论文

展示并论证了算法无损性与其在有限精度算术下实现之间存在差距，主张应从精确生成轨迹与下游任务性能两个层面评估无损投机解码A gap between algorithmic losslessness and its implementation under finite-precision arithmetic is demonstrated and motivated, to motivate evaluating lossless speculative decoding at the level of exact generation trajectories as well as downstream task performance.

ModaLens 是一种配对图像交换审计，用于衡量报告可用性如何改变图像敏感性，并限制对视觉正确性的结论；该方向在另外两个模型系列中得到复现。ModaLens, a paired image-swap audit, measures how report availability changes image sensitivity and limits conclusions about visual correctness; the direction replicates in two further model lineages.

本文在 Argonne Leadership Computing Facility 的 Polaris 超级计算机上对分布式向量数据库性能进行了实证研究，选取 Qdrant 评估在最多 32 个 worker 下的插入、索引构建与 query latency。This work presents an empirical study of distributed vector database performance on the Polaris supercomputer in the Argonne Leadership Computing Facility, and selects Qdrant to evaluate insertion, index construction, and query latency with up to 32 workers.

提出 ESRL，一种架构感知的框架，显式探索 MoE 模型的专家路由空间，将高置信度专家保留为锚点，并把随机路由限制在合理候选池内，从而保留可靠的计算路径。ESRL is introduced, an architecture-aware framework that explicitly explores the expert-routing space of MoE models, and preserves high-confidence experts as anchors, and restricts stochastic routing to a plausible candidate pool, thereby retaining reliable computation paths.

该工作提出了最小干预强化学习（Minimal Intervention Reinforcement Learning, MInTRL），通过在原本的 on-policy rollout 中引入稀疏的局部干预来扩展探索边界，确立了最小干预作为增强 on-policy RL 的有效范式。This work introduces Minimal Intervention Reinforcement Learning (MInTRL), which expands the exploration frontier through sparse, local interventions in otherwise on-policy rollouts, and establishes minimal intervention as an effective paradigm for enhancing on-policy RL.

我们报告了一项历时数月的工程项目，用于构建 Sophea，一个生产级希腊语-英语双语自动语音识别系统。我们依据九个生产门控评估该系统，涵盖希腊语和英语词错误率、语言识别及非语音音频的幻觉问题。经过二十三次训练迭代和两种模型架构，没有任何训练数据组合能同时通过全部九个门控。满足希腊语噪声环境目标需要约1,500步密集领域暴露，而保持英语语言识别仅能容忍约2%We report a multi-month engineering program to build Sophea, a production bilingual Greek-English automatic speech recognition system. We evaluate the system against nine production gates covering Greek and English word error rate, language identification, and hallucinations on non-speech audio. Across twenty-three training iterations and two model architectures, no training-data composition passed all nine gates simultaneously. Meeting the Greek noisy-environment target required about 1,500 steps of dense domain exposure, while preserving English language identification tolerated only about 2

Dynin-Robotics 在 LIBERO 和零样本 LIBERO-Plus 上取得了具有竞争力的性能，并在 Franka Research 3 机器人的四种操作条件下达到了 78.4% 的平均成功率。Dynin-Robotics achieves competitive performance on LIBERO and zero-shot LIBERO-Plus, together with a 78.4% average success rate across four manipulation conditions on a Franka Research 3 robot.

该工作提出了分组值注意力（Grouped Value Attention, GVA），通过存储分组值并在推理时利用可学习的线性映射重构内容键，该映射可被吸收进 query 中，从而在预期解码路径中无需显式物化内容键。This work introduces Grouped Value Attention (GVA), which stores grouped values and reconstructs content keys with a learned linear map at inference, which can be absorbed into the query, eliminating the need to materialize content keys in the intended decode path.

外部评估显示，尽管引用有效性保持稳健，但在领域偏移下证据利用、片段对齐与拒答校准变得更加困难，表明可信的 RAG 系统需要在检索与最终答案交付之间进行显式验证。External evaluation shows that while citation validity remains strong, evidence utilization, span alignment, and refusal calibration become harder under domain shift, indicating that trustworthy RAG systems require explicit validation between retrieval and final answer delivery.

该工作提出了 SCoRE（Selection and Consolidation for Robust Evidence），一个用于显式证据选择与整合的统一 agent 循环，将最终推理与探索式试错解耦，并通过索引化的声明-图像关联确保严格的视觉锚定。This work proposes SCoRE (Selection and Consolidation for Robust Evidence), a unified agent loop for explicit evidence selection and consolidation, which decouples final reasoning from exploratory trial-and-error while ensuring strict visual grounding via indexed claim-to-image linkages.

论文提出了 E2A-Bench，一个面向金融图表推理的 969 查询基准，由 323 个 HS300 成分股在三种输入模态下构建，并附带由 OHLCV 确定性派生的证据锚点；结果表明金融 VLM 评估应追溯从证据到决策的完整链路，而非依赖单一幻觉分数。E2A-Bench is introduced, a 969-query benchmark for financial chart reasoning, constructed from 323 HS300 constituents under three input modalities with deterministic OHLCV-derived evidence anchors, and results show that financial VLM evaluation should trace the full evidence-to-action chain rather than rely on a single hallucination score.

人类认知的一个核心原则是系统性，即理解一个概念本质上与理解该概念的相近变体相关联。推理模型能否稳健地展现这种系统性？如果可以，我们应能预期模型在其结构等价变体任务上表现一致。本文扩展了认知科学中既定的规则归纳任务，以评估当前推理模型思维的系统性。每个任务族都具有组合结构，我们借此通过任务同构创建结构等价的任务变体...A central tenet of human cognition is systematicity, the principle that understanding one concept is inherently tied to understanding close variations of that concept. Do reasoning models robustly exhibit such systematicity? If so, we would expect consistent performance on structurally equivalent variants of the same task. Here, we extend established rule induction tasks from cognitive science to assess the systematicity of thought in current reasoning models. Each task family has compositional structure that we use to create structurally equivalent task variations through task isomorphisms su

提出 Continual Search，一个迭代框架：在多轮对话中持续推动判别器搜索尚未解决的诊断证据，在多个基准测试套件和模型系列上一致提升归因性能Continual Search is introduced, an iterative framework that nudges the judge, over successive turns, to keep searching for unresolved diagnostic evidence, which consistently improves attribution performance across multiple benchmark suites and model families.

本文提出一个无偏的多进程 evaluation 框架，能够有效分散 client 端负载，从而在每秒数千次 query 以上的生产规模下实现对 LLM 的精确、可复现 profiling。This work proposes an unbiased, multi-process evaluation framework that effectively distributes client-side load, enabling accurate, reproducible profiling of LLMs at production scales exceeding thousands of queries per second.

结果表明，模型级对齐并不具备可组合性：单独能力强且看似安全的 agent 在组成系统后，可能随着 AI 的持续性与互联化而产生性质上不同的失效模式。The results suggest that model-level alignment is not compositional: individually capable and apparently safe agents can form systems with qualitatively different failure modes as AI becomes persistent and interconnected.

检索增强生成（RAG）管线通常依赖在预处理阶段确定的固定索引与检索配置。这种一刀切的设计难以适配领域专家场景，因为异构查询需要不同的分块粒度、元数据约束与来源选择策略。因此，针对某一类查询有效的配置，往往在其他类查询上表现欠佳。本文提出 ORDER（Optimal Routing for Dynamic Evidence Retrieval），一种查询条件化的 RAG 框架，可联合自适应地调整索引与……Retrieval-Augmented Generation (RAG) pipelines typically rely on a fixed indexing and retrieval configuration determined at preprocessing time. This one-size-fits-all design is ill-suited to domain-expert settings, where heterogeneous queries require different chunking granularities, metadata constraints, and source-selection strategies. As a result, configurations that are effective for one family of queries often perform poorly for others. In this paper, we introduce ORDER (Optimal Routing for Dynamic Evidence Retrieval), a query-conditioned RAG framework that jointly adapts indexing and ret

论文提出了 InceptionRAG，一种颠覆 RAG 标准攻击范式的隐蔽攻击机制，展现出更优的规避能力，能够有效绕过针对传统单文档注入的既有防御。This paper introduces InceptionRAG, a stealthy attack mechanism that subverts the standard attack paradigm of RAG, and shows superior evasion capabilities, effectively bypassing established defenses that mitigate traditional single-document injections.

论文提出了 ModAR，这是首个在预测动作前以自回归方式对多种未来模态进行去噪的 WAM；研究发现 ModAR 的序列化生成优于现有 WAM 形式，并在所有评估数据规模下取得最高的平均成功率。ModAR is introduced, the first WAM to autoregressively denoise multiple future modalities before predicting actions, and it is found that ModAR's sequential generation outperforms existing WAM formulations, with the highest average success rate at all evaluated data scales.

PhysStream 是一个用于物理驱动图像到视频合成的自回归模型，通过引入结构化场景记忆并支持基于稀疏速度增量信号的细粒度运动控制（编码物理量），使模型能够学习底层动力学。PhysStream is an autoregressive model for physics-grounded image-to-video synthesis that incorporates structured scene memory and supports fine-grained motion control via sparse velocity-increment signals that encode physical quantities, letting the model learn the underlying dynamics.

Mind2Dialogue 框架提出了一个心理学引导的模拟器，在交互过程中保留个人特征并更新心智状态以生成连贯对话；通过对 Oracle 信息充分的回复进行训练，使模型在部署时无需直接访问用户心智状态即可提供帮助。The Mind2Dialogue framework proposes a psychology-guided simulator that preserves personal characteristics while updating mental states through interaction to generate coherent conversations, and trains models on the Oracle's well-informed responses to assist users without direct access to their mental states at deployment.

该工作表明，专用优化隐式地从此潜空间轨迹中进行选择，并建立了一种新的专用训练视角：当缺少 gold reasoning 时，调参选择直接控制传递给下游模型的潜在监督信号。This work shows that specialist optimization implicitly selects from this latent trajectory space, and establishes a new view of specialist training: when gold reasoning is absent, tuning choices directly control the latent supervision passed to downstream models.

该工作利用 Headroom-Closed Index（HCI）揭示现有 LLM 的问题，并提出 RSI 概念及其发展路线图：从改进执行自主性、改进策略自主性、经验获取自主性、环境适应自主性，到递归元改进。This work uses the Headroom-Closed Index (HCI) to reveal the problems of existing LLMs, and introduces the RSI concept and its development roadmap: from improvement-execution autonomy, improvement-strategy autonomy, experience-acquisition autonomy, and environment-adaptation autonomy, to recursive meta-improvement.

论文提出了 ImpossibleRubrics，一个包含 169 个不可能任务的基准，覆盖六类不可能性类别，每项任务都配有可验证的 oracle 证书，规定诚实回答可以与不可以声明的内容，并附带 48 个可回答的对照样本。ImpossibleRubrics is introduced, a benchmark of 169 impossible tasks spanning six impossibility categories, each paired with a verifiable oracle certificate specifying what an honest answer may and may not claim, together with 48 answerable controls.

本工作提出 HarnessVLN，一个零样本、无需训练的框架：通过共享的 Agent Harness 统一指令跟随与物体目标导航，并展示了其在真实世界中两类导航任务上的适用性This work introduces HarnessVLN, a zero-shot, training-free framework that unifies instruction-following and object-goal navigation through a shared Agent Harness and demonstrates its applicability to both navigation tasks in real-world environments.

本文提出 DualPath，一种 inference 系统，通过引入 dual-path KV-Cache loading 打破瓶颈，并实现一条新的 storage-to-decode 路径：KV-Cache 先加载到 decode engine，再通过 compute network 上的 RDMA 高效转发至 prefill engine。DualPath is presented, an inference system that breaks this bottleneck by introducing dual-path KV-Cache loading and enables a novel storage-to-decode path, in which the KV-Cache is loaded into decoding engines and then efficiently transferred to prefill engines via RDMA over the compute network.

论文提出了 ModularRSI，一个与基准解耦、对比式、模块化的可泛化 harness 进化框架，对同一任务下成功与失败的轨迹进行对比，并跨任务聚合证据以识别反复出现的行为缺陷。ModularRSI is proposed, a benchmark-disjoint, contrastive, and modular framework for generalizable harness evolution that contrasts successful and failed trajectories for the same task and aggregates evidence across tasks to identify recurring behavioral deficiencies.

论文报告了来自 OpenAI、Anthropic、xAI 与 Google DeepMind 的六类前沿模型实验，并使用 epistemic jailbreak 一词来指称随请求具体性增加而伴随出现的技术溯源严谨性丧失现象。This paper reports experiments across six frontier model types from OpenAI, Anthropic, xAI, and Google DeepMind, and uses the term epistemic jailbreak for the accompanying loss of discipline in technical provenance as requested specificity increases.

StepAudio 3 Realtime 是一个围绕持续 listen-converse-think-act 循环组织的音频-语言基础模型，在实时语音输出的同时达到与专用推理模型相当的对话与推理性能，并通过 Think-While-Speaking 机制化解深度思考与延迟之间的矛盾。StepAudio 3 Realtime, an audio-language foundation model organized around a continuous listen-converse-think-act loop, achieves dialogue and reasoning performance comparable to dedicated reasoning models while speaking in real time, and resolves the tension between deep deliberation and latency via Think-While-Speaking.

该工作提出了 StepAudio 3 Music，一个支持显式音乐规划与开放域文本控制生成的大规模长篇音乐生成模型，在所评估系统中取得最高的 AudioBox Content Enjoyment、Content Usefulness 与 Production Quality 分数，以及最高的 MuQ-MuLan 相似度。This work introduces StepAudio 3 Music, a large-scale, long-form music generation model that supports explicit musical planning and open-domain text-controlled generation, and achieves the highest AudioBox Content Enjoyment, Content Usefulness, and Production Quality scores and the highest MuQ-MuLan similarity among the evaluated systems.

本文重新审视联合多模态表示学习与生成，旨在产生可直接被生成式解码器使用的线性可插值嵌入，并确保其表示同时充当判别性语义描述符和生成条件。This work revisits joint multimodal representation learning and generation to produce linearly interpolatable embeddings that are directly consumable by generative decoders and ensures its representations function as both discriminative semantic descriptors and generative conditions.

我们推出 LimiX 家族新模型 LimiX-2，通过先前建立的 scaling laws 指导模型与数据规模扩展。LimiX-2 采用上下文机制网络 (CMNs) 范式，并以上下文条件掩码建模 (CCMM) 进行预训练。CMNs 将上下文学习的组织原则从以目标为中心的预测转向以机制为导向的联合建模。它并非围绕传统表格 PFN 的 p(y|x, D_context) 目标设计网络，而是围绕学习 p(x, y|D_context)——一种上下文依赖的表征We introduce LimiX-2, a new model in the LimiX family, developed through model and data scaling guided by our previously established scaling laws. LimiX-2 adopts the Contextual Mechanism Networks (CMNs) paradigm and is pretrained with Context-Conditional Masked Modeling (CCMM). CMNs shifts the organizing principle of in-context learning from target-centric prediction to mechanism-oriented joint modeling. Rather than centering the network on the p(y mid x, D_{context}) objective of conventional tabular PFNs, it is designed around learning p(x, y mid D_{context}), a context-dependent representat

Register被实现为专用的固定位置 token，其连续的隐藏状态被训练用于跨生成块承载推理进度，对有界代码生成尤其有效，因为正确程序通常跨越多个块。Registers are implemented as dedicated fixed-position tokens whose continuous hidden states are trained to carry reasoning progress across generation chunks, which are especially effective for bounded code generation, where correct programs usually span several chunks.

本文提出 OmniHarness，一个通过符号策略学习实现可泛化视觉生成的框架，将已验证的执行抽象为视觉生成任务族的符号策略，捕获共享流程和适用条件，同时去除实例特定的输入。OmniHarness is introduced, a framework for generalizable visual generation via symbolic policy learning that abstracts verified executions into symbolic policies for visual generation task families, capturing shared procedures and applicability conditions while removing instance-specific inputs.

本文提出一个受控的跨模态框架，在多种模态下实例化相同的任务套件以检验 Convergent Emergence Hypothesis，结果显示配对映射的 ICL 在六种模态中出现，超越受控基线，并在其中五种模态上呈现相关的逐任务效应。A controlled cross-modality framework that instantiates the same task suite in a variety of modalities to test the Convergent Emergence Hypothesis and shows that paired-mapping ICL emerges across six modalities, surpasses controlled baselines, and has correlated per-task effects across five of them.

本文提出 RelateAnything，一个 53M 参数的模型，输入一张图像和来自任意来源的区域，针对推理时以字符串形式提供的谓词词汇返回带分数的关系。This work presents RelateAnything, a 53M-parameter model taking an image and regions from any source and returning scored relations over a predicate vocabulary supplied at inference as strings, a 53M-parameter model taking an image and regions from any source and returning scored relations over a predicate vocabulary supplied at inference as strings.

本文探讨了 cloud platform 与 distributed system 在支撑 LLM 可扩展性、效率与优化方面的作用，涵盖数据管理、资源优化，以及对 microservices、autoscaling 与 hybrid cloud-edge 方案的需求。The role of cloud platforms and distributed systems in supporting the scalability, efficiency, and optimization of LLMs is explored, including data management, resource optimization, and the need for microservices, autoscaling, and hybrid cloud-edge solutions.