中文使用 googletrans 翻译，翻译不对的地方以英文为准

目录

法学硕士可以清理你的烂摊子吗？法学硕士应用程序就绪数据准备调查

英文摘要

Data preparation aims to denoise raw datasets, uncover cross-dataset relationships, and extract valuable insights from them, which is essential for a wide range of data-centric applications. Driven by (i) rising demands for application-ready data (e.g., for analytics, visualization, decision-making), (ii) increasingly powerful LLM techniques, and (iii) the emergence of infrastructures that facilitate flexible agent construction (e.g., using Databricks Unity Catalog), LLM-enhanced methods are rapidly becoming a transformative and potentially dominant paradigm for data preparation. By investigating hundreds of recent literature works, this paper presents a systematic review of this evolving landscape, focusing on the use of LLM techniques to prepare data for diverse downstream tasks. First, we characterize the fundamental paradigm shift, from rule-based, model-specific pipelines to prompt-driven, context-aware, and agentic preparation workflows. Next, we introduce a task-centric taxonomy that organizes the field into three major tasks: data cleaning (e.g., standardization, error processing, imputation), data integration (e.g., entity matching, schema matching), and data enrichment (e.g., data annotation, profiling). For each task, we survey representative techniques, and highlight their respective strengths (e.g., improved generalization, semantic understanding) and limitations (e.g., the prohibitive cost of scaling LLMs, persistent hallucinations even in advanced agents, the mismatch between advanced methods and weak evaluation). Moreover, we analyze commonly used datasets and evaluation metrics (the empirical part). Finally, we discuss open research challenges and outline a forward-looking roadmap that emphasizes scalable LLM-data systems, principled designs for reliable agentic workflows, and robust evaluation protocols.

中文摘要

数据准备旨在对原始数据集进行去噪，揭示跨数据集关系，并从中提取有价值的见解，这对于各种以数据为中心的应用程序至关重要。在 (i) 对应用程序就绪数据的需求不断增长（例如，用于分析、可视化、决策）、(ii) 日益强大的 LLM 技术以及 (iii) 促进灵活代理构建的基础设施的出现（例如，使用 Databricks Unity Catalog）的推动下，LLM 增强方法正在迅速成为数据准备的变革性和潜在主导范式。通过调查数百篇最近的文献著作，本文对这一不断发展的前景进行了系统回顾，重点关注使用法学硕士技术为各种下游任务准备数据。首先，我们描述了基本范式的转变，从基于规则的、特定于模型的管道到提示驱动的、上下文感知的和代理的准备工作流程。接下来，我们引入以任务为中心的分类法，将该领域组织为三个主要任务：数据清理（例如标准化、错误处理、插补）、数据集成（例如实体匹配、模式匹配）和数据丰富（例如数据注释、分析）。对于每项任务，我们都会调查代表性技术，并强调它们各自的优势（例如，改进的泛化、语义理解）和局限性（例如，扩展法学硕士的高昂成本、即使在高级代理中也存在持续的幻觉、高级方法与弱评估之间的不匹配）。此外，我们分析了常用的数据集和评估指标（实证部分）。最后，我们讨论了开放研究挑战，并概述了一个前瞻性的路线图，强调可扩展的法学硕士数据系统、可靠的代理工作流程的原则性设计以及强大的评估协议。

​

Idea2Story：将研究概念转化为完整科学叙述的自动化流程

英文摘要

Autonomous scientific discovery with large language model (LLM)-based agents has recently made substantial progress, demonstrating the ability to automate end-to-end research workflows. However, existing systems largely rely on runtime-centric execution paradigms, repeatedly reading, summarizing, and reasoning over large volumes of scientific literature online. This on-the-spot computation strategy incurs high computational cost, suffers from context window limitations, and often leads to brittle reasoning and hallucination. We propose Idea2Story, a pre-computation-driven framework for autonomous scientific discovery that shifts literature understanding from online reasoning to offline knowledge construction. Idea2Story continuously collects peer-reviewed papers together with their review feedback, extracts core methodological units, composes reusable research patterns, and organizes them into a structured methodological knowledge graph. At runtime, underspecified user research intents are aligned to established research paradigms, enabling efficient retrieval and reuse of high-quality research patterns instead of open-ended generation and trial-and-error. By grounding research planning and execution in a pre-built knowledge graph, Idea2Story alleviates the context window bottleneck of LLMs and substantially reduces repeated runtime reasoning over literature. We conduct qualitative analyses and preliminary empirical studies demonstrating that Idea2Story can generate coherent, methodologically grounded, and novel research patterns, and can produce several high-quality research demonstrations in an end-to-end setting. These results suggest that offline knowledge construction provides a practical and scalable foundation for reliable autonomous scientific discovery.

中文摘要

使用基于大语言模型（LLM）的代理的自主科学发现最近取得了实质性进展，展示了自动化端到端研究工作流程的能力。然而，现有系统很大程度上依赖于以运行时为中心的执行范例，对大量在线科学文献进行反复阅读、总结和推理。这种现场计算策略会产生很高的计算成本，受到上下文窗口的限制，并且经常导致脆弱的推理和幻觉。我们提出了 Idea2Story，这是一个用于自主科学发现的预计算驱动框架，它将文献理解从在线推理转变为离线知识构建。Idea2Story不断收集同行评审的论文及其评审反馈，提取核心方法论单元，组成可重用的研究模式，并将它们组织成结构化的方法论知识图谱。在运行时，未明确的用户研究意图与已建立的研究范式保持一致，从而能够有效检索和重用高质量的研究模式，而不是开放式生成和试错。通过将研究计划和执行建立在预先构建的知识图中，Idea2Story 缓解了法学硕士的上下文窗口瓶颈，并大大减少了对文献的重复运行时推理。我们进行定性分析和初步实证研究，证明 Idea2Story 可以生成连贯的、有方法论依据的新颖的研究模式，并可以在端到端的环境中产生一些高质量的研究论证。这些结果表明，离线知识构建为可靠的自主科学发现提供了实用且可扩展的基础。

​

LongCat-Flash-Thinking-2601技术报告

  • 标题: LongCat-Flash-Thinking-2601 Technical Report
  • 作者: Meituan LongCat Team, Anchun Gui, Bei Li, Bingyang Tao, Bole Zhou, Borun Chen, Chao Zhang, Chao Zhang, Chen Gao, Chen Zhang, Chengcheng Han, Chenhui Yang, Chuyu Zhang, Cong Chen, Cunguang Wang, Daoru Pan, Defei Bu, Dengchang Zhao, Di Xiu, Dishan Liu, Dongyu Ru, Dunwei Tu, Fan Wu, Fengcheng Yuan, Fengcun Li, Gang Xu, Guanyu Wu, Guoyuan Lin, Haibin Wang, Hansi Yang, Hao Yang, Haonan Yan, Haoxiang Ma, Haoxing Wen, Hongyan Hao, Hongyin Tang, Hongyu Zang, Hongzhi Ni, Hui Su, Jiacheng Zhang, Jiahong Zhou, Jiahuan Li, Jiaming Wang, Jian Yang, Jianfei Zhang, Jianhao Xu, Jianing Wang, Jiapeng Zhu, Jiaqi Sun, Jiarong Shi, Jiarui Zhao, Jingang Wang, Jinluan Yang, Jinrui Ding, Jinwei Xiao, Jiyuan He, Juncan Xu, Kefeng Zhang, Keheng Wang, Li Wei, Lianhui Ma, Lin Qiu, Lingbing Kong, Lingchuan Liu, Linsen Guo, Mengshen Zhu, Mengxia Shen, Mingyang Zhu, Peiguang Li, Peng Pei, Pengcheng Jia, Pengtao Zhang, Peng Zhao, Qi Gu, Qiong Huang, Qiyuan Duan, Quanchi Weng, Rongxiang Weng, Rongzhi Zhang, Rumei Li, Shanglin Lei, Shengnan An, Shijun Dai, Shuaikang Liu, Shuang Zhou, Shuo Wang, Songyuan Zhao, Tao Liang, Tianhao Hu, Tianze Chen, Wei Liu, Wei Shi, Wei Wang, Weifeng Tang, Wenjie Shi, Wenlong Zhu, Wentao Chen, Wentao Shi, Xi Su, Xiangcheng Liu, Xiandi Ma, Xiangyu Xi, Xiangyuan Liu, Xiangzhou Huang, Xiao Liu, Xiaodong Cai, Xiaolong Chen, Xiaowei Shi, Xiaoyu Li, Xin Chen, Xingchen Liu, Xuan Huang, Xuezhi Cao, Xunliang Cai, Yan Chen, Yang Bai, Yang Liu, Yang Yang, Yang Zheng, Yaoming Wang, Yaoming Zhu, Yaqi Huo, Yanyu Chen, Yaorui Shi, Yerui Sun, Yi Zhang, Yihao Chen, Yi-Kai Zhang, Yifan Lu, Yifan Zhao, Yitao Zhai, Yongjing Yin, Yongwei Zhou, Youshao Xiao, Yuchuan Dai, Yuchen Xie, Yuchen Yu, Yufei Zhang, Yuhuai Wei, Yulei Qian, Yunfan Liang, Yunke Zhao, Yuwei Jiang, Yuxin Bian, Yuxin Chen, Yuxin Liu, Yue Xu, Yueqing Sun, Zeyang Yu, Zhao Yang, Zhengsheng Huang, Zhengyu Chen, Zhijian Liu, Zhikang Xia, Zhimin Lin, Zhiyuan Yao, Zhuofan Chen, Zhuowen Han, Zijian Zhang, Ziran Li, Ziwen Wang, Ziyuan Zhuang
  • 日期: 2026-01-23
  • ArXiv主页: https://arxiv.org/abs/2601.16725
  • 论文链接: https://arxiv.org/pdf/2601.16725
  • 项目链接: https://github.com/meituan-longcat/LongCat-Flash-Thinking-2601
  • gitHub仓库: https://github.com/meituan-longcat/LongCat-Flash-Thinking-2601

英文摘要

We introduce LongCat-Flash-Thinking-2601, a 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model with superior agentic reasoning capability. LongCat-Flash-Thinking-2601 achieves state-of-the-art performance among open-source models on a wide range of agentic benchmarks, including agentic search, agentic tool use, and tool-integrated reasoning. Beyond benchmark performance, the model demonstrates strong generalization to complex tool interactions and robust behavior under noisy real-world environments. Its advanced capability stems from a unified training framework that combines domain-parallel expert training with subsequent fusion, together with an end-to-end co-design of data construction, environments, algorithms, and infrastructure spanning from pre-training to post-training. In particular, the model’s strong generalization capability in complex tool-use are driven by our in-depth exploration of environment scaling and principled task construction. To optimize long-tailed, skewed generation and multi-turn agentic interactions, and to enable stable training across over 10,000 environments spanning more than 20 domains, we systematically extend our asynchronous reinforcement learning framework, DORA, for stable and efficient large-scale multi-environment training. Furthermore, recognizing that real-world tasks are inherently noisy, we conduct a systematic analysis and decomposition of real-world noise patterns, and design targeted training procedures to explicitly incorporate such imperfections into the training process, resulting in improved robustness for real-world applications. To further enhance performance on complex reasoning tasks, we introduce a Heavy Thinking mode that enables effective test-time scaling by jointly expanding reasoning depth and width through intensive parallel thinking.

中文摘要

我们推出了 LongCat-Flash-Thinking-2601，这是一个具有 5600 亿参数的开源混合专家 (MoE) 推理模型，具有卓越的代理推理能力。LongCat-Flash-Thinking-2601 在各种代理基准上实现了开源模型中最先进的性能，包括代理搜索、代理工具使用和工具集成推理。除了基准性能之外，该模型还展示了对复杂工具交互的强大泛化能力以及嘈杂的现实环境下的稳健行为。其先进能力源于统一的训练框架，将领域并行专家训练与后续融合相结合，以及从训练前到训练后的数据构建、环境、算法和基础设施的端到端协同设计。特别是，该模型在复杂工具使用方面的强大泛化能力是由我们对环境扩展和原则性任务构建的深入探索驱动的。为了优化长尾、倾斜生成和多轮代理交互，并实现跨 20 多个领域的 10,000 多个环境的稳定训练，我们系统地扩展了异步强化学习框架 DORA，以实现稳定高效的大规模多环境训练。此外，认识到现实世界的任务本质上是有噪声的，我们对现实世界的噪声模式进行系统分析和分解，并设计有针对性的训练程序，以明确地将这些缺陷纳入训练过程，从而提高现实世界应用的鲁棒性。为了进一步提高复杂推理任务的性能，我们引入了重思考模式，通过密集的并行思维共同扩展推理深度和宽度，从而实现有效的测试时间扩展。

​

推进开源世界模式

英文摘要

We present LingBot-World, an open-sourced world simulator stemming from video generation. Positioned as a top-tier world model, LingBot-World offers the following features. (1) It maintains high fidelity and robust dynamics in a broad spectrum of environments, including realism, scientific contexts, cartoon styles, and beyond. (2) It enables a minute-level horizon while preserving contextual consistency over time, which is also known as “long-term memory”. (3) It supports real-time interactivity, achieving a latency of under 1 second when producing 16 frames per second. We provide public access to the code and model in an effort to narrow the divide between open-source and closed-source technologies. We believe our release will empower the community with practical applications across areas like content creation, gaming, and robot learning.

中文摘要

我们推出 LingBot-World，一个源于视频生成的开源世界模拟器。LingBot-World定位于顶级世界模型，具有以下特点。(1) 它在广泛的环境中保持高保真度和强大的动态，包括现实主义、科学背景、卡通风格等。（2）它能够实现分钟级的视野，同时保持一段时间内的上下文一致性，这也称为“长期记忆”。(3) 支持实时交互，每秒生成16帧时延迟低于1秒。我们提供对代码和模型的公共访问，以努力缩小开源和闭源技术之间的鸿沟。我们相信我们的发布将为社区提供内容创建、游戏和机器人学习等领域的实际应用。

​

daVinci-Dev：软件工程的原生代理中期培训

英文摘要

Recently, the frontier of Large Language Model (LLM) capabilities has shifted from single-turn code generation to agentic software engineering-a paradigm where models autonomously navigate, edit, and test complex repositories. While post-training methods have become the de facto approach for code agents, **agentic mid-training**-mid-training (MT) on large-scale data that mirrors authentic agentic workflows-remains critically underexplored due to substantial resource requirements, despite offering a more scalable path to instilling foundational agentic behaviors than relying solely on expensive reinforcement learning. A central challenge in realizing effective agentic mid-training is the distribution mismatch between static training data and the dynamic, feedback-rich environment of real development. To address this, we present a systematic study of agentic mid-training, establishing both the data synthesis principles and training methodology for effective agent development at scale. Central to our approach is **agent-native data**-supervision comprising two complementary types of trajectories: **contextually-native trajectories** that preserve the complete information flow an agent experiences, offering broad coverage and diversity; and **environmentally-native trajectories** collected from executable repositories where observations stem from actual tool invocations and test executions, providing depth and interaction authenticity. We verify the model’s agentic capabilities on SWE-Bench Verified. We demonstrate our superiority over the previous open software engineering mid-training recipe Kimi-Dev under two post-training settings with an aligned base model and agentic scaffold, while using less than half mid-training tokens (73.1B). Besides relative advantage, our best performing 32B and 72B models achieve **56.1%** and **58.5%** resolution rates, respectively, which are …

中文摘要

最近，大型语言模型 (LLM) 功能的前沿已从单轮代码生成转向代理软件工程——模型自主导航、编辑和测试复杂存储库的范例。虽然训练后方法已经成为代码代理事实上的方法，但由于大量的资源需求，在反映真实代理工作流程的大规模数据上进行代理中期训练（MT）仍然严重不足，尽管它提供了比仅仅依赖昂贵的强化学习更可扩展的途径来灌输基础代理行为。实现有效的代理中期训练的一个核心挑战是静态训练数据与真实开发的动态、反馈丰富的环境之间的分布不匹配。为了解决这个问题，我们对代理中期培训进行了系统研究，建立了大规模有效代理开发的数据合成原则和培训方法。我们方法的核心是**代理原生数据**-监督，包括两种互补类型的轨迹：**上下文原生轨迹**保留代理体验的完整信息流，提供广泛的覆盖范围和多样性；和从可执行存储库收集的**环境原生轨迹**，其中观察源于实际的工具调用和测试执行，提供深度和交互真实性。我们在“SWE-Bench Verified”上验证了模型的代理能力。我们在两种训练后设置下展示了我们相对于之前的开放软件工程中期训练配方“Kimi-Dev”的优越性，其中使用了对齐的基础模型和代理支架，同时使用了不到一半的中期训练令牌（73.1B）。除了相对优势之外，我们性能最佳的 32B 和 72B 型号分别实现了 **56.1%** 和 **58.5%** 分辨率，这是…

​

AgentDoG：AI 代理安全的诊断护栏框架

  • 标题: AgentDoG: A Diagnostic Guardrail Framework for AI Agent Safety and Security

  • 作者: Dongrui Liu, Qihan Ren, Chen Qian, Shuai Shao, Yuejin Xie, Yu Li, Zhonghao Yang, Haoyu Luo, Peng Wang, Qingyu Liu, Binxin Hu, Ling Tang, Jilin Mei, Dadi Guo, Leitao Yuan, Junyao Yang, Guanxu Chen, Qihao Lin, Yi Yu, Bo Zhang, Jiaxuan Guo, Jie Zhang, Wenqi Shao, Huiqi Deng, Zhiheng Xi, Wenjie Wang, Wenxuan Wang, Wen Shen, Zhikai Chen, Haoyu Xie, Jialing Tao, Juntao Dai, Jiaming Ji, Zhongjie Ba, Linfeng Zhang, Yong Liu, Quanshi Zhang, Lei Zhu, Zhihua Wei, Hui Xue, Chaochao Lu, Jing Shao, Xia Hu

  • 日期: 2026-01-26

  • ArXiv主页: https://arxiv.org/abs/2601.18491

  • 论文链接: https://arxiv.org/pdf/2601.18491

  • gitHub仓库: https://github.com/AI45Lab/AgentDoG

英文摘要

The rise of AI agents introduces complex safety and security challenges arising from autonomous tool use and environmental interactions. Current guardrail models lack agentic risk awareness and transparency in risk diagnosis. To introduce an agentic guardrail that covers complex and numerous risky behaviors, we first propose a unified three-dimensional taxonomy that orthogonally categorizes agentic risks by their source (where), failure mode (how), and consequence (what). Guided by this structured and hierarchical taxonomy, we introduce a new fine-grained agentic safety benchmark (ATBench) and a Diagnostic Guardrail framework for agent safety and security (AgentDoG). AgentDoG provides fine-grained and contextual monitoring across agent trajectories. More Crucially, AgentDoG can diagnose the root causes of unsafe actions and seemingly safe but unreasonable actions, offering provenance and transparency beyond binary labels to facilitate effective agent alignment. AgentDoG variants are available in three sizes (4B, 7B, and 8B parameters) across Qwen and Llama model families. Extensive experimental results demonstrate that AgentDoG achieves state-of-the-art performance in agentic safety moderation in diverse and complex interactive scenarios. All models and datasets are openly released.

中文摘要

人工智能代理的兴起带来了由于自主工具使用和环境交互而产生的复杂的安全挑战。目前的护栏模型缺乏主体风险意识和风险诊断的透明度。为了引入涵盖复杂且众多风险行为的代理护栏，我们首先提出了一个统一的三维分类法，该分类法根据代理风险的来源（何处）、故障模式（如何）和后果（什么）对代理风险进行正交分类。在这种结构化和分层分类法的指导下，我们引入了新的细粒度代理安全基准（ATBench）和用于代理安全和保障的诊断护栏框架（AgentDoG）。AgentDoG 提供跨代理轨迹的细粒度和上下文监控。更重要的是，AgentDoG 可以诊断不安全操作和看似安全但不合理的操作的根本原因，提供超越二进制标签的来源和透明度，以促进有效的代理对齐。AgentDoG 变体在 Qwen 和 Llama 模型系列中提供三种尺寸（4B、7B 和 8B 参数）。大量的实验结果表明，AgentDoG 在各种复杂的交互场景中在代理安全调节方面实现了最先进的性能。All models and datasets are openly released.

​

越难越好：通过难度感知 GRPO 和多方面问题重新表述来促进数学推理

英文摘要

Reinforcement Learning with Verifiable Rewards (RLVR) offers a robust mechanism for enhancing mathematical reasoning in large models. However, we identify a systematic lack of emphasis on more challenging questions in existing methods from both algorithmic and data perspectives, despite their importance for refining underdeveloped capabilities. Algorithmically, widely used Group Relative Policy Optimization (GRPO) suffers from an implicit imbalance where the magnitude of policy updates is lower for harder questions. Data-wise, augmentation approaches primarily rephrase questions to enhance diversity without systematically increasing intrinsic difficulty. To address these issues, we propose a two-dual MathForge framework to improve mathematical reasoning by targeting harder questions from both perspectives, which comprises a Difficulty-Aware Group Policy Optimization (DGPO) algorithm and a Multi-Aspect Question Reformulation (MQR) strategy. Specifically, DGPO first rectifies the implicit imbalance in GRPO via difficulty-balanced group advantage estimation, and further prioritizes harder questions by difficulty-aware question-level weighting. Meanwhile, MQR reformulates questions across multiple aspects to increase difficulty while maintaining the original gold answer. Overall, MathForge forms a synergistic loop: MQR expands the data frontier, and DGPO effectively learns from the augmented data. Extensive experiments show that MathForge significantly outperforms existing methods on various mathematical reasoning tasks. The code and augmented data are all available at https://github.com/AMAP-ML/MathForge.

中文摘要

带有可验证奖励的强化学习 (RLVR) 提供了一种强大的机制来增强大型模型中的数学推理。然而，我们从算法和数据的角度发现，现有方法中系统性地缺乏对更具挑战性问题的重视，尽管它们对于完善未开发的能力很重要。从算法上讲，广泛使用的组相对策略优化（GRPO）存在隐性不平衡，即对于更难的问题，策略更新的幅度较低。从数据角度来看，增强方法主要是改写问题以增强多样性，而不会系统地增加内在难度。为了解决这些问题，我们提出了一个双 MathForge 框架，通过从两个角度针对更难的问题来改进数学推理，该框架包括难度感知组策略优化（DGPO）算法和多方面问题重构（MQR）策略。具体来说，DGPO首先通过难度平衡的群体优势估计来纠正GRPO中隐含的不平衡，并进一步通过难度感知的问题级别权重来优先考虑较难的问题。同时，MQR 重新设计了多个方面的问题，以增加难度，同时保持原来的黄金答案。总体而言，MathForge 形成了一个协同循环：MQR 扩展了数据前沿，DGPO 有效地从增强的数据中学习。大量实验表明，MathForge 在各种数学推理任务上显着优于现有方法。代码和增强数据均可在 https://github.com/AMAP-ML/MathForge 上获取。

​

一切就位：文本到图像模型的空间智能基准测试

英文摘要

Text-to-image (T2I) models have achieved remarkable success in generating high-fidelity images, but they often fail in handling complex spatial relationships, e.g., spatial perception, reasoning, or interaction. These critical aspects are largely overlooked by current benchmarks due to their short or information-sparse prompt design. In this paper, we introduce SpatialGenEval, a new benchmark designed to systematically evaluate the spatial intelligence of T2I models, covering two key aspects: (1) SpatialGenEval involves 1,230 long, information-dense prompts across 25 real-world scenes. Each prompt integrates 10 spatial sub-domains and corresponding 10 multi-choice question-answer pairs, ranging from object position and layout to occlusion and causality. Our extensive evaluation of 21 state-of-the-art models reveals that higher-order spatial reasoning remains a primary bottleneck. (2) To demonstrate that the utility of our information-dense design goes beyond simple evaluation, we also construct the SpatialT2I dataset. It contains 15,400 text-image pairs with rewritten prompts to ensure image consistency while preserving information density. Fine-tuned results on current foundation models (i.e., Stable Diffusion-XL, Uniworld-V1, OmniGen2) yield consistent performance gains (+4.2%, +5.7%, +4.4%) and more realistic effects in spatial relations, highlighting a data-centric paradigm to achieve spatial intelligence in T2I models.

中文摘要

文本到图像（T2I）模型在生成高保真图像方面取得了显着的成功，但它们常常无法处理复杂的空间关系，例如空间感知、推理或交互。由于提示设计简短或信息稀疏，当前的基准在很大程度上忽略了这些关键方面。在本文中，我们介绍了 SpatialGenEval，这是一个旨在系统评估 T2I 模型的空间智能的新基准，涵盖两个关键方面：（1）SpatialGenEval 涉及跨 25 个现实世界场景的 1,230 个长的、信息密集的提示。每个提示集成了10个空间子域和相应的10个多项选择问答对，范围从对象位置和布局到遮挡和因果关系。我们对 21 个最先进模型的广泛评估表明，高阶空间推理仍然是一个主要瓶颈。(2) 为了证明我们的信息密集设计的实用性超出了简单的评估，我们还构建了 SpatialT2I 数据集。它包含 15,400 个文本-图像对，并带有重写的提示，以确保图像一致性，同时保留信息密度。对当前基础模型（即 Stable Diffusion-XL、Uniworld-V1、OmniGen2）的微调结果产生了一致的性能增益（+4.2%、+5.7%、+4.4%）和空间关系中更真实的效果，突出了以数据为中心的范例，以在 T2I 模型中实现空间智能。

​

扩展嵌入在语言模型方面的表现优于扩展专家

  • 标题: Scaling Embeddings Outperforms Scaling Experts in Language Models
  • 作者: Hong Liu, Jiaqi Zhang, Chao Wang, Xing Hu, Linkun Lyu, Jiaqi Sun, Xurui Yang, Bo Wang, Fengcun Li, Yulei Qian, Lingtong Si, Yerui Sun, Rumei Li, Peng Pei, Yuchen Xie, Xunliang Cai
  • 日期: 2026-01-29
  • ArXiv主页: https://arxiv.org/abs/2601.21204
  • 论文链接: https://arxiv.org/pdf/2601.21204

英文摘要

While Mixture-of-Experts (MoE) architectures have become the standard for sparsity scaling in large language models, they increasingly face diminishing returns and system-level bottlenecks. In this work, we explore embedding scaling as a potent, orthogonal dimension for scaling sparsity. Through a comprehensive analysis and experiments, we identify specific regimes where embedding scaling achieves a superior Pareto frontier compared to expert scaling. We systematically characterize the critical architectural factors governing this efficacy – ranging from parameter budgeting to the interplay with model width and depth. Moreover, by integrating tailored system optimizations and speculative decoding, we effectively convert this sparsity into tangible inference speedups. Guided by these insights, we introduce LongCat-Flash-Lite, a 68.5B parameter model with ~3B activated trained from scratch. Despite allocating over 30B parameters to embeddings, LongCat-Flash-Lite not only surpasses parameter-equivalent MoE baselines but also exhibits exceptional competitiveness against existing models of comparable scale, particularly in agentic and coding domains.

中文摘要

虽然专家混合 (MoE) 架构已成为大型语言模型中稀疏扩展的标准，但它们越来越面临收益递减和系统级瓶颈。在这项工作中，我们探索嵌入缩放作为缩放稀疏性的有效正交维度。通过全面的分析和实验，我们确定了与专家缩放相比，嵌入缩放实现了优越的帕累托前沿的特定机制。我们系统地描述了控制这种功效的关键架构因素——从参数预算到与模型宽度和深度的相互作用。此外，通过集成定制的系统优化和推测解码，我们有效地将这种稀疏性转化为有形的推理加速。在这些见解的指导下，我们推出了 LongCat-Flash-Lite，这是一个从头开始训练的 68.5B 参数模型，具有约 3B 激活值。尽管为嵌入分配了超过 30B 个参数，LongCat-Flash-Lite 不仅超越了参数等效的 MoE 基线，而且相对于同等规模的现有模型也表现出了卓越的竞争力，特别是在代理和编码领域。

​

SWE-Pruner：编码代理的自适应上下文修剪

英文摘要

LLM agents have demonstrated remarkable capabilities in software development, but their performance is hampered by long interaction contexts, which incur high API costs and latency. While various context compression approaches such as LongLLMLingua have emerged to tackle this challenge, they typically rely on fixed metrics such as PPL, ignoring the task-specific nature of code understanding. As a result, they frequently disrupt syntactic and logical structure and fail to retain critical implementation details. In this paper, we propose SWE-Pruner, a self-adaptive context pruning framework tailored for coding agents. Drawing inspiration from how human programmers “selectively skim” source code during development and debugging, SWE-Pruner performs task-aware adaptive pruning for long contexts. Given the current task, the agent formulates an explicit goal (e.g., “focus on error handling”) as a hint to guide the pruning targets. A lightweight neural skimmer (0.6B parameters) is trained to dynamically select relevant lines from the surrounding context given the goal. Evaluations across four benchmarks and multiple models validate SWE-Pruner’s effectiveness in various scenarios, achieving 23-54% token reduction on agent tasks like SWE-Bench Verified and up to 14.84x compression on single-turn tasks like LongCodeQA with minimal performance impact.

中文摘要

LLM 代理在软件开发方面表现出了卓越的能力，但其性能受到长交互上下文的阻碍，这会导致高昂的 API 成本和延迟。虽然已经出现了 LongLLMLingua 等各种上下文压缩方法来应对这一挑战，但它们通常依赖于 PPL 等固定指标，而忽略了代码理解的特定于任务的性质。因此，它们经常破坏语法和逻辑结构，并且无法保留关键的实现细节。在本文中，我们提出了 SWE-Pruner，一种专为编码代理量身定制的自适应上下文修剪框架。SWE-Pruner 从人类程序员在开发和调试过程中“选择性浏览”源代码的方式中汲取灵感，对长上下文执行任务感知的自适应剪枝。给定当前任务，代理制定明确的目标（例如，“专注于错误处理”）作为指导修剪目标的提示。轻量级神经撇渣器（0.6B 参数）经过训练，可以根据目标从周围环境中动态选择相关行。对四个基准和多个模型的评估验证了 SWE-Pruner 在各种场景中的有效性，在 SWE-Bench Verified 等代理任务上实现了 23-54% 的令牌减少，在 LongCodeQA 等单轮任务上实现了高达 14.84 倍的压缩，同时对性能影响最小。

​

Innovator-VL：用于科学发现的多模态大语言模型

英文摘要

We present Innovator-VL, a scientific multimodal large language model designed to advance understanding and reasoning across diverse scientific domains while maintaining excellent performance on general vision tasks. Contrary to the trend of relying on massive domain-specific pretraining and opaque pipelines, our work demonstrates that principled training design and transparent methodology can yield strong scientific intelligence with substantially reduced data requirements. (i) First, we provide a fully transparent, end-to-end reproducible training pipeline, covering data collection, cleaning, preprocessing, supervised fine-tuning, reinforcement learning, and evaluation, along with detailed optimization recipes. This facilitates systematic extension by the community. (ii) Second, Innovator-VL exhibits remarkable data efficiency, achieving competitive performance on various scientific tasks using fewer than five million curated samples without large-scale pretraining. These results highlight that effective reasoning can be achieved through principled data selection rather than indiscriminate scaling. (iii) Third, Innovator-VL demonstrates strong generalization, achieving competitive performance on general vision, multimodal reasoning, and scientific benchmarks. This indicates that scientific alignment can be integrated into a unified model without compromising general-purpose capabilities. Our practices suggest that efficient, reproducible, and high-performing scientific multimodal models can be built even without large-scale data, providing a practical foundation for future research.

中文摘要

我们推出了 Innovator-VL，这是一种科学多模态大语言模型，旨在促进跨不同科学领域的理解和推理，同时在一般视觉任务上保持出色的性能。与依赖大规模特定领域预训练和不透明管道的趋势相反，我们的工作表明，有原则的训练设计和透明的方法可以在大幅减少数据需求的情况下产生强大的科学情报。(i) 首先，我们提供完全透明、端到端可重复的训练流程，涵盖数据收集、清理、预处理、监督微调、强化学习和评估，以及详细的优化方案。这有利于社区的系统扩展。(ii) 其次，Innovator-VL 表现出卓越的数据效率，在不进行大规模预训练的情况下，使用不到 500 万个精选样本，在各种科学任务中实现了具有竞争力的表现。这些结果强调，有效的推理可以通过有原则的数据选择而不是不加区别的缩放来实现。(iii) 第三，Innovator-VL 表现出很强的泛化能力，在总体视觉、多模态推理和科学基准方面取得了有竞争力的表现。这表明科学对齐可以集成到统一模型中，而不会影响通用功能。我们的实践表明，即使没有大规模数据，也可以建立高效、可重复和高性能的科学多模态模型，为未来的研究提供实践基础。

​

DynamicVLA：用于动态对象操作的视觉-语言-动作模型

英文摘要

Manipulating dynamic objects remains an open challenge for Vision-Language-Action (VLA) models, which, despite strong generalization in static manipulation, struggle in dynamic scenarios requiring rapid perception, temporal anticipation, and continuous control. We present DynamicVLA, a framework for dynamic object manipulation that integrates temporal reasoning and closed-loop adaptation through three key designs: 1) a compact 0.4B VLA using a convolutional vision encoder for spatially efficient, structurally faithful encoding, enabling fast multimodal inference; 2) Continuous Inference, enabling overlapping reasoning and execution for lower latency and timely adaptation to object motion; and 3) Latent-aware Action Streaming, which bridges the perception-execution gap by enforcing temporally aligned action execution. To fill the missing foundation of dynamic manipulation data, we introduce the Dynamic Object Manipulation (DOM) benchmark, built from scratch with an auto data collection pipeline that efficiently gathers 200K synthetic episodes across 2.8K scenes and 206 objects, and enables fast collection of 2K real-world episodes without teleoperation. Extensive evaluations demonstrate remarkable improvements in response speed, perception, and generalization, positioning DynamicVLA as a unified framework for general dynamic object manipulation across embodiments.

中文摘要

操纵动态对象仍然是视觉-语言-动作（VLA）模型的一个开放挑战，尽管静态操纵具有很强的泛化性，但在需要快速感知、时间预测和连续控制的动态场景中却表现不佳。我们提出了 DynamicVLA，一个动态对象操作框架，通过三个关键设计集成了时间推理和闭环自适应：1）紧凑的 0.4B VLA，使用卷积视觉编码器进行空间高效、结构忠实的编码，从而实现快速多模态推理；2）连续推理，实现重叠推理和执行，以降低延迟并及时适应对象运动；3）潜在感知动作流，通过强制时间对齐的动作执行来弥合感知与执行之间的差距。为了填补动态操作数据的缺失基础，我们引入了动态对象操作 (DOM) 基准，该基准从头开始构建，具有自动数据收集管道，可有效收集 2.8K 场景和 206 个对象的 200K 合成片段，并无需远程操作即可快速收集 2K 真实世界片段。广泛的评估证明了响应速度、感知和泛化方面的显着改进，将 DynamicVLA 定位为跨实施例的通用动态对象操作的统一框架。

​

DeepSeek-OCR 2：视觉因果流

英文摘要

We present DeepSeek-OCR 2 to investigate the feasibility of a novel encoder-DeepEncoder V2-capable of dynamically reordering visual tokens upon image semantics. Conventional vision-language models (VLMs) invariably process visual tokens in a rigid raster-scan order (top-left to bottom-right) with fixed positional encoding when fed into LLMs. However, this contradicts human visual perception, which follows flexible yet semantically coherent scanning patterns driven by inherent logical structures. Particularly for images with complex layouts, human vision exhibits causally-informed sequential processing. Inspired by this cognitive mechanism, DeepEncoder V2 is designed to endow the encoder with causal reasoning capabilities, enabling it to intelligently reorder visual tokens prior to LLM-based content interpretation. This work explores a novel paradigm: whether 2D image understanding can be effectively achieved through two-cascaded 1D causal reasoning structures, thereby offering a new architectural approach with the potential to achieve genuine 2D reasoning. Codes and model weights are publicly accessible at http://github.com/deepseek-ai/DeepSeek-OCR-2.

中文摘要

我们提出 DeepSeek-OCR 2 来研究新型编码器 DeepEncoder V2 的可行性，该编码器能够根据图像语义动态重新排序视觉标记。传统的视觉语言模型 (VLM) 在输入 LLM 时总是以严格的光栅扫描顺序（从左上到右下）处理视觉标记，并使用固定的位置编码。然而，这与人类视觉感知相矛盾，人类视觉感知遵循由固有逻辑结构驱动的灵活且语义连贯的扫描模式。特别是对于具有复杂布局的图像，人类视觉表现出因果信息顺序处理。受这种认知机制的启发，DeepEncoder V2 旨在赋予编码器因果推理能力，使其能够在基于 LLM 的内容解释之前智能地重新排序视觉标记。这项工作探索了一种新颖的范式：是否可以通过两个级联的一维因果推理结构有效地实现二维图像理解，从而提供一种新的架构方法，有可能实现真正的二维推理。代码和模型权重可在 http://github.com/deepseek-ai/DeepSeek-OCR-2 上公开访问。

​

MMFineReason：通过开放的以数据为中心的方法缩小多模式推理差距

英文摘要

Recent advances in Vision Language Models (VLMs) have driven significant progress in visual reasoning. However, open-source VLMs still lag behind proprietary systems, largely due to the lack of high-quality reasoning data. Existing datasets offer limited coverage of challenging domains such as STEM diagrams and visual puzzles, and lack consistent, long-form Chain-of-Thought (CoT) annotations essential for eliciting strong reasoning capabilities. To bridge this gap, we introduce MMFineReason, a large-scale multimodal reasoning dataset comprising 1.8M samples and 5.1B solution tokens, featuring high-quality reasoning annotations distilled from Qwen3-VL-235B-A22B-Thinking. The dataset is established via a systematic three-stage pipeline: (1) large-scale data collection and standardization, (2) CoT rationale generation, and (3) comprehensive selection based on reasoning quality and difficulty awareness. The resulting dataset spans STEM problems, visual puzzles, games, and complex diagrams, with each sample annotated with visually grounded reasoning traces. We fine-tune Qwen3-VL-Instruct on MMFineReason to develop MMFineReason-2B/4B/8B versions. Our models establish new state-of-the-art results for their size class. Notably, MMFineReason-4B succesfully surpasses Qwen3-VL-8B-Thinking, and MMFineReason-8B even outperforms Qwen3-VL-30B-A3B-Thinking while approaching Qwen3-VL-32B-Thinking, demonstrating remarkable parameter efficiency. Crucially, we uncover a “less is more” phenomenon via our difficulty-aware filtering strategy: a subset of just 7% (123K samples) achieves performance comparable to the full dataset. Notably, we reveal a synergistic effect where reasoning-oriented data composition simultaneously boosts general capabilities.

中文摘要

视觉语言模型 (VLM) 的最新进展推动了视觉推理领域的重大进展。然而，开源 VLM 仍然落后于专有系统，这很大程度上是由于缺乏高质量的推理数据。现有数据集对 STEM 图和视觉谜题等具有挑战性的领域的覆盖范围有限，并且缺乏对于引发强大推理能力至关重要的一致的长形式思想链 (CoT) 注释。为了弥补这一差距，我们引入了 MMFineReason，这是一个大规模多模态推理数据集，包含 180 万个样本和 5.1B 个解决方案标记，具有从 Qwen3-VL-235B-A22B-Thinking 中提取的高质量推理注释。该数据集通过系统的三阶段流程建立：（1）大规模数据收集和标准化，（2）CoT基本原理生成，（3）基于推理质量和难度意识的综合选择。生成的数据集涵盖 STEM 问题、视觉谜题、游戏和复杂图表，每个样本都用基于视觉的推理轨迹进行注释。我们在 MMFineReason 上微调 Qwen3-VL-Instruct 以开发 MMFineReason-2B/4B/8B 版本。我们的模型在其尺寸级别中建立了新的最先进的结果。值得注意的是，MMFineReason-4B 成功超越了 Qwen3-VL-8B-Thinking，MMFineReason-8B 甚至超越了 Qwen3-VL-30B-A3B-Thinking，同时接近 Qwen3-VL-32B-Thinking，展示了显着的参数效率。至关重要的是，我们通过难度感知过滤策略发现了“少即是多”的现象：仅 7%（123K 样本）的子集就实现了与完整数据集相当的性能。值得注意的是，我们揭示了一种协同效应，即面向推理的数据组合同时提高了一般能力。

​

TwinBrainVLA：通过非对称混合变压器释放通用 VLM 的潜力来完成具体任务

英文摘要

Standard Vision-Language-Action (VLA) models typically fine-tune a monolithic Vision-Language Model (VLM) backbone explicitly for robotic control. However, this approach creates a critical tension between maintaining high-level general semantic understanding and learning low-level, fine-grained sensorimotor skills, often leading to “catastrophic forgetting” of the model’s open-world capabilities. To resolve this conflict, we introduce TwinBrainVLA, a novel architecture that coordinates a generalist VLM retaining universal semantic understanding and a specialist VLM dedicated to embodied proprioception for joint robotic control. TwinBrainVLA synergizes a frozen “Left Brain”, which retains robust general visual reasoning, with a trainable “Right Brain”, specialized for embodied perception, via a novel Asymmetric Mixture-of-Transformers (AsyMoT) mechanism. This design allows the Right Brain to dynamically query semantic knowledge from the frozen Left Brain and fuse it with proprioceptive states, providing rich conditioning for a Flow-Matching Action Expert to generate precise continuous controls. Extensive experiments on SimplerEnv and RoboCasa benchmarks demonstrate that TwinBrainVLA achieves superior manipulation performance compared to state-of-the-art baselines while explicitly preserving the comprehensive visual understanding capabilities of the pre-trained VLM, offering a promising direction for building general-purpose robots that simultaneously achieve high-level semantic understanding and low-level physical dexterity.

中文摘要

标准视觉语言动作 (VLA) 模型通常会针对机器人控制明确调整整体视觉语言模型 (VLM) 主干。然而，这种方法在维持高水平的一般语义理解和学习低水平、细粒度的感觉运动技能之间造成了严重的紧张关系，常常导致模型的开放世界能力的“灾难性遗忘”。为了解决这一冲突，我们引入了 TwinBrainVLA，这是一种新颖的架构，它协调保留通用语义理解的通用 VLM 和致力于联合机器人控制的具体本体感觉的专业 VLM。TwinBrainVLA 通过一种新颖的非对称混合变形金刚 (AsyMoT) 机制，将冻结的“左脑”（保留了强大的一般视觉推理功能）与可训练的“右脑”（专门用于具体感知）进行了协同作用。这种设计允许右脑从冻结的左脑动态查询语义知识，并将其与本体感受状态融合，为流程匹配动作专家提供丰富的调节，以生成精确的连续控制。在 SimplerEnv 和 RoboCasa 基准上进行的大量实验表明，与最先进的基线相比，TwinBrainVLA 实现了卓越的操作性能，同时明确保留了预训练 VLM 的全面视觉理解能力，为构建同时实现高级语义理解和低级物理灵活性的通用机器人提供了一个有希望的方向。

​

脚本就是您所需要的：用于长视野对话到电影视频生成的代理框架

英文摘要

Recent advances in video generation have produced models capable of synthesizing stunning visual content from simple text prompts. However, these models struggle to generate long-form, coherent narratives from high-level concepts like dialogue, revealing a ``semantic gap’’ between a creative idea and its cinematic execution. To bridge this gap, we introduce a novel, end-to-end agentic framework for dialogue-to-cinematic-video generation. Central to our framework is ScripterAgent, a model trained to translate coarse dialogue into a fine-grained, executable cinematic script. To enable this, we construct ScriptBench, a new large-scale benchmark with rich multimodal context, annotated via an expert-guided pipeline. The generated script then guides DirectorAgent, which orchestrates state-of-the-art video models using a cross-scene continuous generation strategy to ensure long-horizon coherence. Our comprehensive evaluation, featuring an AI-powered CriticAgent and a new Visual-Script Alignment (VSA) metric, shows our framework significantly improves script faithfulness and temporal fidelity across all tested video models. Furthermore, our analysis uncovers a crucial trade-off in current SOTA models between visual spectacle and strict script adherence, providing valuable insights for the future of automated filmmaking.

中文摘要

视频生成领域的最新进展已经产生了能够从简单的文本提示合成令人惊叹的视觉内容的模型。然而，这些模型很难从对话等高级概念中生成长篇连贯的叙述，揭示了创意与其电影执行之间的“语义差距”。为了弥补这一差距，我们引入了一种新颖的端到端代理框架，用于对话到电影视频的生成。我们框架的核心是 ScripterAgent，这是一个经过训练的模型，可以将粗略的对话转换为细粒度的、可执行的电影脚本。为了实现这一点，我们构建了 ScriptBench，这是一个具有丰富多模式上下文的新的大规模基准，通过专家指导的管道进行注释。然后生成的脚本指导DirectorAgent，它使用跨场景连续生成策略来编排最先进的视频模型，以确保长期一致性。我们的综合评估采用人工智能驱动的 CriticAgent 和新的视觉脚本对齐 (VSA) 指标，表明我们的框架显着提高了所有测试视频模型的脚本忠实度和时间保真度。此外，我们的分析揭示了当前 SOTA 模型中视觉奇观和严格脚本遵守之间的关键权衡，为自动化电影制作的未来提供了宝贵的见解。

​

OCRVerse：在端到端视觉语言模型中迈向整体 OCR

英文摘要

The development of large vision language models drives the demand for managing, and applying massive amounts of multimodal data, making OCR technology, which extracts information from visual images, increasingly popular. However, existing OCR methods primarily focus on recognizing text elements from images or scanned documents (Text-centric OCR), neglecting the identification of visual elements from visually information-dense image sources (Vision-centric OCR), such as charts, web pages and science plots. In reality, these visually information-dense images are widespread on the internet and have significant real-world application value, such as data visualization and web page analysis. In this technical report, we propose OCRVerse, the first holistic OCR method in end-to-end manner that enables unified text-centric OCR and vision-centric OCR. To this end, we constructe comprehensive data engineering to cover a wide range of text-centric documents, such as newspapers, magazines and books, as well as vision-centric rendered composites, including charts, web pages and scientific plots. Moreover, we propose a two-stage SFT-RL multi-domain training method for OCRVerse. SFT directly mixes cross-domain data to train and establish initial domain knowledge, while RL focuses on designing personalized reward strategies for the characteristics of each domain. Specifically, since different domains require various output formats and expected outputs, we provide sufficient flexibility in the RL stage to customize flexible reward signals for each domain, thereby improving cross-domain fusion and avoiding data conflicts. Experimental results demonstrate the effectiveness of OCRVerse, achieving competitive results across text-centric and vision-centric data types, even comparable to large-scale open-source and closed-source models.

中文摘要

大视觉语言模型的发展推动了对海量多模态数据的管理和应用的需求，使得从视觉图像中提取信息的OCR技术越来越受欢迎。然而，现有的 OCR 方法主要侧重于从图像或扫描文档中识别文本元素（以文本为中心的 OCR），而忽略了从视觉信息密集的图像源中识别视觉元素（以视觉为中心的 OCR），例如图表、网页和科学绘图。事实上，这些视觉信息密集的图像在互联网上广泛存在，并且具有重要的现实应用价值，例如数据可视化和网页分析。在本技术报告中，我们提出了 OCRVerse，这是第一个端到端的整体 OCR 方法，可以实现统一的以文本为中心的 OCR 和以视觉为中心的 OCR。为此，我们构建了全面的数据工程，涵盖各种以文本为中心的文档，例如报纸、杂志和书籍，以及以视觉为中心的渲染复合材料，包括图表、网页和科学绘图。此外，我们还提出了一种针对 OCRVerse 的两阶段 SFT-RL 多域训练方法。SFT直接混合跨领域数据来训练和建立初始领域知识，而RL则侧重于针对每个领域的特点设计个性化的奖励策略。具体来说，由于不同的领域需要不同的输出格式和预期输出，因此我们在 RL 阶段提供足够的灵活性，为每个领域定制灵活的奖励信号，从而提高跨领域融合并避免数据冲突。实验结果证明了 OCRVerse 的有效性，在以文本为中心和以视觉为中心的数据类型上取得了有竞争力的结果，甚至可以与大规模开源和闭源模型相媲美。

​

务实的VLA基础模型

英文摘要

Offering great potential in robotic manipulation, a capable Vision-Language-Action (VLA) foundation model is expected to faithfully generalize across tasks and platforms while ensuring cost efficiency (e.g., data and GPU hours required for adaptation). To this end, we develop LingBot-VLA with around 20,000 hours of real-world data from 9 popular dual-arm robot configurations. Through a systematic assessment on 3 robotic platforms, each completing 100 tasks with 130 post-training episodes per task, our model achieves clear superiority over competitors, showcasing its strong performance and broad generalizability. We have also built an efficient codebase, which delivers a throughput of 261 samples per second per GPU with an 8-GPU training setup, representing a 1.5~2.8times (depending on the relied VLM base model) speedup over existing VLA-oriented codebases. The above features ensure that our model is well-suited for real-world deployment. To advance the field of robot learning, we provide open access to the code, base model, and benchmark data, with a focus on enabling more challenging tasks and promoting sound evaluation standards.

中文摘要

一个强大的视觉-语言-动作 (VLA) 基础模型在机器人操作方面具有巨大的潜力，预计能够忠实地跨任务和平台进行泛化，同时确保成本效率（例如，适应所需的数据和 GPU 时间）。为此，我们开发了 LingBot-VLA，其中包含来自 9 种流行双臂机器人配置的约 20,000 小时的真实世界数据。通过对 3 个机器人平台进行系统评估，每个平台完成 100 个任务，每个任务有 130 个训练后片段，我们的模型比竞争对手取得了明显的优势，展示了其强大的性能和广泛的通用性。我们还构建了一个高效的代码库，在 8 个 GPU 训练设置下，每个 GPU 的吞吐量为每秒 261 个样本，这比现有面向 VLA 的代码库加速了 1.5~2.8 倍（取决于所依赖的 VLM 基本模型）。上述功能确保我们的模型非常适合实际部署。为了推动机器人学习领域的发展，我们提供对代码、基础模型和基准数据的开放访问，重点是实现更具挑战性的任务并促进健全的评估标准。

​

AdaReasoner：用于迭代视觉推理的动态工具编排

英文摘要

When humans face problems beyond their immediate capabilities, they rely on tools, providing a promising paradigm for improving visual reasoning in multimodal large language models (MLLMs). Effective reasoning, therefore, hinges on knowing which tools to use, when to invoke them, and how to compose them over multiple steps, even when faced with new tools or new tasks. We introduce AdaReasoner, a family of multimodal models that learn tool use as a general reasoning skill rather than as tool-specific or explicitly supervised behavior. AdaReasoner is enabled by (i) a scalable data curation pipeline exposing models to long-horizon, multi-step tool interactions; (ii) Tool-GRPO, a reinforcement learning algorithm that optimizes tool selection and sequencing based on end-task success; and (iii) an adaptive learning mechanism that dynamically regulates tool usage. Together, these components allow models to infer tool utility from task context and intermediate outcomes, enabling coordination of multiple tools and generalization to unseen tools. Empirically, AdaReasoner exhibits strong tool-adaptive and generalization behaviors: it autonomously adopts beneficial tools, suppresses irrelevant ones, and adjusts tool usage frequency based on task demands, despite never being explicitly trained to do so. These capabilities translate into state-of-the-art performance across challenging benchmarks, improving the 7B base model by +24.9% on average and surpassing strong proprietary systems such as GPT-5 on multiple tasks, including VSP and Jigsaw.

中文摘要

当人类面临超出其直接能力的问题时，他们会依靠工具，为改进多模态大语言模型（MLLM）中的视觉推理提供有前景的范例。因此，有效的推理取决于知道使用哪些工具、何时调用它们以及如何通过多个步骤组合它们，即使面对新工具或新任务也是如此。我们引入了 AdaReasoner，这是一个多模态模型系列，它将工具使用作为一般推理技能来学习，而不是作为特定于工具或明确监督的行为。AdaReasoner 的实现是通过 (i) 可扩展的数据管理管道，将模型暴露给长期、多步骤的工具交互；(ii) Tool-GRPO，一种强化学习算法，可根据最终任务的成功情况优化工具选择和排序；(iii) 动态调节工具使用的自适应学习机制。这些组件共同允许模型从任务上下文和中间结果推断工具效用，从而实现多个工具的协调以及对未见过的工具的泛化。根据经验，AdaReasoner 表现出强大的工具自适应和泛化行为：它自主地采用有益的工具，抑制不相关的工具，并根据任务需求调整工具使用频率，尽管从未接受过明确的培训。这些功能在具有挑战性的基准测试中转化为最先进的性能，将 7B 基本模型平均提高了 24.9%，并在多项任务（包括 VSP 和 Jigsaw）上超越了强大的专有系统（例如 GPT-5）。

​

通过自蒸馏强化学习

英文摘要

Large language models are increasingly post-trained with reinforcement learning in verifiable domains such as code and math. Yet, current methods for reinforcement learning with verifiable rewards (RLVR) learn only from a scalar outcome reward per attempt, creating a severe credit-assignment bottleneck. Many verifiable environments actually provide rich textual feedback, such as runtime errors or judge evaluations, that explain why an attempt failed. We formalize this setting as reinforcement learning with rich feedback and introduce Self-Distillation Policy Optimization (SDPO), which converts tokenized feedback into a dense learning signal without any external teacher or explicit reward model. SDPO treats the current model conditioned on feedback as a self-teacher and distills its feedback-informed next-token predictions back into the policy. In this way, SDPO leverages the model’s ability to retrospectively identify its own mistakes in-context. Across scientific reasoning, tool use, and competitive programming on LiveCodeBench v6, SDPO improves sample efficiency and final accuracy over strong RLVR baselines. Notably, SDPO also outperforms baselines in standard RLVR environments that only return scalar feedback by using successful rollouts as implicit feedback for failed attempts. Finally, applying SDPO to individual questions at test time accelerates discovery on difficult binary-reward tasks, achieving the same discovery probability as best-of-k sampling or multi-turn conversations with 3x fewer attempts.

中文摘要

大型语言模型越来越多地通过代码和数学等可验证领域的强化学习进行后训练。然而，当前具有可验证奖励的强化学习方法（RLVR）仅从每次尝试的标量结果奖励中学习，从而造成了严重的信用分配瓶颈。许多可验证的环境实际上提供了丰富的文本反馈，例如运行时错误或判断评估，以解释尝试失败的原因。我们将这种设置形式化为具有丰富反馈的强化学习，并引入自蒸馏策略优化（SDPO），它将标记化反馈转换为密集学习信号，无需任何外部教师或显式奖励模型。SDPO 将当前以反馈为条件的模型视为自学模型，并将其基于反馈的下一个代币预测提炼回策略中。通过这种方式，SDPO 利用模型的能力来回顾性地识别自己在上下文中的错误。在 LiveCodeBench v6 上的科学推理、工具使用和竞争性编程方面，SDPO 在强大的 RLVR 基线上提高了样本效率和最终准确性。值得注意的是，SDPO 的性能还优于标准 RLVR 环境中的基线，后者仅通过使用成功的推出作为失败尝试的隐式反馈来返回标量反馈。最后，在测试时将 SDPO 应用于单个问题可以加速发现困难的二元奖励任务，以 3 倍的尝试次数实现与 best-of-k 采样或多轮对话相同的发现概率。

​

Youtu-VL：通过统一视觉语言监督释放视觉潜力

  • 标题: Youtu-VL: Unleashing Visual Potential via Unified Vision-Language Supervision
  • 作者: Zhixiang Wei, Yi Li, Zhehan Kan, Xinghua Jiang, Zuwei Long, Shifeng Liu, Hongze Shen, Wei Liu, Xiaoyu Tan, Haojia Lin, Yubo Zhu, Qianyu Li, Di Yin, Haoyu Cao, Weibo Gu, Xin Li, Yinsong Liu, Deqiang Jiang, Xing Sun, Yunsheng Wu, Mingkong Tang, Shuangyin Liu, Lexiang Tang, Haodong Lin, Junru Lu, Jiarui Qin, Lingfeng Qiao, Ruizhi Qiao, Bo Ke, Jianfeng He, Ke Li, Yangning Li, Yunhang Shen, Mengdan Zhang, Peixian Chen, Kun Yin, Bing Liu, Yunfei Wu, Huang Chen, Zhongpeng Cai, Xiaotian Li
  • 日期: 2026-01-27
  • ArXiv主页: https://arxiv.org/abs/2601.19798
  • 论文链接: https://arxiv.org/pdf/2601.19798
  • 项目链接: https://youtu-tip.com/#llm
  • gitHub仓库: https://github.com/TencentCloudADP/youtu-vl

英文摘要

Despite the significant advancements represented by Vision-Language Models (VLMs), current architectures often exhibit limitations in retaining fine-grained visual information, leading to coarse-grained multimodal comprehension. We attribute this deficiency to a suboptimal training paradigm inherent in prevailing VLMs, which exhibits a text-dominant optimization bias by conceptualizing visual signals merely as passive conditional inputs rather than supervisory targets. To mitigate this, we introduce Youtu-VL, a framework leveraging the Vision-Language Unified Autoregressive Supervision (VLUAS) paradigm, which fundamentally shifts the optimization objective from vision-as-input'' to vision-as-target.‘’ By integrating visual tokens directly into the prediction stream, Youtu-VL applies unified autoregressive supervision to both visual details and linguistic content. Furthermore, we extend this paradigm to encompass vision-centric tasks, enabling a standard VLM to perform vision-centric tasks without task-specific additions. Extensive empirical evaluations demonstrate that Youtu-VL achieves competitive performance on both general multimodal tasks and vision-centric tasks, establishing a robust foundation for the development of comprehensive generalist visual agents.

中文摘要

Despite the significant advancements represented by Vision-Language Models (VLMs), current architectures often exhibit limitations in retaining fine-grained visual information, leading to coarse-grained multimodal comprehension. We attribute this deficiency to a suboptimal training paradigm inherent in prevailing VLMs, which exhibits a text-dominant optimization bias by conceptualizing visual signals merely as passive conditional inputs rather than supervisory targets. To mitigate this, we introduce Youtu-VL, a framework leveraging the Vision-Language Unified Autoregressive Supervision (VLUAS) paradigm, which fundamentally shifts the optimization objective from vision-as-input'' to vision-as-target.‘’ By integrating visual tokens directly into the prediction stream, Youtu-VL applies unified autoregressive supervision to both visual details and linguistic content. Furthermore, we extend this paradigm to encompass vision-centric tasks, enabling a standard VLM to perform vision-centric tasks without task-specific additions. Extensive empirical evaluations demonstrate that Youtu-VL achieves competitive performance on both general multimodal tasks and vision-centric tasks, establishing a robust foundation for the development of comprehensive generalist visual agents.

​

自学教学模式：可学习性边缘的推理

英文摘要

Can a model learn to escape its own learning plateau? Reinforcement learning methods for finetuning large reasoning models stall on datasets with low initial success rates, and thus little training signal. We investigate a fundamental question: Can a pretrained LLM leverage latent knowledge to generate an automated curriculum for problems it cannot solve? To explore this, we design SOAR: A self-improvement framework designed to surface these pedagogical signals through meta-RL. A teacher copy of the model proposes synthetic problems for a student copy, and is rewarded with its improvement on a small subset of hard problems. Critically, SOAR grounds the curriculum in measured student progress rather than intrinsic proxy rewards. Our study on the hardest subsets of mathematical benchmarks (0/128 success) reveals three core findings. First, we show that it is possible to realize bi-level meta-RL that unlocks learning under sparse, binary rewards by sharpening a latent capacity of pretrained models to generate useful stepping stones. Second, grounded rewards outperform intrinsic reward schemes used in prior LLM self-play, reliably avoiding the instability and diversity collapse modes they typically exhibit. Third, analyzing the generated questions reveals that structural quality and well-posedness are more critical for learning progress than solution correctness. Our results suggest that the ability to generate useful stepping stones does not require the preexisting ability to actually solve the hard problems, paving a principled path to escape reasoning plateaus without additional curated data.

中文摘要

模型能否学会摆脱自身的学习高原？用于微调大型推理模型的强化学习方法在初始成功率较低的数据集上停滞不前，因此训练信号很少。我们研究一个基本问题：预训练的法学硕士能否利用潜在知识为它无法解决的问题生成自动化课程？为了探索这一点，我们设计了 SOAR：一个自我改进框架，旨在通过元强化学习呈现这些教学信号。该模型的教师副本为学生副本提出综合问题，并因其对一小部分难题的改进而获得奖励。至关重要的是，SOAR 将课程建立在衡量学生进步的基础上，而不是内在的代理奖励。我们对数学基准最难子集（0/128 成功）的研究揭示了三个核心发现。首先，我们证明可以实现双层元强化学习，通过增强预训练模型的潜在能力来生成有用的踏脚石，从而解锁稀疏二元奖励下的学习。其次，扎根奖励优于之前 LLM 自我博弈中使用的内在奖励方案，可靠地避免了它们通常表现出的不稳定性和多样性崩溃模式。第三，分析生成的问题表明，结构质量和适定性对于学习进度比解决方案的正确性更重要。我们的结果表明，产生有用垫脚石的能力并不需要预先存在的能力来实际解决难题，从而为摆脱推理高原铺平了一条原则性道路，而无需额外的精选数据。

​

ConceptMoE：用于隐式计算分配的自适应令牌到概念压缩

英文摘要

Large language models allocate uniform computation across all tokens, ignoring that some sequences are trivially predictable while others require deep reasoning. We introduce ConceptMoE, which dynamically merges semantically similar tokens into concept representations, performing implicit token-level compute allocation. A learnable chunk module identifies optimal boundaries by measuring inter-token similarity, compressing sequences by a target ratio R before they enter the compute-intensive concept model. Crucially, the MoE architecture enables controlled evaluation: we reallocate saved computation to match baseline activated FLOPs (excluding attention map computation) and total parameters, isolating genuine architectural benefits. Under these conditions, ConceptMoE consistently outperforms standard MoE across language and vision-language tasks, achieving +0.9 points on language pretraining, +2.3 points on long context understanding, and +0.6 points on multimodal benchmarks. When converting pretrained MoE during continual training with layer looping, gains reach +5.5 points, demonstrating practical applicability. Beyond performance, ConceptMoE reduces attention computation by up to R^2times and KV cache by Rtimes. At R=2, empirical measurements show prefill speedups reaching 175% and decoding speedups up to 117% on long sequences. The minimal architectural modifications enable straightforward integration into existing MoE, demonstrating that adaptive concept-level processing fundamentally improves both effectiveness and efficiency of large language models.

中文摘要

大型语言模型在所有标记之间分配统一的计算，忽略了一些序列是可以简单预测的，而另一些序列则需要深度推理。我们引入了 ConceptMoE，它动态地将语义相似的标记合并到概念表示中，执行隐式标记级计算分配。可学习的块模块通过测量标记间相似性来识别最佳边界，在序列进入计算密集型概念模型之前按目标比率 R 压缩序列。至关重要的是，MoE 架构实现了受控评估：我们重新分配保存的计算以匹配基线激活的 FLOP（不包括注意力图计算）和总参数，从而隔离真正的架构优势。在这些条件下，ConceptMoE 在语言和视觉语言任务上始终优于标准 MoE，在语言预训练上获得 +0.9 分，在长上下文理解上获得 +2.3 分，在多模式基准上获得 +0.6 分。在通过层循环进行连续训练期间转换预训练的 MoE 时，增益达到 +5.5 点，证明了实际适用性。除了性能之外，ConceptMoE 将注意力计算减少了 R^2 倍，将 KV 缓存减少了 R 倍。在 R=2 时，经验测量显示长序列上的预填充加速达到 175%，解码加速达到 117%。最小的架构修改可以直接集成到现有的 MoE 中，这表明自适应概念级处理从根本上提高了大型语言模型的有效性和效率。

​

科学图像合成：基准测试、方法论和下游实用程序

英文摘要

While synthetic data has proven effective for improving scientific reasoning in the text domain, multimodal reasoning remains constrained by the difficulty of synthesizing scientifically rigorous images. Existing Text-to-Image (T2I) models often produce outputs that are visually plausible yet scientifically incorrect, resulting in a persistent visual-logic divergence that limits their value for downstream reasoning. Motivated by recent advances in next-generation T2I models, we conduct a systematic study of scientific image synthesis across generation paradigms, evaluation, and downstream use. We analyze both direct pixel-based generation and programmatic synthesis, and propose ImgCoder, a logic-driven framework that follows an explicit “understand - plan - code” workflow to improve structural precision. To rigorously assess scientific correctness, we introduce SciGenBench, which evaluates generated images based on information utility and logical validity. Our evaluation reveals systematic failure modes in pixel-based models and highlights a fundamental expressiveness-precision trade-off. Finally, we show that fine-tuning Large Multimodal Models (LMMs) on rigorously verified synthetic scientific images yields consistent reasoning gains, with potential scaling trends analogous to the text domain, validating high-fidelity scientific synthesis as a viable path to unlocking massive multimodal reasoning capabilities.

中文摘要

While synthetic data has proven effective for improving scientific reasoning in the text domain, multimodal reasoning remains constrained by the difficulty of synthesizing scientifically rigorous images. Existing Text-to-Image (T2I) models often produce outputs that are visually plausible yet scientifically incorrect, resulting in a persistent visual-logic divergence that limits their value for downstream reasoning. Motivated by recent advances in next-generation T2I models, we conduct a systematic study of scientific image synthesis across generation paradigms, evaluation, and downstream use. We analyze both direct pixel-based generation and programmatic synthesis, and propose ImgCoder, a logic-driven framework that follows an explicit “understand - plan - code” workflow to improve structural precision. To rigorously assess scientific correctness, we introduce SciGenBench, which evaluates generated images based on information utility and logical validity. Our evaluation reveals systematic failure modes in pixel-based models and highlights a fundamental expressiveness-precision trade-off. Finally, we show that fine-tuning Large Multimodal Models (LMMs) on rigorously verified synthetic scientific images yields consistent reasoning gains, with potential scaling trends analogous to the text domain, validating high-fidelity scientific synthesis as a viable path to unlocking massive multimodal reasoning capabilities.

​

VisGym：多模式代理的多样化、可定制、可扩展环境

英文摘要

Modern Vision-Language Models (VLMs) remain poorly characterized in multi-step visual interactions, particularly in how they integrate perception, memory, and action over long horizons. We introduce VisGym, a gymnasium of 17 environments for evaluating and training VLMs. The suite spans symbolic puzzles, real-image understanding, navigation, and manipulation, and provides flexible controls over difficulty, input representation, planning horizon, and feedback. We also provide multi-step solvers that generate structured demonstrations, enabling supervised finetuning. Our evaluations show that all frontier models struggle in interactive settings, achieving low success rates in both the easy (46.6%) and hard (26.0%) configurations. Our experiments reveal notable limitations: models struggle to effectively leverage long context, performing worse with an unbounded history than with truncated windows. Furthermore, we find that several text-based symbolic tasks become substantially harder once rendered visually. However, explicit goal observations, textual feedback, and exploratory demonstrations in partially observable or unknown-dynamics settings for supervised finetuning yield consistent gains, highlighting concrete failure modes and pathways for improving multi-step visual decision-making. Code, data, and models can be found at: https://visgym.github.io/.

中文摘要

现代视觉语言模型（VLM）在多步骤视觉交互中的特征仍然很差，特别是在它们如何整合长期视野中的感知、记忆和行动方面。我们推出 VisGym，这是一个拥有 17 个环境的体育馆，用于评估和训练 VLM。该套件涵盖符号谜题、真实图像理解、导航和操作，并提供对难度、输入表示、规划范围和反馈的灵活控制。我们还提供多步求解器来生成结构化演示，从而实现监督微调。我们的评估表明，所有前沿模型都在交互设置中陷入困境，在简单（46.6％）和困难（26.0％）配置中的成功率都很低。我们的实验揭示了显着的局限性：模型难以有效地利用长上下文，在无限历史记录中的表现比在截断窗口中的表现更差。此外，我们发现一些基于文本的符号任务一旦以视觉方式呈现就会变得更加困难。然而，明确的目标观察、文本反馈和在部分可观察或未知动态设置中进行监督微调的探索性演示会产生一致的收益，突出显示具体的故障模式和改进多步骤视觉决策的途径。代码、数据和模型可以在：https://visgym.github.io/ 找到。

​

Qwen3-ASR技术报告

英文摘要

In this report, we introduce Qwen3-ASR family, which includes two powerful all-in-one speech recognition models and a novel non-autoregressive speech forced alignment model. Qwen3-ASR-1.7B and Qwen3-ASR-0.6B are ASR models that support language identification and ASR for 52 languages and dialects. Both of them leverage large-scale speech training data and the strong audio understanding ability of their foundation model Qwen3-Omni. We conduct comprehensive internal evaluation besides the open-sourced benchmarks as ASR models might differ little on open-sourced benchmark scores but exhibit significant quality differences in real-world scenarios. The experiments reveal that the 1.7B version achieves SOTA performance among open-sourced ASR models and is competitive with the strongest proprietary APIs while the 0.6B version offers the best accuracy-efficiency trade-off. Qwen3-ASR-0.6B can achieve an average TTFT as low as 92ms and transcribe 2000 seconds speech in 1 second at a concurrency of 128. Qwen3-ForcedAligner-0.6B is an LLM based NAR timestamp predictor that is able to align text-speech pairs in 11 languages. Timestamp accuracy experiments show that the proposed model outperforms the three strongest force alignment models and takes more advantages in efficiency and versatility. To further accelerate the community research of ASR and audio understanding, we release these models under the Apache 2.0 license.

中文摘要

在本报告中，我们介绍了 Qwen3-ASR 系列，其中包括两个强大的一体式语音识别模型和一种新颖的非自回归语音强制对齐模型。Qwen3-ASR-1.7B和Qwen3-ASR-0.6B是支持语言识别和52种语言和方言的ASR的ASR模型。两者都利用了大规模语音训练数据和基础模型 Qwen3-Omni 强大的音频理解能力。除了开源基准之外，我们还进行了全面的内部评估，因为 ASR 模型在开源基准分数上可能差异不大，但在现实场景中表现出显着的质量差异。实验表明，1.7B 版本在开源 ASR 模型中实现了 SOTA 性能，并且与最强的专有 API 具有竞争力，而 0.6B 版本则提供了最佳的准确性与效率权衡。Qwen3-ASR-0.6B 可以实现低至 92ms 的平均 TTFT，并在 128 并发的情况下在 1 秒内转录 2000 秒语音。Qwen3-ForcedAligner-0.6B 是基于 LLM 的 NAR 时间戳预测器，能够对齐 11 种语言的文本语音对。时间戳精度实验表明，所提出的模型优于三种最强的力对齐模型，并且在效率和通用性方面更具优势。为了进一步加速 ASR 和音频理解的社区研究，我们在 Apache 2.0 许可证下发布了这些模型。

​

自我蒸馏实现持续学习

英文摘要

Continual learning, enabling models to acquire new skills and knowledge without degrading existing capabilities, remains a fundamental challenge for foundation models. While on-policy reinforcement learning can reduce forgetting, it requires explicit reward functions that are often unavailable. Learning from expert demonstrations, the primary alternative, is dominated by supervised fine-tuning (SFT), which is inherently off-policy. We introduce Self-Distillation Fine-Tuning (SDFT), a simple method that enables on-policy learning directly from demonstrations. SDFT leverages in-context learning by using a demonstration-conditioned model as its own teacher, generating on-policy training signals that preserve prior capabilities while acquiring new skills. Across skill learning and knowledge acquisition tasks, SDFT consistently outperforms SFT, achieving higher new-task accuracy while substantially reducing catastrophic forgetting. In sequential learning experiments, SDFT enables a single model to accumulate multiple skills over time without performance regression, establishing on-policy distillation as a practical path to continual learning from demonstrations.

中文摘要

持续学习，使模型能够在不降低现有能力的情况下获得新的技能和知识，仍然是基础模型面临的基本挑战。虽然策略强化学习可以减少遗忘，但它需要通常不可用的显式奖励函数。从专家演示中学习是主要的替代方案，它以监督微调（SFT）为主，这本质上是脱离策略的。我们引入了自蒸馏微调（SDFT），这是一种简单的方法，可以直接从演示中进行在策略学习。SDFT 通过使用演示条件模型作为自己的老师来利用情境学习，生成符合策略的训练信号，在获取新技能的同时保留先前的能力。在技​​能学习和知识获取任务中，SDFT 始终优于 SFT，实现了更高的新任务准确性，同时大幅减少了灾难性遗忘。在顺序学习实验中，SDFT 使单个模型能够随着时间的推移积累多种技能，而不会出现性能回归，从而将策略蒸馏建立为从演示中持续学习的实用路径。

​

DeepPlanning：具有可验证约束的长期代理规划基准

英文摘要

While agent evaluation has shifted toward long-horizon tasks, most benchmarks still emphasize local, step-level reasoning rather than the global constrained optimization (e.g., time and financial budgets) that demands genuine planning ability. Meanwhile, existing LLM planning benchmarks underrepresent the active information gathering and fine-grained local constraints typical of real-world settings. To address this, we introduce DeepPlanning, a challenging benchmark for practical long-horizon agent planning. It features multi-day travel planning and multi-product shopping tasks that require proactive information acquisition, local constrained reasoning, and global constrained optimization. Evaluations on DeepPlanning show that even frontier agentic LLMs struggle with these problems, highlighting the importance of reliable explicit reasoning patterns and parallel tool use for achieving better effectiveness-efficiency trade-offs. Error analysis further points to promising directions for improving agentic LLMs over long planning horizons. We open-source the code and data to support future research.

中文摘要

虽然智能体评估已转向长期任务，但大多数基准仍然强调局部、步骤级推理，而不是需要真正规划能力的全局约束优化（例如时间和财务预算）。与此同时，现有的法学硕士规划基准未能充分体现现实世界环境中典型的主动信息收集和细粒度局部约束。为了解决这个问题，我们引入了 DeepPlanning，这是一个针对实际长期智能体规划的具有挑战性的基准。它的特点是多日旅行规划和多产品购物任务，需要主动信息获取、局部约束推理和全局约束优化。对 DeepPlanning 的评估表明，即使是前沿代理法学硕士也在努力解决这些问题，这凸显了可靠的显式推理模式和并行工具使用对于实现更好的有效性-效率权衡的重要性。错误分析进一步指出了在长期规划范围内改进代理法学硕士的有希望的方向。我们开源代码和数据以支持未来的研究。

​

弹性注意力：高效变压器的测试时间自适应稀疏率

英文摘要

The quadratic complexity of standard attention mechanisms poses a significant scalability bottleneck for large language models (LLMs) in long-context scenarios. While hybrid attention strategies that combine sparse and full attention within a single model offer a viable solution, they typically employ static computation ratios (i.e., fixed proportions of sparse versus full attention) and fail to adapt to the varying sparsity sensitivities of downstream tasks during inference. To address this issue, we propose Elastic Attention, which allows the model to dynamically adjust its overall sparsity based on the input. This is achieved by integrating a lightweight Attention Router into the existing pretrained model, which dynamically assigns each attention head to different computation modes. Within only 12 hours of training on 8xA800 GPUs, our method enables models to achieve both strong performance and efficient inference. Experiments across three long-context benchmarks on widely-used LLMs demonstrate the superiority of our method.

中文摘要

标准注意力机制的二次复杂度给长上下文场景中的大型语言模型（LLM）带来了显着的可扩展性瓶颈。虽然在单个模型中结合稀疏和完整注意力的混合注意力策略提供了可行的解决方案，但它们通常采用静态计算比率（即稀疏与完整注意力的固定比例）并且无法适应推理过程中下游任务不同的稀疏敏感性。为了解决这个问题，我们提出了弹性注意力（Elastic Attention），它允许模型根据输入动态调整其整体稀疏性。这是通过将轻量级注意力路由器集成到现有的预训练模型中来实现的，该模型动态地将每个注意力头分配给不同的计算模式。在 8xA800 GPU 上仅 12 小时的训练内，我们的方法就使模型能够实现强大的性能和高效的推理。在广泛使用的法学硕士上进行的三个长上下文基准的实验证明了我们方法的优越性。

​

iFSQ：用 1 行代码改进图像生成的 FSQ

英文摘要

The field of image generation is currently bifurcated into autoregressive (AR) models operating on discrete tokens and diffusion models utilizing continuous latents. This divide, rooted in the distinction between VQ-VAEs and VAEs, hinders unified modeling and fair benchmarking. Finite Scalar Quantization (FSQ) offers a theoretical bridge, yet vanilla FSQ suffers from a critical flaw: its equal-interval quantization can cause activation collapse. This mismatch forces a trade-off between reconstruction fidelity and information efficiency. In this work, we resolve this dilemma by simply replacing the activation function in original FSQ with a distribution-matching mapping to enforce a uniform prior. Termed iFSQ, this simple strategy requires just one line of code yet mathematically guarantees both optimal bin utilization and reconstruction precision. Leveraging iFSQ as a controlled benchmark, we uncover two key insights: (1) The optimal equilibrium between discrete and continuous representations lies at approximately 4 bits per dimension. (2) Under identical reconstruction constraints, AR models exhibit rapid initial convergence, whereas diffusion models achieve a superior performance ceiling, suggesting that strict sequential ordering may limit the upper bounds of generation quality. Finally, we extend our analysis by adapting Representation Alignment (REPA) to AR models, yielding LlamaGen-REPA. Codes is available at https://github.com/Tencent-Hunyuan/iFSQ

中文摘要

图像生成领域目前分为在离散标记上运行的自回归（AR）模型和利用连续潜伏的扩散模型。这种分歧源于 VQ-VAE 和 VAE 之间的区别，阻碍了统一建模和公平基准测试。有限标量量化 (FSQ) 提供了一个理论桥梁，但普通 FSQ 存在一个严重缺陷：其等间隔量化可能导致激活崩溃。这种不匹配迫使重建保真度和信息效率之间进行权衡。在这项工作中，我们通过简单地用分布匹配映射替换原始 FSQ 中的激活函数来解决这个困境，以强制执行统一的先验。这种简单的策略称为 iFSQ，仅需要一行代码，但在数学上保证了最佳的 bin 利用率和重建精度。利用 iFSQ 作为受控基准，我们发现了两个关键见解：(1) 离散表示和连续表示之间的最佳平衡在于每个维度大约 4 位。（2）在相同的重建约束下，AR模型表现出快速的初始收敛，而扩散模型实现了优越的性能上限，这表明严格的顺序排序可能会限制生成质量的上限。最后，我们通过将表示对齐 (REPA) 应用于 AR 模型来扩展我们的分析，产生 LlamaGen-REPA。代码可见https://github.com/Tencent-Hunyuan/iFSQ

​

# 人工智能 # 机器人
Logo
AtomGit开源社区

AtomGit 是由开放原子开源基金会联合 CSDN 等生态伙伴共同推出的新一代开源与人工智能协作平台。平台坚持“开放、中立、公益”的理念，把代码托管、模型共享、数据集托管、智能体开发体验和算力服务整合在一起，为开发者提供从开发、训练到部署的一站式体验。

更多推荐

cover

AtomGit 首发开源｜光轮智能发布全球首个十万小时级全模态人类行为开源数据集 EgoSuite-Open100K

avatar AtomGit开源社区
cover

构筑赋能金融行业应用 Agent｜20 强决战北京，一场金融 Agent 的硬核答卷

avatar AtomGit开源社区
cover

AtomCode 全新上线万亿代码模型 LongCat-2.0。限时 7 天免费体验！

avatar AtomGit开源社区