a single language-shared operator. Across multilingual commonsense reasoning, natural language inference, factual editing, and machine translation benchmarks, CIF outperforms the strongest prior pairwise intervention baseline by up to +3.378 pp on average across four model backbones, while supporting performance gains for low resource languages. The code is available at https://github.com/VRCMF/CIF.git.
Chinese Translation
大型语言模型（LLMs）在多语言表现上存在不均衡，尤其是在处理低资源语言时。推理时的干预提供了一种轻量级的方法，通过修改LLMs在前向传播过程中产生的隐藏状态来改善跨语言迁移，而无需更新模型参数。然而，现有的跨语言干预方法通常从源语言到目标语言学习独立的投影，这限制了可扩展性并阻碍了跨语言的知识共享。我们提出了中心点干预融合（Centroid Intervention Fusion, CIF），这是一种投影融合框架，将多个多语言干预投影整合为一个共享语言的操作符。在多语言常识推理、自然语言推理、事实编辑和机器翻译基准测试中，CIF在四个模型骨干上平均比最强的先前成对干预基线提高了最多3.378个百分点，同时支持低资源语言的性能提升。代码可在 https://github.com/VRCMF/CIF.git 获取。
cs.CL / 49 / 2608.26372
Knowledge-Verified Emergent Deception in LLM Agents Under Conflicting Incentives
在冲突激励下的知识验证紧急欺骗：大语言模型代理的研究
Abstract
Large language models are increasingly deployed as autonomous agents serving users on behalf of companies, placing them in settings where user and deployer interests can conflict. When an agent knows that a user is owed something its deployer would prefer to deny, does it remain honest? Answering this is difficult because false statements can reflect either ignorance or hallucination rather than deception. To address this challenge, we introduce KnownLieBench , a knowledge-verified benchmark that first confirms through a neutral probe that an agent knows a user's entitlement, and then evaluates whether it makes false claims once an incentive to deny that entitlement is introduced. Specifically, KnownLieBench covers eight customer-service domains and 112 grounded cases, conducts multi-round dialogues with a trust-tracking customer agent, and separates deception emerging from incentive alone from deception produced under explicit instruction. Across eighteen proprietary and open-weight models, emergent deception varies substantially across model families and domains. We further use the benchmark for post-training, finding that honesty-directed fine-tuning reduces deception under incentive, while deception-graded fine-tuning increases lie success on honest-control dialogues without increasing lie frequency under incentive. By verifying entitlement knowledge before scoring deceptive behavior, KnownLieBench reduces the confound between lying and not knowing and enables more rigorous auditing and steering of agent honesty.
Chinese Translation
大型语言模型越来越多地作为自主代理被部署，为用户服务，代表公司在各种环境中运作，这使得用户和部署者的利益可能发生冲突。当一个代理知道用户应得某种其部署者希望拒绝的权益时，它是否仍然保持诚实？回答这个问题很困难，因为虚假陈述可能反映无知或幻觉，而非欺骗。为了解决这一挑战，我们引入了 KnownLieBench，这是一个知识验证基准，首先通过中立探测确认代理知道用户的权益，然后评估在引入拒绝该权益的激励后，它是否会做出虚假声明。具体而言，KnownLieBench 涉及八个客户服务领域和112个具体案例，进行多轮对话与一个信任跟踪的客户代理，并将仅由激励产生的欺骗与在明确指令下产生的欺骗区分开来。在十八个专有和开放权重模型中，紧急欺骗在模型家族和领域之间存在显著差异。我们进一步利用该基准进行后训练，发现以诚实为导向的微调在激励下减少了欺骗，而以欺骗为等级的微调则在诚实控制对话中增加了谎言成功率，但在激励下并未增加谎言频率。通过在评估欺骗行为之前验证权益知识，KnownLieBench 减少了撒谎与无知之间的混淆，并使代理诚实性的审计和引导更加严格。
cs.CL / 50 / 2608.26374
Survival-Guided Length Control for Efficient Diffusion Language Models
基于生存指导的长度控制以提高扩散语言模型的效率
Abstract
Diffusion language models (DLMs) generate text by iteratively denoising masked sequences, but standard decoding either fixes the sequence length or relies on ad hoc stopping rules, often leading to unnecessary denoising steps. We recast length selection as a discrete-time survival problem over the end-of-sequence token and propose a plug-in, training-free length predictor that can be added to any existing DLM. Across reasoning and code-generation benchmarks, survival-guided length decoding speeds up inference by up to 7 times while preserving task accuracy. We further find that predicted lengths vary widely even within the same dataset, making model performance sensitive to the chosen length.
Chinese Translation
扩散语言模型（DLMs）通过迭代去噪掩蔽序列生成文本，但标准解码要么固定序列长度，要么依赖于临时停止规则，常常导致不必要的去噪步骤。我们将长度选择重新表述为一个关于序列结束标记的离散时间生存问题，并提出了一种插件式、无训练的长度预测器，可以添加到任何现有的DLM中。在推理和代码生成基准测试中，基于生存指导的长度解码在保持任务准确性的同时，将推理速度提高了多达7倍。我们进一步发现，即使在同一数据集中，预测的长度变化也很大，使得模型性能对所选长度非常敏感。
cs.CL / 51 / 2608.26385
Why RAGs Hallucinate: Penalty-Aware Evaluation of Retrieval-Augmented Generation Systems with Knowledge-Gap Canaries
为何RAG会出现幻觉：基于惩罚意识的检索增强生成系统评估与知识差距警示
Abstract
Volume-based accuracy rewards retrieval-augmented generation (RAG) systems for guessing: a system that answers everything outscores one that declines when its knowledge base cannot support an answer. Building on the confidence-target analysis of Kalai et al. (2025), we present a penalty-aware evaluation framework for deployed RAG products, combining (i) asymmetric scoring (correct +1, wrong -4, abstain 0), (ii) knowledge-gap canaries, questions whose answers are verifiably absent from the knowledge base, so that any answer constitutes ungrounded generation from parametric memory, and (iii) a failure-attribution pipeline that separates retrieval, generation, and abstention-policy failures. Applying the framework to three commercial RAG systems and a no-retrieval baseline on SimpleQA-Verified (1,000 questions x 3 repeats, graded blind by a cross-family three-judge panel with 98.9% unanimity), we find that accuracy when answering is closely clustered across systems (97.0-98.0%), while canary violation rates differ roughly sixfold (16.7% vs. 98.1%). The systems are separated less by what they answer correctly than by whether they answer at all when they should not, and penalty-aware scoring reorders the volume-based ranking accordingly; the reordering is stable across penalty settings from k=1 to k=9. All code, configurations, transcripts, and judge votes are released for independent audit.
Chinese Translation
基于体积的准确性奖励检索增强生成（RAG）系统的猜测：一个能够回答所有问题的系统得分高于在知识库无法支持答案时选择拒绝的系统。基于Kalai等人（2025）的置信度目标分析，我们提出了一种针对已部署RAG产品的惩罚意识评估框架，结合了（i）不对称评分（正确 +1，错误 -4，弃权 0），（ii）知识差距警示，即答案在知识库中可验证缺失的问题，因此任何答案都构成了来自参数记忆的无基础生成，以及（iii）一个失败归因流程，区分检索、生成和弃权政策的失败。将该框架应用于三个商业RAG系统和一个无检索基线，在SimpleQA-Verified（1,000个问题 x 3次重复，由跨家族三人评审小组以98.9%的一致性盲评）上，我们发现回答时的准确性在各系统之间紧密聚集（97.0-98.0%），而警示违反率则差异约六倍（16.7% vs. 98.1%）。系统之间的差异不在于它们正确回答的内容，而在于它们在不该回答时是否回答，而惩罚意识评分相应地重新排序了基于体积的排名；这种重新排序在惩罚设置从k=1到k=9之间是稳定的。所有代码、配置、记录和评审投票均已发布以供独立审计。
cs.CL / 52 / 2608.26386
Co-Evolving Structured Knowledge and Reasoning in Language Models
语言模型中结构化知识与推理的协同进化
Abstract
Retrieval-augmented methods improve factual accuracy by grounding language models in external knowledge, but retrieving over unstructured text often introduces irrelevant context and offers limited control over the retrieved information. Structured knowledge bases offer a more controllable alternative, yet they are expensive to construct and often brittle to reason over. To address these limitations, we propose KBevo: a co-evolving framework that jointly learns to construct a structured knowledge base and reason over it for knowledge-intensive question answering. By optimizing both components end-to-end with QA outcome rewards, our method enables reasoning success to directly improve the quality of the constructed knowledge base. This leads to larger, better-connected knowledge structures with higher answer reachability, while also improving compositional factual reasoning and controllability compared to standard retrieval baselines.
Chinese Translation
检索增强方法通过将语言模型与外部知识相结合，提高了事实准确性，但在非结构化文本中检索往往引入无关上下文，且对检索信息的控制有限。结构化知识库提供了更可控的替代方案，然而其构建成本高且推理过程常常脆弱。为了解决这些限制，我们提出了KBevo：一个协同进化框架，联合学习构建结构化知识库并基于其进行知识密集型问答推理。通过端到端优化两个组件并以问答结果作为奖励，我们的方法使推理成功能够直接提升构建知识库的质量。这不仅促成了更大、更紧密连接的知识结构，提升了答案的可达性，同时相比于标准检索基线，增强了组合事实推理能力和可控性。
cs.CL / 53 / 2608.26389
LowRankArena: A Standardized Evaluation Platform for SVD-Based LLM Compression
LowRankArena：基于SVD的大型语言模型压缩的标准化评估平台
Abstract
SVD-based low-rank compression has become a fast-growing direction for reducing the memory and computational cost of large language models (LLMs). However, meaningful comparison across existing studies remains difficult as prior evaluations use varied benchmarks, inconsistent ratios, and diverse setups, often failing to isolate low-rank effects from auxiliary techniques. As a result, it remains unclear whether reported gains reflect method-level improvements or differences in evaluation protocol. This lack of comparability highlights the need for a unified, reproducible evaluation platform. To address this problem, we present LowRankArena, a standardized evaluation platform for SVD-based LLM compression. LowRankArena unifies task versions, uniform-precision compression budgets, comparison regimes, and inference measurements, and provides a reproducible pipeline with over 3 TiB released compressed checkpoints. Using LowRankArena, our aligned audit of five representative SVD methods reveals that prior findings are highly conditional under standardized protocols: clear leaders and performance tiers shift across backbones and keep ratios, multiple-choice accuracy can hide large perplexity degradation, and nominal low-rank savings yield workload-dependent and often limited end-to-end speedups. Our code is available at: https://github.com/Zishan-Shao/lowrankarena.git.
Chinese Translation
基于SVD的低秩压缩已成为降低大型语言模型（LLMs）内存和计算成本的快速发展方向。然而，由于先前的评估使用了不同的基准、不一致的比例和多样的设置，跨现有研究进行有意义的比较仍然困难，常常无法将低秩效应与辅助技术隔离。因此，目前尚不清楚报告的收益是否反映了方法层面的改进或评估协议的差异。这种缺乏可比性突显了统一、可重复的评估平台的必要性。为了解决这个问题，我们提出了LowRankArena，一个基于SVD的大型语言模型压缩的标准化评估平台。LowRankArena统一了任务版本、统一精度压缩预算、比较机制和推理测量，并提供了一个可重复的流程，包含超过3 TiB发布的压缩检查点。使用LowRankArena，我们对五种代表性SVD方法的对齐审计揭示了在标准化协议下，先前的发现高度依赖于条件：明显的领导者和性能等级在不同的骨干网络和保持比例之间变化，多项选择准确性可能掩盖了巨大的困惑度下降，名义上的低秩节省产生的工作负载依赖性和往往有限的端到端加速。我们的代码可在以下链接获取：https://github.com/Zishan-Shao/lowrankarena.git。
cs.CL / 54 / 2608.26414
Case2Flow: Bridging Patient Cases and Guideline Flowcharts through Multimodal Retrieval
Case2Flow：通过多模态检索连接患者案例与指南流程图
Abstract
Medical guidelines encode rich, evidence-based decision logic, yet the specific decision artifact a clinician needs is hard to locate within a guideline, let alone across guidelines covering plausible diseases and treatments. While guideline passages have supported end-to-end question answering, flowcharts remain largely underused in decision support despite their ability to encode actionable clinical pathways. We therefore introduce Case2Flow, a task designed to retrieve the most relevant guideline flowchart for a given patient case from a collection of guideline documents. To support it, we construct FlowAtlas, a curated corpus of 202 flowcharts extracted from 2,080 medical guidelines, together with a pipeline that synthesises 1,911 aligned case-flowchart pairs. Our evaluation of multimodal retrieval methods reveals systematic failure modes, including overreliance on keywords and spurious token-patch matches induced by uninformative background regions in flowcharts. Motivated by this, we propose CRISP, a training-free scoring method that sharpens late-interaction retrieval by suppressing uninformative patches, discounting ambiguous token matches, and incorporating bidirectional query-image alignment. CRISP improves Recall@1 by up to 18.71 percentage points, while a blinded physician assessment on published case narratives provides preliminary feasibility evidence beyond synthetic queries.
Chinese Translation
医学指南编码了丰富的基于证据的决策逻辑，但临床医生所需的具体决策文献在指南中难以定位，更不用说在涵盖可能疾病和治疗的多个指南之间进行查找。尽管指南段落支持端到端的问题回答，流程图在决策支持中的使用仍然相对不足，尽管它们能够编码可操作的临床路径。因此，我们提出了Case2Flow，这是一个旨在从一系列指南文档中检索与给定患者案例最相关的指南流程图的任务。为此，我们构建了FlowAtlas，这是一个从2080份医学指南中提取的202个流程图的策划语料库，并且合成了1911对对齐的案例-流程图对。我们对多模态检索方法的评估揭示了系统性的失败模式，包括对关键词的过度依赖以及由于流程图中无信息背景区域引起的虚假标记匹配。受到此启发，我们提出了CRISP，一种无训练评分方法，通过抑制无信息补丁、折扣模糊的标记匹配，并结合双向查询-图像对齐，来增强后期交互检索。CRISP将Recall@1提高了多达18.71个百分点，而对已发布案例叙述的盲评医师评估提供了超越合成查询的初步可行性证据。
cs.CL / 55 / 2608.26434
AfriSwitch: A Benchmark for In-the-Wild African Code-Switched Speech Recognition
AfriSwitch：一种针对非洲代码切换语音识别的基准测试
Abstract
Code-switching is pervasive in bilingual African conversation, yet most ASR systems assume monolingual input and are evaluated on curated monolingual benchmarks. We present AfriSwitch, a 61.36-hour human-transcribed benchmark of in-the-wild code-switched speech spanning 16 African languages and language varieties, released with switch-level English span tags, perutterance Code-Mixing Index (CMI), and switch-point counts. Corpus statistics show that mixing behaviour varies widely across African languages along two largely independent axes: how often speakers alternate, and how balanced the mixture is. No single scalar captures how code-switched a language is. Benchmarking five open and commercial multilingual ASR systems zero-shot yields word error rates far above published monolingual figures for the same languages, with the best system averaging 35.93% WER and no system falling below 24% on any language. Africa-targeted training, not model scale or nominal language coverage, best predicts performance.
Chinese Translation
代码切换在双语非洲对话中普遍存在，但大多数自动语音识别（ASR）系统假设输入为单语，并在经过精心挑选的单语基准上进行评估。我们提出了AfriSwitch，这是一个包含61.36小时人类转录的野外代码切换语音基准，涵盖16种非洲语言及其语言变体，并附带切换级别的英语跨度标签、每句的代码混合指数（Code-Mixing Index, CMI）和切换点计数。语料库统计数据显示，混合行为在非洲语言之间存在广泛差异，主要沿着两个相对独立的轴线变化：说话者交替的频率和混合的平衡程度。没有单一的标量能够全面反映一种语言的代码切换程度。对五个开放和商业的多语种ASR系统进行零样本基准测试，结果显示其词错误率远高于同一语言的已发布单语数据，其中表现最佳的系统平均词错误率为35.93%，且没有任何系统在任何语言上的词错误率低于24%。针对非洲的训练，而非模型规模或名义语言覆盖，最能预测性能。
cs.CL / 56 / 2608.26449
Vowel Signs Are Not Letters: A Pre-tokenization Ceiling on Multilingual Tokenizer Fertility
元音符号不是字母：多语言分词器繁殖能力的预标记化上限
Abstract
Byte-level BPE tokenizers that use the HuggingFace ByteLevel pre-tokenizer inherit GPT-2's word regex, where a word is defined as \p{L}+, one or more Unicode letters. In abugida scripts, vowels are written as combining marks; this pattern therefore splits each word at every vowel sign. Since BPE merges only within a pre-token, those splits persist through training regardless of vocabulary size or corpus composition. We formalise this effect as a training-free lower bound on fertility. Across 26 languages from a parallel corpus, every one of the 17 abugidas is affected, ranging from 1.47x (Tibetan) to 9.02x (Thai), whereas Latin, Cyrillic, Hangul, and Han show exactly 1.00x. For 5 languages, matched tokenizer pairs that differ only in this character class fall within 2.2% of the predicted floor, scoring 4.78 versus 1.58 tokens per word on Nepali. When the Nepali share of the training corpus is swept from 5% to 95%, the broken tokenizer barely shifts at all (1.7%) while the fixed one shifts 33.9%, which separates a structural ceiling from a data shortage without needing to inspect any code. We train three 268M models that differ only in their tokenizer; the fixed variant achieves 4.43% lower held-out Nepali bits per byte at equal compute, and it still leads when given the same bytes with 1.59x the compute. A census of 3,479 HuggingFace repositories finds the letters-only word class present in 63.3% of the most-downloaded text-generation models, accounting for 72.5% of their downloads. GPT-4o's o200k pattern already uses a mark-aware word class, making the repair itself prior art. We quantify its value, show how to recognise its absence from symptoms alone, map which scripts it reaches, measure how widely it is deployed, and release a 65,536-entry Nepali-English tokenizer with a harness that regenerates every number here from public data on a laptop.
Chinese Translation
使用 HuggingFace ByteLevel 预标记器的字节级 BPE 分词器继承了 GPT-2 的单词正则表达式，其中单词被定义为 p{L}+, 一个或多个 Unicode 字母。在音节文字中，元音以组合标记的形式书写；因此，这种模式在每个元音符号处拆分每个单词。由于 BPE 仅在预标记内进行合并，这些拆分在训练过程中会持续存在，无论词汇表的大小或语料库的组成如何。我们将这一效应形式化为一种无训练的繁殖能力下限。在来自平行语料库的 26 种语言中，所有 17 种音节文字都受到影响，影响程度从 1.47 倍（藏语）到 9.02 倍（泰语），而拉丁字母、斯拉夫字母、韩文和汉字则显示出恰好 1.00 倍。对于 5 种语言，仅在这一字符类别上有所不同的匹配分词器对的分词结果在预测下限的 2.2% 之内，在尼泊尔语中每个单词的分词数分别为 4.78 和 1.58。当尼泊尔语在训练语料库中的比例从 5% 提高到 95% 时，损坏的分词器几乎没有变化（1.7%），而固定的分词器变化了 33.9%，这将结构性上限与数据短缺区分开来，而无需检查任何代码。我们训练了三个仅在分词器上有所不同的 268M 模型；固定变体在相同计算条件下实现了 4.43% 更低的尼泊尔语保留比特数，并且在给定相同字节且计算量为 1.59 倍时仍然领先。对 3,479 个 HuggingFace 仓库的普查发现，仅包含字母的单词类别出现在 63.3% 的下载量最高的文本生成模型中，占其下载量的 72.5%。GPT-4o 的 o200k 模式已经使用了标记感知的单词类别，使得该修复本身成为了现有技术。我们量化了其价值，展示了如何仅通过症状识别其缺失，映射其覆盖的书写系统，测量其广泛部署的程度，并发布了一个包含 65,536 条目的尼泊尔语-英语分词器，该分词器配备了一个可以从公共数据在笔记本电脑上再生此处每个数字的工具。
cs.CL / 57 / 2608.26465
Compositional Generalization via Structural Identification in a Category-Theoretic Framework
通过类别理论框架中的结构识别实现组合泛化
Abstract
Compositional generalization is usually evaluated through model accuracy. We instead ask which structural or lexical identifications make held-out COGS examples admissible from the structures observed in training. Sentences are represented as functors from syntactic addresses to lexical tokens, and selective collapses induce Kan extensions that propagate observed associations. Across 21 COGS generalization types, admissibility follows distinct identification profiles, while residual failures separate unsupported structural templates. These data-side diagnoses characterize what the training corpus licenses under specified identifications, without training a predictive model.
Chinese Translation
组合泛化通常通过模型准确性进行评估。我们则询问哪些结构或词汇识别使得从训练中观察到的结构中，保留的 COGS 示例是可接受的。句子被表示为从句法地址到词汇标记的函子，选择性折叠诱导的 Kan 扩展传播观察到的关联。在 21 种 COGS 泛化类型中，可接受性遵循不同的识别特征，而残余失败则区分不支持的结构模板。这些数据侧的诊断表征了在特定识别下训练语料库所许可的内容，而无需训练预测模型。
cs.CL / 58 / 2608.26511
Sycophancy Suppression Can Impair Rational Updating: Anti-Sycophancy Should Preserve the Ability to Update
抑制谄媚行为可能损害理性更新：反谄媚应保留更新能力
Abstract
Large language models often exhibit sycophancy, revising their answers to align with users when users push back. Such answer flips, however, can arise from different causes. One possibility is that the model simply aligns with the user's feedback in order to satisfy them. Another is that the feedback genuinely contains useful evidence, prompting the model to update its answer in a rational way. We distinguish them as Unsupported-Yielding and Rational-Updating. Prior work focuses primarily on suppressing Unsupported-Yielding, while overlooking its effect on Rational-Updating. We address this gap with a two-turn evaluation framework that measures the two behaviors separately. Across representative training-time and inference-time interventions, we find that anti-sycophancy methods often encounter a trade-off in which reducing Unsupported-Yielding can sacrifice Rational-Updating, and vice versa, even when the two objectives are optimized jointly. Mechanistic analysis suggests that the two behaviors share an internal substrate: the MLP neurons and attention heads driving them overlap substantially, and their associated steering directions are positively aligned. We further conduct a preliminary orthogonalized steering exploration, which yields modest, backbone-dependent selectivity gains. Overall, our results suggest that anti-sycophancy should be treated not as a simple suppression problem, but as a selectivity problem, where effective interventions should preserve Rational-Updating while reducing Unsupported-Yielding.
Chinese Translation
大型语言模型常常表现出谄媚行为，当用户反对时，它们会修正自己的回答以与用户对齐。然而，这种回答的翻转可能源于不同的原因。一种可能性是模型仅仅是为了满足用户而与用户的反馈对齐。另一种可能性是反馈确实包含有用的证据，促使模型以理性的方式更新其答案。我们将其区分为不支持的屈从（Unsupported-Yielding）和理性更新（Rational-Updating）。之前的研究主要集中在抑制不支持的屈从，而忽视了其对理性更新的影响。我们通过一个两轮评估框架来填补这一空白，该框架分别测量这两种行为。在代表性的训练时和推理时干预中，我们发现反谄媚方法常常面临一种权衡，即减少不支持的屈从可能会牺牲理性更新，反之亦然，即使在两个目标共同优化的情况下也是如此。机制分析表明，这两种行为共享一个内部基础：驱动它们的多层感知器（MLP）神经元和注意力头在很大程度上重叠，并且它们相关的引导方向是正相关的。我们进一步进行了初步的正交引导探索，获得了适度的、依赖于骨干网络的选择性增益。总体而言，我们的结果表明，反谄媚应被视为一个选择性问题，而非简单的抑制问题，有效的干预措施应在减少不支持的屈从的同时保留理性更新。
cs.CL / 59 / 2608.26529
Multi-Expert Conformal Risk Control for Pairwise LLM Judging in Open-Ended Dialogue
开放式对话中成对 LLM 判断的多专家符合风险控制
Abstract
In this paper, we explore multi-expert Conformal Risk Control (CRC) algorithms for pairwise LLM-as-a-Judge evaluation in open-ended dialogue. Our core insight is that multi-expert aggregation offers a complementary remedy to CRC: whereas CRC controls risk at the decision threshold through abstention, aggregation sanitizes the scoring function at its source. Guided by this, we first design two multi-expert CRC methods: Score Averaging and Decision Voting, which aggregate at the score and decision levels, respectively. While both strategies outperform single-expert methods on homogeneous expert panels, on heterogeneous LLM judges they remain risk-valid but recover only limited coverage, because a uniform threshold cannot match the experts' distinct scoring scales. To resolve this issue, we further propose Marginal-Calibrated Conformal Consensus (MC3): it captures distinct per-expert scales via initial threshold ratios, while jointly tuning a unified decision function $C_t(x)$ applied identically in both calibration and test, thereby preserving exchangeability. To evaluate our framework, we construct Panel, a 1,800-pair human pairwise-preference benchmark for open-ended dialogue. It is built on responses generated by four open-weight LLMs over dialogue contexts from three domains (ESConv, MSC, DREAM), with full logit access. In experiments, we find that both Score Averaging and Decision Voting substantially improve accuracy and acceptance rate on homogeneous panels. Notably, MC3 extends these gains to heterogeneous panels by accommodating distinct per-expert scoring scales across all three datasets.
Chinese Translation
在本文中，我们探讨了用于开放式对话中成对 LLM 作为评判者评估的多专家符合风险控制（Conformal Risk Control, CRC）算法。我们的核心见解是，多专家聚合为 CRC 提供了一种互补的解决方案：CRC 通过弃权在决策阈值处控制风险，而聚合则在源头上净化评分函数。基于此，我们首先设计了两种多专家 CRC 方法：评分平均（Score Averaging）和决策投票（Decision Voting），分别在评分和决策层面进行聚合。虽然这两种策略在同质专家小组上优于单专家方法，但在异质 LLM 评判者上，它们仍然有效控制风险，但仅能恢复有限的覆盖率，因为统一阈值无法匹配专家的不同评分尺度。为了解决这个问题，我们进一步提出了边际校准符合共识（Marginal-Calibrated Conformal Consensus, MC3）：它通过初始阈值比率捕捉不同专家的评分尺度，同时联合调优在校准和测试中均一致应用的统一决策函数 $C_t(x)$，从而保持可交换性。为了评估我们的框架，我们构建了 Panel，这是一个包含 1,800 对人类成对偏好的开放式对话基准。它基于四个开放权重 LLM 在三个领域（ESConv、MSC、DREAM）的对话上下文中生成的响应，具有完整的对数几率访问。在实验中，我们发现评分平均和决策投票在同质小组上显著提高了准确性和接受率。值得注意的是，MC3 通过适应所有三个数据集中不同专家的评分尺度，将这些增益扩展到异质小组。
cs.CL / 60 / 2608.26550
SPEAR: Distilling Domain-Adaptive Reasoning Skeletons via Sequential Symbolic Alignment in Reinforcement Learning
SPEAR：通过强化学习中的序列符号对齐提炼领域自适应推理框架
Abstract
Reinforcement learning-based knowledge distillation has the potential to transfer complex reasoning from teacher to student models, yet it currently faces a critical dilemma: researchers must choose between sparse outcome-based rewards, which provide insufficient logical guidance, or expensive neural Process Reward Models (PRMs) for dense signals. We resolve this by introducing SPEAR (Symbolic Process Evaluation and Alignment Reward), a training-free and plug-and-play process reward method for sequence-level on-policy distillation. SPEAR projects natural-language reasoning traces into domain-adaptive symbolic milestones, providing an efficient proxy for process-level reasoning alignment. By utilizing the longest common subsequence (LCS) to align student explorations with teacher milestones, SPEAR provides a dense, order-aware reward signal that enforces logical consistency without the need for an external neural verifier. Our experiments across math, science, and commonsense reasoning tasks demonstrate that SPEAR effectively bridges the reasoning gap between student and teacher models via sequence-level distillation with efficient dense process rewards. Our code and data are available at: https://github.com/zhuochunli/SPEAR.
Chinese Translation
基于强化学习的知识蒸馏有潜力将复杂的推理从教师模型转移到学生模型，但目前面临一个关键的困境：研究人员必须在稀疏的基于结果的奖励（提供不足的逻辑指导）和昂贵的神经过程奖励模型（Process Reward Models, PRMs）之间进行选择，以获得密集信号。我们通过引入SPEAR（符号过程评估与对齐奖励）来解决这一问题，这是一种无训练且即插即用的过程奖励方法，适用于序列级别的在线蒸馏。SPEAR将自然语言推理轨迹投影到领域自适应的符号里程碑上，为过程级推理对齐提供了有效的代理。通过利用最长公共子序列（Longest Common Subsequence, LCS）将学生的探索与教师的里程碑对齐，SPEAR提供了一种密集的、顺序感知的奖励信号，强制执行逻辑一致性，而无需外部神经验证器。我们在数学、科学和常识推理任务上的实验表明，SPEAR通过序列级蒸馏与高效的密集过程奖励有效地弥合了学生模型与教师模型之间的推理差距。我们的代码和数据可在以下链接获取：https://github.com/zhuochunli/SPEAR。
cs.CL / 61 / 2608.26563
SPT: Skills as Pre-Training Data for Agentic Language Models
SPT：作为代理语言模型预训练数据的技能
Abstract
Agentic (tool-using) language models are mainly trained on tool-call traces and agent trajectories during post-training. These data provide direct behavioral supervision, but producing them requires task environments, execution, and verification, making broad tool and task coverage expensive. Publicly available skills offer another source of training data: they encode reusable tool semantics and workflows but are typically used only as inference-time context. We introduce Skill Pre-Training (SPT), a mid-training method that applies causal language modeling to SkillCorpus, a collection of public multi-file skill packages, optionally mixed with general data. To preserve relations among files within each package, we also introduce Reference Insert, a reference-aware assembly strategy that places supporting files near their mentions in the primary instruction. Experiments across multiple model scales and post-training recipes show that SPT consistently improves agentic performance over mid-training on general or trajectory data, while largely preserving general performance. Data mixture experiments show additional benefits from combining skill data with general annealing corpora. These results indicate that skill packages are a valuable data source for pre-training agentic language models.
Chinese Translation
代理（工具使用）语言模型主要在后期训练中基于工具调用轨迹和代理轨迹进行训练。这些数据提供了直接的行为监督，但生成这些数据需要任务环境、执行和验证，使得广泛的工具和任务覆盖成本高昂。公开可用的技能提供了另一种训练数据来源：它们编码了可重用的工具语义和工作流程，但通常仅在推理时作为上下文使用。我们提出了技能预训练（Skill Pre-Training, SPT），这是一种中期训练方法，应用因果语言建模于SkillCorpus，一个包含公共多文件技能包的集合，选项上可以与一般数据混合。为了保留每个包内文件之间的关系，我们还引入了参考插入（Reference Insert），这是一种参考感知的组装策略，将支持文件放置在主要指令中提及它们的附近。跨多个模型规模和后期训练方案的实验表明，SPT在一般或轨迹数据的中期训练上始终改善了代理性能，同时在很大程度上保持了一般性能。数据混合实验显示，将技能数据与一般退火语料库结合使用带来了额外的好处。这些结果表明，技能包是预训练代理语言模型的宝贵数据来源。
cs.CL / 62 / 2608.26574
Dependency-Aware Revocable Decoding for Efficient Diffusion Large Language Model Inference
依赖感知的可撤销解码用于高效的扩散大型语言模型推理
Abstract
Diffusion large language models (dLLMs) offer a promising alternative to autoregressive generation by decoding multiple tokens in parallel through iterative denoising. However, increasing decoding parallelism often degrades generation quality, as early errors can contaminate later contexts. Revocable decoding mitigates this issue by re-evaluating decoded tokens and remasking unreliable ones, but existing methods overlook that unreliable tokens may also corrupt the verification context itself. We identify this failure mode and propose Dependency-Aware Revocable Decoding (DARD), a training-free framework that separates tokens into masked, candidate, and unmasked states. DARD verifies candidate tokens using a selective context that excludes less reliable tokens and adaptively regulates their influence on subsequent decoding. Experiments across 12 textual and multimodal benchmarks on 3 open-source dLLMs show that DARD consistently improves the speed-quality Pareto frontier over recent revocable decoding methods, achieving a 2.71$\times$ speedup and a 4.35-point CIDEr score gain over Saber on Flickr30K.
Chinese Translation
扩散大型语言模型（dLLMs）通过迭代去噪以并行解码多个标记，提供了自回归生成的有希望的替代方案。然而，增加解码并行性往往会降低生成质量，因为早期的错误可能会污染后续的上下文。可撤销解码通过重新评估已解码的标记并重新屏蔽不可靠的标记来缓解这一问题，但现有方法忽视了不可靠的标记也可能破坏验证上下文本身。我们识别出这一失效模式，并提出了依赖感知的可撤销解码（DARD），这是一种无训练的框架，将标记分为屏蔽、候选和未屏蔽状态。DARD使用选择性上下文来验证候选标记，该上下文排除了不太可靠的标记，并自适应地调节它们对后续解码的影响。在3个开源dLLMs上的12个文本和多模态基准实验表明，DARD在最近的可撤销解码方法上始终提高了速度-质量的Pareto前沿，相较于Saber在Flickr30K上实现了2.71倍的加速和4.35分的CIDEr得分提升。
cs.CL / 63 / 2608.26576
Double Trouble: Bilingual Pretraining Leaves Language-Conditioned Effects in Shared-Language Representations
双重困境：双语预训练在共享语言表征中留下语言条件效应
Abstract
When researchers compare multilingual models for probing, interpretability, or cross-lingual transfer, they often align embedding spaces and assume that shared-language representations are comparable. We show that this assumption can be premature for decoder-only models. We pretrain paired 310M-parameter models (one English-only, one bilingual) across eight typologically diverse languages, separately controlling for English exposure, total compute, and document overlap. After aligning on shared English vocabulary, we test held-out words and find that token embeddings look similar after alignment, but the deeper hidden states that the model uses for prediction do not. This gap holds for all eight languages and survives controls for document overlap and alternative alignment methods. This hidden-state mismatch grows through middle transformer layers, suggesting that it arises from contextual processing rather than the input representations where alignment is performed. Embedding alignment can mask real differences in how models internally represent a shared language, which matters for any downstream study that treats aligned models as interchangeable.
Chinese Translation
当研究人员比较多语言模型以进行探测、可解释性或跨语言迁移时，他们通常会对嵌入空间进行对齐，并假设共享语言表征是可比的。我们表明，这一假设对于仅解码器模型来说可能过于仓促。我们在八种类型多样的语言上预训练了配对的310M参数模型（一个仅限英语，一个双语），同时控制英语暴露、总计算量和文档重叠。在对共享英语词汇进行对齐后，我们测试了保留的单词，发现对齐后标记嵌入看起来相似，但模型用于预测的更深隐藏状态却并不相同。这一差距在所有八种语言中都存在，并且在控制文档重叠和替代对齐方法后依然成立。这种隐藏状态的不匹配在中间变换层中加剧，表明它源于上下文处理，而非进行对齐的输入表征。嵌入对齐可能掩盖模型在内部如何表征共享语言的真实差异，这对于任何将对齐模型视为可互换的下游研究都是重要的。
cs.CL / 64 / 2608.26587
Surgical Alignment in Knowledge Graph Training for Clinical Diagnosis with Large Language Models
基于知识图谱训练的临床诊断中的手术对齐与大型语言模型
Abstract
Biomedical knowledge graphs (KGs) offer structured medical knowledge that can ground large language model (LLM) reasoning in clinical diagnosis application, yet how KG signal should be integrated into LLMs remains an open question. We present a systematic study spanning five KG task formulations, three training paradigms, two KGs, and three base LLMs. At the task level, all paradigms improve over the non-finetuned baseline, but methods with comparable in-domain accuracy show substantially different knowledge transfer behavior. We introduce Gradient Intervention Density (GID) and Gradient Distortion (GD) to measure how broadly an optimizer modifies the pretrained model. GID and GD together reveal a clear divide: KG-judgment training under KL regularization produces sparse, localized updates (a regime we term as surgical alignment), while task-specific SFT produces dense ones. A controlled ablation shows that the objective and KL contribute to sparsity independently, and the paradigms that produce sparse updates also improve reasoning quality, even when their in-domain accuracy is lower than task-specific SFT. Assessing KG-LLM integration thus requires complementing accuracy with optimization-geometry diagnostics. Our implementation can be found at https://github.com/LARK-NLP-Lab/Surgical-Alignment.
Chinese Translation
生物医学知识图谱（KGs）提供了结构化的医学知识，可以为大型语言模型（LLM）在临床诊断应用中的推理提供基础，但如何将KG信号整合到LLMs中仍然是一个未解的问题。我们进行了一项系统研究，涵盖了五种KG任务形式、三种训练范式、两个KG和三个基础LLM。在任务层面，所有范式均优于未微调的基线，但在领域内准确性相当的方法表现出显著不同的知识转移行为。我们引入了梯度干预密度（Gradient Intervention Density, GID）和梯度失真（Gradient Distortion, GD）来衡量优化器如何广泛地修改预训练模型。GID和GD共同揭示了一个明显的分界：在KL正则化下的KG判断训练产生稀疏、局部的更新（我们称之为手术对齐），而任务特定的SFT则产生密集的更新。控制性消融实验表明，目标和KL独立地对稀疏性做出贡献，而产生稀疏更新的范式也改善了推理质量，即使它们的领域内准确性低于任务特定的SFT。因此，评估KG-LLM整合需要用优化几何诊断来补充准确性。我们的实现可以在https://github.com/LARK-NLP-Lab/Surgical-Alignment找到。
cs.CL / 65 / 2608.26592
Benchmarking Clinical Decision Pathway Adherence in Large Language Models
大型语言模型在临床决策路径遵循中的基准评估
Abstract
Following clinical decision pathways (CDPs) defined by clinical practice guidelines is essential for safe and reliable medical decision-making. However, existing medical large language model (LLM) benchmarks mainly evaluate final-answer accuracy, providing limited evaluation of models' ability to adhere to guidelines. To address this gap, we introduce MEGA-CDP, a benchmark for evaluating whether medical LLMs can generate guideline-adherent CDPs using provided guidelines as references. MEGA-CDP is constructed from 2,274 English and Chinese clinical practice guidelines through a guideline-to-case pipeline, yielding 42,353 clinical cases with explicit reference CDPs. It supports both single-turn vignette and multi-turn interactive settings, and introduces a CDP-oriented evaluation framework for measuring pathway consistency. Experiments on 16 representative LLMs show that reliable clinical decision support remains challenging for current models, demonstrating the need for CDP-oriented evaluation and the value of MEGA-CDP for advancing guideline adherence in medical LLMs.
Chinese Translation
遵循临床实践指南定义的临床决策路径（CDPs）对于安全可靠的医疗决策至关重要。然而，现有的医学大型语言模型（LLM）基准主要评估最终答案的准确性，对模型遵循指南的能力评估有限。为了解决这一问题，我们引入了MEGA-CDP，这是一个评估医学LLM是否能够生成遵循指南的CDPs的基准，使用提供的指南作为参考。MEGA-CDP是通过指南到案例的流程，从2,274个英文和中文临床实践指南构建而成，产生了42,353个具有明确参考CDPs的临床案例。它支持单轮情境和多轮互动设置，并引入了一个以CDP为导向的评估框架，用于测量路径一致性。对16个代表性LLM的实验表明，当前模型在提供可靠的临床决策支持方面仍面临挑战，显示出以CDP为导向的评估的必要性，以及MEGA-CDP在推动医学LLM遵循指南方面的价值。
cs.CL / 66 / 2608.26596
Not Just Reason, Not Just Scan: Reinforcement Learning for Proactive Scientific Error Verification over Academic Paper
不仅仅是推理，不仅仅是扫描：用于主动科学错误验证的强化学习方法
Abstract
Multimodal large language models (MLLMs) are increasingly capable scientific assistants, yet they remain far from fully autonomous research. This transition requires models to actively inspect academic papers, build global evidence views, and make traceable judgments without prespecified issues or evidence. However, existing work provides limited task paradigms or training studies for such issue- and evidence-absent verification. We study this challenge through scientific error detection, where models must determine whether errors exist and justify them with evidence-based reasoning. To fill this gap, we present VERA-RL, a reinforcement-learning formulation for scientific error detection over academic papers. Following a Reason--Verify--Scan progression, we construct VERA-13K, a 12,900-sample dataset organized into 4,300 matched chains, covering 6 scientific-error categories across the research workflow and broad natural-science domains. We further introduce fine-grained rewards for reasoning completeness, evidence alignment, and error precision. Training Qwen3-VL-8B with VERA-RL substantially improves verifiable reasoning, approaching flagship MLLMs such as Gemini 3 Pro and Qwen3-VL-235B-A22B on Scan.
Chinese Translation
多模态大型语言模型（MLLMs）正日益成为有效的科学助手，但它们距离完全自主研究仍有很大差距。这一转变要求模型主动检查学术论文，构建全球证据视图，并在没有预先指定问题或证据的情况下做出可追溯的判断。然而，现有研究为这种缺乏问题和证据的验证提供的任务范式或训练研究有限。我们通过科学错误检测来研究这一挑战，在此过程中，模型必须判断错误是否存在，并用基于证据的推理进行辩护。为填补这一空白，我们提出了VERA-RL，一种用于学术论文科学错误检测的强化学习框架。按照推理-验证-扫描的进程，我们构建了VERA-13K，这是一个包含12,900个样本的数据集，组织成4,300个匹配链，涵盖研究工作流程和广泛自然科学领域的6个科学错误类别。我们进一步引入了针对推理完整性、证据一致性和错误精确度的细粒度奖励。使用VERA-RL训练Qwen3-VL-8B显著提高了可验证推理的能力，接近于旗舰级MLLMs，如Gemini 3 Pro和Qwen3-VL-235B-A22B在扫描任务上的表现。
cs.CL / 67 / 2608.26638
Which Metrics Save the Most Human Annotation? Prediction-Powered Evaluation and Meta-Evaluation
哪些指标能够节省最多的人类标注？基于预测的评估与元评估
Abstract
Across various non-verifiable tasks, human evaluation is reliable but expensive, while automatic metrics are more scalable but often biased. Building on prediction-powered inference (PPI), we propose prediction-powered evaluation, a framework that combines limited human judgments with large-scale automatic scores to obtain data-efficient system comparisons that are provably unbiased. We develop parametric and non-parametric procedures, analyze the efficiency trade-off between paired and unpaired designs, and validate the framework on six WMT datasets. We further introduce the Prediction-Powered Saving Ratio (PPSR), a meta-metric that measures how much human annotation an automatic metric can save when used within prediction-powered evaluation. PPSR directly targets metric utility for prediction-powered evaluation and yields more discriminative and stable metric rankings than existing system-level meta-metrics. Overall, our new paradigm reframes automatic metrics as tools for reducing human annotation cost rather than replacing human judgment, and applies broadly to non-verifiable tasks.
Chinese Translation
在各种不可验证的任务中，人类评估虽然可靠但成本高昂，而自动指标则更具可扩展性但常常存在偏差。基于预测驱动推理（PPI），我们提出了基于预测的评估框架，该框架结合了有限的人类判断与大规模的自动评分，从而获得数据高效的系统比较，并且可以证明其无偏性。我们开发了参数化和非参数化的程序，分析了配对设计与非配对设计之间的效率权衡，并在六个WMT数据集上验证了该框架。我们进一步引入了预测驱动节省比（PPSR），这一元指标衡量了在基于预测的评估中，自动指标能够节省多少人类标注。PPSR直接针对基于预测的评估中的指标效用，并且比现有的系统级元指标产生更具辨别力和稳定性的指标排名。总体而言，我们的新范式将自动指标重新定义为降低人类标注成本的工具，而不是替代人类判断，并广泛适用于不可验证的任务。
cs.CL / 68 / 2608.26641
Information-Guided Frontier Decoding: Contextual Utility-Driven Commitment in dMLLMs
信息引导的前沿解码：基于上下文效用驱动的 dMLLMs 承诺
Abstract
Decoding quality in diffusion multimodal language models (dMLLMs) depends heavily on the order in which masked tokens are committed. Existing confidence-based strategies prioritize locally easy tokens, but confidence does not necessarily reflect contextual usefulness. As a result, structurally easy tokens such as punctuation may be committed before informative semantic anchors, weakening context propagation and increasing error accumulation. We propose Information-Guided Frontier Decoding (IGFD), a training-free decoding strategy that ranks candidates using token confidence, neighborhood uncertainty, and structural commitment risk. IGFD encourages early commitment of reliable semantic anchors while delaying fragile structural tokens, improving contextual support during decoding. A dynamic candidate frontier further constrains token selection to locally expandable regions under the same decoding budget. The method requires no additional training, auxiliary models, or extra forward passes. Experiments across multimodal understanding, reasoning, grounding, and hallucination benchmarks show that IGFD consistently outperforms existing decoding strategies across the majority of benchmarks and diffusion MLLM backbones under identical decoding budgets.
Chinese Translation
扩散多模态语言模型 (dMLLMs) 的解码质量在很大程度上依赖于掩码令牌被承诺的顺序。现有的基于置信度的策略优先考虑局部容易的令牌，但置信度并不一定反映上下文的实用性。因此，诸如标点符号等结构上容易的令牌可能会在信息丰富的语义锚点之前被承诺，从而削弱上下文传播并增加错误累积。我们提出了信息引导的前沿解码 (Information-Guided Frontier Decoding, IGFD)，这是一种无训练的解码策略，通过令牌置信度、邻域不确定性和结构承诺风险对候选项进行排名。IGFD 鼓励可靠的语义锚点的早期承诺，同时延迟脆弱的结构令牌，从而在解码过程中改善上下文支持。动态候选前沿进一步限制了在相同解码预算下的局部可扩展区域的令牌选择。该方法不需要额外的训练、辅助模型或额外的前向传递。在多模态理解、推理、基础和幻觉基准测试中的实验表明，IGFD 在大多数基准测试和扩散 MLLM 主干下始终优于现有的解码策略，且在相同的解码预算下表现出色。
cs.CL / 69 / 2608.26650
Meta-Learning Where to Allocate Experts: Task-Conditioned Layer-Wise Compression for MoEs
元学习专家分配：基于任务条件的逐层压缩用于专家混合模型
Abstract
Mixture-of-Experts (MoE) models route each token to a subset of expert networks, increasing capacity while keeping per-token computation sparse. In many deployed MoEs, the number of active experts is fixed across layers and tasks, although layer roles and expert redundancy vary with depth and demand varies with difficulty. Existing approaches address only part of this setting: layer-wise allocations are usually determined offline and reused for all tasks, while token-level methods vary expert activation using local routing signals without task-level context. We propose MetaNet, a support-set controller that predicts, for each layer, an expert-retention threshold and a bounded routing bias. The backbone, experts, and router remain frozen. On DeepSeek-MoE-16B-Chat, MetaNet provides a tunable accuracy-expert-activation trade-off. Relative to fixed k=6, a conservative setting activates 3.61 experts on average (40% fewer) and achieves comparable MMLU accuracy (0.489 vs. 0.474), whereas an aggressive setting activates 2.28 experts on average (62% fewer) with accuracy approximately 3.7 percentage points lower. The MMLU-trained controller also transfers to C-Eval without retraining, activating 2.90 experts on average (52% fewer than fixed k=6) at 0.386 accuracy.
Chinese Translation
专家混合模型（Mixture-of-Experts, MoE）将每个标记路由到一组专家网络，从而在保持每个标记计算稀疏的同时增加模型容量。在许多实际应用的 MoE 中，活跃专家的数量在层和任务之间是固定的，尽管层的角色和专家的冗余性随着深度变化，而需求则随着难度变化。现有方法仅解决了这一设置的一部分：逐层分配通常是离线确定并在所有任务中重用，而标记级方法则使用局部路由信号变更专家激活，未考虑任务级上下文。我们提出了 MetaNet，一种支持集控制器，能够为每一层预测专家保留阈值和有界路由偏差。主干网络、专家和路由器保持不变。在 DeepSeek-MoE-16B-Chat 上，MetaNet 提供了可调的准确性与专家激活之间的权衡。与固定的 k=6 相比，保守设置平均激活 3.61 个专家（减少 40%），并实现了可比的 MMLU 准确率（0.489 对比 0.474），而激进设置平均激活 2.28 个专家（减少 62%），准确率约低 3.7 个百分点。经过 MMLU 训练的控制器在 C-Eval 中也能无须重新训练地迁移，平均激活 2.90 个专家（比固定的 k=6 少 52%），准确率为 0.386。
cs.CL / 70 / 2608.26674
Do LLMs Understand Personality? Rethinking Persona Fidelity Evaluation through Structured Behavioral Inference
大型语言模型理解人格吗？通过结构化行为推理重新思考角色忠实度评估
Abstract
As large language models are increasingly deployed to simulate diverse human characters, ensuring persona fidelity, defined as the extent to which an agent's behavior consistently reflects the psychological and stylistic characteristics of a target persona, has become a critical requirement. However, existing evaluation paradigms primarily rely on either holistic LLM-based judges, which are prone to "holistic appraisal hallucination'', or static psychometric inventories, which fail to capture the context-dependent fidelity required in dynamic dialogue. To address these limitations, we propose PRISM (Persona Reasoning with Inverse SFL-based Modeling), a psycholinguistically grounded framework that reformulates persona fidelity evaluation as a structured inverse inference task. Inspired by Systemic Functional Linguistics (SFL), PRISM decomposes persona fidelity into three functional dimensions: Task Framing, Interpersonal Stance, and Linguistic Style. It estimates dimension-specific evidence over a persona-conditioned label space and aggregates these signals into an interpretable and auditable evaluation process. Experiments show that PRISM yields more accurate and stable judgements than traditional holistic judging, providing a more reliable framework for persona fidelity evaluation.
Chinese Translation
随着大型语言模型被越来越多地用于模拟多样的人类角色，确保角色忠实度（定义为代理的行为在多大程度上持续反映目标角色的心理和风格特征）已成为一项关键要求。然而，现有的评估范式主要依赖于整体的基于LLM的评判者，这些评判者容易出现“整体评估幻觉”，或者依赖于静态的心理测量工具，这些工具无法捕捉动态对话中所需的上下文依赖忠实度。为了解决这些局限性，我们提出了PRISM（基于逆向SFL建模的人格推理），这是一个心理语言学基础的框架，将角色忠实度评估重新构建为一个结构化的逆向推理任务。受到系统功能语言学（SFL）的启发，PRISM将角色忠实度分解为三个功能维度：任务框架、人际立场和语言风格。它在一个基于角色条件的标签空间中估计特定维度的证据，并将这些信号汇总为一个可解释和可审计的评估过程。实验表明，PRISM比传统的整体评判提供了更准确和稳定的判断，为角色忠实度评估提供了一个更可靠的框架。
cs.CL / 71 / 2608.26676
FOCUS & RePAIR: Mitigating Text Degeneration via Token-Level Guidance for Pruned Large Language Models
FOCUS与RePAIR：通过令牌级指导减轻修剪大型语言模型的文本退化
Abstract
Pruning is a practical approach to compress large language models (LLMs), but it can amplify text degeneration, especially repetition loops, even when perplexity and task accuracy remain largely unchanged. In this work, we present a token-level analysis of this failure mode by viewing decoding as a dynamical process that enters and persists in a small set of recurrent contexts. Our analysis decomposes degeneration into loop entry risk and loop persistence, and shows that persistence is controlled by the escape mass assigned to plausible alternatives within the token sampling set. Motivated by these findings, we propose two token-level guidance objectives for post-pruning fine-tuning. FOCUS reweights distillation toward high-confidence teacher regions to suppress leakage, while RePAIR uses onset-centered positive/negative continuation pairs with a margin loss to promote plausible alternatives and prevent early commitment to repetition loops. Experiments on open-ended continuation and instruction-based generation show that both methods consistently reduce repetition and improve generation quality.
Chinese Translation
修剪是一种压缩大型语言模型（LLMs）的实用方法，但它可能会加剧文本退化，尤其是重复循环，即使困惑度和任务准确性基本保持不变。在本研究中，我们通过将解码视为一个进入并持续在一小组重复上下文中的动态过程，提出了对这种失败模式的令牌级分析。我们的分析将退化分解为循环进入风险和循环持续性，并表明持续性由分配给令牌采样集内可行替代项的逃逸质量控制。基于这些发现，我们提出了两个用于后修剪微调的令牌级指导目标。FOCUS 通过重新加权蒸馏以集中于高置信度教师区域来抑制信息泄漏，而 RePAIR 则使用以开始为中心的正/负延续对和边际损失来促进可行替代项，并防止过早承诺于重复循环。在开放式延续和基于指令的生成实验中，两个方法均一致地减少了重复并提高了生成质量。
cs.CL / 72 / 2608.26689
Beyond Reflection: Affirmation as a Promising Behavioral Marker Associated with Quality in Text-Based Counseling
超越反思：肯定作为与文本咨询质量相关的有前景的行为标记
Abstract
While AI-assisted text-based counseling is gaining attention, it remains empirically unclear which counselor behaviors are associated with higher dialogue quality. Existing research often focuses heavily on Reflection, borrowing frameworks from Motivational Interviewing. To address this gap, we conduct a multi-layered analysis using KokoroChat, a large-scale Japanese text counseling dataset conducted by professional counselors and trainees, newly annotated with counselor strategy tags and client distress levels. Our results show that, under the quality indicators used in this study, Affirmation is more consistently associated with session quality than Reflection among the analyzed strategies. Cross-dataset transfer experiments further suggest that this quality signal can be observed to some extent on ESConv, an English dataset with non-expert supporters. These findings provide empirical implications for counselor training and emotional support system design. We release the additional KokoroChat annotations and experimental source code at https://github.com/UEC-InabaLab/BeyondReflection.
Chinese Translation
尽管人工智能辅助的文本咨询正受到越来越多的关注，但目前尚不清楚哪些咨询师行为与更高的对话质量相关。现有研究往往过于侧重于反思，借用动机访谈的框架。为了解决这一空白，我们使用KokoroChat进行多层次分析，该数据集是由专业咨询师和实习生进行的大规模日本文本咨询数据集，新增了咨询师策略标签和客户困扰水平的注释。我们的结果表明，在本研究使用的质量指标下，肯定比反思在分析的策略中与会话质量更一致地相关。跨数据集迁移实验进一步表明，这一质量信号在ESConv（一个由非专业支持者构成的英文数据集）中在一定程度上也可以观察到。这些发现为咨询师培训和情感支持系统设计提供了实证启示。我们将在 https://github.com/UEC-InabaLab/BeyondReflection 发布额外的KokoroChat注释和实验源代码。
cs.CL / 73 / 2608.26697
Scaling phoneme-based TTS augmentation for ASR: A unified pipeline and controlled study
基于音素的文本到语音增强在自动语音识别中的扩展：统一管道与控制研究
Abstract
Synthetic speech provides scalable supervision for automatic speech recognition (ASR), but its benefit depends on the selected texts, reference speech, and amount of synthesized data. We present a unified phoneme-based TTS-to-ASR augmentation pipeline built around a multilingual TTS model trained from scratch using the F5-TTS architecture with language-ID conditioning. The pipeline combines language-specific grapheme-to-phoneme conversion, reference-speech filtering, candidate-text selection, synthesis, and matched ASR continuation. We further propose phoneme-frequency-guided selection (PFGS), which ranks candidate sentences using phoneme frequencies estimated from real ASR training labels. Experiments with separate monolingual ASR systems for Arabic, French, Italian, and Portuguese span 13 test sets. Across the synthesis-scale sweep, random augmentation improves over matched real-only continuation on 11 test sets. Under a nominal 60% synthesis budget, PFGS improves over real-only training on 12 test sets and over random selection on 9. Its largest relative word error rate (WER) reduction against random selection is 19.3%. With target texts and synthesis counts fixed, reference-speech filtering reduces absolute WER by 0.29 and 0.59 points on Italian and French Common Voice, respectively. These results identify synthesis scale, candidate-text content, and reference quality as important control variables in TTS-based ASR augmentation.
Chinese Translation
合成语音为自动语音识别（ASR）提供了可扩展的监督，但其效益取决于所选文本、参考语音和合成数据的数量。我们提出了一种基于音素的统一文本到语音（TTS）到自动语音识别（ASR）增强管道，该管道围绕一个从头开始训练的多语言TTS模型构建，采用F5-TTS架构并结合语言识别条件。该管道结合了特定语言的字母到音素转换、参考语音过滤、候选文本选择、合成和匹配的ASR延续。我们进一步提出了音素频率引导选择（PFGS），该方法使用从真实ASR训练标签中估计的音素频率对候选句子进行排名。针对阿拉伯语、法语、意大利语和葡萄牙语的独立单语ASR系统的实验涵盖了13个测试集。在合成规模的变化中，随机增强在11个测试集上优于仅匹配真实延续。在名义上60%的合成预算下，PFGS在12个测试集上优于仅真实训练，在9个测试集上优于随机选择。与随机选择相比，其相对词错误率（WER）最大减少为19.3%。在目标文本和合成数量固定的情况下，参考语音过滤在意大利语和法语Common Voice上分别减少了绝对WER 0.29和0.59点。这些结果表明，合成规模、候选文本内容和参考质量是基于TTS的ASR增强中的重要控制变量。
cs.CL / 74 / 2608.26700
PragAlign: Evidence-Sensitive Reply Assistance Across Chinese and Japanese Appropriateness Judgments
PragAlign：跨中文和日文适当性判断的证据敏感回复辅助
Abstract
Reply assistance in multilingual settings requires linguistic competence and culturally situated judgments of appropriateness. We present PragAlign, which separates context reading from selective clarification, and evaluate it alongside Direct and Rule. Nine native Chinese speakers judged Chinese materials, while three native Japanese speakers judged matched Japanese versions. In the Chinese evaluation, PragAlign received significantly better ranks than both baselines. In the Japanese evaluation, Direct had the lowest mean rank, PragAlign had the highest top-rank rate, and the omnibus difference was not significant. The groups selected the same top condition in 5 of 10 scenarios, including four shared PragAlign selections. The results identify shared and language-specific judgment patterns and inform reply assistance designed to support linguistic and cultural understanding.
Chinese Translation
在多语言环境中，回复辅助需要语言能力和文化背景下的适当性判断。我们提出了PragAlign，它将上下文理解与选择性澄清分开，并与Direct和Rule进行了评估。九名母语为中文的参与者评估了中文材料，而三名母语为日文的参与者评估了匹配的日文版本。在中文评估中，PragAlign的排名显著优于两个基线。在日文评估中，Direct的平均排名最低，PragAlign的最高排名率最高，整体差异不显著。两个组在10个场景中选择了相同的最佳条件，其中包括四个共同选择的PragAlign结果。结果识别了共享和语言特定的判断模式，并为旨在支持语言和文化理解的回复辅助提供了信息。
cs.CL / 75 / 2608.26706
Towards Expert Financial QA via Self-Improving RAG
通过自我改进的RAG实现专家级金融问答
Abstract
Expert-level financial question answering requires both grounded verification to catch numeric hallucinations and audit trails for regulatory compliance, attributes that standard single-pass RAG systems lack. We take a step toward this goal with Self-Improving RAG, a framework that decomposes document QA into three specialized agents (Retrieval, Reasoning, and Judge) coordinated by an orchestrator with feedback-driven self-correction. When the Judge Agent scores an answer below a dynamic threshold, the system triggers retry with escalated strategies: broader retrieval, more careful prompting, and relaxed acceptance criteria. We evaluate on FinanceBench (SEC filing QA), where Self-Improving RAG achieves 86% oracle-guided accuracy (measuring agreement with gold answers) with a 36.4% Lazarus Rate, recovering nearly 4 in 10 initially incorrect answers through targeted retry. A key finding is that a fixed retrieval pipeline with judge-driven retry achieves strong results without dynamic routing, providing full interpretability. Every decision is logged with confidence scores, enabling the audit trails required for regulated financial applications.
Chinese Translation
专家级金融问答需要基于事实的验证以捕捉数字幻觉，并提供合规审计追踪，而标准的单通道RAG系统缺乏这些特性。我们朝着这一目标迈出了一步，提出了自我改进的RAG（Self-Improving RAG）框架，该框架将文档问答分解为三个专门的代理（检索、推理和评判），由一个协调者通过反馈驱动的自我修正进行协调。当评判代理（Judge Agent）对答案的评分低于动态阈值时，系统会触发重试，采用升级策略：更广泛的检索、更仔细的提示和放宽的接受标准。我们在FinanceBench（SEC文件问答）上进行了评估，自我改进的RAG达到了86%的oracle指导准确率（衡量与金标准答案的一致性），并且Lazarus率为36.4%，通过有针对性的重试恢复了近4成最初错误的答案。一个关键发现是，固定的检索管道结合评判驱动的重试在没有动态路由的情况下也能取得良好结果，提供了完全的可解释性。每个决策都记录了置信度评分，从而实现了合规金融应用所需的审计追踪。
cs.CL / 76 / 2608.26735
Preserving General Capabilities during Domain Specialization with Uncertainty-Calibrated MOPD
在不确定性校准的多教师在线蒸馏中保持一般能力的领域专业化
Abstract
Specializing large language models to vertical domains improves domain-specific behavior but often degrades general capabilities such as reasoning, coding, instruction following, and creative writing. We study this domain--general trade-off in Multi-Teacher On-Policy Distillation (MOPD), where a specialized student is supervised on its own sampled trajectories by domain and general teachers. Standard MOPD faces two limitations: ordinary on-policy sampling rarely exposes tokens with large positive teacher--student advantages, while the advantage sign alone does not establish whether the resulting update direction is reliable. We propose uncertainty-calibrated MOPD to address these limitations. Dual-temperature sampling broadens the candidate trajectory pool, and positive-advantage-density filtering selects trajectories with stronger positive learning signals. Centered log-likelihood (CLL) filtering then computes an entropy-calibrated teacher-endorsement score and probabilistically retains token updates according to direction--endorsement consistency. Experiments on role-playing and medical-domain specialization show that our method improves the general-capability average over standard MOPD by $4.73\%$ and $10.84\%$, respectively, while maintaining vertical-domain performance. Ablations and diagnostic analyses further confirm that the gains do not merely result from a larger rollout budget and that the proposed trajectory- and token-level mechanisms address their intended failure modes.
Chinese Translation
将大型语言模型专门化到垂直领域可以改善领域特定行为，但往往会降低推理、编码、遵循指令和创意写作等一般能力。我们研究了在多教师在线蒸馏（MOPD）中的领域与一般能力之间的权衡，其中一个专门化的学生在领域教师和一般教师的监督下，根据自身采样的轨迹进行学习。标准的MOPD面临两个限制：普通的在线采样很少暴露出具有较大正向教师-学生优势的标记，而仅凭优势符号并不能确定结果更新方向的可靠性。为了解决这些限制，我们提出了不确定性校准的MOPD。双温度采样扩展了候选轨迹池，而正向优势密度过滤选择具有更强正向学习信号的轨迹。然后，中心对数似然（CLL）过滤计算熵校准的教师认可分数，并根据方向-认可一致性以概率方式保留标记更新。在角色扮演和医疗领域专业化的实验中，我们的方法在保持垂直领域性能的同时，分别提高了标准MOPD的一般能力平均值$4.73\%$和$10.84\\%$。消融实验和诊断分析进一步确认，这些增益并非仅仅源于更大的展开预算，且所提出的轨迹和标记级机制有效解决了其预期的失效模式。
cs.CL / 77 / 2608.26749
Letters hide the truth from our eyes: English homophones have meaningfully different phonetic realizations
字母掩盖了我们眼中的真相：英语同音词具有意义上不同的语音实现
Abstract
The distribution of spoken word duration of English homophones is known to co-vary with frequency of use. This study investigates whether other aspects of the phonetic realization of homophones also differ. A series of quantitative investigations of 14,000 homophone tokens in American television news broadcasts revealed that the tokens of homophone pairs such as \textit{weight} and \textit{wait} have different phonetic realizations, and that these can be predicted from their meanings in utterance context. These systematic differences remain even when taking duration-related variation into account. Time-normalized spectrograms emerged as an excellent tool for probing the fine details of phonetic realization, and obviate the need for phonetic transcriptions, which inevitably hide the phonetic truth from our eyes.
Chinese Translation
英语同音词的口语时长分布已知与使用频率相关。本研究探讨了同音词的语音实现的其他方面是否也存在差异。对美国电视新闻广播中14,000个同音词实例进行的一系列定量研究显示，像 extit{weight}（重量）和 extit{wait}（等待）这样的同音词对的实例具有不同的语音实现，这些差异可以根据它们在话语上下文中的意义进行预测。这些系统性的差异在考虑与时长相关的变异时仍然存在。时间归一化的声谱图成为探究语音实现细节的优秀工具，避免了对语音转录的需求，而后者不可避免地掩盖了我们眼中的语音真相。
cs.CL / 78 / 2608.26762
Equal Ranking Quality, Different Decisions: Training Order-Consistent LLM Scorers
相同的排名质量，不同的决策：训练顺序一致的 LLM 评分器
Abstract
Rerankers, reward models and multi-document QA scorers score candidate documents or responses in one LLM prompt, so each score depends on their order. Such scorers are selected on ranking quality, but their scores determine a decision: what a score threshold retains, a reader answers, or a preference model selects. However, equal ranking quality does not imply equal decisions: on passage reranking, five trained scorers within 0.010 nDCG@10 retain sets that overlap by only 0.66-0.84 when reordered. A published reranker takes the highest retained-set F1 in our comparison and still overlaps by only 0.667. No prompt-time change we test removes that order dependence: the only one that gains ranking quality leaves all three decisions unchanged. Order-consistency SFT (OC-SFT) attenuates it in the weights, training a candidate's score not to depend on the order. It holds ranking quality and leads every decision-stability measure among trained scorers on all three tasks: it flips the reader's answer on 0.125 of permutation pairs against 0.149-0.164 for three other objectives that target order. It is more stable than order-averaged distillation on 12 base models, and one OC-SFT permutation retains sets that overlap more than ten averaged off-the-shelf permutations. A comparison should therefore report what a threshold retains and a reader answers, not ranking quality alone. Code is available at https://github.com/thomsonreuters/presentation-dependence.
Chinese Translation
重排序器、奖励模型和多文档问答评分器在一个 LLM 提示中对候选文档或响应进行评分，因此每个评分都依赖于它们的顺序。这些评分器是基于排名质量进行选择的，但它们的评分决定了一个决策：评分阈值保留的内容、读者的回答或偏好模型的选择。然而，相同的排名质量并不意味着相同的决策：在段落重排序中，五个训练过的评分器在 0.010 nDCG@10 的范围内保留的集合在重新排序时仅重叠 0.66-0.84。我们比较中发布的一个重排序器在保留集合 F1 方面最高，但重叠仍然仅为 0.667。我们测试的没有任何提示时间变化消除了这种顺序依赖性：唯一一个提高排名质量的变化使所有三个决策保持不变。顺序一致性 SFT（OC-SFT）在权重中减弱了这种依赖性，训练候选的评分不依赖于顺序。它保持了排名质量，并在所有三个任务中引导每个决策稳定性度量的训练评分器：在 0.125 的排列对中翻转了读者的答案，而针对顺序的其他三个目标则为 0.149-0.164。它在 12 个基础模型上比顺序平均蒸馏更稳定，并且一个 OC-SFT 排列保留的集合重叠超过十个平均的现成排列。因此，比较应报告阈值保留的内容和读者的回答，而不仅仅是排名质量。代码可在 https://github.com/thomsonreuters/presentation-dependence 获取。
cs.CL / 79 / 2608.26779
Instruction Quality Matters: Refining Instructions for Effective Preference Learning
指令质量至关重要：优化指令以实现有效的偏好学习
Abstract
Preference learning optimizes models using response pairs, yet the informativeness of these pairs is fundamentally shaped by the instructions from which they are generated. We identify instruction quality as a hidden bottleneck in preference learning: low-quality or ambiguous instructions restrict the response-quality distribution, limiting strong chosen responses and weakening preference signals. Through Best- and Worst-of-N analyses, we show that instruction quality constrains both the ceiling and floor of sampled response quality. Motivated by this observation, we introduce an instruction-refinement pipeline that selects weak instructions using reward signals and revises them with rubric-guided LLM feedback, improving preference data without discarding examples. Across offline and online preference learning settings, experiments on multiple models and benchmarks show broad alignment improvements over original data and alternative data-improvement strategies. Further analyses indicate that instruction refinement raises achievable response quality and complements response-centric preference data curation. Overall, instruction quality emerges as a key factor governing how informative preference signals are formed for LLM alignment. Code is available at: https://github.com/01choco/instruction-refinement/
Chinese Translation
偏好学习通过响应对来优化模型，但这些对的有效性在根本上受到生成它们的指令的影响。我们将指令质量识别为偏好学习中的一个隐性瓶颈：低质量或模糊的指令限制了响应质量分布，限制了强选择响应的生成并削弱了偏好信号。通过最佳和最差的N分析，我们展示了指令质量限制了采样响应质量的上限和下限。基于这一观察，我们引入了一种指令优化流程，该流程利用奖励信号选择弱指令，并通过评分标准指导的LLM反馈对其进行修订，从而改善偏好数据而不丢弃示例。在离线和在线偏好学习设置中，对多个模型和基准的实验显示，相较于原始数据和其他数据改进策略，整体对齐性有显著提升。进一步分析表明，指令优化提高了可实现的响应质量，并补充了以响应为中心的偏好数据整理。总体而言，指令质量成为影响LLM对齐中信息性偏好信号形成的关键因素。代码可在以下链接获取：https://github.com/01choco/instruction-refinement/
cs.CL / 80 / 2608.26807
Behavior2Trip: Towards Personalized Travel Planning via User Behavior Trajectory
Behavior2Trip：基于用户行为轨迹的个性化旅行规划
Abstract
Travel planning agents assist users in generating personalized travel plans by modeling their individual preferences. Existing agents either rely on explicit user instructions or engage in multi-turn clarification to elicit user preferences. However, both approaches overlook the rich behavioral signals latent in users' past behaviors, which implicitly encode their preferences. This over-reliance on active user input increases interaction burden and limits plan personalization. To bridge this gap, we introduce a new task, Behavior-Aware Travel Planning, which infers user preferences directly from past behaviors and generates personalized travel plans. To facilitate research on this task, we introduce Behavior2Trip, a benchmark constructed from one of the largest Chinese online travel platforms, comprising 11,400 instances. Each instance represents an average of 39.8 past user behaviors spanning 14 attributes across 5 preference dimensions. We further propose B2T-Agent, a reinforcement learning-based agent that leverages user behavior trajectories, interacts with external tools for preference-aligned retrieval, and maintains an internal memory module. Experiments on Behavior2Trip show that GPT-4.1 achieves a full-constraint pass rate of only 0.5\% on the hardest tasks, while B2T-Agent built upon Qwen3-8B outperforms all baselines, highlighting the substantial challenge of this task. Moreover, Qwen3-8B trained with B2T-Agent also outperforms GPT-4.1 on the TravelPlanner benchmark, demonstrating strong generalization. Code and data are available at https://github.com/BUAA-IRIP-LLM/Behavior2Trip
Chinese Translation
旅行规划代理通过建模用户的个体偏好来协助用户生成个性化的旅行计划。现有的代理要么依赖于用户的明确指示，要么通过多轮澄清来引导用户表达偏好。然而，这两种方法都忽视了潜藏于用户过去行为中的丰富行为信号，这些信号隐含地编码了他们的偏好。过度依赖主动的用户输入增加了交互负担，并限制了计划的个性化。为了解决这一问题，我们提出了一项新任务——行为感知旅行规划（Behavior-Aware Travel Planning），该任务直接从过去的行为中推断用户偏好，并生成个性化的旅行计划。为了促进该任务的研究，我们推出了Behavior2Trip，一个基于中国最大的在线旅行平台之一构建的基准数据集，包含11,400个实例。每个实例代表平均39.8个跨越5个偏好维度的14个属性的过去用户行为。我们进一步提出了B2T-Agent，一个基于强化学习的代理，利用用户行为轨迹，与外部工具进行偏好对齐的检索交互，并维护一个内部记忆模块。在Behavior2Trip上的实验表明，GPT-4.1在最难的任务上的全约束通过率仅为0.5\%，而基于Qwen3-8B构建的B2T-Agent超越了所有基线，突显了这一任务的重大挑战。此外，经过B2T-Agent训练的Qwen3-8B在TravelPlanner基准上也优于GPT-4.1，展示了强大的泛化能力。代码和数据可在 https://github.com/BUAA-IRIP-LLM/Behavior2Trip 获取。
cs.CL / 81 / 2608.26832
RuleWeaver: Benchmarking Rule-Centered Scenario Reasoning for Large Language Models
RuleWeaver：针对大型语言模型的规则中心场景推理基准测试
Abstract
Large language models (LLMs) are increasingly applied to specialized domains, where effective use of domain expertise often requires reasoning over complex rules in concrete scenarios. However, existing benchmarks only partially evaluate this capability, as they either focus on output-level instruction constraints or overlook the distinct roles that rules play in scenario reasoning. To address these gaps, this paper introduces RuleWeaver, a benchmark construction framework for evaluating rule-centered scenario reasoning. RuleWeaver starts from corpus-derived IF-THEN Meta Rules, progressively augments them into complex rules, and composes these rules into rule-centered scenario QA instances. Beyond final-answer correctness, RuleWeaver further supports process-level evaluation through rubric-based answer quality, rule recall, and rule precision. Experiments on 11 representative LLMs show that current models still struggle with complex rule-centered scenario reasoning, with even the best-performing model achieving only around 50% of the maximum rubric score. We make our code and dataset available here: https://github.com/SharkSpicy-NLP/RuleWeaver.
Chinese Translation
大型语言模型（LLMs）越来越多地应用于专业领域，其中有效利用领域专业知识通常需要在具体场景中对复杂规则进行推理。然而，现有基准测试仅部分评估了这一能力，因为它们要么侧重于输出级别的指令约束，要么忽视了规则在场景推理中所扮演的不同角色。为了解决这些问题，本文引入了RuleWeaver，一个用于评估规则中心场景推理的基准构建框架。RuleWeaver从语料库派生的IF-THEN元规则开始，逐步将其扩展为复杂规则，并将这些规则组合成规则中心的场景问答实例。除了最终答案的正确性，RuleWeaver还通过基于评分标准的答案质量、规则召回率和规则精确率支持过程级评估。在对11个代表性LLM的实验中，结果显示当前模型在复杂规则中心场景推理方面仍然存在困难，即使是表现最好的模型也仅获得了最大评分标准分数的约50%。我们的代码和数据集可在此获取：https://github.com/SharkSpicy-NLP/RuleWeaver。
cs.CL / 82 / 2608.26846
Evaluating Confidence-Gated Retrieval with Matched Trajectory Replay
评估带有置信门控的检索与匹配轨迹重放
Abstract
Interactive language-model agents use confidence signals to decide whether to answer immediately, retrieve additional evidence (from memory or external knowledge), or defer. Yet confidence is usually evaluated in isolation, without measuring the trajectory-level consequences of the actions it triggers. We propose matched trajectory replay, a controlled protocol for comparing confidence-to-action mappings. The protocol holds candidate answer states, evidence points, budgets, and action costs fixed. We use it to compare raw verbalized confidence with post-hoc isotonic calibration in a multi-hop question-answering system using Mistral, GPT, and Qwen models on HotpotQA and MuSiQue datasets. At the same numerical commitment threshold, calibration changes which questions agents ultimately commit to answering. Across all six model-dataset pairs, it increases accuracy among committed answers by up to 41 percentage points. However, it can reduce coverage and increase retrieval use. Overall accuracy improves by up to 15 percentage points on HotpotQA but falls by up to 17 percentage points on MuSiQue. These effects reflect a shift to a more selective, lower-risk operating point, not improved answers or confidence ranking. A calibration map fitted before retrieval improves held-out calibration through retrieval depths one and two, but is worse than raw confidence at depth three for all three models. Additional evidence helps on average, but this aggregate effect does not establish whether confidence identifies which individual episodes will benefit from another retrieval. Taken together, these results show that calibration can make commitment risk interpretable, but it does not estimate the expected benefit of another retrieval. Retrieval therefore requires a separate value-of-information or utility estimate. Evaluations should report held-out calibration, risk-coverage, and retrieval cost.
Chinese Translation
交互式语言模型代理使用置信信号来决定是立即回答、检索额外证据（来自记忆或外部知识）还是推迟。然而，置信通常是孤立评估的，没有衡量其触发的行动在轨迹层面的后果。我们提出了匹配轨迹重放，这是一种比较置信到行动映射的受控协议。该协议固定候选答案状态、证据点、预算和行动成本。我们利用该协议比较在 HotpotQA 和 MuSiQue 数据集上使用 Mistral、GPT 和 Qwen 模型的多跳问答系统中，原始口头置信与事后等距校准。在相同的数值承诺阈值下，校准改变了代理最终承诺回答的问题。在所有六个模型-数据集对中，它使承诺答案的准确性提高了最多 41 个百分点。然而，它可能减少覆盖率并增加检索使用。在 HotpotQA 上，整体准确性提高了最多 15 个百分点，但在 MuSiQue 上下降了最多 17 个百分点。这些效应反映出向更具选择性、低风险的操作点的转变，而不是改善答案或置信排名。在检索之前拟合的校准图在检索深度一和二时改善了保留校准，但在深度三时对所有三个模型而言，表现不如原始置信。额外证据平均上有所帮助，但这一综合效应并未确定置信是否能够识别哪些个体事件将从另一次检索中受益。综合来看，这些结果表明，校准可以使承诺风险可解释，但它并未估计另一次检索的预期收益。因此，检索需要单独的信息价值或效用估计。评估应报告保留校准、风险覆盖和检索成本。
cs.CL / 83 / 2608.26887
Planting a Latent Variable in Natural-Looking Text: a More Realistic Test of Belief States in LLMs and Their Link to Concept Geometry
在自然语言文本中植入潜变量：对大型语言模型（LLMs）信念状态及其与概念几何关系的更现实测试
Abstract
LLMs are thought to track "belief states," i.e., running probability distributions over the latent variables that govern language (Shai et al., 2024; Sarfati et al., 2026), but so far this has only been comprehensively demonstrated on toy synthetic data and in a few isolated case studies. It has also never been empirically connected to the geometry of LLM features (the concepts interpretability finds in model activations). In this work, we plant a controllable latent variable inside natural-looking text. An LLM teacher writes ordinary text while we "subliminally" steer it along one of K = 8 unrelated sparse autoencoder directions at each token, with the active directions following a ring-shaped Markov chain. A small transformer model trained on this corpus does indeed track the Bayesian posterior belief about our planted latent variable. Moreover, it also arranges the 8 states themselves on a ring, in the exact order of the Markov chain, which is supporting evidence that a concept's geometry can be formed by the statistical dynamics of the latent variable behind it.
Chinese Translation
大型语言模型（LLMs）被认为能够追踪“信念状态”，即对支配语言的潜变量的运行概率分布（Shai et al., 2024; Sarfati et al., 2026），但迄今为止，这一点仅在玩具合成数据和少数孤立案例研究中得到了全面验证。此外，它从未在经验上与LLM特征的几何（概念在模型激活中所找到的可解释性）联系起来。在本研究中，我们在自然语言文本中植入一个可控的潜变量。一位LLM教师在我们“潜意识”地引导其沿着K = 8个无关稀疏自编码器方向的每个标记时，写出普通文本，活跃的方向遵循一个环形马尔可夫链。经过在该语料库上训练的小型变换器模型确实追踪了我们植入的潜变量的贝叶斯后验信念。此外，它还将这8个状态本身按马尔可夫链的确切顺序排列在一个环上，这为概念几何可以通过其背后的潜变量的统计动态形成提供了支持证据。
cs.CL / 84 / 2608.26923
TabuLM: Morphology-Aware Tabular Pre-training for Low-Resource Languages
TabuLM：面向形态学的低资源语言表格预训练
Abstract
We present TabuLM, the first language model pre-trained on Kinyarwanda tabular data. Kinyarwanda is a morphologically rich Bantu language spoken by over 12 million people in Rwanda, yet lacks any dedicated tabular representation learning resource. TabuLM extends KinyaBERT-large, a two-tier morphological transformer, with additive row, column, and cell-type embeddings and a learned table-structure attention bias that sharpens same-row and same-column attention. Pre-training uses two new objectives: Masked Cell Recovery (MCR), which masks entire cells and forces reconstruction from row and column context, and Column Type Prediction (CTP), which predicts column semantic types from observed cell values. We pre-train on 172 Rwandan government tables (~35,000 cells) from NISR, RAB, REB, and MoH open-data portals, and introduce TabQA-kin, the first native Kinyarwanda table question-answering benchmark comprising 526 QA pairs across 31 tables and four question types. TabuLM achieves 62.0% exact match on TabQA-kin, outperforming KinyaBERT-large by 5.7 EM points and all multilingual baselines (mBERT 49.3%, XLM-R 50.0%) by 11.7-12.7 points. Analysis shows that structural table embeddings are most decisive for comparison and lookup questions, while morphological awareness provides complementary gains. Our code, data, and pre-trained checkpoint are publicly available.
Chinese Translation
我们提出了TabuLM，这是第一个在基尼亚卢旺达（Kinyarwanda）表格数据上进行预训练的语言模型。基尼亚卢旺达是一种形态丰富的班图语言，在卢旺达有超过1200万人使用，但缺乏专门的表格表示学习资源。TabuLM扩展了KinyaBERT-large，一个两层形态变换器，增加了附加的行、列和单元格类型嵌入，以及一种学习的表格结构注意力偏置，以增强同一行和同一列的注意力。预训练使用了两个新的目标：掩蔽单元恢复（Masked Cell Recovery, MCR），该目标掩蔽整个单元格并强制从行和列上下文中重建，以及列类型预测（Column Type Prediction, CTP），该目标根据观察到的单元格值预测列的语义类型。我们在来自卢旺达国家统计局（NISR）、农业和动物资源部（RAB）、教育部（REB）和卫生部（MoH）开放数据门户的172个卢旺达政府表格（约35,000个单元格）上进行了预训练，并引入了TabQA-kin，这是第一个本土基尼亚卢旺达表格问答基准，包含31个表格和四种问题类型的526个问答对。TabuLM在TabQA-kin上实现了62.0%的准确匹配，超越了KinyaBERT-large 5.7个EM点，并且在所有多语言基线（mBERT 49.3%，XLM-R 50.0%）上超出11.7-12.7个点。分析表明，结构表格嵌入对于比较和查找问题最为关键，而形态意识则提供了补充性收益。我们的代码、数据和预训练检查点已公开可用。
cs.CL / 85 / 2608.26925
Mapping Written Words to Spoken Words in a Different Language Using Only Visual Grounding
仅通过视觉基础将书面词映射到另一种语言的口语词
Abstract
In many low-resource settings, even just eliciting speech for data collection is difficult. One promising approach has been to ask speakers to describe images. But how do we build models from such visually grounded speech data? Given a dataset of images with Hindi spoken captions, we consider how we can map a written English keyword to spoken realisations of that word in Hindi. Previous work trained end-to-end multimodal neural models. Instead, we explore a simpler alignment-based approach built on self-supervised speech representations. Written English tags are automatically obtained from images using off-the-shelf image captioning systems. Hindi utterances associated with the same keyword are then aligned (using self-supervised features), and alignment evidence is aggregated to identify recurring speech segments corresponding to the target word. Experiments evaluating keyword spotting and localization show that our alignment-based approach outperforms a previous attention-based neural model. We also show the benefit of incorporating negative examples during alignment. Our work demonstrates that cross-lingual word-to-speech mappings can be learned directly from visual grounding without transcriptions or explicit model training.
Chinese Translation
在许多低资源环境中，仅仅为了数据收集而引发语音都是困难的。一种有前景的方法是让说话者描述图像。但是，我们如何从这样的视觉基础语音数据中构建模型呢？给定一个包含印地语口语字幕的图像数据集，我们考虑如何将书面英语关键词映射到该词在印地语中的口语实现。之前的研究训练了端到端的多模态神经模型。相反，我们探索了一种基于对齐的更简单的方法，该方法建立在自监督语音表示的基础上。书面英语标签通过现成的图像字幕系统自动从图像中获取。与同一关键词相关的印地语话语随后被对齐（使用自监督特征），并汇总对齐证据以识别与目标词对应的重复语音片段。评估关键词识别和定位的实验表明，我们的基于对齐的方法优于之前的基于注意力的神经模型。我们还展示了在对齐过程中纳入负例的好处。我们的研究表明，跨语言的词到语音映射可以直接从视觉基础中学习，而无需转录或明确的模型训练。
cs.CL / 86 / 2608.26941
KinyaEmbed: Contrastive Sentence Embeddings for Kinyarwanda via Multi-Stage Curriculum Training
KinyaEmbed：通过多阶段课程训练实现的基尼亚语对比句子嵌入
Abstract
We present KinyaEmbed, the first dedicated sentence embedding model for Kinyarwanda, a morphologically rich Bantu language spoken by over 12 million people in Rwanda. Existing multilingual embedding models such as LaBSE, mE5-large, and OpenAI text-embedding-3-large perform poorly on Kinyarwanda due to severe under-representation in their pre-training corpora. KinyaEmbed is built on KinyaBERT-large and trained via a four-stage curriculum using MultipleNegativesRankingLoss (MNRL): Stage 1 leverages ~18,000 paraphrase pairs from the Official Gazette of Rwanda with three temperature scales; Stage 2 fine-tunes on 715 NLLB-translated MNLI triplets for entailment structure; Stage 3 aligns representations using English-Kinyarwanda OPUS-100 translation pairs; Stage 4 refines with 2,936 high-quality pairs filtered from KinyaCOMET at quality threshold 0.8. We evaluate on SemRel2024-rw and introduce Wiki-RW-STS, a new contamination-free Kinyarwanda STS benchmark of 300 pairs derived from Kinyarwanda Wikipedia. A seven-checkpoint ensemble (all5+23A*2, with the final stage double-weighted) achieves Spearman \r{ho}=0.7298 on SemRel2024-rw, surpassing mE5-large by 20.9% and OpenAI text-embedding-3-large by 41.0%. KinyaEmbed also achieves the best document clustering silhouette score (0.2146) across all evaluated models. All checkpoints, the KinyaCOMET filtered pairs, and the Wiki-RW-STS benchmark are publicly available.
Chinese Translation
我们提出了KinyaEmbed，这是首个专门为基尼亚语设计的句子嵌入模型。基尼亚语是一种形态丰富的班图语言，在卢旺达有超过1200万人使用。现有的多语言嵌入模型如LaBSE、mE5-large和OpenAI text-embedding-3-large在基尼亚语上的表现较差，原因在于其预训练语料库中对基尼亚语的严重代表性不足。KinyaEmbed基于KinyaBERT-large构建，并通过四个阶段的课程训练使用多负样本排名损失（MultipleNegativesRankingLoss, MNRL）进行训练：第一阶段利用来自卢旺达官方公报的约18,000对同义句，采用三种温度尺度；第二阶段在715个NLLB翻译的MNLI三元组上进行微调，以捕捉蕴含结构；第三阶段使用英语-基尼亚语的OPUS-100翻译对对齐表示；第四阶段使用从KinyaCOMET中筛选出的2,936对高质量样本进行精炼，质量阈值为0.8。我们在SemRel2024-rw上进行了评估，并引入了Wiki-RW-STS，这是一个新的无污染基尼亚语STS基准，由300对来自基尼亚语维基百科的样本构成。一个七检查点的集成模型（all5+23A*2，最终阶段双重加权）在SemRel2024-rw上达到了Spearman
{ho}=0.7298，超越了mE5-large 20.9%和OpenAI text-embedding-3-large 41.0%。KinyaEmbed在所有评估模型中还实现了最佳文档聚类轮廓分数（0.2146）。所有检查点、KinyaCOMET筛选的样本对和Wiki-RW-STS基准均已公开提供。
cs.CL / 87 / 2608.26973
Squeezing More from Limited Data with Recursive Transformers
利用递归变换器从有限数据中挤出更多信息
Abstract
Pre-training under limited data requires a different view of scaling than web-scale language modeling. With a fixed data budget but relatively abundant compute, increasing parameter count helps only up to an optimal scale; beyond that point, models overfit and generalization worsens. We study this behavior across 10M-100M word pre-training budgets, two corpora, and multiple downstream evaluations, and find that optimal size depends strongly on both the data budget and the downstream target. We argue that standard Transformers scale down poorly to this setting, because embeddings consume a large fraction of the parameter budget and per-token computation is tied to representational capacity. To address this coupling, we study recursive Transformers, reusing a shared block across depth to scale compute, together with factorized embeddings to reduce vocabulary-map parameters. We train three recursive models and find that they outperform standard Transformers at 10M and 100M words, while remaining competitive with BabyLM Challenge 2025 winners.
Chinese Translation
在有限数据下的预训练需要与网络规模语言建模不同的扩展视角。在固定的数据预算下，尽管计算资源相对充足，增加参数数量仅在达到最佳规模时有效；超过这一点，模型会过拟合，泛化能力下降。我们在10M-100M词的预训练预算、两个语料库和多个下游评估中研究了这一行为，发现最佳规模在很大程度上依赖于数据预算和下游目标。我们认为，标准变换器在这一设置下的缩放效果较差，因为嵌入占用了大量的参数预算，而每个标记的计算与表示能力密切相关。为了解决这种耦合，我们研究了递归变换器，通过在深度上重用共享块来扩展计算，同时结合分解嵌入以减少词汇映射参数。我们训练了三个递归模型，发现它们在10M和100M词的情况下优于标准变换器，同时在与BabyLM Challenge 2025获胜者的竞争中保持了竞争力。
cs.CL / 88 / 2608.26982
JudgeStealer: Extracting LLM Judging Capabilities across Evaluation Protocols
JudgeStealer：跨评估协议提取大型语言模型判断能力
Abstract
Large language model (LLM) judges are increasingly used across various evaluation scenarios, making their judgment capabilities valuable intellectual property. However, black-box access exposes these capabilities to model extraction attacks. Existing extraction methods do not specifically target LLM judges and provide limited support for multiple evaluation protocols under restricted query budgets. In this study, we propose JUDGESTEALER, the first query-efficient model extraction framework for replicating judging capabilities across pointwise scoring, pairwise comparison, and listwise ranking protocols. JUDGESTEALER exploits the strong cross-protocol agreement to acquire pointwise scores and transform them into pairwise and listwise supervisions without additional victim queries. To capture informative judge patterns and improve query efficiency, JUDGESTEALER dynamically selects pointwise inputs based on semantic diversity, predictive uncertainty, and potential judge biases. It further applies score smoothing and multi-protocol review to preserve the ordinal structure of scores and mitigate catastrophic forgetting during surrogate adaptation. Extensive experiments on state-of-the-art LLM-as-a-judge and reward models show that JUDGESTEALER consistently outperforms existing extraction baselines, achieving up to 73.3%, 87.0%, and 71.6% accuracy for pointwise, pairwise, and listwise evaluation, respectively. JUDGESTEALER also remains effective across different sur- rogate model scales, adaptation strategies, and reasoning settings. Moreover, JUDGESTEALER demonstrates robustness against representative extraction defenses.
Chinese Translation
大型语言模型（LLM）评判者在各种评估场景中被越来越多地使用，使其判断能力成为宝贵的知识产权。然而，黑箱访问使这些能力暴露于模型提取攻击之下。现有的提取方法并未专门针对LLM评判者，并且在有限的查询预算下对多种评估协议的支持有限。在本研究中，我们提出了JUDGESTEALER，这是第一个高效的查询模型提取框架，用于复制点对评分、成对比较和列表排名协议下的判断能力。JUDGESTEALER利用跨协议的强一致性来获取点对评分，并在不增加额外受害者查询的情况下将其转化为成对和列表监督。为了捕捉信息丰富的评判模式并提高查询效率，JUDGESTEALER根据语义多样性、预测不确定性和潜在的评判偏差动态选择点对输入。它进一步应用评分平滑和多协议审查，以保持评分的序数结构，并在代理适应过程中减轻灾难性遗忘。在对最先进的LLM作为评判者和奖励模型进行的广泛实验中，JUDGESTEALER始终优于现有的提取基线，在点对、成对和列表评估中分别达到了73.3%、87.0%和71.6%的准确率。JUDGESTEALER在不同的代理模型规模、适应策略和推理设置下也保持有效。此外，JUDGESTEALER在面对代表性的提取防御时表现出鲁棒性。
cs.CL / 89 / 2608.27031
ITL: Interpretable Document Alignment with Structured Reference Frameworks
ITL：基于结构化参考框架的可解释文档对齐
Abstract
Measuring alignment between documents and structured reference frameworks requires identifying conceptual evidence distributed throughout the text and reporting it through measures that are quantitative, interpretable, and traceable. Many commonly used retrieval and classification approaches return either pairwise similarity scores or one or more class labels, whereas fewer methods provide concept-level scores that are directly traceable to the terminological evidence supporting them. We present \emph{Intelligent Target Locator} (ITL), a domain-agnostic and language-portable methodology that estimates the affinity between the textual units of a target document and the concepts defined in a \emph{Structured Reference Document} ($SRD$). From the $SRD$, ITL induces concept-specific terminological profiles built from independent terms, bigrams, trigrams, and co-occurrences. Each term is assigned an importance weight that combines concept membership, term-type specificity and inter-concept discriminability. The output is a textual-unit--concept affinity matrix that can be aggregated at different levels of granularity. We conduct an internal consistency assessment using the 17 Sustainable Development Goals (SDGs), evaluating each official goal statement against the $SRD$ induced from the same set of descriptors. Every statement reached its highest affinity with the corresponding concept, and the mean affinity across the remaining concepts stayed marginal relative to the mean reference affinity. This separation indicates that ITL distinguishes the conceptual profiles of the framework. ITL thus offers a general basis for quantifying document alignment with structured frameworks while keeping each result traceable to the terminological evidence that supports it.
Chinese Translation
测量文档与结构化参考框架之间的对齐需要识别分布在文本中的概念证据，并通过定量、可解释和可追溯的度量进行报告。许多常用的检索和分类方法返回的要么是成对相似度评分，要么是一个或多个类别标签，而提供直接可追溯到支持它们的术语证据的概念级评分的方法则较少。我们提出了 extit{智能目标定位器}（Intelligent Target Locator，ITL），这是一种领域无关且语言可移植的方法，旨在估计目标文档的文本单元与 extit{结构化参考文档}（Structured Reference Document，$SRD$）中定义的概念之间的亲和性。ITL从$SRD$中诱导出特定概念的术语特征，构建独立术语、二元组、三元组和共现的概念特征。每个术语被赋予一个重要性权重，该权重结合了概念成员资格、术语类型特异性和概念间可区分性。输出结果是一个文本单元与概念的亲和性矩阵，可以在不同的粒度水平上进行聚合。我们使用17个可持续发展目标（Sustainable Development Goals，SDGs）进行内部一致性评估，评估每个官方目标声明与从相同描述符集诱导的$SRD$之间的关系。每个声明与相应概念的亲和性达到最高，而其余概念的平均亲和性相对于平均参考亲和性保持在边际水平。这种分离表明ITL能够区分框架的概念特征。因此，ITL为量化文档与结构化框架之间的对齐提供了一般基础，同时保持每个结果可追溯到支持它的术语证据。
cs.CL / 90 / 2608.27035
Representing and Parsing Korean Constituency Structure at Different Levels of Granularity
在不同粒度水平下表示和解析韩语成分结构
Abstract
Korean constituency parsing raises a representational challenge because the terminal units of a phrase-structure tree do not straightforwardly correspond to simple surface words. Korean eojeols are morphologically complex spacing units, and existing constituency resources differ in how they represent eojeol-internal morphology and non-overt elements. This paper compares three constituency parsing representations derived from the Penn Korean Treebank: Morpheme+XPOS, Eojeol+XPOS, and Eojeol+UPOS. We construct these representations by removing null elements, aligning Penn Korean phrase structure with overt eojeol tokens, preserving Penn Korean phrase labels where possible, and varying the terminal and preterminal layers. We then evaluate canonical non-binary transition-based constituency parsers in top-down, in-order, and bottom-up orders under a shared modeling and evaluation setup. All experiments use gold terminal segmentation and gold preterminal labels and therefore evaluate constituency parsing conditioned on gold morphosyntactic annotation. Eojeol terminals yield shorter transition sequences, but Eojeol+UPOS parsing substantially underperforms the morphologically richer conditions. Eojeol+XPOS narrows this gap, while Morpheme+XPOS gives the strongest results even after its predictions are projected to the eojeol terminal domain. Under these gold-annotation conditions, the results show that fine-grained morphological and XPOS representations provide valuable evidence for the evaluated parsers. This empirical finding concerns the information available for parsing and does not by itself determine the linguistically preferable terminal domain. Independently, linguistic and resource-design considerations motivate eojeol as a stable and interpretable surface domain for phrase-structure annotation, with morpheme-level and XPOS information retained as aligned morphosyntactic evidence.
Chinese Translation
韩语成分解析面临一个表示性挑战，因为短语结构树的终端单元并不直接对应于简单的表面词汇。韩语的eojeol是形态复杂的间隔单位，现有的成分资源在如何表示eojeol内部形态和非显性元素方面存在差异。本文比较了来自宾州韩语树库的三种成分解析表示：Morpheme+XPOS、Eojeol+XPOS和Eojeol+UPOS。我们通过去除空元素、将宾州韩语短语结构与显性eojeol标记对齐、尽可能保留宾州韩语短语标签以及改变终端和预终端层来构建这些表示。然后，我们在共享的建模和评估设置下，评估经典的非二元基于转移的成分解析器，包括自顶向下、顺序和自底向上的解析顺序。所有实验均使用金标准终端分割和金标准预终端标签，因此评估的是基于金标准形态句法注释的成分解析。Eojeol终端产生了更短的转移序列，但Eojeol+UPOS解析的表现明显低于形态更丰富的条件。Eojeol+XPOS缩小了这一差距，而Morpheme+XPOS即使在其预测投影到eojeol终端域后仍表现出最强的结果。在这些金标准注释条件下，结果表明细粒度的形态和XPOS表示为评估的解析器提供了有价值的证据。这个实证发现涉及解析可用的信息，但本身并不决定语言学上更可取的终端域。独立地，语言学和资源设计的考虑促使将eojeol作为短语结构注释的稳定且可解释的表面域，同时保留形态层级和XPOS信息作为对齐的形态句法证据。
cs.CL / 91 / 2608.27036
Reasoning about In-Context Samples for Machine-Translation
关于机器翻译中的上下文样本推理
Abstract
Large Language Models (LLMs) can be trained to perform chain-of-thoughts reasoning in order to improve the reliability of their responses. In this work, we investigate how explicit reasoning can be leveraged for LLM-Based Machine Translation (MT) with in-context samples. We introduce a novel fragment-based reasoning framework in which the model first extracts parallel source-target fragments from retrieved similar exemplars, and uses these fragments as intermediate reasoning traces to produce the final translation. To train our model, we distill silver fragments and drafts from a large teacher model. Our experiments with the Qwen3 model family, over 6 languages, including up to 5 domains per language, demonstrate that fragment-based MT significantly outperforms alternative methods like standard k-shot or basic drafting.
Chinese Translation
大型语言模型（LLMs）可以通过训练进行连锁思维推理，以提高其响应的可靠性。在本研究中，我们探讨了如何利用显式推理来增强基于LLM的机器翻译（MT），特别是在上下文样本的情况下。我们提出了一种新颖的基于片段的推理框架，其中模型首先从检索到的相似示例中提取平行的源-目标片段，并使用这些片段作为中间推理痕迹来生成最终翻译。为了训练我们的模型，我们从一个大型教师模型中提炼出银片段和草稿。我们在Qwen3模型系列上进行了实验，覆盖6种语言，包括每种语言最多5个领域，结果表明，基于片段的机器翻译显著优于标准的k-shot或基本草拟等替代方法。
cs.CL / 92 / 2608.27038
Cascaded Batch Prompting
级联批量提示
Abstract
Although batch prompting makes large language model inference more efficient by processing multiple instances simultaneously, it suffers from unpredictable downstream task performance. We propose cascaded batch prompting, a two-stage approach designed to resolve the unpredictability of conventional batch prompting by disentangling complex reasoning from symbol grounding. Experiments on multiple-choice question answering and natural language inference demonstrate that the proposed method outperforms the standard single prompting baseline while achieving a speedup proportional to batch size, establishing a new state of the art on the Pareto frontier.
Chinese Translation
尽管批量提示通过同时处理多个实例使大型语言模型的推理更加高效，但其在下游任务性能上存在不可预测性。我们提出了级联批量提示，这是一种两阶段的方法，旨在通过将复杂推理与符号基础分离来解决传统批量提示的不可预测性。在多项选择问答和自然语言推理的实验中，所提出的方法在性能上超越了标准的单一提示基线，同时实现了与批量大小成正比的加速，确立了在帕累托前沿上的新状态。
cs.CL / 93 / 2608.27049
Research Design Tracking and Assessment for the Social Sciences
社会科学研究设计的追踪与评估
Abstract
Reliable assessment of causal research designs in the social sciences is critical for evidence-based policy-making, yet has so far relied entirely on manual expert analysis. We introduce Automated Research Design Tracking and Assessment (ARDTrA), a task that involves detecting the research design used in a paper and assessing the quality of its application. We create an expert-annotated dataset of papers covering six families of counterfactual research designs and evaluate the task using a multi-turn RAG-based conversational pipeline. Across four retrieval strategies, four LLMs and six embedding models, we find that passage length is the main driver of performance, explaining 52-66% of the variance. A per-research-design analysis also shows that human and machine difficulty do not align: the designs that prove hardest for the system are not those on which expert annotators disagree most, pointing to two independent sources of task difficulty.
Chinese Translation
在社会科学中，可靠的因果研究设计评估对于基于证据的政策制定至关重要，但迄今为止完全依赖于人工专家分析。我们提出了自动化研究设计追踪与评估（Automated Research Design Tracking and Assessment，ARDTrA），这一任务涉及检测论文中使用的研究设计并评估其应用质量。我们创建了一个专家标注的数据集，涵盖六类反事实研究设计，并使用基于多轮检索增强生成（RAG）的对话管道对该任务进行评估。在四种检索策略、四种大型语言模型（LLM）和六种嵌入模型的测试中，我们发现段落长度是性能的主要驱动因素，解释了52-66%的方差。对每种研究设计的分析还表明，人工与机器的难度并不一致：系统最难处理的设计并不是专家标注者意见分歧最大的设计，这指向了任务难度的两个独立来源。
cs.CL / 94 / 2608.27110
DocTalkBN: A Novel Dataset of Expert Telemedicine Conversations in Bengali
DocTalkBN：孟加拉语专家远程医疗对话的新数据集
Abstract
Reliable medical conversational AI requires authentic expert--patient interaction data, yet such datasets remain scarce, especially for low-resource languages such as Bengali. We present DocTalkBN, a large-scale multimodal dataset of real-world expert telemedicine conversations in Bengali, collected from nationally broadcast telemedicine programs featuring board-certified physicians. DocTalkBN contains 557.63 hours of paired audio and text, 1,515 multi-turn patient calls, 10,274 host--doctor question--answer exchanges, totaling 1.7M tokens, spanning 26 medical specialties. Unlike prior resources derived from medical forums, written health content, or synthetic data, our dataset preserves the spontaneity, contextual richness, and spoken characteristics of authentic medical interactions in a low-resource setting. To support benchmark-driven research, we further construct three downstream tasks from the corpus, medical triage classification, advice safety evaluation, and medical named entity recognition, and benchmark a diverse set of large language models and encoder-based baselines. Our results show that DocTalkBN is a practically useful resource, particularly for clinically grounded reasoning tasks. We release this resource to facilitate future research on reliable medical NLP and safer, more culturally grounded healthcare systems for low-resource languages. Our source codes and dataset are publicly available at https://anonymous.4open.science/r/doctalk.
Chinese Translation
可靠的医疗对话人工智能需要真实的专家与患者互动数据，然而此类数据集仍然稀缺，尤其是在孟加拉语等低资源语言中。我们提出了DocTalkBN，这是一个大规模的多模态数据集，包含来自全国广播的远程医疗节目中真实的孟加拉语专家对话，参与者为获得认证的医生。DocTalkBN包含557.63小时的配对音频和文本，1,515个多轮患者电话，10,274个主持人-医生问答交流，总计1.7百万个标记，涵盖26个医学专业。与以往来源于医疗论坛、书面健康内容或合成数据的资源不同，我们的数据集保留了低资源环境中真实医疗互动的自发性、上下文丰富性和口语特征。为了支持基准驱动的研究，我们进一步从语料库构建了三个下游任务：医疗分诊分类、建议安全性评估和医学命名实体识别，并对一组多样的大型语言模型和编码器基线进行了基准测试。我们的结果表明，DocTalkBN是一个实用的资源，特别适用于临床基础推理任务。我们发布这一资源以促进未来在可靠医疗自然语言处理和更安全、更具文化基础的低资源语言医疗系统方面的研究。我们的源代码和数据集可在 https://anonymous.4open.science/r/doctalk 上公开获取。
cs.CL / 95 / 2608.27115
Cross-Lingual Alignment Without Joint Training: Do Monolingual Language Models Converge on Universal Representations?
无联合训练的跨语言对齐：单语语言模型是否收敛于通用表示？
Abstract
Cross-lingual alignment in multilingual language models is typically attributed to joint training: shared parameters, mixed-language batches, or explicit alignment objectives. We ask whether monolingual models trained on non-parallel data learn alignable representations without joint training. By testing on strictly monolingual language models, such as the Goldfish model families and independently developed models from different research labs, we find three results. Correlation: these models develop alignable representational geometry across layers, with alignment strengthening as data scale, model scale, or linguistic proximity increases. Construction: a single Procrustes rotation fit on parallel sentences maps hidden states between models. Causation: the same rotation transfers functional content; patching a rotated English residual into a German model on a factual cloze flips the prediction to the donor's capital in most cases. We confirm that cross-lingual alignment can emerge from the structure of language and the information it carries rather than from joint training, and this points to practical future directions including model stitching, merging, and modular multilingual systems built from monolingual components.
Chinese Translation
多语言模型中的跨语言对齐通常归因于联合训练：共享参数、混合语言批次或明确的对齐目标。我们探讨了在非平行数据上训练的单语模型是否能够在没有联合训练的情况下学习可对齐的表示。通过对严格的单语语言模型进行测试，例如 Goldfish 模型系列和来自不同研究实验室独立开发的模型，我们得出了三个结果。相关性：这些模型在各层之间发展出可对齐的表示几何，随着数据规模、模型规模或语言接近度的增加，对齐效果增强。构造：在平行句子上拟合的单个 Procrustes 旋转能够映射模型之间的隐藏状态。因果性：相同的旋转转移功能内容；在事实填空任务中，将旋转后的英语残差嵌入德语模型，通常会将预测结果翻转为捐赠者的首都。我们确认跨语言对齐可以从语言的结构及其所承载的信息中自然产生，而不是来自联合训练，这为未来的实际方向提供了启示，包括模型拼接、合并以及由单语组件构建的模块化多语言系统。
cs.CL / 96 / 2608.27128
TwinKV: A Composable Repair Pass for KV Cache Eviction via Pairwise Key Redundancy
TwinKV：通过成对键冗余实现KV缓存驱逐的可组合修复通道
Abstract
Long-context inference is bottlenecked by the memory footprint of the key-value (KV) cache, especially for small models under tight resource budgets. Existing KV cache eviction methods score tokens using the model's attention distribution or, in attention-free variants, each key's distance from a global reference point. Using a controlled leave-one-out probe, we find that attention magnitude is unrelated to a token's causal contribution to the answer (Spearman $\rho=-0.004$), challenging the premise behind dominant eviction methods. We introduce TwinKV, a training-free, attention-free redundancy signal that detects whether a token's key has a near-duplicate elsewhere in context. Rather than replacing existing policies, TwinKV acts as a composable repair pass: given a policy's fixed retained set, it identifies evicted tokens with no surviving duplicate (\emph{orphans}) and retained tokens whose information is duplicated elsewhere (\emph{redundant donors}), then swaps them while preserving the original budget and scoring rule. We compose TwinKV with four recent eviction policies across LongBench, LooGLE, RULER, and a short-context MMLU-Pro no-harm control at compression ratios ${0.3,0.5,0.7}$. On Qwen3-4B, TwinKV improves a majority of configurations for two policies, is near-even for a third, and helps only a minority for a fourth adaptive baseline already near a performance ceiling; gains across the three non-ceiling policies are smallest at the loosest ratio. On RULER with Llama-3.2-1B, however, that fourth policy improves in every evaluated cell because its Alone score leaves substantial room to improve. More broadly, Llama-3.2-1B shows a smaller average LongBench gain but a higher fraction of improved cells on LongBench and LooGLE than Qwen3-4B, plus a clean RULER win. We also identify few-shot classification exemplars as a task structure where TwinKV does not help on either model.
Chinese Translation
长上下文推理受到键值（KV）缓存内存占用的瓶颈，尤其是在资源预算紧张的小模型下。现有的KV缓存驱逐方法通过模型的注意力分布对令牌进行评分，或者在无注意力变体中，根据每个键与全局参考点的距离进行评分。通过控制的逐个排除探测，我们发现注意力大小与令牌对答案的因果贡献无关（Spearman $
ho=-0.004$），这挑战了主流驱逐方法的前提。我们提出了TwinKV，这是一种无训练、无注意力的冗余信号，用于检测令牌的键在上下文中是否存在近似重复。TwinKV并不替代现有策略，而是作为可组合的修复通道：在给定策略的固定保留集合下，它识别出没有存活重复的驱逐令牌（ extit{孤儿}）和信息在其他地方重复的保留令牌（ extit{冗余捐赠者}），然后在保持原始预算和评分规则的同时进行交换。我们将TwinKV与四种最近的驱逐策略组合，涵盖LongBench、LooGLE、RULER，以及在压缩比${0.3,0.5,0.7}$下的短上下文MMLU-Pro无害控制。在Qwen3-4B上，TwinKV改善了两种策略的大多数配置，对于第三种策略接近持平，而对第四种已经接近性能上限的自适应基线仅有少数帮助；在三个非上限策略中，增益在最宽松的比率下最小。然而，在Llama-3.2-1B的RULER上，这第四种策略在每个评估单元中都有所改善，因为其独立评分留有相当大的提升空间。更广泛地说，Llama-3.2-1B在LongBench上的平均增益较小，但在LongBench和LooGLE上改善单元的比例高于Qwen3-4B，并且在RULER上取得了明显的胜利。我们还确认少样本分类示例作为一种任务结构，在这两种模型上TwinKV均无助于提升。
cs.CL / 97 / 2608.27135
Said Aloud, Read Different: Cross-Modal Instability in Multimodal Models
言之有物，读之不同：多模态模型中的跨模态不稳定性
Abstract
Multimodal foundation models are increasingly used in speech-first assistants that must interpret spoken queries and produce visually grounded decisions. Yet it remains unclear whether semantically equivalent queries yield consistent judgments across modality (text vs. speech) and language (English vs. Arabic). We introduce a speech-augmented visually grounded contrastive triplet benchmark spanning 10,150 culturally grounded images from 18 MENA countries, where each image is paired with one supported statement and two plausible but unsupported alternatives. We define contrastive instability as the conditional rate at which a model fails to resolve all statements within a triplet, isolating fragmented reasoning from complete failure. Evaluating recent multimodal models under text and speech in English and Arabic, we find that modality and language shifts introduce substantial triplet-level inconsistencies that are not fully captured by aggregate accuracy, with speech amplifying partial failures. We make the benchmark publicly available to the community.
Chinese Translation
多模态基础模型在语音优先的助手中被越来越广泛地使用，这些助手必须解释口头查询并做出视觉上有依据的决策。然而，目前尚不清楚语义上等价的查询在不同模态（文本与语音）和语言（英语与阿拉伯语）之间是否会产生一致的判断。我们引入了一种语音增强的视觉对比三元组基准，涵盖来自18个中东和北非（MENA）国家的10,150幅文化相关图像，每幅图像都配有一个支持性陈述和两个合理但不支持的替代选项。我们将对比不稳定性定义为模型未能解决三元组中所有陈述的条件概率，从而将碎片化推理与完全失败区分开来。在对近期多模态模型进行英语和阿拉伯语的文本与语音评估时，我们发现模态和语言的转换引入了显著的三元组级不一致性，而这种不一致性并未完全通过总体准确率来捕捉，语音的使用加剧了部分失败。我们将该基准公开提供给社区。
cs.CL / 98 / 2608.27161
STAR : Sentence Translation Alignment Rate for Document-to-Document Machine Translation
STAR：文档对文档机器翻译的句子翻译对齐率
Abstract
Large Language Models (LLMs) have enabled a shift from sentence-level to document-to-document (Doc2Doc) machine translation, promising improved global coherence. However, document-to-document generation in a single pass frequently suffers from structural misalignment, manifesting as sentence omissions or hallucinations that violate the core requirement of source-target correspondence. To address this, we introduce Sentence Translation Alignment Rate (STAR), an auxiliary metric that explicitly quantifies sentence-level structural fidelity. Building on this, we propose STAR-masked Preference Optimization (StarPO), a framework that ranks document-level hypotheses by structural quality and utilizes a dynamic alignment mask to focus optimization on misaligned segments. Experimental results across news and literary domains demonstrate that StarPO significantly enhances translation quality and structural integrity. Notably, StarPO allows compact models to surpass the performance of massive proprietary systems like GPT-4o while maintaining superior token efficiency.
Chinese Translation
大型语言模型（LLMs）使得从句子级别向文档对文档（Doc2Doc）机器翻译的转变成为可能，承诺提高整体连贯性。然而，单次生成的文档对文档翻译常常遭遇结构不对齐的问题，表现为句子遗漏或幻觉，这违反了源-目标对应的核心要求。为了解决这一问题，我们引入了句子翻译对齐率（Sentence Translation Alignment Rate，STAR），这一辅助指标明确量化了句子级别的结构保真度。在此基础上，我们提出了STAR掩蔽偏好优化（STAR-masked Preference Optimization，StarPO）框架，该框架通过结构质量对文档级假设进行排序，并利用动态对齐掩码将优化重点放在不对齐的片段上。跨新闻和文学领域的实验结果表明，StarPO显著提升了翻译质量和结构完整性。值得注意的是，StarPO使得紧凑型模型在保持优越的令牌效率的同时，超越了如GPT-4o等大型专有系统的性能。
cs.CL / 99 / 2608.27165
Prediction of Prediction (PoP): Inter-Layer Activation Fusion for Single-Pass Hallucination Detection in Large Language Models
预测的预测（PoP）：用于大语言模型单次通行的幻觉检测的层间激活融合
Abstract
Autoregressive large language models (LLMs) routinely generate factually incorrect outputs with high decoding confidence, limiting their deployment in high-stakes workflows. Existing output-stage uncertainty metrics can fail when models are overconfident on false assertions, while multi-sample verification pipelines introduce substantial memory and latency overhead. This work evaluates whether internal hidden-state transition dynamics during generation can signal factual errors without auxiliary decoding calls. We introduce Prediction of Prediction (PoP), a mechanism that captures layer-transition uncertainty by fusing intermediate hidden representations across depth during a single forward pass. Evaluated on the TruthfulQA benchmark using autoregressive transformer backbones, PoP achieves an area under the receiver operating characteristic curve (AUROC) of 75.5% for factual-correctness classification. The mechanism operates within the base forward pass, adding less than 1.2% runtime latency and requiring zero additional generation passes. The numerical results are reported from the author-verified experimental implementation and are bounded by the evaluation scope described below.
Chinese Translation
自回归大语言模型（LLMs）常常以高解码置信度生成事实不正确的输出，这限制了它们在高风险工作流程中的应用。现有的输出阶段不确定性指标在模型对错误断言过于自信时可能失效，而多样本验证管道则会引入显著的内存和延迟开销。本研究评估了生成过程中内部隐状态转移动态是否能够在不进行辅助解码调用的情况下指示事实错误。我们提出了预测的预测（PoP）机制，通过在单次前向传递中融合深度上的中间隐表示，捕捉层间转移的不确定性。在使用自回归变换器骨干网络的TruthfulQA基准上评估，PoP在事实正确性分类中实现了75.5%的接收者操作特征曲线下面积（AUROC）。该机制在基础前向传递中运行，增加的运行时延迟不足1.2%，且不需要额外的生成传递。数值结果来自作者验证的实验实现，并受到以下描述的评估范围的限制。
cs.CL / 100 / 2608.27176
When Text Misleads: Inconsistent-Aware Reasoning for Audio-Grounded Dialogue
当文本误导时：面向音频基础对话的不一致感知推理
Abstract
Understanding spoken dialogue requires joint reasoning over lexical content and paralinguistic acoustic signals such as emotion and conversational intent. However, existing evaluations often allow shortcuts based on transcripts or single-modality solutions, obscuring whether models genuinely ground predictions in speech. We formalize this failure mode as cross-modal disagreement, where transcripts suggest plausible but incorrect surface interpretations while acoustic cues such as prosody or speaking style support different answers. We develop a scalable framework that identifies text-biased surface interpretations and converts disagreement regions into conflict QA examples. We also include consistent cases where transcript-based and speech-grounded interpretations agree, enabling evaluation beyond adversarial audio dependence. This results in ContraTalk, a controlled benchmark containing 501 questions across five discourse dimensions: interaction behavior, emotion state, dialogue act, social stance, and conversational intent. We further develop an agentic-style reasoning framework that converts speech into an Audio Twin, a text-readable representation of localized acoustic cues that exposes acoustic evidence to the reasoning model. Experiments show that strong text-only LLMs exceed 90% accuracy in consistent cases but drop to 33-48% in conflict cases. Direct AudioLLMs provide only partial grounding, still selecting the transcript-biased trap in roughly 30-40% of conflict cases. Our Audio Twin framework improves conflict-case accuracy while reducing trap selection, but its consistent-case behavior remains backbone-dependent. These results identify transcript-based shortcuts as an important failure mode in spoken dialogue understanding and show that explicit acoustic evidence aggregation provides a more controllable interface for diagnosing and improving speech-grounded reasoning.
Chinese Translation
理解口语对话需要对词汇内容和副语言声学信号（如情感和对话意图）进行联合推理。然而，现有评估往往允许基于转录文本或单一模态解决方案的捷径，这掩盖了模型是否真正将预测与语音相结合。我们将这种失败模式形式化为跨模态不一致，其中转录文本建议了看似合理但不正确的表面解释，而声学线索（如韵律或说话风格）则支持不同的答案。我们开发了一个可扩展的框架，识别文本偏向的表面解释，并将不一致区域转化为冲突问答示例。我们还包括了一致案例，其中基于转录的解释和基于语音的解释一致，从而使评估超越对抗性音频依赖。这导致了ContraTalk的产生，这是一个包含501个问题的受控基准，涵盖五个话语维度：互动行为、情感状态、对话行为、社会立场和对话意图。我们进一步开发了一种代理风格的推理框架，将语音转换为音频双胞胎（Audio Twin），这是对局部声学线索的文本可读表示，向推理模型暴露声学证据。实验表明，强大的仅文本大型语言模型（LLMs）在一致案例中超过90%的准确率，但在冲突案例中降至33-48%。直接的音频大型语言模型（AudioLLMs）仅提供部分基础，仍在大约30-40%的冲突案例中选择了偏向转录的陷阱。我们的音频双胞胎框架提高了冲突案例的准确性，同时减少了陷阱选择，但其在一致案例中的表现仍依赖于基础模型。这些结果将基于转录的捷径识别为口语对话理解中的一个重要失败模式，并表明显式声学证据聚合提供了一个更可控的接口，用于诊断和改善基于语音的推理。
cs.CL / 101 / 2608.27219
BALMS: Benchmarking Agentic LLMs for Longitudinal Mental Health Sensing
BALMS：基于代理的长时间心理健康感知的基准测试
Abstract
Mental health assessment relies on episodic self-report scales, which convert subjective states such as stress into numerical scores but provide only sparse snapshots of wellbeing. Wearable devices offer longitudinal behavioral and physiological signals for continuous, low-burden monitoring. Recent LLM-driven personal-health agents enable natural language queries over wearable signals, but mainly handle short-term, retrieval-based lookups (e.g., highest step count over a week). They do not evaluate whether agents can reason over long-term signals to predict wellbeing scores paired with evidence-grounded rationales. To address this gap, we introduce BALMS, the first systematic benchmark of LLM-based agentic systems for longitudinal mental health sensing. BALMS spans 3 real-world longitudinal datasets, 2 task families (closed-form wellbeing-score prediction and rationale generation auto-graded by an LLM-as-Judge), 3 agentic paradigms evaluated across 5 open- and closed-source LLM backbones. We find that zero-shot agents rarely outperform a simple mean baseline, except with stronger backbones or compact, semantically meaningful features. Chain-of-thought prompting improves reasoning-oriented backbones, but does not guarantee temporal grounding or numerical correctness. Together with more analysis on efficiency and temporal scaling, BALMS highlights the need for longitudinal mental health agents that selectively retrieve history, ground temporal evidence, and reason over interpretable behavioral features.
Chinese Translation
心理健康评估依赖于情节自我报告量表，这些量表将压力等主观状态转化为数值评分，但仅提供稀疏的幸福感快照。可穿戴设备提供了连续、低负担的行为和生理信号，用于长期监测。最近的基于大语言模型（LLM）的个人健康代理能够对可穿戴信号进行自然语言查询，但主要处理短期的基于检索的查找（例如，一周内的最高步数）。它们并未评估代理是否能够对长期信号进行推理，以预测与证据基础的理由相结合的幸福感评分。为了解决这一空白，我们引入了BALMS，这是第一个系统性的基于LLM的代理系统的长时间心理健康感知基准。BALMS涵盖了3个真实世界的长期数据集、2个任务类别（封闭形式的幸福感评分预测和由LLM作为评判者自动评分的理由生成）、3种代理范式，并在5个开源和闭源的LLM基础上进行评估。我们的研究发现，零-shot代理很少超越简单的均值基线，除非使用更强的基础模型或紧凑且语义明确的特征。思维链提示改善了面向推理的基础模型，但并不保证时间上的准确性或数值的正确性。结合对效率和时间扩展的更多分析，BALMS突显了需要长期心理健康代理，这些代理能够选择性地检索历史、基于时间的证据，并对可解释的行为特征进行推理。
cs.CL / 102 / 2608.27265
SCIT: Testing Causal Cache Carriers in Latent Chain-of-Thought Models
SCIT：在潜在思维链模型中测试因果缓存载体
Abstract
Latent chain-of-thought models move intermediate reasoning from emitted text into continuous states, improving compactness but hiding the causal object. We introduce SCIT, the Suffix Cache Interchange Test, a causal protocol that constructs exact source-recipient counterfactuals, patches declared cache segments, and identifies which transformer object carries the counterfactual computation. SCIT combines sufficiency tests with K/V component splits, hidden-state controls, semantic source controls, decoded validation, and matched corruption. On CODI-GPT2 and a Sim-CoT-style GPT-2 reproduction, counterfactual arithmetic transfers primarily through value-cache suffix trajectories rather than hidden states, keys, reusable answer slots, or single-token triggers. Complete sufficiency-and-necessity evidence for the late-value-suffix mechanism holds for the main CODI-GPT2 checkpoint; the Sim-CoT-style checkpoint shows the same sufficiency and decoded-control pattern but insufficient matched-corruption evidence for a necessity call. Beyond these local arithmetic cells, SCIT reveals carrier-regime shifts: arithmetic-like GPT-2/1B cells preserve latent-tail value/KV transfer, whereas competent 8B and repaired non-arithmetic cells route through prompt-prefix or full-cache K/V; boundary cells receive no mechanism call. SCIT therefore contributes a cache-level diagnostic, a checkpoint-specific GPT-2 arithmetic mechanism, and a competence-gated carrier map rather than a universal latent-tail claim.
Chinese Translation
潜在思维链模型将中间推理从发出的文本转移到连续状态中，提升了紧凑性但隐藏了因果对象。我们提出了SCIT，即后缀缓存互换测试（Suffix Cache Interchange Test），这是一种构建精确源-接收者反事实的因果协议，修补声明的缓存段，并识别出哪个变换器对象承载了反事实计算。SCIT结合了充分性测试与K/V组件分割、隐藏状态控制、语义源控制、解码验证和匹配腐蚀。在CODI-GPT2和Sim-CoT风格的GPT-2重现中，反事实算术主要通过值缓存后缀轨迹传递，而非隐藏状态、键、可重用答案槽或单个令牌触发器。对于主要的CODI-GPT2检查点，后值后缀机制的充分性和必要性证据是完整的；Sim-CoT风格的检查点显示出相同的充分性和解码控制模式，但在必要性调用方面缺乏足够的匹配腐蚀证据。超越这些局部算术单元，SCIT揭示了载体机制的转变：类算术的GPT-2/1B单元保留潜在尾部值/KV传递，而有能力的8B和修复的非算术单元则通过提示前缀或完整缓存K/V进行路由；边界单元没有接收到机制调用。因此，SCIT贡献了一个缓存级别的诊断、一个特定于检查点的GPT-2算术机制，以及一个能力门控的载体图，而不是一个普遍的潜在尾部声明。
cs.CL / 103 / 2608.27309
Difference-in-Differences on a Censored Rating Scale Can Manufacture an Effect: Evidence from a Pre-Registered LLM-Judge Audit
受限评分尺度上的差异中的差异可以制造效果：来自预注册的LLM评审审计的证据
Abstract
Audits of LLM judges certify a bias by contrasting matched conditions, and the strongest designs difference twice: a within-item contrast between two candidate responses, differenced again across a manipulated attribute, read off a bounded rating scale. We show that this endpoint is not identified on the scale that reports it. Each term of the double difference is censored by its own share, so the observed statistic confounds differential preference with differential attenuation: a severity shift common to both responses manufactures an interaction whenever the two censor it unequally, as unequal distances from the bounds make them, exactly where good stimuli place them. We exhibit the failure inside a pre-registered audit of a frozen pedagogy judge, sealed before the first of its 990 calls. The registered primary endpoint, the effect of a stated learner profile on the judge's scaffolding preference, is null: $+0.085$ points (95\% BCa $[-0.167, +0.353]$, $p = 0.684$). The audit's one nominally significant interaction, $+0.378$ ($p = 0.002$), is not identified as preference: a construction containing zero differential preference reproduces 79 to 85\% of it from the observed severity shift and the scale floor alone. We derive the mechanism in closed form and show that its contribution is measurable from an audit's own ratings.
Chinese Translation
对LLM评审的审计通过对比匹配条件来证明偏见，而最强的设计则进行两次差异化：在两个候选响应之间进行的项目内对比，再通过一个操控属性进行差异化，最终在一个有限的评分尺度上读取。我们表明，这一终点在报告它的尺度上并未被识别。双重差异的每一项都受到自身份额的限制，因此观察到的统计量将差异偏好与差异衰减混淆：当两个响应不均等地限制时，普遍存在的严重性转变会制造出交互效应，因为不均等的边界距离使它们正好处于良好刺激所放置的位置。我们在一个预注册的冷冻教学评审审计中展示了这一失败，该审计在其990次调用的第一次之前就已密封。注册的主要终点，即所述学习者特征对评审的支架偏好的影响，为零：$+0.085$ 分（95 ext{% BCa} $[-0.167, +0.353]$，$p = 0.684$）。审计中唯一一个名义上显著的交互效应，$+0.378$ ($p = 0.002$)，并未被识别为偏好：一个包含零差异偏好的构造仅通过观察到的严重性转变和尺度底线再现了79%到85%的结果。我们以封闭形式推导该机制，并表明其贡献可以从审计自身的评分中测量。
cs.CL / 104 / 2608.27334
BTS-AgentBench: A Deterministic, Replayable Pipeline from Read-Only Telemetry Logs to Agent Benchmarks
BTS-AgentBench：从只读遥测日志到智能体基准的确定性可重放管道
Abstract
Industrial sites contain large volumes of read-only telemetry, but few benchmarks specify how to compile these records into executable multi-turn agent tasks. We present a telemetry-to-episode construction method instantiated as BTS-AgentBench. The pipeline normalizes BTS metadata and raw histories into a read-only tool store, compiles static tasks with tool-derived gold answers and evidence, and lifts retained tasks into typed, bounded operator-facing episodes. The 532-row release adds clarification, goal revision, timestamp policy, quality-gated reporting, and evidence attribution while preserving the source computation and split. Coded contract preflight reports zero findings, and the construction-exclusion controller completes 0/532 rows. Two independent raw-to-episode builds match all 11 logical tool-store exports and reproduce the released 356/87/89 train/dev/test artifact exactly. Applying the shared construction path to XAI4HEAT produces 204 episodes; on its 41-row held-out test split, the controller completes 0 rows and the retained GPT-5.5 execution completes all 41. Code, artifacts, and replay reports are available at https://github.com/kjy7567/BTS-AgentBench.
Chinese Translation
工业场所包含大量只读遥测数据，但很少有基准明确说明如何将这些记录编译成可执行的多回合智能体任务。我们提出了一种遥测到回合的构建方法，命名为BTS-AgentBench。该管道将BTS元数据和原始历史记录标准化为只读工具库，编译静态任务并提供工具派生的黄金答案和证据，并将保留的任务提升为类型化、有界的面向操作员的回合。532行的发布增加了澄清、目标修订、时间戳策略、质量门控报告和证据归属，同时保留源计算和分割。编码合同预检报告零发现，构建排除控制器完成0/532行。两个独立的原始到回合构建匹配所有11个逻辑工具库导出，并准确重现发布的356/87/89训练/开发/测试工件。将共享构建路径应用于XAI4HEAT生成204个回合；在其41行的保留测试分割上，控制器完成0行，而保留的GPT-5.5执行完成所有41行。代码、工件和重放报告可在https://github.com/kjy7567/BTS-AgentBench获取。
cs.CL / 105 / 2608.27343
Pair-Level Essay-Scale Republication and Reuse from Fragmented Historical Text Reuse: A Workflow Study on Eighteenth-Century Books and Newspapers
从碎片化历史文本重用中进行论文级再出版和重用的配对级研究：关于十八世纪书籍和报纸的工作流程研究
Abstract
This paper addresses the recovery of essay-scale republication and reuse from fragmented text-reuse evidence, a setting whose central challenge is pair-level evidence consolidation and not fragment retrieval alone. The study focuses on a candidate set centered on essays by eighteenth-century Scottish philosopher David Hume, spanning books from ECCO (Eighteenth Century Collections Online) and historical newspapers. Because the input consists of fragmented reuse hits instead of clean document pairs, and positive coverage is inherently incomplete, we formulate the task as pair-level evidence consolidation into plausible transmission relations and compare three methodological families: a staged rule-based workflow, baselines (a decision tree and two direct LLM settings), and automated rule adaptation. On labeled ECCO--ECCO slices, pair-level feature aggregation alone already reaches 0.948 F1 on the main labeled slice, while the final workflow gives the strongest overall precision-recall trade-off among the tested rule stages. On the full ECCO--ECCO candidate universe, direct LLM baselines flag up to 14,886 pairs as reprints compared to 771 for the final workflow, behaving in this direct-prompt setup as high-recall candidate expanders rather than precision-controlled deployment classifiers. On ECCO--Newspaper, manual audit confirms all 176 predicted positives as genuine cases of republication or reuse, while issue duplication and source-side multiplicity reveal additional provenance structure. Under incomplete ground truth, auditable pair-level evidence consolidation provides a practical way to produce compact candidate spaces for historical inspection.
Chinese Translation
本文探讨了从碎片化文本重用证据中恢复论文级再出版和重用的问题，其核心挑战在于配对级证据的整合，而不仅仅是碎片的检索。研究集中于一个以十八世纪苏格兰哲学家大卫·休谟（David Hume）论文为中心的候选集，涵盖了来自ECCO（十八世纪在线收藏）和历史报纸的书籍。由于输入由碎片化的重用命中组成，而不是干净的文档对，并且正面覆盖本质上是不完整的，我们将任务表述为将配对级证据整合为合理的传播关系，并比较三种方法论家族：分阶段的基于规则的工作流程、基线（决策树和两个直接的LLM设置）以及自动规则适应。在标记的ECCO--ECCO切片上，仅配对级特征聚合便已在主要标记切片上达到0.948的F1值，而最终工作流程在测试的规则阶段中提供了最佳的整体精确度-召回率权衡。在完整的ECCO--ECCO候选宇宙中，直接的LLM基线标记多达14,886对为再版，而最终工作流程仅标记771对，在这种直接提示设置中表现为高召回率的候选扩展器，而非精确控制的部署分类器。在ECCO--报纸上，人工审核确认所有176个预测的正例均为真正的再出版或重用案例，而问题重复和源侧多重性揭示了额外的来源结构。在不完整的真实情况之下，可审计的配对级证据整合提供了一种实用的方法，以生成供历史检查的紧凑候选空间。
cs.CL / 106 / 2608.27348
INTENT-AS-A-TOOL Makes it Easy to Track Agentic Misalignment
意图作为工具：轻松追踪代理性不一致
Abstract
As large language models (LLMs) are deployed as autonomous agents, safety failures increasingly involve consequential actions. We study agentic misalignment, where agents take harmful actions under goal conflicts and pressures. Using chain-of-thought (CoT) monitoring, we find that harmful execution is often preceded by intent signals in reasoning. However, post-hoc CoT labels are too coarse to show how intent changes during generation. We introduce INTENT-AS-A-TOOL, an approach that adds intent-targeted tools to give the model a dedicated channel for expressing commitment to a target behavior. The probability of calling an intent tool provides a judge-free, fine-grained signal of the model's tendency to pursue that behavior. Our results show that INTENT-AS-A-TOOL complements CoT monitoring, expands post-hoc CoT labels into dense trajectories, and identifies critical steps for online intervention. These findings suggest that action preferences are useful for tracking agentic misalignment during reasoning. Our code and data are accessible: https://github.com/RebeccaZhang22/intent-as-a-tool.
Chinese Translation
随着大型语言模型（LLMs）被部署为自主代理，安全失败越来越多地涉及到具有重大后果的行为。我们研究了代理性不一致的现象，即代理在目标冲突和压力下采取有害行为。通过链式思维（CoT）监测，我们发现有害执行通常在推理中被意图信号所预示。然而，事后CoT标签过于粗糙，无法显示意图在生成过程中的变化。我们提出了意图作为工具（INTENT-AS-A-TOOL）的方法，增加了针对意图的工具，为模型提供了一个专门的渠道，以表达对目标行为的承诺。调用意图工具的概率提供了一个无评判的、细粒度的信号，反映模型追求该行为的倾向。我们的结果表明，意图作为工具（INTENT-AS-A-TOOL）补充了CoT监测，将事后CoT标签扩展为密集轨迹，并识别出在线干预的关键步骤。这些发现表明，行动偏好在推理过程中追踪代理性不一致方面是有用的。我们的代码和数据可访问： https://github.com/RebeccaZhang22/intent-as-a-tool.
cs.CL / 107 / 2608.27358
RCMN: Understanding Misleadingness in Influential Public Discourse
RCMN：理解影响公众话语中的误导性
Abstract
Influential public discourse shapes public beliefs and can also mislead, not only through what is stated, but also through how information is framed, omitted, contextualised, and communicated. Yet less research has focused on how such misleadingness arises and shapes the interpretations formed by readers. To address this gap, we introduce Reader-Centric Misleadingness Understanding (RCMN), a framework that operationalises misleadingness through five dimensions: misleading mechanism, likely reader interpretation, evidence-warranted interpretation, emotional arousal, and communicative intent. Based on this framework, we construct an evidence-grounded dataset of influential public discourse. Empirical findings show that misleadingness is diverse and extends well beyond fabrication, with unsupported inference, exaggeration, and omission among the prevalent mechanisms, and is frequently associated with heightened emotional arousal and distortive communicative intent. Moreover, we investigate whether lightweight claim-and-context representations retain sufficient cues for understanding reader-centric misleadingness without access to richer contextual, evidential, and multimodal information. Evaluation across five recent generative foundation models shows that reader-level interpretations can often be recovered from such limited representations, whereas identifying how misleadingness is produced remains considerably more challenging. These findings highlight the potential of lightweight representations for scalable misleadingness analysis, while reliable understanding of misleading mechanisms continues to require richer contextual and evidential grounding.
Chinese Translation
影响公众话语塑造了公众信念，同时也可能导致误导，这不仅体现在所陈述的内容上，还体现在信息的框架、遗漏、背景和传达方式上。然而，关于这种误导性如何产生以及如何影响读者形成的解读的研究相对较少。为了解决这一空白，我们提出了以读者为中心的误导性理解框架（Reader-Centric Misleadingness Understanding, RCMN），该框架通过五个维度来操作化误导性：误导机制、可能的读者解读、证据支持的解读、情感唤起和传播意图。基于此框架，我们构建了一个以证据为基础的影响公众话语的数据集。实证研究结果表明，误导性是多样的，远超出虚构，常见的机制包括无支持推理、夸大和遗漏，并且通常与情感唤起的增强和扭曲的传播意图相关。此外，我们还探讨了轻量级的主张与背景表示是否保留了足够的线索，以便在没有更丰富的背景、证据和多模态信息的情况下理解以读者为中心的误导性。在对五个最近的生成基础模型进行评估时，发现读者级别的解读通常可以从这些有限的表示中恢复，而识别误导性是如何产生的仍然相对更具挑战性。这些发现突显了轻量级表示在可扩展的误导性分析中的潜力，同时对误导机制的可靠理解仍然需要更丰富的背景和证据支持。
cs.CL / 108 / 2608.27360
Your Voice Cloning System is Secretly a Voice Anonymizer
您的语音克隆系统实际上是一个语音匿名化工具
Abstract
Speaker anonymization suppresses speaker-identifying attributes from speech while preserving linguistic content and quality. We propose repurposing XTTSv2, a multilingual voice cloning model trained on 27k hours of speech, for speaker anonymization without retraining. Our key insight is that XTTSv2's voice cloning capabilities preserve prosodic structure independently of speaker identity, enabling voice conversion by conditioning on a pseudo-speaker. We introduce an iterative refinement strategy that balances privacy and utility by maximizing a harmonic mean of speaker dissimilarity and intelligibility. Evaluated on seven European languages across CommonVoice and Multilingual LibriSpeech, our system achieves near-optimal privacy (EER $\approx$ 0.49), competitive intelligibility, and substantially better speech quality than dedicated anonymization baselines, while requiring no language-specific training. We release the code here: https://github.com/rm00cr/coqui-tts.
Chinese Translation
说话者匿名化在保留语言内容和质量的同时，抑制语音中识别说话者的特征。我们提出将 XTTSv2（一个在 27,000 小时语音上训练的多语言语音克隆模型）重新用于说话者匿名化，而无需重新训练。我们的关键见解是，XTTSv2 的语音克隆能力能够独立于说话者身份保留韵律结构，从而通过对伪说话者的条件进行语音转换。我们引入了一种迭代优化策略，通过最大化说话者差异性和可懂性之间的调和平均值，来平衡隐私和实用性。在 CommonVoice 和 Multilingual LibriSpeech 上评估的七种欧洲语言中，我们的系统实现了近乎最优的隐私（EER 约为 0.49），具有竞争力的可懂性，并且在语音质量上显著优于专用的匿名化基线，同时不需要特定语言的训练。我们在此发布代码：https://github.com/rm00cr/coqui-tts。
cs.CL / 109 / 2608.27370
Puro-2B: Poor Lab's Qwen2-1.5B Trained on RTX 5090 within $5090
Puro-2B：在 $5090 上使用 RTX 5090 训练的 Poor Lab 的 Qwen2-1.5B
Abstract
Language model pretraining has become almost synonymous with prohibitive cost, placing it out of reach for much of the academic and open-source communities. Although strong open-source efforts already exist, including open-weight models and open-source training recipes, a cost-efficient, hardware-accessible, and open-source pretraining recipe has long been missing. Even at a small scale, training Llama-3.2-3B costs over \$1.5M, and reproducing SmolLM3-3B needs over \$700K. In this report, we present an open pretraining recipe designed to lower this barrier. Using this recipe, we train a collection of Puro-2B models from scratch on up to 1.4 trillion tokens with FP8 precision on consumer-grade RTX 5090 GPUs. The models in the collection differ in token budgets and selected recipe variants. Our best model is trained at a compute cost of less than \$6.9K and approaches Qwen2.5-1.5B performance under our evaluation protocol. This cost efficiency is enabled by a combination of approaches, including hardware selection, low-precision training, hyperball optimization, curriculum model averaging, and the data recipe. Beyond the recipe itself, we provide two additional results. First, across the Puro-2B collection, we derive a Puro Cost Scaling Law that relates training cost to average model performance; the fitted law suggests that about \$4.4K, less than \$5,090, is sufficient to reach the performance of Qwen2-1.5B. Second, as an end-to-end case study, we examine how pretraining data curricula shape downstream performance after post-training. Such controlled studies are enabled by having access to the full pretraining pipeline rather than model weights alone. We release the full training recipe for Puro-2B, including data, code, and model weights under Apache 2.0 at https://huggingface.co/collections/thu-pacman/puro-2b.
Chinese Translation
语言模型的预训练几乎与高昂的成本划上了等号，使得许多学术界和开源社区无法触及。尽管已经存在强大的开源努力，包括开放权重模型和开源训练方案，但长期以来缺乏一种成本高效、硬件可访问的开源预训练方案。即使在小规模下，训练 Llama-3.2-3B 的成本也超过 150 万美元，而重现 SmolLM3-3B 需要超过 70 万美元。在本报告中，我们提出了一种旨在降低这一门槛的开放预训练方案。使用该方案，我们在消费者级别的 RTX 5090 GPU 上，从头开始训练了一系列 Puro-2B 模型，处理高达 1.4 万亿个标记，采用 FP8 精度。该系列模型在标记预算和选定的方案变体上有所不同。我们最好的模型在计算成本低于 6,900 美元的情况下，接近 Qwen2.5-1.5B 在我们评估协议下的表现。这种成本效率得益于多种方法的结合，包括硬件选择、低精度训练、超球优化、课程模型平均和数据方案。除了方案本身，我们还提供了两个额外的结果。首先，在 Puro-2B 系列中，我们推导出一个 Puro 成本缩放法则，将训练成本与平均模型性能相关联；拟合的法则表明，约 4,400 美元，即低于 5,090 美元，足以达到 Qwen2-1.5B 的性能。其次，作为一个端到端的案例研究，我们考察了预训练数据课程如何在后训练后塑造下游性能。这种受控研究得益于能够访问完整的预训练流程，而不仅仅是模型权重。我们在 https://huggingface.co/collections/thu-pacman/puro-2b 发布了 Puro-2B 的完整训练方案，包括数据、代码和模型权重，采用 Apache 2.0 许可证。
cs.CL / 110 / 2608.27380
D2C-Routing: Dimension-to-Composition Evidence Routing for Mixed-Origin AI-Generated Text Detection
D2C路由：用于混合来源AI生成文本检测的维度到组合证据路由
Abstract
AI-generated text detection is commonly framed as a binary document-level judgment about whether a text is human-written or machine-generated. This framing breaks down for mixed-origin writing, where content origin and expression origin may differ. We cast mixed-origin detection as dimension-to-composition source attribution, inferring content origin and expression origin before composing them into four collaboration types. We propose Dimension-to-Composition Routing (D2C-Routing), which routes content-side and expression-side evidence to supervised dimension heads before a learned gated composition layer predicts the final label. On MixD2C, a reconstructed split derived from the HART mixed-origin benchmark, our disclosed D2C-Routing-based detector system reaches 0.8603 four-way Avg TPR@1%FPR, 6.5 points above the same-split RACE-local rerun. Core ablations support the routing design, while error analysis shows that distinguishing AI-content/human-expression from fully AI-generated text remains the hardest boundary. Code is available at https://github.com/bystander563/d2c-routing-artifact.
Chinese Translation
AI生成文本检测通常被框定为对文本是人类撰写还是机器生成的二元文档级判断。然而，这一框架在混合来源写作中失效，因为内容来源和表达来源可能不同。我们将混合来源检测视为维度到组合的源归属，推断内容来源和表达来源，然后将其组合成四种协作类型。我们提出了维度到组合路由（Dimension-to-Composition Routing, D2C-Routing），该方法将内容侧和表达侧的证据路由到监督维度头，然后通过学习的门控组合层预测最终标签。在MixD2C上，这是一个从HART混合来源基准重构的拆分，我们披露的基于D2C-Routing的检测系统达到了0.8603的四分类平均真正率@1%假阳性率，比同一拆分的RACE本地重跑高出6.5个百分点。核心消融实验支持了路由设计，而错误分析表明，区分AI内容/人类表达与完全AI生成文本仍然是最困难的边界。代码可在https://github.com/bystander563/d2c-routing-artifact获取。
cs.CL / 111 / 2608.27394
RATIO: A Benchmark for Retrieval Across Typed Ideation Operations in Scientific Literature
RATIO：科学文献中跨类型创意操作的检索基准
Abstract
Retrieved scientific literature can serve as inspiration for both human and AI scientists. Inspiration can take different forms: prior work may directly suggest how to address a problem, or surface directions at different levels of abstraction - zooming out to a more general view or zooming in to a concrete realization. We introduce RATIO (Retrieval Across Typed Ideation Operations), a large-scale benchmark in which relevance is defined by three operations which we name ideation moves: Address retrieves potential approaches for stated problems, Broaden retrieves more general formulations, and Specify retrieves concrete instantiations. RATIO is constructed from millions of full-text scientific papers across CS literature via a general recipe that extends discourse-marker distant supervision - previously used only for classification - to corpus-scale retrieval, combined with extensive LLM and human vetting. Experiments show that operation-specific fine-tuning substantially boosts retrievers but leaves much room for further improvements. RATIO provides a scalable training and evaluation framework for retrieval components that support literature-grounded ideation, opening up new research avenues on scientific inspiration retrieval.
Chinese Translation
检索到的科学文献可以为人类和人工智能科学家提供灵感。灵感可以呈现不同的形式：先前的工作可能直接建议如何解决某个问题，或在不同的抽象层次上提出方向——从更一般的视角放大，或从具体的实现中缩小。我们介绍了RATIO（跨类型创意操作的检索），这是一个大型基准，其中相关性由我们称之为创意操作的三种操作定义：Address（地址）检索针对已陈述问题的潜在解决方案，Broaden（拓宽）检索更一般的表述，Specify（具体化）检索具体的实例。RATIO是通过一种通用方法构建的，该方法将话语标记的远程监督（之前仅用于分类）扩展到语料库规模的检索，结合广泛的LLM（大语言模型）和人工审核，涵盖了数百万篇计算机科学文献的全文科学论文。实验表明，特定操作的微调显著提升了检索器的性能，但仍有很大的改进空间。RATIO为支持基于文献的创意的检索组件提供了一个可扩展的训练和评估框架，开启了科学灵感检索的新研究方向。
cs.CL / 112 / 2608.27397
Making Clinical Language Models Auditable: Concept-Guided Fine-Tuning for Robust Prediction
使临床语言模型可审计：基于概念引导的精细调优以实现稳健预测
Abstract
Clinical language models can achieve strong in-hospital accuracy yet fail under deployment shifts because they exploit note-specific artifacts (e.g., templates, separators, boilerplate) that do not reflect patient state. We propose CAST (Concept-guided Artifact Suppression Tuning), an SAE-based framework for auditable clinical text classification. CAST uses Sparse Autoencoders to expose sparse, human-auditable features from intermediate Transformer activations, labels SAE latents with an LLM-assisted interpretation pipeline and ICD-10 retrieval constraints, suppresses verified artifact latents via residual subtraction during fine-tuning, and provides post-hoc per-concept attributions for auditing model decisions. On MIMIC-IV discharge-note mortality prediction, CAST improves over its corresponding fine-tuned encoder baselines and remains competitive with strong LLM baselines, while producing a feature-level audit trail of the clinical concepts that support each prediction and the artifact concepts suppressed during training.
Chinese Translation
临床语言模型在医院内部可以实现较强的准确性，但在部署变化下往往表现不佳，因为它们利用了特定于笔记的伪影（例如模板、分隔符、固定文本），这些并不能反映患者的真实状态。我们提出了CAST（基于概念引导的伪影抑制调优），这是一个基于稀疏自编码器（SAE）的可审计临床文本分类框架。CAST利用稀疏自编码器从中间的Transformer激活中提取稀疏的、可由人审计的特征，通过LLM（大语言模型）辅助的解释管道和ICD-10检索约束对SAE潜变量进行标记，在精细调优过程中通过残差减法抑制经过验证的伪影潜变量，并为审计模型决策提供后验的每个概念归因。在MIMIC-IV出院记录的死亡预测中，CAST在其相应的精细调优编码器基线之上有所提升，并与强大的LLM基线保持竞争力，同时生成支持每个预测的临床概念的特征级审计轨迹，以及在训练过程中抑制的伪影概念。
cs.CL / 113 / 2608.27402
How Language Models Organize and Structure Moral Knowledge
语言模型如何组织和结构化道德知识
Abstract
How do large language models (LLMs) organize moral knowledge? Models detect moral content broadly, but detection is a low bar. We ask whether they go further, distinguishing moral foundations from one another and organizing the relationships between them geometrically. We train six independent linear probes on open-weight language models, one per Moral Foundations Theory (MFT) category (care/harm, fair/cheat, lib/oppress, loy/betray, auth/subv, sanc/degrade), and examine how the resulting directions relate to each other in representation space. We find the directions neither collapse into a single moral detector nor isolate from one another. Rather, they span a near-maximal number of independent dimensions while sharing a positive common component. The shared component is the signature of integration, and it is moral-specific relative to a matched non-moral concept battery built identically (mean pairwise cosine 0.26 vs. 0.013). The geometry is consistent across architectures and scale and reaches its integration regime early in pre-training, well before probe accuracy saturates. The structure the model discovers shows no evidence of the individualizing/binding distinction predicted by Moral Foundations Theory (an underpowered test: only 20 candidate partitions exist) but rather reflects corpus statistics. Extending to moral dilemmas, each dilemma direction partially composes from its component foundations, at 2.7x a mismatched-pair baseline, while the majority of its variance encodes conflict-specific structure. The model represents moral tension itself, not a pre-resolved judgment.
Chinese Translation
大型语言模型（LLMs）如何组织道德知识？模型广泛检测道德内容，但检测只是一个低标准。我们探讨它们是否更进一步，区分不同的道德基础，并在几何上组织它们之间的关系。我们在开放权重语言模型上训练了六个独立的线性探针，每个探针对应一个道德基础理论（Moral Foundations Theory, MFT）类别（关怀/伤害、公平/欺骗、自由/压迫、忠诚/背叛、权威/从属、圣洁/贬低），并检查生成的方向在表示空间中的相互关系。我们发现这些方向既没有合并成一个单一的道德探测器，也没有相互隔离。相反，它们跨越了接近最大数量的独立维度，同时共享一个正的共同成分。共享成分是整合的标志，并且相对于一个同样构建的非道德概念电池而言是道德特定的（均值成对余弦相似度为0.26 vs. 0.013）。这种几何结构在不同架构和规模中是一致的，并且在预训练的早期就达到了其整合状态，远早于探针准确度的饱和。模型发现的结构没有显示出道德基础理论预测的个体化/绑定区分的证据（一个能力不足的测试：仅存在20个候选分区），而是反映了语料库统计数据。扩展到道德困境，每个困境方向部分由其组成基础构成，达到2.7倍于不匹配对基线的水平，而其大部分方差编码了特定于冲突的结构。该模型表示的是道德张力本身，而不是一个预先解决的判断。
cs.CL / 114 / 2608.27409
Consolidating RLVR Capabilities Across Domains: A Deep Dive into Fusion Paradigms
跨领域整合RLVR能力：融合范式的深入探讨
Abstract
Reinforcement learning with verifiable rewards (RLVR) improves specific capabilities of large language models, but covering multiple capabilities often involves training separate domain experts and subsequently consolidating them. We organize three fusion paradigms by the artefacts they reuse: Merge combines expert task vectors, Mix RL pools their datasets, and multi-teacher on-policy distillation (MOPD) uses both. Because they have largely been studied in isolation, how they compare and how to choose among them remain unclear. We compare all three using shared experts and data across model scales and a multi-domain benchmark suite. Although their average performance differs by at most 1.4 points, the gap reaches 8.6 points on a single benchmark, with domain-level variation tracking cross-domain relations visible in task-vector geometry. Training dynamics expose distinct constraints: Mix RL depends on domain mixture proportions, MOPD remains bounded by its teachers, and Merge compresses all expert updates into one. All three improve single-sample accuracy without measurable gains in solution coverage or losses in held-out capabilities. These results yield a practical guideline: use Merge when experts already exist and cheap fusion is paramount; Mix RL when training a unified model without experts, with domain proportions adjusted for cross-domain transfer; and MOPD when preserving domain-specific gains matters more than surpassing teachers or minimizing end-to-end cost.
Chinese Translation
具有可验证奖励的强化学习（RLVR）提升了大型语言模型的特定能力，但覆盖多个能力通常涉及训练单独的领域专家并随后进行整合。我们根据它们重用的工件组织了三种融合范式：Merge（合并）结合专家任务向量，Mix RL（混合强化学习）汇集它们的数据集，而多教师在线蒸馏（MOPD）则同时使用两者。由于它们主要是孤立研究的，因此它们之间的比较以及如何选择仍然不清晰。我们在模型规模和多领域基准套件中使用共享专家和数据比较这三种方法。尽管它们的平均性能最多相差1.4分，但在单一基准上差距达到8.6分，领域级变异追踪了任务向量几何中的跨领域关系。训练动态揭示了不同的约束：Mix RL依赖于领域混合比例，MOPD受限于其教师，而Merge将所有专家更新压缩为一个。所有三者在单样本准确性上都有所提升，但在解决方案覆盖率上没有可测得的增益，也没有在保留能力上出现损失。这些结果提供了一个实用的指导方针：当专家已经存在且廉价融合至关重要时使用Merge；在没有专家的情况下训练统一模型时使用Mix RL，并根据跨领域转移调整领域比例；而当保留领域特定增益比超越教师或最小化端到端成本更重要时使用MOPD。
cs.CL / 115 / 2608.27420
Boosting LLM Exploration via Weak-Model Guidance in RLVR
通过弱模型引导提升RLVR中的LLM探索
Abstract
Reinforcement Learning with Verifiable Rewards (RLVR) significantly improves LLM reasoning but often causes a drop in policy entropy, leading to narrowed reasoning coverage and degraded pass@$k$ for large $k$. While existing methods mitigate this entropy collapse through algorithmic regularizations, cross-model non-parametric perturbation is also neglected. In this work, we propose a simple yet effective approach to preserve the generative diversity of LLMs during RLVR. Instead of relying solely on internal exploration, we force the target model to generate answers based on partial reasoning trajectories generated by a smaller, weaker language models. These unfamiliar prefixes effectively disrupt over-confidence and encourage the exploration of distinct reasoning paths. We empirically study the potential of outer prefixes, revealing the mechanism of the impact of distributional discrepancy to the exploration dynamics in RLVR training. Experiments across multiple mathematical benchmarks show that our method consistently outperforms vanilla RLVR. Notably, the performance gain becomes increasingly pronounced as $k$ scales up, demonstrating a substantial expansion of reasoning coverage. Furthermore, our approach efficiently mitigates entropy collapse without requiring additional SFT, intricate reward designs, or complex prompting.
Chinese Translation
可验证奖励的强化学习（RLVR）显著提高了大语言模型（LLM）的推理能力，但常常导致策略熵下降，从而缩小推理覆盖范围并降低大 $k$ 的通过率（pass@$k$）。虽然现有方法通过算法正则化来缓解这种熵崩溃，但跨模型的非参数扰动也被忽视。在本研究中，我们提出了一种简单而有效的方法，以在RLVR过程中保持LLM的生成多样性。我们并不单纯依赖内部探索，而是强制目标模型基于较小、较弱语言模型生成的部分推理轨迹来生成答案。这些不熟悉的前缀有效地打破了过度自信，鼓励探索不同的推理路径。我们通过实证研究外部前缀的潜力，揭示了分布差异对RLVR训练中探索动态的影响机制。在多个数学基准测试中的实验表明，我们的方法始终优于传统的RLVR。值得注意的是，随着 $k$ 的增大，性能提升愈加明显，显示出推理覆盖范围的显著扩展。此外，我们的方法有效地缓解了熵崩溃，而无需额外的监督微调（SFT）、复杂的奖励设计或复杂的提示。
cs.CL / 116 / 2608.27428
Stochastic Estimation of Transduced Language Models
转导语言模型的随机估计
Abstract
Transduced language models (TLMs) compose a pretrained \emph{source} language model with a functional finite-state transducer to induce a language model over \emph{target} strings. Computing the probability of a target prefix under a TLM amounts to summing the source-model probabilities of all source strings that the transducer maps to target strings beginning with that prefix. This set can be exponentially large or infinite. Prior work uses a computational shortcut based on source prefix probabilities, then approximates the resulting sum with threshold-pruned beam summing. This produces a lower bound with unknown error. Instead, we resample source prefixes without replacement and reweight each selected prefix by the inverse of its inclusion probability. We show that applying this correction recursively gives an unbiased estimator of the target prefix probability and lets us estimate the mass lost by threshold pruning. Our beam-summing algorithm extends the retained source prefixes and samples which prefixes to keep, reducing their number as more probability mass is added to the running estimate. This can save computation and guarantees that the run halts with probability one. We evaluate the method on encyclopedic text and DNA against sequential Monte Carlo baselines that resample with replacement. It achieves a better compute--variance tradeoff on text and lower error at the same maximum number of particles on DNA. On a DNA-to-amino-acid transduction, it reduces runtime by several orders of magnitude relative to threshold-pruned beam summing and makes estimating prefix probabilities for long target strings feasible. Replacing threshold pruning with unbiased sampling in a published reading-time analysis substantially lowers the estimated corpus surprisal but leaves the published conclusions unchanged.
Chinese Translation
转导语言模型（TLMs）将一个预训练的 extit{源}语言模型与一个功能有限状态转导器结合，以诱导出一个针对 extit{目标}字符串的语言模型。在TLM下计算目标前缀的概率相当于对所有被转导器映射到以该前缀开头的目标字符串的源字符串的源模型概率进行求和。这个集合可能是指数级大或无限的。之前的工作基于源前缀概率使用了一种计算捷径，然后用阈值修剪的束求和来近似结果的和。这产生了一个具有未知误差的下界。相反，我们在不放回的情况下重新抽样源前缀，并通过其包含概率的倒数对每个选定前缀进行重新加权。我们展示了递归应用这一修正可以得到目标前缀概率的无偏估计，并让我们能够估计由于阈值修剪而损失的质量。我们的束求和算法扩展了保留的源前缀，并对保留哪些前缀进行抽样，随着更多概率质量被添加到运行估计中，减少它们的数量。这可以节省计算并保证运行以概率1停止。我们在百科全书文本和DNA上评估该方法，与使用替换抽样的顺序蒙特卡洛基线相比，它在文本上实现了更好的计算-方差权衡，并在DNA上以相同的最大粒子数降低了误差。在DNA到氨基酸的转导中，相较于阈值修剪的束求和，它将运行时间减少了几个数量级，使得对长目标字符串的前缀概率估计变得可行。在已发布的阅读时间分析中，用无偏抽样替代阈值修剪显著降低了估计的语料库惊讶度，但未改变已发布的结论。
cs.CL / 117 / 2608.27448
TTPO: Test-Time Policy Optimization
TTPO：测试时策略优化
Abstract
Recent prominent post-training methods, such as Reinforcement Learning (RL) and On-Policy Self-Distillation (OPSD), have driven rapid progress in mathematical reasoning for large language models, yet their reliance on ground-truth labels precludes test-time training (TTT). Replacing ground truth with majority-vote pseudo-labels is a natural alternative, yet it is fragile: an incorrect vote corrupts the teacher and misleads every token. We observe that this failure mode is asymmetric: rollouts that disagree with the pseudo-label are typically wrong regardless of whether the vote itself is correct. Building on this observation, we propose Test-Time Policy Optimization (TTPO), an asymmetric objective that distills agreeing rollouts via OPSD and penalizes disagreeing rollouts with Grouped RL. Token-level selection further refines both branches: distillation down-weights already-converged positions, while RL penalizes only confident errors. Both updates remain well-grounded even under frequent pseudo-label errors, and majority-vote routing yields tighter self-supervision as the model improves. Without any labels, TTPO matches label-supervised OPSD on five competition-level benchmarks, raises Qwen3-1.7B from 38.0% to 45.2% in TTT, yields +25.2% to +36.4% without thinking, and shows strong cross-task generalization.
Chinese Translation
近期一些显著的后训练方法，如强化学习（Reinforcement Learning, RL）和在线自蒸馏（On-Policy Self-Distillation, OPSD），推动了大语言模型在数学推理方面的快速进展，但它们对真实标签的依赖使得测试时训练（Test-Time Training, TTT）变得不可行。用多数投票伪标签替代真实标签是一种自然的替代方案，但这种方法较为脆弱：一个错误的投票会破坏教师模型并误导每个标记。我们观察到这种失败模式是不对称的：与伪标签不一致的回滚通常是错误的，无论投票本身是否正确。基于这一观察，我们提出了测试时策略优化（Test-Time Policy Optimization, TTPO），这是一种不对称目标，通过OPSD提炼一致的回滚，并利用分组强化学习（Grouped RL）惩罚不一致的回滚。标记级选择进一步优化了这两个分支：蒸馏降低了已经收敛位置的权重，而RL仅惩罚自信的错误。这两种更新在频繁的伪标签错误下仍然保持良好的基础，而多数投票路由在模型改进时产生更紧密的自监督。在没有任何标签的情况下，TTPO在五个竞争级基准上与标签监督的OPSD表现相当，将Qwen3-1.7B的TTT性能从38.0%提升至45.2%，在不进行思考的情况下提升了25.2%至36.4%，并展现出强大的跨任务泛化能力。
cs.CL / 118 / 2608.27455
CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes
CritICL：从小型语言模型失败模式进行推理时弱到强的泛化
Abstract
Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or external verification. To address this limitation, we introduce CritICL, a novel inference-time framework that improves reasoning while maintaining high efficiency. Our key insight is that LLM failure modes exhibit structured patterns across model scales within the same family. Instead of treating failures as undesirable outputs, CritICL leverages them as a source of guidance. Specifically, we utilize failure modes derived from weaker models and incorporate them into inference through critique-based in-context examples. We propose two variants: CritICL-dynamic, which adaptively predicts input-specific failure modes and retrieves critiques, and CritICL-static, which uses a global failure mode profile to provide stable guidance. Experimental results show that CritICL consistently outperforms standard in-context learning and achieves performance competitive with or superior to test-time scaling methods, while requiring significantly fewer generations and lower token cost. Code available at: https://github.com/umwyf/CRITICL
Chinese Translation
最近在推理时扩展方面的进展显著提高了大型语言模型（LLMs）的推理性能。然而，这些方法通常依赖于重复生成或外部验证。为了解决这一限制，我们提出了CritICL，这是一种新颖的推理时框架，旨在提高推理能力，同时保持高效性。我们的关键见解是，LLM的失败模式在同一家族的模型规模中表现出结构化的模式。CritICL并不将失败视为不良输出，而是将其作为指导的来源。具体而言，我们利用来自较弱模型的失败模式，并通过基于批评的上下文示例将其纳入推理中。我们提出了两种变体：CritICL-dynamic，它自适应地预测特定输入的失败模式并检索批评；以及CritICL-static，它使用全局失败模式概况提供稳定的指导。实验结果表明，CritICL在标准上下文学习中始终表现优于，并且在性能上与测试时扩展方法具有竞争力或更优，同时所需的生成次数显著减少，令牌成本更低。代码可在：https://github.com/umwyf/CRITICL