夜雨聆风1、大模型与Agent
① ELF: Embedded Language Flows
作者：Keya Hu, Linlu Qiu, Yiyang Lu, Hanhong Zhao, Tianhong Li, Yoon Kim et al. arXiv: 2605.10938v1 链接: https://arxiv.org/abs/2605.10938v1
中文摘要：扩散和基于流的模型已成为生成连续数据的事实方法，例如在图像和视频等领域。他们的成功吸引了越来越多的人将其应用于语言建模。与图像域模型不同，当今领先的扩散语言模型（ DLM ）主要在离散令牌上运行。在本文中，我们证明了连续DLM可以在最小程度地适应离散域的情况下有效。我们提出嵌入式语言
English Abstract：Diffusion and flow-based models have become the de facto approaches for generating continuous data, e.g., in domains such as images and videos. Their success has attracted growing interest in applying them to language modeling. Unlike their image-domain counterparts, today's leading diffusion language models (DLMs) primarily operate over discrete tokens. In this paper, we show that continuous DLMs can be made effective with minimal adaptation to the discrete domain. We propose Embedded Language Flows (ELF), a class of diffusion models in continuous embedding space based on continuous-time Flow Matching. Unlike existing DLMs, ELF predominantly stays within the continuous embedding space until the final time step, where it maps to discrete tokens using a shared-weight network. This formulati...
② DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices
作者：Chenyang Song, Weilin Zhao, Xu Han, Chaojun Xiao, Yingfa Chen, Zhiyuan Liu arXiv: 2605.10933v1 链接: https://arxiv.org/abs/2605.10933v1
中文摘要：虽然Mixture-of-Experts （ MoE ）在不按比例增加计算的情况下扩展模型容量，但其庞大的总参数占用量会产生显着的存储和内存访问瓶颈，这阻碍了同时需要高性能、低计算成本和小存储开销的高效端部署。为了实现这些特性，我们提出了DECO ，这是一种稀疏的MoE架构，旨在匹配相同总参数b下密集变压器的性能
English Abstract：While Mixture-of-Experts (MoE) scales model capacity without proportionally increasing computation, its massive total parameter footprint creates significant storage and memory-access bottlenecks, which hinder efficient end-side deployment that simultaneously requires high performance, low computational cost, and small storage overhead. To achieve these properties, we present DECO, a sparse MoE architecture designed to match the performance of dense Transformers under identical total parameter budgets and training tokens. DECO utilizes the differentiable and flexible ReLU-based routing enhanced by learnable expert-wise scaling, which adaptively balances the contributions of routed and shared experts. Furthermore, we introduce NormSiLU, an activation function that normalizes inputs prior to...
③ Dynamic Skill Lifecycle Management for Agentic Reinforcement Learning
作者：Junhao Shen, Teng Zhang, Xiaoyan Zhao, Hong Cheng arXiv: 2605.10923v1 链接: https://arxiv.org/abs/2605.10923v1
中文摘要：大型语言模型代理越来越依赖外部技能来解决复杂的任务，其中技能充当模块化单元，将其功能扩展到仅参数内存支持的范围之外。现有的方法假设外部技能要么作为持久的指导积累，要么内化到政策中，最终导致零技能推断。我们认为这种假设过于严格，因为在参数能力有限和不均衡的技能边际贡献的情况下，
English Abstract：Large language model agents increasingly rely on external skills to solve complex tasks, where skills act as modular units that extend their capabilities beyond what parametric memory alone supports. Existing methods assume external skills either accumulate as persistent guidance or internalized into the policy, eventually leading to zero-skill inference. We argue this assumption is overly restrictive, since with limited parametric capacity and uneven marginal contribution across skills, the optimal active skill set is non-monotonic, task- and stage-dependent. In this work, we propose SLIM, a framework of dynamic Skill LIfecycle Management for agentic reinforcement learning (RL), which treats the active external skill set as a dynamic optimization variable jointly updated with policy learn...
④ RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards
作者：Gaotang Li, Bhavana Dalvi Mishra, Zifeng Wang, Jun Yan, Yanfei Chen, Chun-Liang Li et al. arXiv: 2605.10899v1 链接: https://arxiv.org/abs/2605.10899v1
中文摘要：培训深度研究代理，即计划、搜索、评估证据和综合长式报告的系统，将强化学习推向可验证奖励制度之外。他们的输出缺乏真实的答案，他们的轨迹跨越了许多工具增强的决策，而标准的后期培训几乎没有提供将过去的尝试转化为可重复使用的经验的机制。在这项工作中，我们认为，评分细则表不仅应作为最终答案评估者，还应作为
English Abstract：Training deep research agents, namely systems that plan, search, evaluate evidence, and synthesize long-form reports, pushes reinforcement learning beyond the regime of verifiable rewards. Their outputs lack ground-truth answers, their trajectories span many tool-augmented decisions, and standard post-training offers little mechanism for turning past attempts into reusable experience. In this work, we argue that rubrics should serve not merely as final-answer evaluators, but as the shared interface that structures policy execution, judge feedback, and agent memory. Based on this view, we introduce RubricEM, a rubric-guided reinforcement learning framework that combines stagewise policy decomposition with reflection-based meta-policy evolution. RubricEM first makes research trajectories sta...
⑤ Compute Where it Counts: Self Optimizing Language Models
作者：Yash Akhauri, Mohamed S. Abdelfattah arXiv: 2605.10875v1 链接: https://arxiv.org/abs/2605.10875v1
中文摘要：有效的LLM推理研究主要集中在降低每个解码步骤的成本（例如，使用量化、修剪或稀疏注意力） ，通常对每个生成的令牌应用统一的计算预算。在实践中，令牌难度差异很大，因此静态压缩可能会在简单步骤上过度计算，而在硬步骤上计算不足。我们研究自回归解码的动态预算分配：学习单个模型中每个令牌花费的计算量。
English Abstract：Efficient LLM inference research has largely focused on reducing the cost of each decoding step (e.g., using quantization, pruning, or sparse attention), typically applying a uniform computation budget to every generated token. In practice, token difficulty varies widely, so static compression can over-compute on easy steps and under-compute on hard ones. We study dynamic budget allocation for autoregressive decoding: learning how much computation to spend per token from within a single model. Self-Optimizing Language Models (SOL) pair a frozen LLM with a lightweight policy network that reads the LLM hidden state and selects a discrete efficiency action at each decode step. Actions can jointly control (i) token-level attention sparsity, (ii) structured activation pruning in the MLP, and ...
⑥ The Generalized Turing Test: A Foundation for Comparing Intelligence
作者：Daniel Mitropolsky, Susan S. Hong, Riccardo Neumarker, Emanuele Rimoldi, Tomaso Poggio arXiv: 2605.10851v1 链接: https://arxiv.org/abs/2605.10851v1
中文摘要：我们引入了广义图灵测试（ GTT ） ，这是一个通过不可区分性来比较任意代理能力的正式框架。对于代理A和代理B ，我们定义了图灵比较器A $\ geq $ B ，如果B作为区分器，不能可靠地区分与A的交互（指示模仿B ）和B的另一个实例，则保持相对智能的数据集和任务无关的概念。我们研究比较器的结构，包括
English Abstract：We introduce the Generalized Turing Test (GTT), a formal framework for comparing the capabilities of arbitrary agents via indistinguishability. For agents A and B, we define the Turing comparator A $\geq$ B to hold if B, acting as a distinguisher, cannot reliably distinguish between interactions with A (instructed to imitate B) and another instance of B. This yields a dataset- and task-agnostic notion of relative intelligence. We study the comparator's structure, including conditions under which it is transitive and therefore induces an ordering over equivalence classes, and we define and analyze variants with querying, bounded interaction, and fixed distinguishers. To complement the theory, we instantiate the framework on a collection of modern models, empirically evaluating pairwise indi...
⑦ Rethinking Agentic Search with Pi-Serini: Is Lexical Retrieval Sufficient?
作者：Tz-Huan Hsu, Jheng-Hong Yang, Jimmy Lin arXiv: 2605.10848v1 链接: https://arxiv.org/abs/2605.10848v1
中文摘要：随着大型语言模型（ LLM ）在代理循环中变得更有能力，词汇检索器是否足够？在构建深度研究系统时，自然会出现这个问题。我们通过将BM25与具有更好推理和工具使用能力的前沿LLM配对来重新审视它。为了支持研究人员提出同样的问题，我们推出了Pi-Serini ，这是一种配备三种工具来检索、浏览和阅读文档的搜索代理。我们的结果表明，在BrowseComp-Plus上，
English Abstract：Does a lexical retriever suffice as large language models (LLMs) become more capable in an agentic loop? This question naturally arises when building deep research systems. We revisit it by pairing BM25 with frontier LLMs that have better reasoning and tool-use abilities. To support researchers asking the same question, we introduce Pi-Serini, a search agent equipped with three tools for retrieving, browsing, and reading documents. Our results show that, on BrowseComp-Plus, a well-configured lexical retriever with sufficient retrieval depth can support effective deep research when paired with more capable LLMs. Specifically, Pi-Serini with gpt-5.5 achieves 83.1% answer accuracy and 94.7% surfaced evidence recall, outperforming released search agents that use dense retrievers. Controlled ab...
⑧ Training-Free Cultural Alignment of Large Language Models via Persona Disagreement
作者：Huynh Trung Kiet, Dao Sy Duy Minh, Tuan Nguyen, Chi-Nguyen Tran, Phu-Hoa Pham, Nguyen Lam Phu Quy et al. arXiv: 2605.10843v1 链接: https://arxiv.org/abs/2605.10843v1
中文摘要：大型语言模型越来越多地调解转向道德判断的决策，但越来越多的证据表明，它们的隐性偏好在文化上并不中立。现有的文化调整方法要么需要每个国家的偏好数据和微调预算，要么假设白盒访问商业API不会暴露的模型内部。在这项工作中，我们专注于这种现实的黑盒子，仅公开数据的制度，并观察到国内社会人口统计学上的不满
English Abstract：Large language models increasingly mediate decisions that turn on moral judgement, yet a growing body of evidence shows that their implicit preferences are not culturally neutral. Existing cultural alignment methods either require per-country preference data and fine-tuning budgets or assume white-box access to model internals that commercial APIs do not expose. In this work, we focus on this realistic black-box, public-data-only regime and observe that within-country sociodemographic disagreement, not consensus, is the primary steering signal. We introduce DISCA (Disagreement-Informed Steering for Cultural Alignment), an inference-time method that instantiates each country as a panel of World-Values-Survey-grounded persona agents and converts their disagreement into a bounded, loss-averse...
2、强化学习
① Variational Inference for Lévy Process-Driven SDEs via Neural Tilting
作者：Yaman Kindap, Manfred Opper, Benjamin Dupuis, Umut Simsekli, Tolga Birdal arXiv: 2605.10934v1 链接: https://arxiv.org/abs/2605.10934v1
中文摘要：模拟极端事件和重尾现象对于在金融、气候科学和安全关键型人工智能等领域建立可靠的预测系统至关重要。虽然Lévy过程为捕获跳跃和重尾提供了一个自然的数学框架，但Lévy驱动的随机微分方程（ SDE ）的贝叶斯推理仍然难以使用现有方法：蒙特卡罗方法是严格的，但缺乏可扩展性，而神经变分推理方法是有效的
English Abstract：Modelling extreme events and heavy-tailed phenomena is central to building reliable predictive systems in domains such as finance, climate science, and safety-critical AI. While Lévy processes provide a natural mathematical framework for capturing jumps and heavy tails, Bayesian inference for Lévy-driven stochastic differential equations (SDEs) remains intractable with existing methods: Monte Carlo approaches are rigorous but lack scalability, whereas neural variational inference methods are efficient but rely on Gaussian assumptions that fail to capture discontinuities. We address this tension by introducing a neural exponential tilting framework for variational inference in Lévy-driven SDEs. Our approach constructs a flexible variational family by exponentially reweighting the Lévy measu...
② Quantifying Concentration Phenomena of Mean-Field Transformers in the Low-Temperature Regime
作者：Albert Alcalde, Leon Bungert, Konstantin Riedl, Tim Roith arXiv: 2605.10931v1 链接: https://arxiv.org/abs/2605.10931v1
中文摘要：以自我注意模块为核心组件的变压器已成为现代大型语言和基础模型中不可或缺的架构。在本文中，我们研究了仅深度编码器变压器中令牌在推理时间的演变，该演变由均值场连续性方程在大令牌极限中描述。利用交互多粒子系统收敛性分析的思想，与令牌对应的粒子，我们证明了令牌分布快速
English Abstract：Transformers with self-attention modules as their core components have become an integral architecture in modern large language and foundation models. In this paper, we study the evolution of tokens in deep encoder-only transformers at inference time which is described in the large-token limit by a mean-field continuity equation. Leveraging ideas from the convergence analysis of interacting multi-particle systems, with particles corresponding to tokens, we prove that the token distribution rapidly concentrates onto the push-forward of the initial distribution under a projection map induced by the key, query, and value matrices, and remains metastable for moderate times. Specifically, we show that the Wasserstein distance of the two distributions scales like $\sqrt{{\log(β+1)}/β}\exp(Ct)+\e...
③ Optimal and Scalable MAPF via Multi-Marginal Optimal Transport and Schrödinger Bridges
作者：Usman A. Khan, Joseph W. Durham arXiv: 2605.10917v1 链接: https://arxiv.org/abs/2605.10917v1
中文摘要：我们考虑匿名多智能体路径查找（ MAPF ） ，其中一组机器人的任务是在有限的连通图上行进到一组目标。我们证明了MAPF可以作为一类特殊的多边缘最优传输（ MMOT ）问题，具有潜在的马尔可夫结构，在该结构下，指数大的MMOT坍缩为线性规划（ LP ）多项式的大小。专注于匿名设置，我们建立相应的LP可行的条件，完全unim
English Abstract：We consider anonymous multi-agent path finding (MAPF) where a set of robots is tasked to travel to a set of targets on a finite, connected graph. We show that MAPF can be cast as a special class of multi-marginal optimal transport (MMOT) problems with an underlying Markovian structure, under which the exponentially large MMOT collapses to a linear program (LP) polynomial in size. Focusing on the anonymous setting, we establish conditions under which the corresponding LP is feasible, totally unimodular, and consequently, yields min-cost, integral $({0,1})$ transports that do not overlap in both space and time. To adapt the approach to large-scale problems, we cast the MAPF-MMOT in a probabilistic framework via Schrödinger bridges. Under standard assumptions, we show that the Schrödinger b...
④ Equivariant Reinforcement Learning for Clifford Quantum Circuit Synthesis
作者：Richie Yeung, Aleks Kissinger, Rob Cornish arXiv: 2605.10910v1 链接: https://arxiv.org/abs/2605.10910v1
中文摘要：我们考虑了为具有全量子位连接的设备合成Clifford量子电路的问题。我们将此任务视为强化学习问题，其中智能体学习发现基本Clifford门序列，该序列将Clifford电路的给定辛矩阵表示简化为恒等式。这个公式允许基于身份的随机游走的简单学习课程。我们引入了一种新的神经网络架构，
English Abstract：We consider the problem of synthesizing Clifford quantum circuits for devices with all-to-all qubit connectivity. We approach this task as a reinforcement learning problem in which an agent learns to discover a sequence of elementary Clifford gates that reduces a given symplectic matrix representation of a Clifford circuit to the identity. This formulation permits a simple learning curriculum based on random walks from the identity. We introduce a novel neural network architecture that is equivariant to qubit relabelings of the symplectic matrix representation, and which is size-agnostic, allowing a single learned policy to be applied across different qubit counts without circuit splicing or network reparameterization. On six-qubit Clifford circuits, the largest regime for which optimal re...
⑤ Revisiting Policy Gradients for Restricted Policy Classes: Escaping Myopic Local Optima with $k$-step Policy Gradients
作者：Alex DeWeese, Guannan Qu arXiv: 2605.10909v1 链接: https://arxiv.org/abs/2605.10909v1
中文摘要：这项工作重新审视了受限策略类上使用的标准策略梯度方法，已知这些方法会陷入次优临界点。我们确定了这种现象的一个重要原因是政策梯度本身从根本上来说是短视的，即它只是基于一步$ Q $ -函数来改进政策。在这项工作中，我们提出了一种广义的$ k $ -step策略梯度方法，该方法在$ k $ -step时间窗口内耦合随机性，并可以在
English Abstract：This work revisits standard policy gradient methods used on restricted policy classes, which are known to get stuck in suboptimal critical points. We identify an important cause for this phenomenon to be that the policy gradient is itself fundamentally myopic, i.e. it only improves the policy based on the one-step $Q$-function. In this work, we propose a generalized $k$-step policy gradient method that couples the randomness within a $k$-step time window and can escape the myopic local optima in MDPs with restricted policy classes. We show this new method is theoretically guaranteed to converge to a solution that is exponentially close in performance to the optimal deterministic policy with respect to $k$. Further, we show projected gradient descent and mirror descent with this $k$-step po...
3、具身智能
① HarmoWAM: Harmonizing Generalizable and Precise Manipulation via Adaptive World Action Models
作者：Qiuxuan Feng, Jiale Yu, Jiaming Liu, Yueru Jia, Zhuangzhe Wu, Hao Chen et al. arXiv: 2605.10942v1 链接: https://arxiv.org/abs/2605.10942v1
中文摘要：世界动作模型（ WAM ）已通过物理动力学建模成为机器人控制的一个有前途的范例。当前的WAM通常遵循两种范式： “Imagine-then-Execute”方法，该方法使用视频预测通过逆动态推断动作，以及“Joint Modeling”方法，该方法联合建模动作和视频表示。基于系统实验，我们观察到这些范式之间的基本权衡：前者明确地利用世界模型来生成
English Abstract：World Action Models (WAMs) have emerged as a promising paradigm for robot control by modeling physical dynamics. Current WAMs generally follow two paradigms: the "Imagine-then-Execute" approach, which uses video prediction to infer actions via inverse dynamics, and the "Joint Modeling" approach, which jointly models actions and video representations. Based on systematic experiments, we observe a fundamental trade-off between these paradigms: the former explicitly leverages world models for generalizable transit but lacks interaction precision, whereas the latter enables fine-grained, temporally coherent action generation but is constrained by the exploration space of the training distribution. Motivated by these findings, we propose HarmoWAM, an end-to-end WAM that fully leverages a world ...
② RoboMemArena: A Comprehensive and Challenging Robotic Memory Benchmark
作者：Huashuo Lei, Wenxuan Song, Huarui Zhang, Jieyuan Pei, Jiayi Chen, Haodong Yan et al. arXiv: 2605.10921v1 链接: https://arxiv.org/abs/2605.10921v1
中文摘要：记忆是机器人智能的关键组成部分，因为机器人必须依靠过去的观察和行动，才能在部分可观察的环境中完成长时间的任务。然而，现有的机器人记忆基准仍然缺乏用于记忆形成的多模态注释，提供有限的任务覆盖和结构复杂性，并且仍然局限于没有真实世界评估的模拟。我们通过RoboMemArena来解决这一差距， RoboMemArena是一个由26项任务组成的大规模基准，
English Abstract：Memory is a critical component of robotic intelligence, as robots must rely on past observations and actions to accomplish long-horizon tasks in partially observable environments. However, existing robotic memory benchmarks still lack multimodal annotations for memory formation, provide limited task coverage and structural complexity, and remain restricted to simulation without real-world evaluation. We address this gap with RoboMemArena, a large-scale benchmark of 26 tasks, with average trajectory lengths exceeding 1,000 steps per task and 68.9% of subtasks being memory-dependent. The generation pipeline leverages a vision-language model (VLM) to design and compose subtasks, generates full trajectories through atomic functions, and provides memory-related annotations, including subtask in...
③ Embodied AI in Action: Insights from SAE World Congress 2026 on Safety, Trust, Robotics, and Real-World Deployment
作者：Jan-Mou Li, Paul Schmitt, Wei Tong, Majed Mohammed, Akshay Chalana, Arpan Kusari et al. arXiv: 2605.10653v1 链接: https://arxiv.org/abs/2605.10653v1
中文摘要：嵌入式人工智能正在迅速从研究转向现实世界的系统，如自动驾驶汽车、移动机器人和工业机器。随着这些系统在动态环境中变得更具感知、决策和行动的能力，它们也在安全、信任、治理和运营可靠性方面带来了新的挑战。本白皮书总结了SAE 2026世界大会小组会议\ textit {Embodied AI in Action}的主要见解，该会议汇集了专家
English Abstract：Embodied artificial intelligence is rapidly moving from research into real-world systems such as autonomous vehicles, mobile robots, and industrial machines. As these systems become more capable of perceiving, deciding, and acting in dynamic environments, they also introduce new challenges in safety, trust, governance, and operational reliability. This white paper summarizes key insights from the SAE World Congress 2026 panel session \textit{Embodied AI in Action}, which brought together experts from automotive, robotics, artificial intelligence, and safety engineering. The discussion highlighted the need to treat embodied AI as a systems challenge requiring engineering rigor, lifecycle governance, human-centered design, and evolving standards. The paper provides practical perspectives for...
④ VISOR: A Vision-Language Model-based Test Oracle for Testing Robot
作者：Prasun Saurabh, Pablo Valle, Aitor Arrieta, Shaukat Ali, Paolo Arcaini arXiv: 2605.10408v1 链接: https://arxiv.org/abs/2605.10408v1
中文摘要：测试机器人需要评估它们是否正确、可靠和高质量地执行预期任务，这一挑战被称为软件测试中的测试预言机问题。传统上，这种评估依赖于任务正确性的特定任务符号预言和人工评估机器人行为，这是耗时、主观且容易出错的。为了解决这个问题，我们提出了VISOR ，这是一种基于视觉语言模型（ VLM ）的自动测试预言机评估方法
English Abstract：Testing robots requires assessing whether they perform their intended tasks correctly, dependably, and with high quality, a challenge known as the test oracle problem in software testing. Traditionally, this assessment relies on task-specific symbolic oracles for task correctness and on human manual evaluation of robot behavior, which is time-consuming, subjective, and error-prone. To address this, we propose VISOR, a Vision-Language Model (VLM)-based approach for automated test oracle assessment that eliminates the need of expensive human evaluations. VISOR performs automated evaluation of task correctness and quality, addressing the limitations of existing symbolic test oracles, which are task-specific and provide pass/fail judgments without explicitly quantifying task quality. Given the...
⑤ Nano-U: Efficient Terrain Segmentation for Tiny Robot Navigation
作者：Federico Pizzolato, Francesco Pasti, Nicola Bellotto arXiv: 2605.10210v1 链接: https://arxiv.org/abs/2605.10210v1
中文摘要：地形分割是在非结构化室外环境中运行的自主移动机器人的基本能力。然而，最先进的模型与微控制器典型的内存和计算约束不兼容，限制了小型机器人平台的可扩展部署。为了解决这一差距，我们开发了一个完整的框架，用于在低成本微控制器上进行强大的二进制地形分割。在我们方法的核心，我们设计了Nano-U ，这是一种高度紧凑的二进制分段
English Abstract：Terrain segmentation is a fundamental capability for autonomous mobile robots operating in unstructured outdoor environments. However, state-of-the-art models are incompatible with the memory and compute constraints typical of microcontrollers, limiting scalable deployment in small robotics platforms. To address this gap, we develop a complete framework for robust binary terrain segmentation on a low-cost microcontroller. At the core of our approach we design Nano-U, a highly compact binary segmentation network with a few thousand parameters. To compensate for the network's minimal capacity, we train Nano-U via Quantization-Aware Distillation (QAD), combining knowledge distillation and quantization-aware training. This allows the final quantized model to achieve excellent results on the Bo...
⑥ HeteroGenManip: Generalizable Manipulation For Heterogeneous Object Interactions
作者：Zhenhao Shen, Zeming Yang, Yue Chen, Yuran Wang, Shengqiang Xu, Mingleyang Li et al. arXiv: 2605.10201v1 链接: https://arxiv.org/abs/2605.10201v1
中文摘要：涉及跨类型对象交互的可推广操作是机器人技术中一项关键但具有挑战性的能力。为了可靠地完成这些任务，机器人必须解决两个基本挑战： “在哪里操纵” （接触点定位）和“如何操纵” （后续交互轨迹规划）。现有的基础模型方法通常采用端到端学习，模糊了这些阶段之间的区别，加剧了长时间的误差积累。
English Abstract：Generalizable manipulation involving cross-type object interactions is a critical yet challenging capability in robotics. To reliably accomplish such tasks, robots must address two fundamental challenges: <code>where to manipulate'' (contact point localization) and </code>how to manipulate'' (subsequent interaction trajectory planning). Existing foundation-model-based approaches often adopt end-to-end learning that obscures the distinction between these stages, exacerbating error accumulation in long-horizon tasks. Furthermore, they typically rely on a single uniform model, which fails to capture the diverse, category-specific features required for heterogeneous objects. To overcome these limitations, we propose HeteroGenManip, a task-conditioned, two-stage framework designed to decouple initial gr...
⑦ Data-Asymmetric Latent Imagination and Reranking for 3D Robotic Imitation Learning
作者：Lianghao Luo, Xizhou Bu, Ruyan Liu, Qingqiu Huang, Chufeng Tang, Xiaoshuai Hao et al. arXiv: 2605.10166v1 链接: https://arxiv.org/abs/2605.10166v1
中文摘要：机器人模仿学习通常假设可以获得最佳演示，但真实世界的数据收集往往会产生次优、探索性甚至失败的轨迹。丢弃这些数据会浪费有关环境动态和故障模式的宝贵信息，而这些信息可以用来改善决策。虽然3D策略通过强大的空间泛化减少了对高质量演示的依赖，但它们仍然需要大规模数据才能实现高任务成功率。
English Abstract：Robotic imitation learning typically assumes access to optimal demonstrations, yet real-world data collection often yields suboptimal, exploratory, or even failed trajectories. Discarding such data wastes valuable information about environment dynamics and failure modes, which can instead be leveraged to improve decision-making. While 3D policies reduce reliance on high-quality demonstrations through strong spatial generalization, they still require large-scale data to achieve high task success. To address this, we propose DALI-R, a Data-Asymmetric Latent Imagination and Reranking framework for 3D robotic imitation learning from mixed-quality trajectories. It learns a Latent World Model over 3D point clouds for imagined rollouts and a Task Completion Scorer that reranks candidate action ch...
⑧ Plan in Sandbox, Navigate in Open Worlds: Learning Physics-Grounded Abstracted Experience for Embodied Navigation
作者：Zhixuan Shen, Jiawei Du, Ziyu Guo, Han Luo, Lilan Peng, Joey Tianyi Zhou et al. arXiv: 2605.10118v1 链接: https://arxiv.org/abs/2605.10118v1
中文摘要：视觉语言模型（ VLM ）已展示出卓越的一般推理能力。然而，由于缺乏对齐的开放世界视觉和机器人控制数据，它们在嵌入式导航中的性能仍然受到阻碍。尽管模拟器为数据收集提供了一种具有成本效益的替代方案，但对逼真模拟的固有依赖通常会限制所学策略的可转移性。为此，我们建议\ textit {\ textbf {S} andbox-\ textbf {A} bstracted\ textbf {G}舍入\ t
English Abstract：Vision-Language Models (VLMs) have demonstrated exceptional general reasoning capabilities. However, their performance in embodied navigation remains hindered by a scarcity of aligned open-world vision and robot control data. Despite simulators providing a cost-effective alternative for data collection, the inherent reliance on photorealistic simulations often limits the transferability of learned policies. To this end, we propose \textit{\textbf{S}andbox-\textbf{A}bstracted \textbf{G}rounded \textbf{E}xperience} (\textbf{\textit{SAGE}}), a framework that enables agents to learn within a physics-grounded semantic abstraction rather than a photorealistic simulation, mimicking the human capacity for mental simulation where plans are rehearsed in simplified physics abstractions before executi...
4、安全对齐
① ConQuR: Corner Aligned Activation Quantization via Optimized Rotations for LLMs
作者：Chayne Thrash, Ali Abbasi, Soheil Kolouri arXiv: 2605.10793v1 链接: https://arxiv.org/abs/2605.10793v1
中文摘要：大型语言模型（ LLM ）由于其大内存占用和高推理成本而部署成本很高。权重激活量化可以降低这些成本，但低比特激活量化仍然很困难，因为激活离群值会导致较大的量化误差。最近基于旋转的方法通过应用正交变换来解决这个问题，正交变换可以跨维度重新分配激活幅度，但现有方法要么需要昂贵的端到端旋转训练，要么
English Abstract：Large language models (LLMs) are costly to deploy due to their large memory footprint and high inference cost. Weight-activation quantization can reduce these costs, but low-bit activation quantization remains difficult because activation outliers induce large quantization error. Recent rotation-based methods address this by applying orthogonal transformations that redistribute activation magnitude across dimensions, but existing approaches either require expensive end-to-end rotation training or rely on stored activation corpora, introducing significant compute or storage overhead. We propose a lightweight post-training rotation calibration method for LLM activation quantization. Our method learns orthogonal rotations that align normalized activations with the corners of an inscribed hype...
② Compander-Aligned Query Geometry for Quantized Zeroth-Order Optimization
作者：Yao Shu, Zilin Zhu arXiv: 2605.10673v1 链接: https://arxiv.org/abs/2605.10673v1
中文摘要：低位前向评估是实现内存高效零阶（ ZO ）自适应的一条有吸引力的途径：优化器只需要标量损失，并且可以在部署精度附近查询模型。障碍在于，量化ZO查询不是连续的有限差分，其次是无害的存储舍入。查询选择端点，低精度引擎对其进行四舍五入，并沿着四舍五入的弦测量损耗差。对于非均匀压缩量化器，这使得代码
English Abstract：Low-bit forward evaluation is an attractive route to memory-efficient zeroth-order (ZO) adaptation: the optimizer needs only scalar losses, and the model can be queried near deployment precision. The obstacle is that a quantized ZO query is not a continuous finite difference followed by harmless storage rounding. The query chooses endpoints, the low-precision engine rounds them, and the loss difference is measured along the rounded chord. For nonuniform companding quantizers, this makes the codebook insufficient to predict ZO behavior: a fixed weight-space radius can collapse in dense cells, over-span sparse cells, or assign a rounded chord to an unrounded update direction. We identify the missing object as query geometry and model scalar nonuniform quantization as $Q = φ^{-1} \circ U \cir...
③ Signature Approach for Contextual Bandits with Nonlinear and Path-dependent Rewards
作者：Xin Guo, Grace He, Xinyu Li arXiv: 2605.10313v1 链接: https://arxiv.org/abs/2605.10313v1
中文摘要：我们通过一种新的基于签名变换的方法来研究具有非线性和路径依赖性奖励的情境土匪。利用特征的普遍非线性性质，我们通过特征空间中的线性函数来近似连续路径依赖的奖励函数。这种表示能够使用高效的线性上下文强盗方法，同时保持表达顺序结构。在此框架的基础上，我们提出\ texttt {DisSigUCB} ，这是一个基于签名的数据
English Abstract：We study contextual bandits with nonlinear and path-dependent rewards through a novel signature-transform-based approach. Leveraging the universal nonlinearity property of signatures, we approximate continuous path-dependent reward functionals by linear functionals in the signature space. This representation enables the use of efficient linear contextual bandit methods while preserving expressive sequential structure. Building on this framework, we propose \texttt{DisSigUCB}, a signature-based disjoint upper confidence bound (UCB) algorithm. Under boundedness and non-degeneracy assumptions, we prove a high-probability data-dependent sublinear regret bound of order (\tilde{\mathcal O}(\sqrt{(d+m)KT})) where (d) is the context dimension and (m) is the signature feature dimension. Synth...
④ Breaking the Reward Barrier: Accelerating Tree-of-Thought Reasoning via Speculative Exploration
作者：Shuzhang Zhong, Haochen Huang, Shengxuan Qiu, Pengfei Zuo, Runsheng Wang, Meng Li arXiv: 2605.10195v1 链接: https://arxiv.org/abs/2605.10195v1
中文摘要：思想树（ Tree-of-Thought ， ToT ）推理结构将大型语言模型（ Large Language Model ， LLM ）推理作为基于树的搜索，显示出解决复杂数学和编程任务的强大潜力。然而，它的效率受到奖励依赖障碍的限制，这是由连续奖励引导的探索引起的同步瓶颈，限制了搜索并行性并引入了大量延迟。先前的系统优化，主要设计用于线性思维链（ CoT ）推理，
English Abstract：Tree-of-Thought (ToT) reasoning structures Large Language Model (LLM) inference as a tree-based search, demonstrating strong potential for solving complex mathematical and programming tasks. However, its efficiency is constrained by the reward dependency barrier -- a synchronization bottleneck caused by sequential reward-guided exploration that limits search parallelism and introduces substantial latency. Prior system optimizations, mainly designed for linear Chain-of-Thought (CoT) reasoning, cannot address these challenges, leaving the efficiency of ToT underexplored. To enhance ToT reasoning efficiency, we observe that the reasoning paths can be explored speculatively to break the reward synchronization barrier. Therefore, in this paper, we propose SPEX and introduce three key techniqu...
⑤ TRACE: Distilling Where It Matters via Token-Routed Self On-Policy Alignment
作者：Jiaxuan Wang, Xuan Ouyang, Zhiyu Chen, Yulan Hu, Zheng Pan, Xin Li et al. arXiv: 2605.10194v1 链接: https://arxiv.org/abs/2605.10194v1
中文摘要：策略自馏（ self-OPD ）通过让策略在特权语境下自学，以可验证的奖励（ RLVR ）密集强化学习。我们发现，当该指导跨越整个响应时，全代币KL将梯度花费在主要是冗余的位置上，并放大特权信息泄漏，从而导致熵升高、推理缩短和长视野数学训练中的分布退化。我们为Critical rEasoning （ TRACE ）提出Token-Routed Alignment ，
English Abstract：On-policy self-distillation (self-OPD) densifies reinforcement learning with verifiable rewards (RLVR) by letting a policy teach itself under privileged context. We find that when this guidance spans the full response, all-token KL spends gradients on mostly redundant positions and amplifies privileged-information leakage, causing entropy rise, shortened reasoning, and out-of-distribution degradation in long-horizon math training. We propose Token-Routed Alignment for Critical rEasoning (TRACE), which distills only on annotator-marked critical spans: forward KL on key spans of correct rollouts, optional reverse KL on localized error spans, and GRPO on all remaining tokens, with the KL channel annealed away after a short warm-up. Our analysis explains TRACE through two effects: forward KL p...
⑥ ProteinOPD: Towards Effective and Efficient Preference Alignment for Protein Design
作者：Yulin Zhang, He Cao, Zihao Jiang, Chenyi Zi, Zhipeng Zhou, Zijing Liu et al. arXiv: 2605.10189v1 链接: https://arxiv.org/abs/2605.10189v1
中文摘要：设计具有所需功能或特性的蛋白质代表了合成生物学和药物发现的核心目标。蛋白质语言模型（ PLM ）的最新进展已经能够生成高度可设计的蛋白质序列，而偏好比对提供了一种有希望的方式来指导设计实现所需的功能和属性。尽管如此，它们往往会引发灾难性的遗忘预先训练的知识，降低基本的可设计性，无法平衡多重竞争
English Abstract：Designing proteins with desired functions or properties represents a core goal in synthetic biology and drug discovery. Recent advances in protein language models (PLMs) have enabled the generation of highly designable protein sequences, while preference alignment provides a promising way to steer designs toward desired functions and properties. Nevertheless, they often trigger catastrophic forgetting of pretrained knowledge, degrading basic designability and failing to balance multiple competing objectives. To address these issues, we draw inspiration from On-Policy Distillation (OPD), an advanced post-training method renowned for mitigating catastrophic forgetting through its mode-seeking nature. In this work, we propose ProteinOPD, a multi-objective preference alignment framework that c...
⑦ Unsupervised Process Reward Models
作者：Artyom Gadetsky, Maxim Kodryan, Siba Smarak Panigrahi, Hang Guo, Maria Brbic arXiv: 2605.10158v1 链接: https://arxiv.org/abs/2605.10158v1
中文摘要：流程奖励模型（ PRM ）是一种强大的机制，通过提供细粒度、步骤级的监督来引导大型语言模型推理。然而，这种有效性需要付出巨大的代价： PRM的每个推理步骤都需要专家注释，这使得它们成本高昂且难以扩展。在这里，我们提出了一种训练无监督PRM （ uPRM ）的方法，该方法不需要人工监督，既不需要逐步注释，也不需要通过FIN的地面真实性验证
English Abstract：Process Reward Models (PRMs) are a powerful mechanism for steering large language model reasoning by providing fine-grained, step-level supervision. However, this effectiveness comes at a significant cost: PRMs require expert annotations for every reasoning step, making them costly and difficult to scale. Here, we propose a method for training unsupervised PRMs (uPRM) that requires no human supervision, neither at the level of step-by-step annotations nor through ground-truth verification of final answers. The key idea behind our approach is to define a scoring function, derived from LLM next-token probabilities, that jointly assesses candidate positions of first erroneous steps across a batch of reasoning trajectories. We demonstrate the effectiveness of uPRM across diverse scenarios: (i)...
