AI Agent - arXivDaily 专题

2406.06469 2024-06-11 cs.AI cs.CL cs.LG 83%

Husky: A Unified, Open-Source Language Agent for Multi-Step Reasoning

Joongwon Kim, Bhargavi Paranjape, Tushar Khot, Hannaneh Hajishirzi

专题命中规划决策：agent(title,abstract)；分类 cs.AI、cs.CL、cs.LG

Comments 50 pages, 42 figures. Project webpage available [here](https://agent-husky.github.io/)

2402.12741 2024-05-27 cs.CV 83%

MuLan: Multimodal-LLM Agent for Progressive and Interactive Multi-Object Diffusion

Sen Li, Ruochen Wang, Cho-Jui Hsieh, Minhao Cheng, Tianyi Zhou

专题命中规划决策：agent(title,abstract)；planning(abstract)

Comments Added the application to human-agent interaction; added discussion with concurrent work

2402.00296 2024-02-02 cs.RO 83%

High-Level, Collaborative Task Planning Grammar and Execution for Heterogeneous Agents

Amy Fang, Hadas Kress-Gazit

专题命中规划决策：planning(title)；agent(abstract)；multi-agent(abstract)；autonomous agent(comments)

Comments To appear in the Proceedings of the 2024 International Conference on Autonomous Agents and Multiagent Systems (AAMAS)

2305.04614 2023-05-09 cs.RO 83%

Reducing Onboard Processing Time for Path Planning in Dynamically Evolving Polygonal Maps

Aditya Shirwatkar, Aman Singh, Jana Ravi Kiran

专题命中规划决策：planning(title,abstract)；autonomous agent(abstract)

Comments Course Project Report for CP:230 Motion Planning and Autonomous Navigation by Dr. Debasish Ghose

2608.03502 2026-08-20 cs.AI cs.LG cs.MA 版本更新 82%

Hybrid LLM-Augmented Reinforcement Learning Agents for Complex Sequential Decision Tasks

用于复杂序列决策任务的混合大语言模型增强强化学习智能体

Christophe D. Hounwanou, John Emeka Eze, Yaé Ulrich Gaba

专题命中规划决策：agent(abstract)；autonomous agent(abstract)；tool-use(abstract)；planning(abstract)

AI总结本文提出混合LLM增强RL智能体，融合LLM规划与RL动作优化，经实验验证其在序列决策任务上优于仅RL、仅LLM的基线方法，为构建更强自主系统提供了新方向。

Comments This submission is withdrawn because the uploaded manuscript does not accurately reflect the intended structure or results. Several components referenced in the text are incomplete or not represented in the PDF, and the current version may mislead readers. The work is therefore withdrawn to maintain clarity of the record

2607.06008 2026-08-18 cs.AI cs.CL 版本更新 82%

PolyWorkBench: Benchmarking LLM Agents for Cross-Lingual Long-Horizon Workflows

PolyWorkBench：多语言长视野语言模型智能体基准测试

Hongliang Li, Yijin Liu, Zhiwei Zhang, Zihe Liu, Xinyue Lou, Jinan Xu, Fandong Meng, Kaiyu Huang

机构 * Beijing Jiaotong University（北京交通大学）； Weixin AI, Tencent Inc（腾讯微云人工智能实验室）

专题命中规划决策：agent(abstract)；tool use(abstract)；tool-use(abstract)；planning(abstract)

AI总结研究多语言长视野工作流程中LLM智能体的表现，提出PolyWorkBench基准测试及结合多种评估方式的混合框架，发现最先进LLM智能体在多语言设置中性能降，强调联合建模语言变化和程序决策对智能体评估的重要性。

Comments 17 Pages, 5 figures

2606.18947 2026-06-18 cs.AI cs.CL cs.IR cs.MA 新提交 82%

Decoupling Search from Reasoning: A Vendor-Agnostic Grounding Architecture for LLM Agents

将搜索与推理解耦：面向LLM Agent的供应商无关的接地架构

Emmanuel Aboah Boateng, Kyle MacDonald, Amardeep Kumar, Siddharth Kodwani, Sudeep Das

机构 * DoorDash, Inc.（DoorDash公司）

专题命中规划决策：agent(title_cn,abstract_cn)；agentic(abstract)；分类 cs.AI、cs.CL

AI总结提出解耦搜索接地（DSG）架构，将搜索接地从推理模型中分离，通过MCP兼容网关实现供应商路由、缓存等控制，在降低成本和延迟的同时保持或提升准确性。

Comments 15 pages, Figure 8

2606.07682 2026-06-09 cs.SE cs.AI 新提交 82%

SWE-Marathon: Can Agents Autonomously Complete Ultra-Long-Horizon Software Work?

SWE-Marathon: 智能体能否自主完成超长时程软件工作？

Rishi Desai, Jesse Hu, Joan Cabezas, Neel Harsola, Pratyush Shukla, Roey Ben Chaim, Adnan El Assadi, Omkaar Mukund Kamath, Fenil Faldu, Prannay Hebbar, Jiankai Sun, Yiyuan Li, Pramod Srinivasan, Ishan Gupta, Christopher Settles, Daniel Wang, Derek Chen, Pranav Raja, Albert Liu, Marek Šuppa, Nevasini Sasikumar, Luyang Kong, Erik Quintanilla, Xiangyi Li, Ivan Bercovich, Steven Dillmann

机构 * Abundant ； Zenity ； Harvard University（哈佛大学）； University of Waterloo（滑铁卢大学）； Gujarat Technological University（古吉拉特技术大学）； Warping ； Stanford University（斯坦福大学）； UNC-Chapel Hill（北卡罗来纳大学教堂山分校）； Independent（独立）； Refresh ； Soleda AI ； Near AI ； Georgia Tech（佐治亚理工学院）； Comenius University in Bratislava（布拉迪斯拉发Comenius大学）； UC San Diego（圣地亚哥大学）； BenchFlow ； UC Santa Barbara（圣巴巴拉大学）

专题命中规划决策：agent(abstract)；AI agent(abstract)；planning(abstract)；workflow(abstract)

AI总结提出SWE-Marathon基准，包含20个超长时程任务，平均消耗2720万token，评估智能体在规划、长上下文理解和记忆方面的能力，当前前沿编码智能体解决率低于30%。

2604.14084 2026-05-22 cs.LG cs.AI 82%

TIP: Token Importance in On-Policy Distillation

TIP: on-policy distillation 中的 token 重要性

Yuanda Xu, Hejian Sang, Zhengze Zhou, Ran He, Zhipeng Wang, Alborz Geramifard

专题命中规划决策：planning(abstract,abstract_cn)；agentic(abstract,abstract_cn)；分类 cs.AI、cs.LG

AI总结本研究探讨了在 on-policy 知识蒸馏中哪些 token 对学习信号最有用，提出了一种基于学生熵和教师-学生分歧的双轴分类方法，并通过实验验证了在有限内存条件下使用少量 token 进行蒸馏的有效性。

2605.09217 2026-05-12 cs.AI cs.LG cs.MA 82%

Learning the Preferences of a Learning Agent

学习学习代理的偏好

Karim Abdel Sadek, Mark Bedaywi, Rhys Gould, Stuart Russell

机构 * University of California, Berkeley（加州大学伯克利分校）

专题命中规划决策：agent(title,abstract)；分类 cs.AI、cs.LG；multi-agent(comments)

AI总结本文研究学习代理偏好学习问题，通过分析在线学习者行为推断潜在奖励函数，探讨无 regrets 或收敛到最优 Boltzmann 策略的理论保证。

Comments Published at ICLR 2026, Workshop on Multi-Agent Learning and Its Opportunities in the Era of Generative AI. 9 pages main text

2605.08257 2026-05-12 cs.CR cs.AI cs.LG 82%

Research on Security Enhancement Methods for Adversarial Robust Large Language Model Intelligent Agents for Medical Decision-Making Tasks

对抗鲁棒性增强方法研究：用于医疗决策任务的对抗鲁棒大语言模型智能体

Saisai Hu

机构 * Pace University（帕克大学）

专题命中规划决策：agent(summary_cn,abstract)；分类 cs.AI、cs.LG

AI总结本文提出ARSM-Agent框架，通过多模块协同提升医疗决策智能体的对抗鲁棒性与安全性，实验表明其在多种攻击下攻击成功率降低至8.7%。

Comments 5 pages, 2 figures, 1 table.Accepted for oral presentation at AINIT 2026

2605.05632 2026-05-08 cs.CR cs.CL cs.LG 82%

Architecture Matters: Comparing RAG Systems under Knowledge Base Poisoning

架构至关重要：在知识库中毒情况下比较RAG系统

Samuel Korn

专题命中规划决策：agentic(abstract,abstract_cn)；agent(abstract)；multi-agent(abstract)；分类 cs.CL、cs.LG

AI总结研究探讨了在知识库中毒情况下不同RAG架构的鲁棒性，发现架构设计对攻击成功率影响显著，MADAM-RAG在矛盾检测上表现最佳但仍有不足。

2605.00798 2026-05-04 cs.LG cs.CL cs.MA 82%

RunAgent: Interpreting Natural-Language Plans with Constraint-Guided Execution

RunAgent：通过约束引导执行解释自然语言计划

Arunabh Srivastava, Mohammad A., Khojastepour, Srimat Chakradhar, Sennur Ulukus

机构 * University of Maryland, College Park（马里兰大学学院公园分校）； NEC Laboratories America, Inc.（NEC美国实验室）

专题命中规划决策：agent(abstract)；workflow(abstract)；agentic(abstract)；multi-agent(abstract)

AI总结 RunAgent通过约束引导执行解释自然语言计划，结合代理语言的显式控制构造，提升结构化工作流执行的可靠性。

2604.07791 2026-04-21 cs.AI cs.LG 82%

SEARL: Joint Optimization of Policy and Tool Graph Memory for Self-Evolving Agents

SEARL：自进化智能体中策略与工具图记忆的联合优化

Xinshun Feng, Xinhao Song, Lijun Li, Gongshen Liu, Jing Shao

机构 * Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）； Shanghai Jiaotong University（上海交通大学）

专题命中规划决策：agent(abstract)；planning(abstract)；agentic(abstract)；multi-agent(abstract)

AI总结 SEARL通过构建结构化经验记忆实现策略与工具图记忆的联合优化，提升智能体在知识推理和数学任务中的实用性和效率。

Comments ACL 2026

2603.12230 2026-04-07 cs.LG cs.AI cs.CR 82%

Security Considerations for Artificial Intelligence Agents

人工智能代理的安全性考虑

Ninghui Li, Kaiyuan Zhang, Kyle Polley, Jerry Ma

机构 * Perplexity

专题命中规划决策：agent(abstract)；AI agent(abstract)；agentic(abstract)；multi-agent(abstract)

AI总结本文探讨了前沿AI代理的安全性问题，分析了代理架构对代码-数据分离、权限边界和执行可预测性的影响，并提出了针对间接提示注入、混淆代理行为和长流程级联故障的防护措施。

Comments This article is adapted from Perplexity's response to NIST/CAISI Request for Information 2025-0035. 91 Fed. Reg. 698 (Jan. 8, 2026). The originally submitted response can be found on the public docket at https://www.regulations.gov/comment/NIST-2025-0035-0505

2604.00324 2026-04-02 cs.LG cs.AI 82%

The Persistent Vulnerability of Aligned AI Systems

对齐AI系统持续的脆弱性

Aengus Lynch

机构 * University College London（伦敦大学学院）

专题命中规划决策：agent(abstract)；AI agent(abstract)；planning(abstract)；agentic(abstract)

AI总结本文探讨了自主AI代理在文件系统访问、电子邮件控制和多步骤规划中的部署问题，提出ACDC、LAT等方法以解决AI安全四大难题，通过实验展示对抗性攻击的成功率及模型对齐测试结果。

Comments PhD thesis, University College London, 2025. 157 pages. Supervised by Ricardo Silva

2603.26458 2026-03-30 cs.SE cs.AI 82%

Can AI Models Direct Each Other? Organizational Structure as a Probe into Training Limitations

AI模型能否相互指导？组织结构作为训练限制的探测器

Rui Liu

机构 * Independent Researcher（独立研究员）

专题命中规划决策：agent(abstract)；tool use(abstract)；planning(abstract)；multi-agent(abstract)

AI总结本文研究AI模型通过ManagerWorker双agent流水线相互指导的有效性，发现强管理者指导弱工作者与单体强模型性能相近，但弱管理者指导弱工作者效果更差，揭示了多agent指导的潜力与限制。

2603.19248 2026-03-23 cs.CL cs.AI 82%

DuCCAE: A Hybrid Engine for Immersive Conversation via Collaboration, Augmentation, and Evolution

DuCCAE：通过协作、增强和进化实现沉浸式对话的混合引擎

Xin Shen, Zhishu Jiang, Jiaye Yang, Haibo Liu, Yichen Wan, Jiarui Zhang, Tingzhi Dai, Luodong Xu, Shuchen Wu, Guanqiang QI, Chenxi Miao, Jiahui Liang, Yang Li, Weikang Li, Deguo Xia, Jizhou Huang

机构 * Baidu Inc（百度公司）

专题命中规划决策：agent(abstract)；planning(abstract)；agentic(abstract)；multi-agent(abstract)

AI总结 DuCCAE通过解耦实时响应与异步代理执行，提升沉浸式对话的响应速度与任务能力，减少延迟并增强用户留存与复杂任务完成率。

2507.20796 2026-03-16 econ.GN cs.AI cs.LG q-fin.EC 82%

Aligning Large Language Model Agents with Rational and Moral Preferences: A Supervised Fine-Tuning Approach

对齐大型语言模型代理的理性与道德偏好：一种监督微调方法

Wei Lu, Amit Dhanda, Daniel L. Chen, Christian B. Hansen

机构 * Zicklin School of Business, CUNY Baruch College（纽约大学法学院）； Amazon（亚马逊）； Toulouse School of Economics（图卢兹经济学院）

专题命中规划决策：agent(abstract)；AI agent(abstract)；autonomous agent(abstract)；multi-agent(abstract)

AI总结本文通过监督微调方法对齐LLM代理的理性与道德偏好，揭示了代理在经济游戏中的系统性偏差，并展示了不同偏好结构对均衡结果的影响。

2512.04276 2025-12-05 cs.AI cs.LG math.ST stat.TH 82%

The Geometry of Benchmarks: A New Path Toward AGI

基准的几何学：通向通用人工智能的新路径

Przemyslaw Chojecki

机构 * Przemyslaw Chojecki（独立研究者）

专题命中规划决策：agent(abstract)；AI agent(abstract)；tool use(abstract)；planning(abstract)

AI总结本文提出了一种基于几何框架的基准评估方法，通过自主AI尺度和GVU操作符，为通用人工智能的发展提供了新的理论视角。

2510.01654 2025-10-03 cs.CL cs.AI 82%

SoK: Measuring What Matters for Closed-Loop Security Agents

Mudita Khurana, Raunak Jain

机构 * IEEE

专题命中规划决策：autonomous agent(abstract)；tool use(abstract)；planning(abstract)；agentic(abstract)

2401.07128 2024-10-07 cs.CL cs.AI 82%

EHRAgent: Code Empowers Large Language Models for Few-shot Complex Tabular Reasoning on Electronic Health Records

Wenqi Shi, Ran Xu, Yuchen Zhuang, Yue Yu, Jieyu Zhang, Hang Wu, Yuanda Zhu, Joyce Ho, Carl Yang, May D. Wang

专题命中规划决策：agent(abstract)；autonomous agent(abstract)；tool-use(abstract)；planning(abstract)

Comments Accepted in EMNLP 2024 main conference

2402.01680 2024-04-22 cs.CL cs.AI cs.MA 82%

Large Language Model based Multi-Agents: A Survey of Progress and Challenges

Taicheng Guo, Xiuying Chen, Yaqi Wang, Ruidi Chang, Shichao Pei, Nitesh V. Chawla, Olaf Wiest, Xiangliang Zhang

专题命中规划决策：agent(abstract)；autonomous agent(abstract)；planning(abstract)；multi-agent(abstract)

Comments This work is ongoing and we welcome your contribution!

2309.14365 2023-09-27 cs.CL cs.AI 82%

An In-depth Survey of Large Language Model-based Artificial Intelligence Agents

Pengyu Zhao, Zijian Jin, Ning Cheng

专题命中规划决策：agent(abstract)；AI agent(abstract)；tool use(abstract)；planning(abstract)

2103.09568 2022-04-22 cs.AI cs.LG 82%

A Practical Guide to Multi-Objective Reinforcement Learning and Planning

Conor F. Hayes, Roxana Rădulescu, Eugenio Bargiacchi, Johan Källström, Matthew Macfarlane, Mathieu Reymond, Timothy Verstraeten, Luisa M. Zintgraf, Richard Dazeley, Fredrik Heintz, Enda Howley, Athirai A. Irissappane, Patrick Mannion, Ann Nowé, Gabriel Ramos, Marcello Restelli, Peter Vamplew, Diederik M. Roijers

专题命中规划决策：planning(title,abstract)；分类 cs.AI、cs.LG；agent(journal_ref)；multi-agent(journal_ref)

Journal ref Auton Agent Multi-Agent Syst 36, 26 (2022)

2509.13309 2025-09-23 cs.CL 82%

WebResearcher: Unleashing unbounded reasoning capability in Long-Horizon Agents

Zile Qiao, Guoxin Chen, Xuanzhong Chen, Donglei Yu, Wenbiao Yin, Xinyu Wang, Zhen Zhang, Baixuan Li, Huifeng Yin, Kuan Li, Rui Min, Minpeng Liao, Yong Jiang, Pengjun Xie, Fei Huang, Jingren Zhou

机构 * Tongyi Lab , Alibaba Group（通义实验室，阿里巴巴集团）

专题命中规划决策：agent(abstract,comments)；AI agent(abstract)；tool-use(abstract)；multi-agent(abstract)

Comments https://tongyi-agent.github.io/blog/introducing-tongyi-deep-research/

2104.05773 2021-08-30 cs.RO cs.AI cs.ET 82%

Approximate Computing for Robotic path planning -- Experimentation, Case Study and Practical Implications

Hrishav Bakul Barua

专题命中规划决策：planning(title,abstract)；分类 cs.AI；agent(comments)；multi-agent(comments)

Comments Approximate Computing, Multi-robot Systems, Multi-agent Systems, Good Enough Computing, Green Computing, Robot Path Planning, Energy-Efficient Computing

2508.19114 2026-08-20 cs.RO cs.MA 82%

DELIVER: A System for LLM-Guided Coordinated Multi-Robot Pickup and Delivery using Voronoi-Based Relay Planning

Alkesh K. Srivastava, Jared Michael Levin, Alexander Derrico, Philip Dames

机构 * Department of Mechanical Engineering, Temple University（机械工程系， Temple 大学）

专题命中规划决策：planning(title,abstract)；agent(abstract)

Comments Submission under review at the 2026 IEEE/SICE International Symposium on System Integration (SII 2026)

2608.16798 2026-08-18 cs.CL cs.AI cs.LG 新提交 82%

ClawGym II: Exploring Black-Box RL on Agent Harness

ClawGym II：探索智能体控制框架下的黑盒强化学习

Huatong Song, Fei Bai, Ming Yang, Renyuan Li, Jia Deng, Jujie He, Zhange Zhang, Daixuan Cheng, Yan Xing, Qi Yun, Xuxing Chen, Danyang Li, Feng Chang, Chuan Hao, Ran Tao, Jian Yang, Bryan Dai, Wayne Xin Zhao, Mingjie Tang, Ji-Rong Wen

机构 * Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学高瓴人工智能学院）； IQuest Research（智臻研究）