嘘~ 正在从服务器偷取页面 . . .

Agent


⚠️ 以下所有内容总结都来自于 大语言模型的能力,如有错误,仅供参考,谨慎使用
🔴 请注意:千万不要用于严肃的学术场景,只能用于论文阅读前的初筛!
💗 如果您觉得我们的项目对您有帮助 ChatPaperFree ,还请您给我们一些鼓励!⭐️ HuggingFace免费体验

2025-10-02 更新

OceanGym: A Benchmark Environment for Underwater Embodied Agents

Authors:Yida Xue, Mingjun Mao, Xiangyuan Ru, Yuqi Zhu, Baochang Ren, Shuofei Qiao, Mengru Wang, Shumin Deng, Xinyu An, Ningyu Zhang, Ying Chen, Huajun Chen

We introduce OceanGym, the first comprehensive benchmark for ocean underwater embodied agents, designed to advance AI in one of the most demanding real-world environments. Unlike terrestrial or aerial domains, underwater settings present extreme perceptual and decision-making challenges, including low visibility, dynamic ocean currents, making effective agent deployment exceptionally difficult. OceanGym encompasses eight realistic task domains and a unified agent framework driven by Multi-modal Large Language Models (MLLMs), which integrates perception, memory, and sequential decision-making. Agents are required to comprehend optical and sonar data, autonomously explore complex environments, and accomplish long-horizon objectives under these harsh conditions. Extensive experiments reveal substantial gaps between state-of-the-art MLLM-driven agents and human experts, highlighting the persistent difficulty of perception, planning, and adaptability in ocean underwater environments. By providing a high-fidelity, rigorously designed platform, OceanGym establishes a testbed for developing robust embodied AI and transferring these capabilities to real-world autonomous ocean underwater vehicles, marking a decisive step toward intelligent agents capable of operating in one of Earth’s last unexplored frontiers. The code and data are available at https://github.com/OceanGPT/OceanGym.

我们推出了OceanGym,这是针对海洋水下智能体的首个全面基准测试平台,旨在在最苛刻的现实环境之一中推动人工智能的发展。与陆地或空中领域不同,水下环境带来了极端的感知和决策挑战,包括低能见度、动态洋流,使得有效部署智能体变得异常困难。OceanGym包含八个现实任务领域和一个由多模态大型语言模型(MLLM)驱动的统一智能体框架,该框架集成了感知、记忆和序列决策。智能体需要理解光学和声纳数据,自主探索复杂环境,并在这些恶劣条件下完成长期目标。大量实验揭示了最先进的MLLM驱动的智能体与人类专家之间的巨大差距,突出了海洋水下环境中感知、规划和适应性的持久性困难。通过提供高保真、严格设计的平台,OceanGym为开发稳健的嵌入式人工智能并将这些能力转移到实际的水下自动驾驶车辆建立了测试床,这是朝着能够操作地球最后一个未开发前沿之一的智能体迈出的决定性一步。代码和数据可在https://github.com/OceanGPT/OceanGym找到。

论文及项目相关链接

PDF Work in progress

Summary

OceanGym是一个针对海洋水下智能体设计的全面基准测试平台,旨在推动人工智能在极具挑战性的水下环境中的发展。该平台涵盖八个现实任务领域,采用多模态大型语言模型驱动的统一智能体框架,包括感知、记忆和序列决策制定等功能。该平台为开发稳健的水下人工智能提供了一个严格设计的高保真测试环境,标志着将智能体应用于地球最后一个未被探索的前沿——海洋的重要一步。

Key Takeaways

  1. OceanGym是首个针对海洋水下智能体的全面基准测试平台。
  2. 平台旨在推动人工智能在极具挑战性的水下环境中的应用和发展。
  3. OceanGym涵盖八个现实任务领域,包括感知和决策制定等功能。
  4. 采用多模态大型语言模型驱动的统一智能体框架。
  5. 智能体需要处理光学和声纳数据,自主探索复杂环境,完成长期目标。
  6. 实验显示,现有技术与人专家之间存在显著差距,特别是在感知、规划和适应性方面。

Cool Papers

点此查看论文截图

VitaBench: Benchmarking LLM Agents with Versatile Interactive Tasks in Real-world Applications

Authors:Wei He, Yueqing Sun, Hongyan Hao, Xueyuan Hao, Zhikang Xia, Qi Gu, Chengcheng Han, Dengchang Zhao, Hui Su, Kefeng Zhang, Man Gao, Xi Su, Xiaodong Cai, Xunliang Cai, Yu Yang, Yunke Zhao

As LLM-based agents are increasingly deployed in real-life scenarios, existing benchmarks fail to capture their inherent complexity of handling extensive information, leveraging diverse resources, and managing dynamic user interactions. To address this gap, we introduce VitaBench, a challenging benchmark that evaluates agents on versatile interactive tasks grounded in real-world settings. Drawing from daily applications in food delivery, in-store consumption, and online travel services, VitaBench presents agents with the most complex life-serving simulation environment to date, comprising 66 tools. Through a framework that eliminates domain-specific policies, we enable flexible composition of these scenarios and tools, yielding 100 cross-scenario tasks (main results) and 300 single-scenario tasks. Each task is derived from multiple real user requests and requires agents to reason across temporal and spatial dimensions, utilize complex tool sets, proactively clarify ambiguous instructions, and track shifting user intent throughout multi-turn conversations. Moreover, we propose a rubric-based sliding window evaluator, enabling robust assessment of diverse solution pathways in complex environments and stochastic interactions. Our comprehensive evaluation reveals that even the most advanced models achieve only 30% success rate on cross-scenario tasks, and less than 50% success rate on others. Overall, we believe VitaBench will serve as a valuable resource for advancing the development of AI agents in practical real-world applications. The code, dataset, and leaderboard are available at https://vitabench.github.io/

随着基于大型语言模型(LLM)的代理在现实生活场景中的部署越来越多,现有的基准测试无法捕捉到它们处理大量信息、利用不同资源和处理动态用户交互的内在复杂性。为了弥补这一空白,我们推出了VitaBench,这是一个具有挑战性的基准测试,旨在评估代理在现实环境中的通用交互任务表现。VitaBench借鉴了食品配送、店内消费和在线旅行服务等日常应用,为代理提供了迄今为止最复杂的生活服务模拟环境,包含66种工具。通过一个消除特定领域策略的框架,我们能够灵活地组合这些场景和工具,产生100个跨场景任务和300个单一场景任务(主要结果)。每个任务都来源于多个真实用户请求,要求代理在时间和空间维度上进行推理,利用复杂的工具集,主动澄清模糊指令,并在多轮对话中跟踪变化的用户意图。此外,我们提出了一种基于条目的滑动窗口评估器,能够对复杂环境和随机交互中的不同解决方案路径进行稳健评估。我们的综合评估发现,即使在跨场景任务中,最先进的模型也仅达到30%的成功率,其他任务的成功率也低于50%。总的来说,我们相信VitaBench将成为推动人工智能代理在实际现实世界应用中的发展的宝贵资源。相关代码、数据集和排行榜可在https://vitabench.github.io/找到。

论文及项目相关链接

PDF The code, dataset, and leaderboard are available at https://vitabench.github.io/

Summary

基于大型语言模型(LLM)的代理在实际场景中部署越来越多,但现有基准测试无法捕捉到其处理大量信息、利用多样资源和处理动态用户交互的内在复杂性。为解决这一差距,我们推出VitaBench基准测试,它评估代理在现实环境中的通用交互任务表现。VitaBench模拟了食品配送、店内消费和在线旅行服务等日常应用,为代理提供了迄今为止最复杂的生活服务模拟环境,包含66种工具。我们通过一个框架来消除特定领域的政策,使这些场景和工具具有灵活的组合性,产生了100个跨场景任务和300个单一场景任务。每个任务都源于多个真实用户需求,要求代理在时间和空间维度上进行推理,使用复杂的工具集,主动澄清模糊指令,并在多轮对话中跟踪变化的用户意图。我们的综合评估发现,最先进的模型在跨场景任务上的成功率仅为30%,其他任务上的成功率也低于50%。总体而言,我们相信VitaBench将为推进AI代理在实用现实应用中的发展提供了宝贵的资源。

Key Takeaways

  1. 现有基准测试无法充分捕捉LLM代理在处理现实场景中的复杂性。
  2. VitaBench是一个新的基准测试,旨在评估代理在现实世界环境下的表现。
  3. VitaBench包含66种工具,模拟了复杂的生活环境。
  4. 框架允许灵活的场景和工具组合,生成多种任务。
  5. 每个任务源于真实用户需求,要求代理进行复杂的推理和交互。
  6. 最先进的模型在跨场景任务上的成功率较低。

Cool Papers

点此查看论文截图

CreAgentive: An Agent Workflow Driven Multi-Category Creative Generation Engine

Authors:Yuyang Cheng, Linyue Cai, Changwei Peng, Yumiao Xu, Rongfang Bie, Yong Zhao

We present CreAgentive, an agent workflow driven multi-category creative generation engine that addresses four key limitations of contemporary large language models in writing stories, drama and other categories of creatives: restricted genre diversity, insufficient output length, weak narrative coherence, and inability to enforce complex structural constructs. At its core, CreAgentive employs a Story Prototype, which is a genre-agnostic, knowledge graph-based narrative representation that decouples story logic from stylistic realization by encoding characters, events, and environments as semantic triples. CreAgentive engages a three-stage agent workflow that comprises: an Initialization Stage that constructs a user-specified narrative skeleton; a Generation Stage in which long- and short-term objectives guide multi-agent dialogues to instantiate the Story Prototype; a Writing Stage that leverages this prototype to produce multi-genre text with advanced structures such as retrospection and foreshadowing. This architecture reduces storage redundancy and overcomes the typical bottlenecks of long-form generation. In extensive experiments, CreAgentive generates thousands of chapters with stable quality and low cost (less than $1 per 100 chapters) using a general-purpose backbone model. To evaluate performance, we define a two-dimensional framework with 10 narrative indicators measuring both quality and length. Results show that CreAgentive consistently outperforms strong baselines and achieves robust performance across diverse genres, approaching the quality of human-authored novels.

我们推出了CreAgentive,这是一款以代理工作流程驱动的跨类别创意生成引擎,解决了当代大型语言模型在写作故事、戏剧和其他类别创意时面临的四大局限:有限的体裁多样性、输出长度不足、叙事连贯性弱以及无法执行复杂的结构构建。CreAgentive的核心采用故事原型,这是一种基于知识图谱的叙事表示方法,通过编码角色、事件和环境为语义三元组,将故事逻辑与风格实现相分离。CreAgentive采用三阶段代理工作流程,包括:构建用户指定的叙事骨架的初始化阶段;长短期目标指导多代理对话以实例化故事原型的生成阶段;以及利用该原型产生具有反思和预示等高级结构的多体裁文本的写作阶段。这种架构减少了存储冗余,克服了长形式生成的典型瓶颈。在广泛的实验中,CreAgentive使用通用骨干模型以稳定的品质和低成本(每100章不到1美元)生成了成千上万章的内容。为了评估性能,我们定义了一个二维框架,包含10个衡量质量和长度的叙事指标。结果表明,CreAgentive持续超越强劲的基线,并在各种体裁中表现稳健,接近人类创作的小说质量。

论文及项目相关链接

PDF

摘要

CrAgentive是一款针对当代大型语言模型写作故事、戏剧和其他类别创意所存在的四大局限性问题,而开发的多类别创意生成引擎。它能有效解决创意领域中面临的四大限制:局限的流派多样性、输出的不足长度、叙事连贯性的弱化以及对复杂结构建构的执行能力不足。其核心采用了故事原型,这是一种基于知识图谱的叙事表示方法,通过编码角色、事件和环境为语义三元组,将故事逻辑与风格实现相分离。CrAgentive采用三阶段代理工作流程,包括构建用户指定的叙事骨架的初始化阶段;长短期目标指导多代理对话以实例化故事原型的生成阶段;以及利用此原型产生具有诸如反思和预示等高级结构的多流派文本的写作阶段。这种架构减少了存储冗余,并克服了长形式生成的典型瓶颈。在广泛的实验中,CrAgentive使用通用骨干模型以低于每100章不到一美元的成本生成了数千章的文本并保持稳定的品质。为评估性能,我们采用了一个包含十个叙事指标的二维框架来衡量文本的质量和长度。实验结果显示,CrAgentive持续超越强大的基线模型并在多种流派中表现稳健,接近人类作者撰写的小说质量。

关键见解

  1. CrAgentive解决了当代大型语言模型在创意写作领域的四大核心问题:流派多样性、输出长度、叙事连贯性和复杂结构建构能力。
  2. 采用故事原型作为核心,通过知识图谱叙事表示法,将故事逻辑与风格实现分离。
  3. 通过三阶段代理工作流程,实现用户叙事骨架的构建、故事原型的实例化和多流派文本的生成。
  4. 架构减少了存储冗余,并能克服长文本生成的典型瓶颈。
  5. 实验显示CrAgentive能以较低的成本生成大量章节文本并保持稳定品质。
  6. CrAgentive的性能评估采用包含质量和长度的二维框架,表现超越基线模型并在多种流派中稳健。

Cool Papers

点此查看论文截图

PANDA: Towards Generalist Video Anomaly Detection via Agentic AI Engineer

Authors:Zhiwei Yang, Chen Gao, Mike Zheng Shou

Video anomaly detection (VAD) is a critical yet challenging task due to the complex and diverse nature of real-world scenarios. Previous methods typically rely on domain-specific training data and manual adjustments when applying to new scenarios and unseen anomaly types, suffering from high labor costs and limited generalization. Therefore, we aim to achieve generalist VAD, i.e., automatically handle any scene and any anomaly types without training data or human involvement. In this work, we propose PANDA, an agentic AI engineer based on MLLMs. Specifically, we achieve PANDA by comprehensively devising four key capabilities: (1) self-adaptive scene-aware strategy planning, (2) goal-driven heuristic reasoning, (3) tool-augmented self-reflection, and (4) self-improving chain-of-memory. Concretely, we develop a self-adaptive scene-aware RAG mechanism, enabling PANDA to retrieve anomaly-specific knowledge for anomaly detection strategy planning. Next, we introduce a latent anomaly-guided heuristic prompt strategy to enhance reasoning precision. Furthermore, PANDA employs a progressive reflection mechanism alongside a suite of context-aware tools to iteratively refine decision-making in complex scenarios. Finally, a chain-of-memory mechanism enables PANDA to leverage historical experiences for continual performance improvement. Extensive experiments demonstrate that PANDA achieves state-of-the-art performance in multi-scenario, open-set, and complex scenario settings without training and manual involvement, validating its generalizable and robust anomaly detection capability. Code is released at https://github.com/showlab/PANDA.

视频异常检测(VAD)是一项至关重要且具有挑战性的任务,因为现实世界的场景复杂多变。以前的方法通常依赖于特定领域的训练数据和在应用到新场景和未见异常类型时的手动调整,导致劳动成本高昂和有限的泛化能力。因此,我们的目标是实现通用VAD,即能够自动处理任何场景和任何异常类型,无需训练数据或人工参与。在这项工作中,我们提出了基于MLLMs的PANDA智能体AI工程师。具体来说,我们通过全面设计四个核心能力来实现PANDA:(1)自适应场景感知策略规划,(2)目标驱动启发式推理,(3)工具增强自我反思,以及(4)自我改进的记忆链。具体来说,我们开发了一种自适应场景感知的RAG机制,使PANDA能够为异常检测策略规划检索特定异常的知识。接下来,我们引入了一种潜在异常引导的启发式提示策略,以提高推理精度。此外,PANDA采用了一种渐进的反思机制和一系列上下文感知工具,以迭代优化复杂场景中的决策制定。最后,记忆链机制使PANDA能够利用历史经验不断改善性能。大量实验表明,PANDA在无训练和人工参与的多场景、开放集和复杂场景设置中实现了最先进的性能,验证了其泛化和鲁棒的异常检测能力。代码已发布在https://github.com/showlab/PANDA。

论文及项目相关链接

PDF Accepted by NeurIPS 2025

Summary
本文提出一种基于多模态轻量级模型(MLLMs)的智能代理PANDA,旨在实现无需训练数据和人工参与的通用视频异常检测(VAD)。PANDA通过四个关键能力实现自我适应的场景感知策略规划、目标驱动启发式推理、工具辅助的自我反思以及自我提升的记忆链。实验表明,PANDA在多场景、开放集和复杂场景设置中实现了最先进的性能,验证了其通用性和鲁棒性的异常检测能力。

Key Takeaways

  1. 视频异常检测(VAD)是一个具有挑战性和实际意义的研究领域,由于现实场景的复杂性和多样性。
  2. 此前的方法通常依赖于特定领域的训练数据和手动调整,在新场景和未见异常类型中的应用存在高劳动成本有限泛化能力的问题。
  3. 本文旨在实现通用VAD,即无需训练数据或人工参与即可自动处理任何场景和任何异常类型。
  4. 提出一种基于多模态轻量级模型(MLLMs)的智能代理PANDA,通过四个关键能力实现:自我适应的场景感知策略规划、目标驱动启发式推理、工具辅助的自我反思以及自我提升的记忆链。
  5. PANDA采用自适应场景感知RAG机制,能够检索异常特定的知识来规划异常检测策略。
  6. PANDA通过潜在异常引导的启发式提示策略提高推理精度,并采用渐进反射机制和一系列上下文感知工具来迭代优化复杂场景中的决策制定。

Cool Papers

点此查看论文截图

Efficient and Transferable Agentic Knowledge Graph RAG via Reinforcement Learning

Authors:Jinyeop Song, Song Wang, Julian Shun, Yada Zhu

Knowledge-graph retrieval-augmented generation (KG-RAG) couples large language models (LLMs) with structured, verifiable knowledge graphs (KGs) to reduce hallucinations and expose reasoning traces. However, many KG-RAG systems compose multiple LLM modules (e.g planning, reasoning, and responding), inflating inference cost and binding behavior to a specific target KG. To address this, we introduce KG-R1, an agentic KG retrieval-augmented generation (KG-RAG) framework through reinforcement learning (RL). KG-R1 utilizes a single agent that interacts with KGs as its environment, learning to retrieve at each step and incorporating the retrieved information into its reasoning and generation. The process is optimized through end-to-end RL. In controlled experiments across Knowledge-Graph Question Answering (KGQA) benchmarks, our method demonstrates both efficiency and transferability: Using Qwen-2.5-3B, KG-R1 improves answer accuracy with fewer generation tokens than prior multi-module workflow methods that use larger foundation or fine-tuned models. Furthermore, KG-R1 enables plug and play: after training, it maintains strong accuracy on new KGs without modification. These properties make KG-R1 a promising KG-RAG framework for real-world deployment. Our code is publicly available at https://github.com/Jinyeop3110/KG-R1.

知识图谱检索增强生成(KG-RAG)将大型语言模型(LLM)与结构化的、可验证的知识图谱(KG)相结合,以减少幻觉并暴露推理痕迹。然而,许多KG-RAG系统由多个LLM模块组成(例如规划、推理和响应),增加了推理成本,并且使行为绑定到特定的目标知识图谱。为了解决这一问题,我们引入了KG-R1,这是一个通过强化学习(RL)实现的知识图谱检索增强生成(KG-RAG)框架。KG-R1利用单个智能体与其环境(知识图谱)进行交互,学习在每一步进行检索,并将检索到的信息融入其推理和生成过程中。该过程通过端到端的RL进行优化。在知识图谱问答(KGQA)基准测试上的受控实验表明,我们的方法既有效率也有迁移性:使用Qwen-2.5-3B,KG-R1在提高答案准确性的同时,生成的标记符号少于使用更大基础或微调模型的多模块工作流程方法。此外,KG-R1实现了即插即用:训练后,它在新的知识图谱上无需修改就能保持高准确性。这些特性使KG-R1成为现实世界部署中前景广阔的知识图谱检索增强生成框架。我们的代码公开在https://github.com/Jinyeop3110/KG-R1

论文及项目相关链接

PDF 10 pages, 5 figures. Submitted to ICLR 2026

Summary

KG-R1是一种基于强化学习(RL)的知识图谱检索增强生成(KG-RAG)框架。它使用一个单一智能体与环境(知识图谱)进行交互,学习在每一步进行检索,并将检索到的信息融入其推理和生成过程中。KG-R1解决了传统KG-RAG系统存在的问题,如推理模块过多、推理成本高昂以及行为特定于目标知识图谱的束缚。实验证明,KG-R1在知识图谱问答(KGQA)基准测试中具有较高的效率和可移植性,可在不使用大型基础模型或精细调整模型的情况下,提高答案的准确性并减少生成标记。此外,KG-R1还具备“即插即用”的特性,在训练后能够在新知识图谱上保持高精确度而无需进行修改。

Key Takeaways

  1. KG-R1结合了大型语言模型(LLMs)和知识图谱(KGs),以减少幻想并暴露推理轨迹。
  2. 传统KG-RAG系统存在模块过多、推理成本高昂的问题。
  3. KG-R1通过强化学习(RL)实现单一智能体与知识图谱的交互,简化了流程并提高了效率。
  4. KG-R1在知识图谱问答(KGQA)基准测试中表现出色,提高了答案准确性并减少了生成标记。
  5. KG-R1具备“即插即用”特性,可适应不同的知识图谱而无需修改。
  6. KG-R1优化了知识图谱检索过程,使其更加高效且适应性强。

Cool Papers

点此查看论文截图

Your Agent May Misevolve: Emergent Risks in Self-evolving LLM Agents

Authors:Shuai Shao, Qihan Ren, Chen Qian, Boyi Wei, Dadi Guo, Jingyi Yang, Xinhao Song, Linfeng Zhang, Weinan Zhang, Dongrui Liu, Jing Shao

Advances in Large Language Models (LLMs) have enabled a new class of self-evolving agents that autonomously improve through interaction with the environment, demonstrating strong capabilities. However, self-evolution also introduces novel risks overlooked by current safety research. In this work, we study the case where an agent’s self-evolution deviates in unintended ways, leading to undesirable or even harmful outcomes. We refer to this as Misevolution. To provide a systematic investigation, we evaluate misevolution along four key evolutionary pathways: model, memory, tool, and workflow. Our empirical findings reveal that misevolution is a widespread risk, affecting agents built even on top-tier LLMs (e.g., Gemini-2.5-Pro). Different emergent risks are observed in the self-evolutionary process, such as the degradation of safety alignment after memory accumulation, or the unintended introduction of vulnerabilities in tool creation and reuse. To our knowledge, this is the first study to systematically conceptualize misevolution and provide empirical evidence of its occurrence, highlighting an urgent need for new safety paradigms for self-evolving agents. Finally, we discuss potential mitigation strategies to inspire further research on building safer and more trustworthy self-evolving agents. Our code and data are available at https://github.com/ShaoShuai0605/Misevolution . Warning: this paper includes examples that may be offensive or harmful in nature.

自然语言模型(LLM)的进步催生了一种新型的自进化智能体,它们能够通过与环境互动自主改进,展现出强大的能力。然而,自我进化也带来了当前安全研究尚未关注的新型风险。在这项工作中,我们研究了智能体自我进化出现意外偏离的情况,导致出现不良甚至有害的结果。我们称之为“劣化进化”(Misevolution)。为了进行系统的研究,我们从模型、记忆、工具和工作流程四个方面对劣化进化进行了评估。我们的实证研究发现,劣化进化是一种广泛存在的风险,甚至影响顶尖的自然语言模型构建的智能体(例如Gemini-2.5-Pro)。在自我进化的过程中观察到不同的新兴风险,如记忆积累后安全对齐的退化,或在工具和重用的创建过程中意外引入的漏洞。据我们所知,这是首次系统地提出并实证了劣化进化的概念及其存在,突显出自进化智能体对新型安全范式的迫切需求。最后,我们讨论了潜在的缓解策略,以激发关于构建更安全、更可靠的自我进化智能体的进一步研究。我们的代码和数据可在 https://github.com/ShaoShuai0605/Misevolution 中找到。警告:本文包含可能具有冒犯性或有害性质的示例。

论文及项目相关链接

PDF Preprint. Under Review

Summary
大型语言模型(LLM)的进步催生了一种新型的自进化代理,它们通过与环境互动实现自主改进,展现出强大的能力。然而,自进化也带来了新的风险,被当前的安全研究所忽视。在本文中,我们探讨了代理自进化产生非预期结果的情况,我们称之为“不良进化”(Misevolution)。为了系统地研究这一问题,我们沿着模型、记忆、工具和工作流程四条关键进化路径对不良进化进行了评估。实证结果表明,不良进化是一种广泛存在的风险,甚至影响顶尖的大型语言模型(如Gemini-2.5-Pro)。在自进化过程中观察到多种新兴风险,例如记忆积累后安全性的对齐程度下降,以及工具创建和再利用过程中意外引入的脆弱性。据我们所知,这是首个系统地概念化不良进化并提供实证证据的研究,突显了为自进化代理开发新的安全范式紧迫性。最后,我们讨论了缓解策略以激发更安全、更可靠的自进化代理的研究。

Key Takeaways

  1. 大型语言模型的进步使得自进化代理具备强大的能力,并能通过与环境互动自主改进。
  2. 自进化代理存在不良进化的风险,可能导致非预期和不希望的结果。
  3. 不良进化影响顶尖的大型语言模型,包括在模型、记忆、工具和工作流程方面的关键进化路径上。
  4. 在自进化过程中观察到多种新兴风险,如安全性和脆弱性问题。
  5. 当前研究是首个系统地概念化不良进化并提供实证证据的研究。
  6. 需要新的安全范式来应对自进化代理的不良进化风险。

Cool Papers

点此查看论文截图

LLM Agents for Knowledge Discovery in Atomic Layer Processing

Authors:Andreas Werbrouck, Marshall B. Lindsay, Matthew Maschmann, Matthias J. Young

Large Language Models (LLMs) have garnered significant attention for several years now. Recently, their use as independently reasoning agents has been proposed. In this work, we test the potential of such agents for knowledge discovery in materials science. We repurpose LangGraph’s tool functionality to supply agents with a black box function to interrogate. In contrast to process optimization or performing specific, user-defined tasks, knowledge discovery consists of freely exploring the system, posing and verifying statements about the behavior of this black box, with the sole objective of generating and verifying generalizable statements. We provide proof of concept for this approach through a children’s parlor game, demonstrating the role of trial-and-error and persistence in knowledge discovery, and the strong path-dependence of results. We then apply the same strategy to show that LLM agents can explore, discover, and exploit diverse chemical interactions in an advanced Atomic Layer Processing reactor simulation using intentionally limited probe capabilities without explicit instructions.

近年来,大型语言模型(LLMs)已经引起了人们多年的关注。最近,有人提出了将其用作独立推理代理。在这项工作中,我们测试了此类代理在材料科学中发现知识的潜力。我们重新利用LangGraph的工具功能,为代理提供一个黑箱函数进行询问。与流程优化或执行特定、用户定义的任务不同,知识发现包括自由探索系统,提出并验证有关黑箱行为的陈述,以生成和验证可推广的陈述为唯一目标。我们通过儿童游乐游戏证明了这种方法的概念,展示了在知识发现中尝试和错误的重要性、毅力的作用以及结果的强烈路径依赖性。然后,我们应用相同的策略来展示LLM代理能够在先进的原子层处理反应器模拟中探索、发现和利用各种化学相互作用,即使使用有意限制的探测能力且没有明确的指令。

论文及项目相关链接

PDF Accepted submission to the AI4MAT workshop@NEURIPS 2025. As submitted, except author names added

Summary

大型语言模型(LLM)近年备受瞩目,最近人们提出将其用作独立推理的代理。本研究旨在测试此类代理在材料科学领域知识发现的潜力。研究使用LangGraph的工具功能,为代理提供黑箱函数以进行交互。与过程优化或执行特定的用户定义任务不同,知识发现是对系统进行自由探索,对黑箱的行为提出和验证陈述,以生成和验证可概括的陈述为主要目标。本研究通过儿童游戏证明了该方法的可行性,展示了知识发现中尝试和错误的重要性、持久性以及结果的强烈路径依赖性。随后,研究应用相同的策略展示了LLM代理在先进的原子层处理反应器模拟中探索、发现和利用各种化学交互的能力,即使使用了有限的探测能力且无明确的指令。

Key Takeaways

  1. 大型语言模型(LLM)作为独立推理代理在知识发现领域具有潜力。
  2. 使用LangGraph工具功能为代理提供黑箱函数进行交互测试。
  3. 知识发现不同于任务优化或特定用户任务,旨在自由探索系统并生成可概括的陈述。
  4. 通过儿童游戏证明了知识发现方法的可行性,强调了尝试和错误的重要性以及结果的路径依赖性。
  5. LLM代理能够在复杂的化学交互环境中进行探索、发现和利用。
  6. 即使在有限的探测能力和无明确指令的情况下,LLM代理仍能够展现其知识发现的潜力。

Cool Papers

点此查看论文截图

Toward an Unbiased Collective Memory for Efficient LLM-Based Agentic 6G Cross-Domain Management

Authors:Hatim Chergui, Miguel Catalan Cid, Pouria Sayyad Khodashenas, Daniel Camps Mur, Christos Verikoukis

This paper introduces a novel framework for proactive cross-domain resource orchestration in 6G RAN-Edge networks, featuring large language model (LLM)-augmented agents. The system comprises specialized RAN (energy efficiency) and Edge (latency assurance) agents that engage in iterative negotiation, supported by advanced reasoning and planning capabilities. Agents dynamically interact with a digital twin (DT) to test their proposals and leverage a long-term collective memory where their joint successful and failed agreements along with the related network contexts are distilled into strategies to either follow or avoid and subsequently stored. Given that agents are subject to a plethora of cognitive distortions when retrieving those past experiences – such as primacy, recency, confirmation and availability biases – we propose in this work a novel unbiased memory design (A reusable mockup version of the unbiased memory source code is available for non-commercial use at https://github.com/HatimChergui/unbiased-collective-memory). featuring (i) semantic retrieval of past strategies via Jaccard similarity; (ii) learning from failures through amplified weighting of SLA violations and mandatory inclusion of failed negotiation cases to mitigate confirmation bias; (iii) diversity enforcement to minimize availability bias and (iv) recency and primacy weighting with slow decay to counteract temporal biases. Evaluation results showcase the impact of existing biases and how the unbiased memory allows to tackle them by learning from both successful and failed strategies, either present or old, resulting in $\times 4.5$ and $\times 3.5$ reductions of unresolved negotiations compared to non-memory and vanilla memory baselines, respectively, while totally mitigating SLA violations as well as improving latency and energy saving distributions.

本文介绍了一种用于6G RAN-Edge网络中主动跨域资源编排的新型框架,该框架具有使用大型语言模型(LLM)增强的代理。该系统包括专门从事能源效率(RAN)和延迟保证(边缘)的代理,这些代理参与迭代协商,并得到先进推理和规划能力的支持。代理与数字双胞胎(DT)动态交互以测试其提案,并利用长期集体记忆功能,将他们的联合成功和失败的协议以及相关网络上下文提炼成要遵循或避免的策略,然后存储起来。考虑到代理在检索这些过去经验时会受到大量认知扭曲的影响,例如首因效应、近因效应、确认偏见和可得性偏见等,我们在本工作中提出了一种新型的无偏见内存设计。(一个可重复使用的无偏见内存源代码的模拟版本可供非商业使用,链接为:[https://github.com/HatimChergui/unbiased-collective-memory)。该设计特点包括:(i)通过Jaccard相似性进行过去策略的语义检索;(ii)通过放大服务级别协议违规的权重以及强制包含失败的谈判案例来学习从失败中汲取教训,以减轻确认偏见;(iii)强制实施多样性以最小化可得性偏见;(iv)通过缓慢衰减来平衡近因和首因效应,以对抗时间偏见。评估结果展示了现有偏见的影响,以及无偏见内存如何通过从成功和失败的策略中学习来解决问题。与无记忆基准和常规记忆基准相比,未解决的谈判减少了×4.5和×3.5,同时完全减轻了服务级别协议违规行为,并改善了延迟和节能分布。](https://github.com/HatimChergui/unbiased-collective-memory%EF%BC%89%E8%AF%A5%E8%AE%BE%E8%AE%A1%E7%89%B9%E7%82%B9%E5%8C%BA%E5%9C%B0%E5%AD%97%E5%AE%BD%E5%B9%B3%E6%AF%94%EF%BC%8C%E5%A6%82%E6%AD%A3%E5%AF%BC%E5%BC%BA%$ %E7 %9A % 84 % E7 %BB % 8F % E9 %AA % 8C % E5 % B9 % B6 % E4 % B8 % 8D % E5 % A4 % A7 % E5 % B0 % BD % E6 % B2 % A2 % E5 % A4 % A7 % E7 %BA % A1 % EFF )

论文及项目相关链接

PDF 12 pages, 8 figures

Summary

本文介绍了一种用于6G RAN-Edge网络中的主动跨域资源编排的新型框架。该系统包括专业化的RAN(能源效率)和边缘(延迟保证)代理,这些代理支持高级推理和规划能力,并参与迭代谈判。代理通过数字双胞胎进行测试其提议并动态交互,同时利用长期集体记忆存储成功和失败的协议以及相关网络上下文以制定策略。鉴于代理在检索过去的经验时会受到认知偏差的影响,本文提出了一种新型的无偏见记忆设计,包括语义检索过去策略、从失败中学习、多样性和执行权重的控制等策略。评估结果显示无偏见记忆有助于解决现有偏见问题,通过学习成功和失败的策略来减少未解决的谈判并提高网络性能。

Key Takeaways

  1. 论文提出了一种基于大型语言模型的新型框架用于主动跨域资源编排在6G RAN-Edge网络中。
  2. 系统包括专业化的RAN和边缘代理,通过数字双胞胎进行测试并参与迭代谈判。
  3. 代理具有高级推理和规划能力,并从过去成功的经验和失败中学习,制定出有效的策略来提高网络性能。这些策略会存储在长期集体记忆中以便以后使用。
  4. 论文指出代理在检索过去经验时存在的认知偏见问题,并介绍了一种新型无偏见记忆设计来应对这些偏见。无偏见记忆包括语义检索、从失败中学习、多样性和执行权重的控制等策略。评估结果显示无偏见记忆能够减少未解决的谈判并提高网络性能。该设计的源代码已公开发布供非商业使用。
  5. 通过数字双胞胎技术,代理能够模拟并测试其提议,从而提高网络性能和可靠性。同时,该技术也有助于减少物理测试的成本和时间。

Cool Papers

点此查看论文截图

Beyond the Algorithm: A Field Guide to Deploying AI Agents in Clinical Practice

Authors:Jack Gallifant, Katherine C. Kellogg, Matt Butler, Amanda Centi, Patrick F. Doyle, Sayon Dutta, Joyce Guo, Matthew J. Hadfield, Esther H. Kim, David E. Kozono, Hugo JWL Aerts, Adam B. Landman, Raymond H. Mak, Rebecca G. Mishuris, Tanna L. Nelson, Guergana K. Savova, Elad Sharon, Benjamin C. Silverman, Umit Topaloglu, Jeremy L. Warner, Danielle S. Bitterman

Large language models (LLMs) integrated into agent-driven workflows hold immense promise for healthcare, yet a significant gap exists between their potential and practical implementation within clinical settings. To address this, we present a practitioner-oriented field manual for deploying generative agents that use electronic health record (EHR) data. This guide is informed by our experience deploying the “irAE-Agent”, an automated system to detect immune-related adverse events from clinical notes at Mass General Brigham, and by structured interviews with 20 clinicians, engineers, and informatics leaders involved in the project. Our analysis reveals a critical misalignment in clinical AI development: less than 20% of our effort was dedicated to prompt engineering and model development, while over 80% was consumed by the sociotechnical work of implementation. We distill this effort into five “heavy lifts”: data integration, model validation, ensuring economic value, managing system drift, and governance. By providing actionable solutions for each of these challenges, this field manual shifts the focus from algorithmic development to the essential infrastructure and implementation work required to bridge the “valley of death” and successfully translate generative AI from pilot projects into routine clinical care.

大型语言模型(LLM)集成到代理驱动的工作流中,在医疗保健领域具有巨大的潜力。然而,在临床环境中实现其潜力和实际应用之间存在着显著的差距。为了解决这一问题,我们编制了一本面向实践的领域手册,介绍如何部署使用电子健康记录(EHR)数据的生成代理。该指南的编写参考了我们部署“irAE-Agent”(一种用于从Mass General Brigham的临床笔记中检测免疫相关不良事件的自动化系统)的经验以及与参与该项目的20名临床医生、工程师和信息技术领导进行的结构化访谈。我们的分析揭示了在临床人工智能发展中存在关键的不匹配问题:我们的努力中只有不到20%投入到即时工程和模型开发中,而超过80%的精力却耗费在实施的社会技术工作上。我们将这些努力归结为五个“艰巨的任务”:数据集成、模型验证、确保经济价值、管理系统漂移和治理。这本领域手册针对这些挑战提供了可行的解决方案,从而将重点从算法开发转向必要的基础设施和实施工作,以填补“死亡之谷”,成功地将生成式人工智能从试点项目转化为常规临床护理。

论文及项目相关链接

PDF Under review. 5 Tables, 2 Figures

Summary
大型语言模型在医疗领域具有巨大潜力,但其在临床环境中的实际应用与其潜力之间存在较大差距。为解决这一问题,本文提供了面向实践的领域手册,介绍如何部署使用电子健康记录数据的生成式智能体。通过部署“irAE-Agent”系统的经验和与相关项目参与者的结构化访谈,本文揭示了在临床人工智能开发中的关键误区,并提出了五个重要挑战的解决方案,包括数据集成、模型验证、确保经济价值、管理系统漂移和治理等方面。此手册旨在将重点从算法开发转向必要的基础设施和实施工作,以成功将生成式人工智能从试点项目转化为常规的临床关怀。

Key Takeaways

  1. 大型语言模型在医疗领域有巨大潜力,但实际应用与潜力之间存在差距。
  2. 部署生成式智能体时,需要关注数据集成、模型验证等五个重要方面。
  3. 数据集成是智能体部署的关键环节之一,需要解决数据质量和标准化问题。
  4. 模型验证是确保智能体准确性和可靠性的重要步骤。
  5. 在实施智能体系统时,需要确保经济价值和长期可持续性。
  6. 管理系统漂移是智能体部署中需要解决的挑战之一,以确保系统的稳定性和准确性。

Cool Papers

点此查看论文截图

DyFlow: Dynamic Workflow Framework for Agentic Reasoning

Authors:Yanbo Wang, Zixiang Xu, Yue Huang, Xiangqi Wang, Zirui Song, Lang Gao, Chenxi Wang, Xiangru Tang, Yue Zhao, Arman Cohan, Xiangliang Zhang, Xiuying Chen

Agent systems based on large language models (LLMs) have shown great potential in complex reasoning tasks, but building efficient and generalizable workflows remains a major challenge. Most existing approaches rely on manually designed processes, which limits their adaptability across different tasks. While a few methods attempt automated workflow generation, they are often tied to specific datasets or query types and make limited use of intermediate feedback, reducing system robustness and reasoning depth. Moreover, their operations are typically predefined and inflexible. To address these limitations, we propose DyFlow, a dynamic workflow generation framework that adaptively constructs and adjusts reasoning procedures based on task requirements and real-time intermediate feedback, thereby enhancing cross-task generalization. DyFlow consists of two core components: a designer and an executor. The designer decomposes complex problems into a sequence of sub-goals defined by high-level objectives and dynamically plans the next steps based on intermediate outputs and feedback. These plans are then carried out by the executor, which executes each operation using dynamic operators with context-aware parameterization, enabling flexible and semantically grounded reasoning. We systematically evaluate DyFlow across diverse domains, including social reasoning, biomedical tasks, mathematical problem solving, and code generation. Results demonstrate that DyFlow significantly outperforms existing baselines, achieving substantial Pass@k improvements and exhibiting robust generalization across diverse domains. The code is publicly available at https://github.com/wyf23187/DyFlow.

基于大语言模型(LLM)的Agent系统在复杂的推理任务中显示出巨大的潜力,但构建高效且可推广的工作流程仍然是一个主要挑战。大多数现有方法依赖于手动设计的过程,这限制了它们在不同任务中的适应性。尽管有一些方法尝试进行自动工作流程生成,但它们通常与特定数据集或查询类型相关联,并且对中间反馈的利用有限,降低了系统的稳健性和推理深度。而且,它们的操作通常是预先定义的,不够灵活。为了解决这些局限性,我们提出了DyFlow,一个动态工作流程生成框架,它可以根据任务需求和实时中间反馈自适应地构建和调整推理过程,从而提高跨任务的通用性。DyFlow由两个核心组件组成:设计师和执行者。设计师将复杂的问题分解为由高级目标定义的一系列子目标,并基于中间输出和反馈动态地规划下一步。然后,执行者根据这些计划执行操作,使用具有上下文感知参数化的动态操作符,实现灵活且语义丰富的推理。我们在社会推理、生物医学任务、数学问题解决和代码生成等多个领域对DyFlow进行了系统评估。结果表明,DyFlow显著优于现有基线,实现了实质性的Pass@k改进,并在多个领域中表现出稳健的泛化能力。代码可在https://github.com/wyf23187/DyFlow公开获取。

论文及项目相关链接

PDF

Summary

基于大型语言模型(LLM)的Agent系统在复杂推理任务中展现出巨大潜力,但构建高效且可推广的工作流程仍是主要挑战。现有方法大多依赖手动设计过程,限制了其在不同任务中的适应性。针对这一问题,本文提出DyFlow动态工作流程生成框架,根据任务需求和实时中间反馈自适应构建和调整推理流程,提高跨任务泛化能力。DyFlow包含设计师和执行官两个核心组件,设计师将复杂问题分解为一系列子目标,并根据中间输出和反馈动态规划下一步操作;执行官则负责执行每个操作,采用具有上下文感知的参数化动态操作符,实现灵活且语义丰富的推理。实验结果表明,DyFlow在社交推理、生物医学任务、数学问题求解和代码生成等多个领域均显著优于现有基线方法,实现了显著的Pass@k提升和稳健的跨域泛化能力。

Key Takeaways

  1. Agent系统基于大型语言模型(LLM)在复杂推理任务中具有巨大潜力。
  2. 构建高效、可推广的工作流程是当前的挑战。
  3. 现有方法主要依赖手动设计过程,限制了其在不同任务中的适应性。
  4. DyFlow是一个动态工作流程生成框架,可以自适应地构建和调整推理流程。
  5. DyFlow包括设计师和执行官两个核心组件,分别负责问题分解与动态规划、操作执行。
  6. DyFlow通过实时中间反馈和任务需求调整推理流程,提高跨任务泛化能力。
  7. DyFlow在多个领域均表现出显著优于现有基线方法的性能。

Cool Papers

点此查看论文截图

RE-Searcher: Robust Agentic Search with Goal-oriented Planning and Self-reflection

Authors:Daocheng Fu, Jianbiao Mei, Licheng Wen, Xuemeng Yang, Cheng Yang, Rong Wu, Tao Hu, Siqi Li, Yufan Shen, Xinyu Cai, Pinlong Cai, Botian Shi, Yong Liu, Yu Qiao

Large language models (LLMs) excel at knowledge-intensive question answering and reasoning, yet their real-world deployment remains constrained by knowledge cutoff, hallucination, and limited interaction modalities. Augmenting LLMs with external search tools helps alleviate these issues, but it also exposes agents to a complex search environment in which small, plausible variations in query formulation can steer reasoning into unproductive trajectories and amplify errors. We present a systematic analysis that quantifies how environmental complexity induces fragile search behaviors and, in turn, degrades overall performance. To address this challenge, we propose a simple yet effective approach to instantiate a search agent, RE-Searcher. During search, RE-Searcher explicitly articulates a concrete search goal and subsequently reflects on whether the retrieved evidence satisfies that goal. This combination of goal-oriented planning and self-reflection enables RE-Searcher to resist spurious cues in complex search environments and perform robust search. Extensive experiments show that our method improves search accuracy and achieves state-of-the-art results. Perturbation studies further demonstrate substantial resilience to noisy or misleading external signals, mitigating the fragility of the search process. We believe these findings offer practical guidance for integrating LLM-powered agents into more complex interactive environments and enabling more autonomous decision-making.

大型语言模型(LLM)在知识密集型问答和推理方面表现出色,但它们在现实世界中的应用仍受到知识截断、幻觉和有限交互模式的制约。通过外部搜索工具增强LLM有助于缓解这些问题,但同时也将代理暴露于复杂的搜索环境中,其中查询表述的微小、看似合理的变化可能导致推理偏离正确的方向并放大错误。我们对环境复杂性如何导致脆弱的搜索行为进行了系统分析,并转而降低了整体性能。为了应对这一挑战,我们提出了一种简单有效的方法来实例化搜索代理RE-Searcher。在搜索过程中,RE-Searcher明确表述了具体的搜索目标,然后反思检索到的证据是否满足该目标。目标导向规划和自我反思的结合使RE-Searcher能够抵抗复杂搜索环境中的虚假线索,并执行稳健的搜索。大量实验表明,我们的方法提高了搜索准确性,并达到了最新水平的结果。扰动研究进一步证明了其对嘈杂或误导性外部信号的显著抗性,减轻了搜索过程的脆弱性。我们相信这些发现为将LLM驱动的代理集成到更复杂的交互式环境中,并实现更自主的决策提供了实际指导。

论文及项目相关链接

PDF 16 pages, 7 figures

Summary

大型语言模型(LLMs)在知识密集型问答和推理方面表现出色,但其在实际部署中仍面临知识截断、幻象和交互模式有限等问题。通过增援外部搜索工具可缓解这些问题,但这也使代理面临复杂的搜索环境,查询表述中的微小且合理的变化可能导致推理偏离正确方向并放大错误。本文进行了系统分析,量化环境复杂性如何导致搜索行为变得脆弱,进而降低整体性能。为解决这一挑战,我们提出了一种简单而有效的方法来实例化搜索代理RE-Searcher。RE-Searcher在搜索过程中明确具体搜索目标,并反思检索到的证据是否满足目标。这种目标导向的规划和自我反思的结合,使RE-Searcher能够在复杂的搜索环境中抵御错误线索,实现稳健搜索。实验表明,我们的方法提高了搜索准确性并达到了最新结果。扰动研究进一步证明了其对嘈杂或误导性外部信号的强大抗性,减轻了搜索过程的脆弱性。

Key Takeaways

  1. 大型语言模型(LLMs)在知识密集型任务上表现出色,但存在知识截断、幻象和交互限制等问题。
  2. 外部搜索工具增援可以缓解这些问题,但引入复杂搜索环境,微小查询变化可能影响结果。
  3. 环境复杂性导致搜索行为脆弱,影响整体性能。
  4. RE-Searcher通过明确具体搜索目标和反思检索证据来满足目标来应对这一挑战。
  5. RE-Searcher结合目标导向规划及自我反思,能在复杂环境中稳健搜索。
  6. 实验证明RE-Searcher提高了搜索准确性并达到最新结果。

Cool Papers

点此查看论文截图

Authors:Huihao Jing, Wenbin Hu, Hongyu Luo, Jianhui Yang, Wei Fan, Haoran Li, Yangqiu Song

Multi-agent systems (MAS), leveraging the remarkable capabilities of Large Language Models (LLMs), show great potential in addressing complex tasks. In this context, integrating MAS with legal tasks is a crucial step. While previous studies have developed legal benchmarks for LLM agents, none are specifically designed to consider the unique advantages of MAS, such as task decomposition, agent specialization, and flexible training. In fact, the lack of evaluation methods limits the potential of MAS in the legal domain. To address this gap, we propose MASLegalBench, a legal benchmark tailored for MAS and designed with a deductive reasoning approach. Our benchmark uses GDPR as the application scenario, encompassing extensive background knowledge and covering complex reasoning processes that effectively reflect the intricacies of real-world legal situations. Furthermore, we manually design various role-based MAS and conduct extensive experiments using different state-of-the-art LLMs. Our results highlight the strengths, limitations, and potential areas for improvement of existing models and MAS architectures.

多智能体系统(MAS)借助大型语言模型(LLM)的出色能力,在应对复杂任务方面显示出巨大潜力。在此背景下,将MAS与法律任务集成是重要的一步。尽管先前的研究已经为LLM代理开发了法律基准,但没有一个是专门考虑MAS的独特优势,如任务分解、代理专业化和灵活训练。事实上,缺乏评估方法限制了MAS在法律领域的潜力。为了解决这一空白,我们提出MASLegalBench,这是一个专为MAS定制的法律基准,采用演绎推理方法设计。我们的基准以GDPR为应用场景,包含丰富的背景知识,涵盖有效的复杂推理过程,充分反映了现实法律情况的复杂性。此外,我们手动设计了各种基于角色的MAS,并使用不同的最新LLM进行广泛实验。我们的结果突显了现有模型和MAS架构的优势、局限性和潜在的改进领域。

论文及项目相关链接

PDF

Summary

基于多代理系统(MAS)利用大型语言模型(LLM)的出色能力,在解决复杂任务方面显示出巨大潜力。本文将MAS与法务任务相结合是关键一步。虽然已有法务基准用于评估LLM代理,但没有特定考虑MAS的独特优势,如任务分解、代理专业化和灵活训练等。为解决此评估方法缺失的问题,本文提出MASLegalBench,一个专为MAS设计的法务基准,采用演绎推理方法。此基准以GDPR为应用场景,包含丰富的背景知识,并能有效反映现实法律情境的复杂性。同时,本文通过不同先进的LLM进行大量实验,揭示了现有模型和MAS架构的优势、局限性和潜在的改进方向。

Key Takeaways

  1. 多代理系统(MAS)与大型语言模型(LLM)结合在解决复杂任务方面具有巨大潜力。
  2. 将MAS应用于法律任务是关键一步,但现有的法律基准未充分考虑MAS的独特优势。
  3. 缺少针对MAS的评估方法限制了其在法律领域的应用潜力。
  4. 提出MASLegalBench,一个专为MAS设计的法律基准,采用演绎推理方法。
  5. MASLegalBench以GDPR为应用场景,包含丰富的背景知识,有效反映现实法律情境的复杂性。
  6. 通过大量实验揭示了现有模型和MAS架构的优势和局限性。

Cool Papers

点此查看论文截图

Sequence Pathfinder for Multi-Agent Pickup and Delivery in the Warehouse

Authors:Zeyuan Zhao, Chaoran Li, Shao Zhang, Ying Wen

Multi-Agent Pickup and Delivery (MAPD) is a challenging extension of Multi-Agent Path Finding (MAPF), where agents are required to sequentially complete tasks with fixed-location pickup and delivery demands. Although learning-based methods have made progress in MAPD, they often perform poorly in warehouse-like environments with narrow pathways and long corridors when relying only on local observations for distributed decision-making. Communication learning can alleviate the lack of global information but introduce high computational complexity due to point-to-point communication. To address this challenge, we formulate MAPF as a sequence modeling problem and prove that path-finding policies under sequence modeling possess order-invariant optimality, ensuring its effectiveness in MAPD. Building on this, we propose the Sequential Pathfinder (SePar), which leverages the Transformer paradigm to achieve implicit information exchange, reducing decision-making complexity from exponential to linear while maintaining efficiency and global awareness. Experiments demonstrate that SePar consistently outperforms existing learning-based methods across various MAPF tasks and their variants, and generalizes well to unseen environments. Furthermore, we highlight the necessity of integrating imitation learning in complex maps like warehouses.

多智能体接送任务(MAPD)是多智能体路径查找(MAPF)的一个具有挑战性的扩展,其中要求智能体按顺序完成具有固定位置接送需求的任务。尽管基于学习的方法在MAPD中取得了进展,但在仓库等环境中,当仅依靠局部观察进行分布式决策时,它们在狭窄通道和长走廊中的表现往往不佳。通信学习可以缓解缺乏全局信息的问题,但由于点对点通信而引入了较高的计算复杂度。为了应对这一挑战,我们将MAPF制定为序列建模问题,并证明序列建模下的路径查找策略具有顺序不变的最优性,确保其适用于MAPD。在此基础上,我们提出了Sequential Pathfinder(SePar),它利用Transformer范式实现隐式信息交换,将决策复杂性从指数级降低到线性,同时保持效率和全局意识。实验表明,在各种MAPF任务及其变种中,SePar始终优于现有的基于学习的方法,并且能很好地推广到未见过的环境。此外,我们强调了在像仓库这样的复杂地图中集成模仿学习的必要性。

论文及项目相关链接

PDF Preprint Under Review

Summary

文本主要讨论了在多智能体路径寻找(MAPF)基础上提出的挑战性扩展任务——多智能体拾取与配送(MAPD)。针对仓库等环境,文本提出将MAPF建模为序列建模问题,并利用序列建模的优势设计了一种名为Sequential Pathfinder(SePar)的新方法。该方法结合Transformer范式实现隐性信息交换,降低了决策复杂性,同时保持了效率和全局意识。实验表明,SePar在各种MAPF任务及其变种中表现优异,并能很好地泛化到未见过的环境。此外,文章强调了复杂地图如仓库中集成模仿学习的必要性。

Key Takeaways

  1. 多智能体拾取与配送(MAPD)是多智能体路径寻找(MAPF)的挑战性扩展。
  2. 在仓库等环境中,基于局部观察的学习方法表现不佳。
  3. 通信学习可以缓解缺乏全局信息的问题,但计算复杂度较高。
  4. 将MAPF建模为序列建模问题,路径查找策略具有顺序不变的最优性。
  5. Sequential Pathfinder(SePar)利用Transformer范式实现隐性信息交换,降低决策复杂性。
  6. SePar在各种MAPF任务中表现优异,并能很好地泛化到未见过的环境。

Cool Papers

点此查看论文截图

InfiAgent: Self-Evolving Pyramid Agent Framework for Infinite Scenarios

Authors:Chenglin Yu, Yang Yu, Songmiao Wang, Yucheng Wang, Yifan Yang, Jinjia Li, Ming Li, Hongxia Yang

Large Language Model (LLM) agents have demonstrated remarkable capabilities in organizing and executing complex tasks, and many such agents are now widely used in various application scenarios. However, developing these agents requires carefully designed workflows, carefully crafted prompts, and iterative tuning, which requires LLM techniques and domain-specific expertise. These hand-crafted limitations hinder the scalability and cost-effectiveness of LLM agents across a wide range of industries. To address these challenges, we propose \textbf{InfiAgent}, a Pyramid-like DAG-based Multi-Agent Framework that can be applied to \textbf{infi}nite scenarios, which introduces several key innovations: a generalized “agent-as-a-tool” mechanism that automatically decomposes complex agents into hierarchical multi-agent systems; a dual-audit mechanism that ensures the quality and stability of task completion; an agent routing function that enables efficient task-agent matching; and an agent self-evolution mechanism that autonomously restructures the agent DAG based on new tasks, poor performance, or optimization opportunities. Furthermore, InfiAgent’s atomic task design supports agent parallelism, significantly improving execution efficiency. This framework evolves into a versatile pyramid-like multi-agent system capable of solving a wide range of problems. Evaluations on multiple benchmarks demonstrate that InfiAgent achieves 9.9% higher performance compared to ADAS (similar auto-generated agent framework), while a case study of the AI research assistant InfiHelper shows that it generates scientific papers that have received recognition from human reviewers at top-tier IEEE conferences.

大型语言模型(LLM)代理在组织和执行复杂任务方面表现出了卓越的能力,并且许多这样的代理现在已广泛应用于各种应用场景。然而,开发这些代理需要精心设计的工作流程、精心制作的提示和迭代调整,这需要LLM技术和特定领域的专业知识。这些手工制作的局限性阻碍了LLM代理在广泛行业中的可扩展性和成本效益。为了解决这些挑战,我们提出了InfiAgent,这是一个基于金字塔式的DAG多代理框架,可应用于无限场景,它引入了几项关键创新:一种通用的“代理作为工具”机制,可自动将复杂代理分解为分层的多代理系统;一种双重审计机制,确保任务完成的品质和稳定性;一个代理路由功能,能够实现高效的任务-代理匹配;以及一个代理自我进化机制,能够基于新任务、性能不佳或优化机会自主重组代理DAG。此外,InfiAgent的原子任务设计支持代理并行性,大大提高了执行效率。该框架演变成为一个通用的金字塔式多代理系统,能够解决一系列广泛的问题。在多个基准测试上的评估表明,InfiAgent与ADAS(类似的自动生成代理框架)相比,性能提高了9.9%。InfiHelper(人工智能研究助理的案例分析)生成的科学论文得到了顶级IEEE会议的人类评审者的认可。

论文及项目相关链接

PDF 9 pages of main content and 32 pages of others, 2 figures, under review as a conference paper at ICLR 2026

Summary

大型语言模型(LLM)代理在组织执行复杂任务方面展现出卓越的能力,并广泛应用于多种应用场景。然而,开发这些代理需要精心设计的工作流程、提示和迭代调整,这需要LLM技术和特定领域的专业知识。为解决手动构建的局限性,我们提出了InfiAgent,一个基于金字塔型DAG的多代理框架,可应用于无限场景。它引入了几个关键创新点:通用的“代理即工具”机制,自动将复杂代理分解为分层的多代理系统;双审计机制确保任务完成的品质和稳定性;代理路由功能实现高效的任务-代理匹配;代理自我进化机制可基于新任务、性能不佳或优化机会自主重组代理DAG。评估表明,InfiAgent相较于类似的自动生成的代理框架ADAS,性能提高了9.9%。

Key Takeaways

  1. LLM代理在组织执行复杂任务方面具有卓越能力,广泛应用于多个行业。
  2. 开发LLM代理需要精心设计的工作流程、提示和迭代调整,需具备LLM技术和特定领域的专业知识。
  3. 手动构建的局限性影响了LLM代理的可扩展性和成本效益。
  4. InfiAgent框架被提出解决这些挑战,它是一个基于金字塔型DAG的多代理框架,适用于无限场景。
  5. InfiAgent引入了几个关键创新,包括“代理即工具”机制、双审计机制、代理路由功能和代理自我进化机制。
  6. InfiAgent框架具有自动分解复杂代理、保证任务完成质量、高效任务-代理匹配和自主重组等特性。

Cool Papers

点此查看论文截图

VerlTool: Towards Holistic Agentic Reinforcement Learning with Tool Use

Authors:Dongfu Jiang, Yi Lu, Zhuofeng Li, Zhiheng Lyu, Ping Nie, Haozhe Wang, Alex Su, Hui Chen, Kai Zou, Chao Du, Tianyu Pang, Wenhu Chen

Reinforcement Learning with Verifiable Rewards (RLVR) has demonstrated success in enhancing LLM reasoning capabilities, but remains limited to single-turn interactions without tool integration. While recent Agentic Reinforcement Learning with Tool use (ARLT) approaches have emerged to address multi-turn tool interactions, existing works develop task-specific codebases that suffer from fragmentation, synchronous execution bottlenecks, and limited extensibility across domains. These inefficiencies hinder broader community adoption and algorithmic innovation. We introduce VerlTool, a unified and modular framework that addresses these limitations through systematic design principles. VerlTool provides four key contributions: (1) upstream alignment with VeRL ensuring compatibility and simplified maintenance, (2) unified tool management via standardized APIs supporting diverse modalities including code execution, search, SQL databases, and vision processing, (3) asynchronous rollout execution achieving near 2$\times$ speedup by eliminating synchronization bottlenecks, and (4) comprehensive evaluation demonstrating competitive performance across 6 ARLT domains. Our framework formalizes ARLT as multi-turn trajectories with multi-modal observation tokens (text/image/video), extending beyond single-turn RLVR paradigms. We train and evaluate models on mathematical reasoning, knowledge QA, SQL generation, visual reasoning, web search, and software engineering tasks, achieving results comparable to specialized systems while providing unified training infrastructure. The modular plugin architecture enables rapid tool integration requiring only lightweight Python definitions, significantly reducing development overhead and providing a scalable foundation for tool-augmented RL research. Our code is open-sourced at https://github.com/TIGER-AI-Lab/verl-tool.

强化学习与可验证奖励(RLVR)已成功提高了大型语言模型的推理能力,但仍局限于没有工具集成的单轮交互。虽然最近出现了工具使用强化学习(ARLT)的方法来解决多轮工具交互问题,但现有工作开发了任务特定的代码库,存在碎片化、同步执行瓶颈以及跨域扩展性有限等缺点。这些低效性阻碍了更广泛的社区采纳和算法创新。我们引入了VerlTool,这是一个统一和模块化的框架,通过系统设计原则来解决这些限制。VerlTool提供了四个主要贡献:(1)与VeRL的上游对齐,确保兼容性并简化维护;(2)通过标准化API进行统一工具管理,支持多种模式,包括代码执行、搜索、SQL数据库和视觉处理等;(3)实现异步推出执行,通过消除同步瓶颈实现近2倍的速度提升;(4)全面评估,在6个ARLT领域展示具有竞争力的性能。我们的框架将ARLT形式化为多轮轨迹,具有多模态观察令牌(文本/图像/视频),超越了单轮RLVR范式。我们在数学推理、知识问答、SQL生成、视觉推理、网络搜索和软件工程任务上对模型进行训练和评估,虽然结果可与专业系统相比,但提供了统一的训练基础设施。模块化的插件架构可以快速集成工具,只需要轻量级的Python定义,大大降低了开发成本,为工具增强型RL研究提供了可扩展的基础。我们的代码已公开在https://github.com/TIGER-AI-Lab/verl-tool。

论文及项目相关链接

PDF 32 pages, 5 figures, 13 tables

Summary
强化学习可验证奖励(RLVR)在提升大型语言模型推理能力方面取得了成功,但仍局限于没有工具集成的单轮交互。虽然最近出现了工具使用的代理强化学习(ARLT)方法来解决多轮工具交互问题,但现有工作开发的任务特定代码库存在碎片化、同步执行瓶颈和跨域扩展性有限等缺点。我们引入VerlTool,这是一个统一和模块化的框架,通过系统的设计原则来解决这些限制。VerlTool提供四个关键贡献:与VeRL的上游对齐确保兼容性并简化维护、通过标准化API进行统一工具管理支持多种模式包括代码执行、搜索、SQL数据库和视觉处理、异步推出执行实现近两倍速的提高通过消除同步瓶颈、全面的评估在6个ARLT领域表现出有竞争力的性能。我们的框架将ARLT形式化为多轮轨迹与多模态观察令牌(文本/图像/视频),超越了单轮RLVR范式。

Key Takeaways

  1. RLVR在提升LLM推理能力上取得进展,但仅限于单轮交互。
  2. ARLT方法解决了多轮工具交互问题,但现有工作存在碎片化、同步执行瓶颈和跨域扩展性问题。
  3. VerlTool框架通过统一和模块化设计解决这些问题,与VeRL对齐、统一工具管理、异步执行以提高速度。
  4. VerlTool提供标准化API支持多种工具模式,包括代码执行、搜索、SQL数据库和视觉处理。
  5. VerlTool实现全面的评估,在多个领域表现出竞争力,包括数学推理、知识问答、SQL生成、视觉推理、网络搜索和软件工程任务。
  6. VerlTool的模块化插件架构便于快速工具集成,只需轻量级Python定义,减少开发成本。

Cool Papers

点此查看论文截图

GraphCogent: Mitigating LLMs’ Working Memory Constraints via Multi-Agent Collaboration in Complex Graph Understanding

Authors:Rongzheng Wang, Shuang Liang, Qizhi Chen, Yihong Huang, Muquan Li, Yizhuo Ma, Dongyang Zhang, Ke Qin, Man-Fai Leung

Large language models (LLMs) show promising performance on small-scale graph reasoning tasks but fail when handling real-world graphs with complex queries. This phenomenon arises from LLMs’ working memory constraints, which result in their inability to retain long-range graph topology over extended contexts while sustaining coherent multi-step reasoning. However, real-world graphs are often structurally complex, such as Web, Transportation, Social, and Citation networks. To address these limitations, we propose GraphCogent, a collaborative agent framework inspired by human Working Memory Model that decomposes graph reasoning into specialized cognitive processes: sense, buffer, and execute. The framework consists of three modules: Sensory Module standardizes diverse graph text representations via subgraph sampling, Buffer Module integrates and indexes graph data across multiple formats, and Execution Module combines tool calling and tool creation for efficient reasoning. We also introduce Graph4real, a comprehensive benchmark that contains four domains of real-world graphs (Web, Transportation, Social, and Citation) to evaluate LLMs’ graph reasoning capabilities. Our Graph4real covers 21 different graph reasoning tasks, categorized into three types (Structural Querying, Algorithmic Reasoning, and Predictive Modeling tasks), with graph scales up to 10 times larger than existing benchmarks. Experiments show that Llama3.1-8B based GraphCogent achieves a 50% improvement over massive-scale LLMs like DeepSeek-R1 (671B). Compared to state-of-the-art agent-based baseline, our framework outperforms by 20% in accuracy while reducing token usage by 80% for in-toolset tasks and 30% for out-toolset tasks. Code will be available after review.

大型语言模型(LLMs)在小规模图推理任务上表现出有前景的性能,但在处理具有复杂查询的真实世界图时却失败了。这一现象源于LLMs的工作记忆约束,导致它们无法在长期上下文中保留长程图拓扑结构,同时维持连贯的多步推理。然而,真实世界的图通常是结构复杂的,如网页、交通、社交和引文网络。为了解决这些局限性,我们提出了GraphCogent,这是一个受人类工作记忆模型启发的协作代理框架,它将图推理分解成专门的认知过程:感知、缓冲和执行。该框架由三个模块组成:感官模块通过子图采样标准化各种图形文本表示,缓冲模块整合和索引跨多种格式的图数据,执行模块结合工具调用和工具创建进行高效推理。我们还引入了Graph4real,这是一个全面的基准测试,包含四个真实世界图的领域(网页、交通、社交和引文),以评估LLMs的图推理能力。我们的Graph4real涵盖了21种不同的图推理任务,分为三类(结构查询、算法推理和预测建模任务),图的规模高达现有基准测试的10倍。实验表明,基于Llama3.1-8B的GraphCogent在大型LLMs(如DeepSeek-R1 671B)上实现了50%的改进。与最先进的基于代理的基线相比,我们的框架在准确性上提高了20%,同时减少了工具内任务令牌使用量的80%和工具外任务令牌使用量的30%。代码将在审查后公布。

论文及项目相关链接

PDF

Summary

大型语言模型在小规模图推理任务上表现良好,但在处理具有复杂查询的真实世界图时表现不佳。这是由于大型语言模型存在工作记忆约束,无法在长期内保持图的拓扑结构并在扩展的上下文中维持连贯的多步推理。为解决这些限制,本文提出了GraphCogent框架,该框架由感官模块、缓冲模块和执行模块三个部分组成,旨在通过分解图推理任务为特定的认知过程来解决大型语言模型在处理复杂图推理任务时的局限性。同时,本文还介绍了Graph4real基准测试集,用于评估大型语言模型的图推理能力。实验表明,基于Llama3.1-8B的GraphCogent框架在大型语言模型上实现了50%的改进。

Key Takeaways

  1. 大型语言模型在小规模图推理任务上表现良好,但在处理真实世界的复杂图时存在局限性。
  2. 局限性的主要原因是大型语言模型的工作记忆约束,无法长期保持图的拓扑结构并进行连贯的多步推理。
  3. GraphCogent框架通过模拟人类工作记忆模型来解决这些限制,分解为感官模块、缓冲模块和执行模块三个专门化的认知过程。
  4. Graph4real基准测试集包含四个真实世界图的领域,用于评估大型语言模型的图推理能力。
  5. GraphCogent框架在大型语言模型上实现了显著的性能改进,相对于其他先进的基于代理的基线模型,准确度高出了20%。
  6. GraphCogent在内部任务上减少了80%的令牌使用,在外部任务上减少了30%。

Cool Papers

点此查看论文截图

Structured Agent Distillation for Large Language Model

Authors:Jun Liu, Zhenglun Kong, Peiyan Dong, Changdi Yang, Tianqi Li, Hao Tang, Geng Yuan, Wei Niu, Wenbin Zhang, Pu Zhao, Xue Lin, Dong Huang, Yanzhi Wang

Large language models (LLMs) exhibit strong capabilities as decision-making agents by interleaving reasoning and actions, as seen in ReAct-style frameworks. Yet, their practical deployment is constrained by high inference costs and large model sizes. We propose Structured Agent Distillation, a framework that compresses large LLM-based agents into smaller student models while preserving both reasoning fidelity and action consistency. Unlike standard token-level distillation, our method segments trajectories into {[REASON]} and {[ACT]} spans, applying segment-specific losses to align each component with the teacher’s behavior. This structure-aware supervision enables compact agents to better replicate the teacher’s decision process. Experiments on ALFWorld, HotPotQA-ReAct, and WebShop show that our approach consistently outperforms token-level and imitation learning baselines, achieving significant compression with minimal performance drop. Scaling and ablation results further highlight the importance of span-level alignment for efficient and deployable agents.

大型语言模型(LLM)通过推理和行动交织展现出强大的作为决策代理的能力,如在ReAct风格框架中所见。然而,它们的实际应用受到高推理成本和大型模型规模的限制。我们提出了结构化代理蒸馏框架,该框架将大型LLM代理压缩成较小的学生模型,同时保留推理保真度和行动一致性。不同于标准的令牌级蒸馏,我们的方法将轨迹分割为{[REASON]}和{[ACT]}跨度,对每种组件应用特定的损失来与教师的行为对齐。这种结构感知的监督使紧凑的代理能够更好地复制教师的决策过程。在ALFWorld、HotPotQA-ReAct和WebShop上的实验表明,我们的方法始终优于令牌级和模仿学习基线,实现了显著的压缩和最小的性能下降。规模和消融结果进一步强调了跨度级对齐对于高效和可部署代理的重要性。

论文及项目相关链接

PDF

Summary

大型语言模型(LLMs)通过融合推理与行动展现出强大的决策能力,如ReAct框架所示。然而,其实践部署受限于高推理成本和庞大的模型规模。我们提出结构化代理蒸馏框架,它能将大型LLM代理压缩成较小的学生模型,同时保持推理的保真度和行动的连贯性。不同于标准的token级蒸馏,我们的方法将轨迹分段为推理和行动区间,并应用特定区间的损失来使每个组件与教师的行为对齐。这种结构化的监督使得紧凑的代理能更好地复制教师的决策过程。在ALFWorld、HotPotQA-ReAct和WebShop的实验表明,我们的方法始终优于token级和模仿学习的基线,实现了显著压缩和极小性能损失。比例和消融结果进一步突显了区间级对齐对于高效和可部署代理的重要性。

Key Takeaways

  1. 大型语言模型(LLMs)展现强大的决策能力,通过融合推理与行动,但面临高推理成本和庞大模型规模的问题。
  2. 提出结构化代理蒸馏框架,能在保持推理的保真度和行动的连贯性的同时,将大型LLM代理压缩成较小的学生模型。
  3. 与标准的token级蒸馏不同,该方法通过分段轨迹进行推理和行动区间的特定损失应用,实现与教师行为的对齐。
  4. 结构化的监督使得紧凑的代理能更好地复制教师的决策过程。
  5. 在多个实验环境中,该方法的性能始终优于token级和模仿学习基线,实现了显著压缩和极小性能损失。
  6. 实验结果证明,区间级对齐对于高效和可部署的代理至关重要。
  7. 提出的框架为未来大型语言模型在现实世界中的应用提供了可能的解决方案。

Cool Papers

点此查看论文截图


文章作者: Kedreamix
版权声明: 本博客所有文章除特別声明外,均采用 CC BY 4.0 许可协议。转载请注明来源 Kedreamix !
 上一篇
MMT MMT
MMT 方向最新论文已更新,请持续关注 Update in 2025-10-02 A Culturally-diverse Multilingual Multimodal Video Benchmark & Model
2025-10-02
下一篇 
LLM LLM
LLM 方向最新论文已更新,请持续关注 Update in 2025-10-02 MLA A Multisensory Language-Action Model for Multimodal Understanding and Forecasting in Robotic Manipulation
2025-10-02
  目录