[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-thinkwee--AgentsMeetRL":3,"tool-thinkwee--AgentsMeetRL":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",143909,2,"2026-04-07T11:33:18",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107888,"2026-04-06T11:32:50",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":10,"last_commit_at":59,"category_tags":60,"status":17},4487,"LLMs-from-scratch","rasbt\u002FLLMs-from-scratch","LLMs-from-scratch 是一个基于 PyTorch 的开源教育项目，旨在引导用户从零开始一步步构建一个类似 ChatGPT 的大型语言模型（LLM）。它不仅是同名技术著作的官方代码库，更提供了一套完整的实践方案，涵盖模型开发、预训练及微调的全过程。\n\n该项目主要解决了大模型领域“黑盒化”的学习痛点。许多开发者虽能调用现成模型，却难以深入理解其内部架构与训练机制。通过亲手编写每一行核心代码，用户能够透彻掌握 Transformer 架构、注意力机制等关键原理，从而真正理解大模型是如何“思考”的。此外，项目还包含了加载大型预训练权重进行微调的代码，帮助用户将理论知识延伸至实际应用。\n\nLLMs-from-scratch 特别适合希望深入底层原理的 AI 开发者、研究人员以及计算机专业的学生。对于不满足于仅使用 API，而是渴望探究模型构建细节的技术人员而言，这是极佳的学习资源。其独特的技术亮点在于“循序渐进”的教学设计：将复杂的系统工程拆解为清晰的步骤，配合详细的图表与示例，让构建一个虽小但功能完备的大模型变得触手可及。无论你是想夯实理论基础，还是为未来研发更大规模的模型做准备",90106,"2026-04-06T11:19:32",[35,15,13,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":72,"owner_name":73,"owner_avatar_url":74,"owner_bio":75,"owner_company":76,"owner_location":77,"owner_email":78,"owner_twitter":79,"owner_website":80,"owner_url":81,"languages":82,"stars":87,"forks":88,"last_commit_at":89,"license":78,"difficulty_score":90,"env_os":91,"env_gpu":92,"env_ram":92,"env_deps":93,"category_tags":96,"github_topics":97,"view_count":32,"oss_zip_url":78,"oss_zip_packed_at":78,"status":17,"created_at":111,"updated_at":112,"faqs":113,"releases":114},5113,"thinkwee\u002FAgentsMeetRL","AgentsMeetRL","Awesome List for Agentic RL","AgentsMeetRL 是一个专注于“大语言模型智能体与强化学习结合”的开源项目精选清单。随着大模型在复杂任务中需要更强的自主决策能力，如何让智能体通过试错和自我进化来掌握工具使用、多轮交互及逻辑推理，成为当前技术攻关的难点。AgentsMeetRL 正是为了解决这一痛点而生，它系统性地梳理了全球范围内利用强化学习训练 LLM 智能体的优质开源代码库。\n\n这份清单不仅涵盖了通用的强化学习训练框架（如 veRL、OpenRLHF），还细致地分类整理了涉及搜索增强、网页操作、代码工程、多智能体协作、记忆管理以及安全对齐等垂直领域的具体实现。其独特亮点在于深入剖析了各个项目所依赖的技术栈，包括具体的 RL 算法、奖励机制设计以及训练环境选择，并提供了交互式仪表盘供用户直观浏览技术细节。\n\nAgentsMeetRL 非常适合 AI 研究人员、大模型应用开发者以及对智能体前沿技术感兴趣的技术爱好者使用。无论是希望寻找合适的基线模型进行二次开发，还是想要了解业界最新的技术选型趋势，都能在这里获得极具价值的参考指引，从而加速高质量智能体系统的构建与迭代。","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fthinkwee_AgentsMeetRL_readme_020ab9c761ce.png\" alt=\"NOVER Logo\" width=\"500\">\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n\n![Base Framework](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FBase_Framework-18-BFA2DB?style=for-the-badge)\n![General](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGeneral-16-4E6813?style=for-the-badge)\n![Search & RAG](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSearch_&_RAG-25-845C40?style=for-the-badge)\n![Web & GUI](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FWeb_&_GUI-20-A259FF?style=for-the-badge)\n\u003Cbr>\n![Tool](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTool-12-D89F7B?style=for-the-badge)\n![Code & SWE](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCode_&_SWE-19-A47B67?style=for-the-badge)\n![Reasoning](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FReasoning-15-FF69B4?style=for-the-badge)\n![Multi-Agent](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMulti--Agent-7-1F4CAD?style=for-the-badge)\n\u003Cbr>\n![Memory](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMemory-3-007a88?style=for-the-badge)\n![Embodied](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FEmbodied-2-C0C5CE?style=for-the-badge)\n![Domain-Specific](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDomain--Specific-5-ffc884?style=for-the-badge)\n![Reward & Training](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FReward_&_Training-5-9B59B6?style=for-the-badge)\n\u003Cbr>\n![Safety](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSafety-5-E74C3C?style=for-the-badge)\n![VLM Agent](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FVLM_Agent-7-2ECC71?style=for-the-badge)\n![Self-Evolution](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSelf--Evolution-6-F39C12?style=for-the-badge)\n![Environment](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FEnvironment-40-FA5A4C?style=for-the-badge)\n\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n\n[![Interactive Dashboard](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F📊_Interactive_Dashboard-Visit_Website-blue?style=for-the-badge)](https:\u002F\u002Fthinkwee.top\u002Famr\u002F)\n\n\u003C\u002Fdiv>\n\n# When LLM Agents Meet Reinforcement Learning\n\n**AgentsMeetRL** is an awesome list that summarizes **open-source repositories** for training LLM Agents using reinforcement learning:\n - 🤖 The criteria for identifying an agent project are that it must have at least one of the following: multi-turn interactions or tool use (so TIR projects, Tool-Integrated Reasoning, are considered in this repo).\n - ⚠️ This project is based on code analysis from open-source repositories using LLM coding agents, which may contain unfaithful cases. Although manually reviewed, there may still be omissions. If you find any errors, please don't hesitate to let us know immediately through issues or PRs - we warmly welcome them!\n - 🚀 We particularly focus on the reinforcement learning frameworks, RL algorithms, rewards, and environments that projects depend on, for everyone's reference on how these excellent open-source projects make their technical choices. See [Click to view technical details] under each table.\n - 📅 Last updated: 2026-03-24\n - 🤗 Feel free to submit your own projects anytime - we welcome contributions!\n\nTaxonomy:\n - **Base Framework**: General-purpose RL training frameworks for LLM agents (e.g., veRL, OpenRLHF, trl)\n - **General\u002FMultiTask**: Agent systems trained\u002Fevaluated across multiple tasks or environments\n - **Search & RAG**: Search-augmented reasoning agents that use retrieval tools to enhance LLM reasoning\n - **Web & GUI**: Agents that interact with web browsers, mobile\u002Fdesktop GUIs, or operating systems\n - **Tool-Use**: Agents trained to invoke external tools (APIs, code executors, MCP, etc.)\n - **Code & SWE**: Software engineering and code generation agents\n - **Reasoning**: Reasoning agents with tool-integrated or multi-turn reasoning (math, QA, visual)\n - **Multi-Agent RL**: Multi-agent collaboration, negotiation, or credit assignment via RL\n - **Memory**: Agents that learn to manage, retrieve, or evolve memory\n - **Embodied**: Agents operating in embodied\u002Fphysical simulation environments\n - **Domain-Specific**: RL agents for specialized domains (medical, OS tuning, etc.)\n - **Reward & Training**: Process\u002Foutcome reward models and training methodologies for agents\n - **Safety**: RL for agent safety alignment, adversarial red-teaming, and jailbreak defense\u002Fattack\n - **VLM Agent**: Vision-language model agents trained with RL for multimodal interaction\n - **Self-Evolution**: Agents that self-evolve via RL feedback loops (⚠️ definition still evolving in the community)\n - **Environment**: Benchmarks, gyms, and sandbox environments for agent training\u002Fevaluation\n\nSome Enumeration:\n - Enumeration for Reward Type:\n   - External Verifier: e.g., a compiler or math solver\n   - Rule-Based: e.g., a LaTeX parser with exact match scoring\n   - Model-Based: e.g., a trained verifier LLM or reward LLM\n   - Custom\n\n---\n\n## Updates\n- 📢 **2026-03 Update**: Restructured taxonomy from 12 to 16 categories. Added ~70 new repositories covering Sep 2025 – Mar 2026. New categories include Multi-Agent RL, Reward & Training, Safety, VLM Agent, Self-Evolution, and Domain-Specific. Merged the old GUI and Web into Web & GUI, retired TextGame and Biomedical as standalone categories. Total repos grew from ~134 to 205.\n\n## 🔧 Base Framework\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link |\n| :----: | :----: | :----: |  :----: | :----: |\n| [Open-AgentRL](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FOpen-AgentRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGen-Verse\u002FOpen-AgentRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | Gen-Verse | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.02488) |\n| [OpenClaw-RL](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FOpenClaw-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGen-Verse\u002FOpenClaw-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | Gen-Verse | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2603.10165) |\n| [Claw-R1](https:\u002F\u002Fgithub.com\u002FAgentR1\u002FClaw-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAgentR1\u002FClaw-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | USTC | -- |\n| [prime-rl](https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fprime-rl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FPrimeIntellect-ai\u002Fprime-rl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Prime Intellect | -- |\n| [NeMo-RL](https:\u002F\u002Fgithub.com\u002FNVIDIA-NeMo\u002FRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNVIDIA-NeMo\u002FRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | NVIDIA | -- |\n| [RLinf](https:\u002F\u002Fgithub.com\u002FRLinf\u002FRLinf) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRLinf\u002FRLinf?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | Tsinghua\u002FInfinigence AI\u002FPKU | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.15965) |\n| [siiRL](https:\u002F\u002Fgithub.com\u002Fsii-research\u002FsiiRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsii-research\u002FsiiRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | Shanghai Innovation Institute | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.13833) |\n| [slime](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fslime) | ![](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002Fslime?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700) | 2025.6 | Tsinghua University (THUDM) | [blog](https:\u002F\u002Flmsys.org\u002Fblog\u002F2025-07-09-slime\u002F) |\n| [agent-lightning](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fagent-lightning) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fagent-lightning?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Microsoft Research | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.03680) |\n| [AReaL](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FAReaL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FinclusionAI\u002FAReaL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | AntGroup\u002FTsinghua | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.24298) |\n| [ROLL](https:\u002F\u002Fgithub.com\u002Falibaba\u002FROLL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Falibaba\u002FROLL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Alibaba | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.06122) |\n| [MARTI](https:\u002F\u002Fgithub.com\u002FTsinghuaC3I\u002FMARTI) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTsinghuaC3I\u002FMARTI?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Tsinghua | -- |\n| [RL2](https:\u002F\u002Fgithub.com\u002FChenmienTan\u002FRL2) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FChenmienTan\u002FRL2?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | Accio | – |\n| [verifiers](https:\u002F\u002Fgithub.com\u002Fwillccbb\u002Fverifiers) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fwillccbb\u002Fverifiers?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | Individual | -- |\n| [oat](https:\u002F\u002Fgithub.com\u002Fsail-sg\u002Foat) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsail-sg\u002Foat?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.11 | NUS\u002FSea AI | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2411.01493) |\n| [veRL](https:\u002F\u002Fgithub.com\u002Fvolcengine\u002Fverl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvolcengine\u002Fverl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.10 | ByteDance | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2409.19256) |\n| [OpenRLHF](https:\u002F\u002Fgithub.com\u002FOpenRLHF\u002FOpenRLHF) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenRLHF\u002FOpenRLHF?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2023.7 | OpenRLHF | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2405.11143) |\n| [trl](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftrl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhuggingface\u002Ftrl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2019.11 | HuggingFace | -- |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [Open-AgentRL](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FOpen-AgentRL) | GRPO-TCR | Single | Both | Multi | Reasoning\u002FGUI\u002FCoding | Model (PRM) | Yes (SandboxFusion) |\n| [OpenClaw-RL](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FOpenClaw-RL) | GRPO\u002FOPD | Both | Both | Multi | Terminal\u002FGUI\u002FSWE\u002FTool-call | Model\u002FExternal | Yes |\n| [Claw-R1](https:\u002F\u002Fgithub.com\u002FAgentR1\u002FClaw-R1) | Generic RL Framework | Multi | Both | Multi | General Agent | All | Yes (Framework-agnostic) |\n| [prime-rl](https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fprime-rl) | GRPO\u002FPPO | Multi | Outcome | Multi | Math\u002FCode\u002FSearch | Model\u002FExternal | Yes |\n| [NeMo-RL](https:\u002F\u002Fgithub.com\u002FNVIDIA-NeMo\u002FRL) | GRPO\u002FDAPO\u002FGDPO\u002FDPO | Single | Outcome | Multi | Math\u002FReasoning\u002FCode | Rule\u002FExternal | No |\n| [RLinf](https:\u002F\u002Fgithub.com\u002FRLinf\u002FRLinf) | PPO\u002FGRPO\u002FDAPO\u002FSAC\u002FREINFORCE++\u002FCrossQ\u002FRLPD | Both | Both | Multi | Robotics\u002FMath\u002FCode\u002FQA\u002FVQA | All (Rule\u002FModel\u002FExternal) | Yes |\n| [siiRL](https:\u002F\u002Fgithub.com\u002Fsii-research\u002FsiiRL) | PPO\u002FGRPO\u002FCPGD\u002FMARFT | Multi | Both | Multi | LLM\u002FVLM\u002FLLM-MAS PostTraining | Model\u002FRule | Planned |\n| [slime](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fslime) | GRPO\u002FGSPO\u002FREINFORCE++ | Single | Both | Both | Math\u002FCode | External Verifier | Yes |\n| [agent-lightning](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fagent-lightning) | PPO\u002FCustom\u002FAutomatic Prompt Optimization | Multi | Outcome | Multi | Calculator\u002FSQL | Model\u002FExternal\u002FRule | Yes |\n| [AReaL](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FAReaL) | PPO | Both | Outcome | Both | Math\u002FCode | External | Yes |\n| [ROLL](https:\u002F\u002Fgithub.com\u002Falibaba\u002FROLL) | PPO\u002FGRPO\u002FReinforce++\u002FTOPR\u002FRAFT++ | Multi | Both | Multi | Math\u002FQA\u002FCode\u002FAlignment | All | Yes |\n| [MARTI](https:\u002F\u002Fgithub.com\u002FTsinghuaC3I\u002FMARTI) | PPO\u002FGRPO\u002FREINFORCE++\u002FTTRL | Multi | Both | Multi | Math | All | Yes |\n| [RL2](https:\u002F\u002Fgithub.com\u002FChenmienTan\u002FRL2) | Dr. GRPO\u002FPPO\u002FDPO | Single | Both | Both | QA\u002FDialogue | Rule\u002FModel\u002FExternal | Yes |\n| [verifiers](https:\u002F\u002Fgithub.com\u002Fwillccbb\u002Fverifiers) | GRPO | Multi | Outcome | Both | Reasoning\u002FMath\u002FCode | All | Code |\n| [oat](https:\u002F\u002Fgithub.com\u002Fsail-sg\u002Foat) | PPO\u002FGRPO | Single | Outcome | Multi | Math\u002FAlignment | External | No |\n| [veRL](https:\u002F\u002Fgithub.com\u002Fvolcengine\u002Fverl) | PPO\u002FGRPO | Single | Outcome | Both | Math\u002FQA\u002FReasoning\u002FSearch | All | Yes |\n| [OpenRLHF](https:\u002F\u002Fgithub.com\u002FOpenRLHF\u002FOpenRLHF) | PPO\u002FREINFORCE++\u002FGRPO\u002FDPO\u002FIPO\u002FKTO\u002FRLOO | Multi | Both | Both | Dialogue\u002FChat\u002FCompletion | Rule\u002FModel\u002FExternal | Yes |\n| [trl](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftrl) | PPO\u002FGRPO\u002FDPO | Single | Both | Single | QA | Custom | No |\n\n\u003C\u002Fdetails>\n\n## 💪 General\u002FMultiTask\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [MetaClaw](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FMetaClaw) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Faiming-lab\u002FMetaClaw?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | UNC-Chapel Hill (AIMING Lab) | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2603.17187) | Custom |\n| [SkillRL](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FSkillRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Faiming-lab\u002FSkillRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | UNC-Chapel Hill (AIMING Lab) | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.08234) | Custom |\n| [LLM-in-Sandbox](https:\u002F\u002Fgithub.com\u002Fllm-in-sandbox\u002Fllm-in-sandbox) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fllm-in-sandbox\u002Fllm-in-sandbox?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | RUC\u002FMSRA\u002FTHU | [Paper](https:\u002F\u002Fhuggingface.co\u002Fpapers\u002F2601.16206) | rllm (w\u002F veRL) |\n| [youtu-agent](https:\u002F\u002Fgithub.com\u002FTencentCloudADP\u002Fyoutu-agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTencentCloudADP\u002Fyoutu-agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.12 | Tencent Youtu Lab | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2512.24615) | Custom |\n| [DEPO](https:\u002F\u002Fgithub.com\u002FOpenCausaLab\u002FDEPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenCausaLab\u002FDEPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | HKUST\u002FSJTU | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.15392) | LLaMA-Factory |\n| [SPEAR](https:\u002F\u002Fgithub.com\u002FTencentYoutuResearch\u002FSPEAR) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTencentYoutuResearch\u002FSPEAR?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | Tencent Youtu Lab | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.22601) | veRL\u002Fverl-agent |\n| [DeepAgent](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FDeepAgent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUC-NLPIR\u002FDeepAgent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | RUC\u002FXiaohongshu | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.21618) | Custom |\n| [AgentRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FAgentRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002FAgentRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | Tsinghua | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.04206) | veRL |\n| [AgentGym-RL](https:\u002F\u002Fgithub.com\u002FWooooDyy\u002FAgentGym-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FWooooDyy\u002FAgentGym-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | Fudan University | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.08755) | veRL |\n| [Agent_Foundation_Models](https:\u002F\u002Fgithub.com\u002FOPPO-PersonalAI\u002FAgent_Foundation_Models) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOPPO-PersonalAI\u002FAgent_Foundation_Models?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | OPPO Personal AI Lab | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.13167) | veRL |\n| [Trinity-RFT](https:\u002F\u002Fgithub.com\u002Fmodelscope\u002FTrinity-RFT) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmodelscope\u002FTrinity-RFT?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Alibaba | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.17826) | veRL |\n| [SPA-RL-Agent](https:\u002F\u002Fgithub.com\u002FWangHanLinHenry\u002FSPA-RL-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FWangHanLinHenry\u002FSPA-RL-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | PolyU | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.20732) | TRL |\n| [verl-agent](https:\u002F\u002Fgithub.com\u002FlangfengQ\u002Fverl-agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FlangfengQ\u002Fverl-agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | NTU\u002FSkywork | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.10978) | veRL |\n| [VAGEN](https:\u002F\u002Fgithub.com\u002FRAGEN-AI\u002FVAGEN) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRAGEN-AI\u002FVAGEN?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | RAGEN-AI | [Paper](https:\u002F\u002Fwww.notion.so\u002FVAGEN-Training-VLM-Agents-with-Multi-Turn-Reinforcement-Learning-1bfde13afb6e80b792f6d80c7c2fcad0) | veRL |\n| [ART](https:\u002F\u002Fgithub.com\u002FOpenPipe\u002FART) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenPipe\u002FART?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | OpenPipe | [Paper](https:\u002F\u002Fgithub.com\u002FOpenPipe\u002FART#-citation) | TRL |\n| [OpenManus-RL](https:\u002F\u002Fgithub.com\u002FOpenManus\u002FOpenManus-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenManus\u002FOpenManus-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | UIUC\u002FMetaGPT | -- | Custom |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [MetaClaw](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FMetaClaw) | GRPO (LoRA) | Single | Process | Multi | General Agentic | Model (PRM) | Yes (Skill-augmented) |\n| [SkillRL](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FSkillRL) | GRPO | Single | Outcome | Multi | ALFWorld\u002FWebShop\u002FSearch | Rule | Yes (Web search, actions) |\n| [LLM-in-Sandbox](https:\u002F\u002Fgithub.com\u002Fllm-in-sandbox\u002Fllm-in-sandbox) | GRPO++ | Single | Outcome | Multi | Math\u002FPhysics\u002FChemistry\u002FBiomedicine\u002FLong-context\u002FIF\u002FSWE | Rule | Yes (Code Sandbox w\u002F Terminal, File, Internet) |\n| [youtu-agent](https:\u002F\u002Fgithub.com\u002FTencentCloudADP\u002Fyoutu-agent) | Training-Free GRPO | Single | Outcome | Multi | Deep Research\u002FData Analysis\u002FTool-use | Model\u002FExternal | Yes (Web search, code, file) |\n| [DEPO](https:\u002F\u002Fgithub.com\u002FOpenCausaLab\u002FDEPO) | KTO + Efficiency Loss | Single | Both | Multi | Agent (BabyAI\u002FWebShop) | Rule | Yes |\n| [SPEAR](https:\u002F\u002Fgithub.com\u002FTencentYoutuResearch\u002FSPEAR) | GRPO\u002FGiGPO + SIL | Single | Both | Multi | Math\u002FAgent | Rule\u002FExternal | Yes (Search, Sandbox, Browser) |\n| [DeepAgent](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FDeepAgent) | ToolPO | Single | Outcome | Multi | ToolBench\u002FALFWorld\u002FWebShop\u002FGAIA\u002FHLE | Model | Yes (16,000+ RapidAPIs) |\n| [AgentRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FAgentRL) | GRPO\u002FREINFORCE++\u002FRLOO\u002FReMax\u002FGAE | Single | Outcome | Multi | Agent Tasks | External | Yes |\n| [AgentGym-RL](https:\u002F\u002Fgithub.com\u002FWooooDyy\u002FAgentGym-RL) | PPO\u002FGRPO\u002FRLOO\u002FREINFORCE++ | Single | Outcome | Multi | Web\u002FSearch\u002FGame\u002FEmbodied\u002FScience | Rule\u002FModel\u002FExternal | Yes (Web, Search, Env APIs) |\n| [Agent_Foundation_Models](https:\u002F\u002Fgithub.com\u002FOPPO-PersonalAI\u002FAgent_Foundation_Models) | DAPO\u002FPPO | Single | Outcome | Single | QA\u002FCode\u002FMath | Rule\u002FExternal | Yes |\n| [Trinity-RFT](https:\u002F\u002Fgithub.com\u002Fmodelscope\u002FTrinity-RFT) | PPO\u002FGRPO | Single | Outcome | Both | Math\u002FTextGame\u002FWeb | All | Yes |\n| [SPA-RL-Agent](https:\u002F\u002Fgithub.com\u002FWangHanLinHenry\u002FSPA-RL-Agent) | PPO | Single | Process | Multi | Navigation\u002FWeb\u002FTextGame | Model | No |\n| [verl-agent](https:\u002F\u002Fgithub.com\u002FlangfengQ\u002Fverl-agent) | PPO\u002FGRPO\u002FGiGPO\u002FDAPO\u002FRLOO\u002FREINFORCE++ | Multi | Both | Multi | Phone Use\u002FMath\u002FCode\u002FWeb\u002FTextGame | All | Yes |\n| [VAGEN](https:\u002F\u002Fgithub.com\u002FRAGEN-AI\u002FVAGEN) | PPO\u002FGRPO | Single | Both | Multi | TextGame\u002FNavigation | All | Yes |\n| [ART](https:\u002F\u002Fgithub.com\u002FOpenPipe\u002FART) | GRPO | Multi | Both | Multi | TextGame | All | Yes |\n| [OpenManus-RL](https:\u002F\u002Fgithub.com\u002FOpenManus\u002FOpenManus-RL) | PPO\u002FDPO\u002FGRPO | Multi | Outcome | Multi | TextGame | All | Yes |\n\n\u003C\u002Fdetails>\n\n## 🔍 Search & RAG Agent\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [ProRAG](https:\u002F\u002Fgithub.com\u002Flilinwz\u002FProRAG) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flilinwz\u002FProRAG?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | RUC | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.21912) | Custom |\n| [MemSearcher](https:\u002F\u002Fgithub.com\u002Ficip-cas\u002FMemSearcher) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ficip-cas\u002FMemSearcher?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | CAS | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.02805) | Custom |\n| [ReSeek](https:\u002F\u002Fgithub.com\u002FTencentBAC\u002FReSeek) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTencentBAC\u002FReSeek?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | Tencent PCG BAC\u002FTsinghua University | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.00568) | veRL |\n| [AutoGraph-R1](https:\u002F\u002Fgithub.com\u002FHKUST-KnowComp\u002FAutoGraph-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FHKUST-KnowComp\u002FAutoGraph-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | HKUST KnowComp | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.15339) | Custom |\n| [Tree-GRPO](https:\u002F\u002Fgithub.com\u002FAMAP-ML\u002FTree-GRPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAMAP-ML\u002FTree-GRPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | AMAP | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.21240) | veRL |\n| [ASearcher](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FASearcher) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FinclusionAI\u002FASearcher?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | Ant Research RL Lab \u003Cbr> Tsinghua University & UW | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.07976) | RealHF\u002FAReaL |\n| [Graph-R1](https:\u002F\u002Fgithub.com\u002FLHRLAB\u002FGraph-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FLHRLAB\u002FGraph-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | BUPT\u002FNTU\u002FNUS | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.21892) | veRL |\n| [Kimi-Researcher](https:\u002F\u002Fgithub.com\u002Fmoonshotai\u002FKimi-Researcher) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmoonshotai\u002FKimi-Researcher?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Moonshot AI | [blog](https:\u002F\u002Fmoonshotai.github.io\u002FKimi-Researcher\u002F) | Custom |\n| [R-Search](https:\u002F\u002Fgithub.com\u002FQingFei1\u002FR-Search) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FQingFei1\u002FR-Search?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Individual | -- | veRL |\n| [R1-Searcher-plus](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher-plus) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUCAIBox\u002FR1-Searcher-plus?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | RUC | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.17005) | Custom |\n| [StepSearch](https:\u002F\u002Fgithub.com\u002FZillwang\u002FStepSearch) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FZillwang\u002FStepSearch?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | SenseTime | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.15107) | veRL |\n| [AutoRefine](https:\u002F\u002Fgithub.com\u002Fsyr-cn\u002FAutoRefine) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsyr-cn\u002FAutoRefine?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | USTC | [Paper](https:\u002F\u002Fwww.arxiv.org\u002Fpdf\u002F2505.11277) | veRL |\n| [ZeroSearch](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FZeroSearch) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FZeroSearch?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Alibaba |[Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.04588) | veRL |\n| [ReasonRAG](https:\u002F\u002Fgithub.com\u002Fwlzhang2020\u002FReasonRAG) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fwlzhang2020\u002FReasonRAG?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | CityU HK \u002F Huawei | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.14069) | Custom |\n| [Agentic-RAG-R1](https:\u002F\u002Fgithub.com\u002Fjiangxinke\u002FAgentic-RAG-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fjiangxinke\u002FAgentic-RAG-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.12 | PKU | -- | Custom |\n| [WebThinker](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FWebThinker) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUC-NLPIR\u002FWebThinker?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | RUC | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.21776) | Custom |\n| [DeepResearcher](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FDeepResearcher) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGAIR-NLP\u002FDeepResearcher?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | SJTU | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.03160) | veRL |\n| [Search-R1](https:\u002F\u002Fgithub.com\u002FPeterGriffinJin\u002FSearch-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FPeterGriffinJin\u002FSearch-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | UIUC\u002FGoogle | [paper1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.09516), [paper2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.15117) | veRL |\n| [R1-Searcher](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUCAIBox\u002FR1-Searcher?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | RUC | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.05592) | OpenRLHF |\n| [C-3PO](https:\u002F\u002Fgithub.com\u002FChen-GX\u002FC-3PO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FChen-GX\u002FC-3PO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Alibaba | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2502.06205) | OpenRLHF |\n| [DeepRetrieval](https:\u002F\u002Fgithub.com\u002Fpat-jj\u002FDeepRetrieval) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpat-jj\u002FDeepRetrieval?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | UIUC | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.00223) | veRL |\n| [SSRL](https:\u002F\u002Fgithub.com\u002FTsinghuaC3I\u002FSSRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTsinghuaC3I\u002FSSRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | Tsinghua | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.10874) | Custom |\n| [Research-Venus](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fantgroup\u002FResearch-Venus?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | Ant Group | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.12800) | Custom |\n| [DeepResearch](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | Alibaba\u002FTongyi Lab | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.24701) | Custom |\n| [DeepDive](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FDeepDive) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002FDeepDive?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | Tsinghua\u002FTHUDM | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.10446) | Custom |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [ProRAG](https:\u002F\u002Fgithub.com\u002Flilinwz\u002FProRAG) | GRPO + DGA (dual-granularity advantage) | Single | Both | Multi | Multi-hop RAG | Model (PRM via MCTS) | Yes (Retrieval) |\n| [MemSearcher](https:\u002F\u002Fgithub.com\u002Ficip-cas\u002FMemSearcher) | Multi-context GRPO | Single | Outcome | Multi | Search\u002FQA + Memory | Rule\u002FModel | Yes (Web search + Memory) |\n| [ReSeek](https:\u002F\u002Fgithub.com\u002FTencentBAC\u002FReSeek) | GRPO\u002FPPO | Single | Both | Multi | QA\u002FSearch | Rule | Search\u002FJUDGE |\n| [AutoGraph-R1](https:\u002F\u002Fgithub.com\u002FHKUST-KnowComp\u002FAutoGraph-R1) | GRPO (via VeRL) | Single | Outcome | Multi | KG Construction for QA | Rule | Yes (Graph retrieval) |\n| [Tree-GRPO](https:\u002F\u002Fgithub.com\u002FAMAP-ML\u002FTree-GRPO) | GRPO\u002FTree-GRPO | Single | Outcome | Multi | Search | Rule | Search |\n| [ASearcher](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FASearcher) | PPO\u002FGRPO + Decoupled PPO | Single | Outcome | Multi | Math\u002FCode\u002FSearchQA | External\u002FRule | Yes |\n| [Graph-R1](https:\u002F\u002Fgithub.com\u002FLHRLAB\u002FGraph-R1) | GRPO\u002FREINFORCE++\u002FPPO | Single | Outcome | Multi | KGQA | Rule (EM\u002FF1) | Yes (Graph retrieval) |\n| [Kimi-Researcher](https:\u002F\u002Fgithub.com\u002Fmoonshotai\u002FKimi-Researcher) | REINFORCE | Single | Outcome | Multi | Research | Outcome | Search, Browse, Coding |\n| [R-Search](https:\u002F\u002Fgithub.com\u002FQingFei1\u002FR-Search) | PPO\u002FGRPO | Single | Both | Multi | QA\u002FSearch | All | Yes |\n| [R1-Searcher-plus](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher-plus) | Custom | Single | Outcome | Multi | Search | Model | Search |\n| [StepSearch](https:\u002F\u002Fgithub.com\u002FZillwang\u002FStepSearch) | PPO | Single | Process | Multi | QA | Model | Search |\n| [AutoRefine](https:\u002F\u002Fgithub.com\u002Fsyr-cn\u002FAutoRefine) | PPO\u002FGRPO | Multi | Both | Multi | RAG QA | Rule | Search |\n| [ZeroSearch](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FZeroSearch) | PPO\u002FGRPO\u002FREINFORCE | Single | Outcome | Multi | QA\u002FSearch | Rule | Yes |\n| [ReasonRAG](https:\u002F\u002Fgithub.com\u002Fwlzhang2020\u002FReasonRAG) | DPO + MCTS-based PRM | Single | Process | Multi | Multi-hop QA | Model (PRM) | Yes (Wikipedia search) |\n| [Agentic-RAG-R1](https:\u002F\u002Fgithub.com\u002Fjiangxinke\u002FAgentic-RAG-R1) | GRPO | Single | Outcome | Multi | Knowledge-intensive QA | Rule\u002FModel | Yes (Wiki\u002FDoc search) |\n| [WebThinker](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FWebThinker) | DPO | Single | Outcome | Multi | Reasoning\u002FQA\u002FResearch | Model\u002FExternal | Web Browsing |\n| [DeepResearcher](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FDeepResearcher) | PPO\u002FGRPO | Multi | Outcome | Multi | Research | All | Yes |\n| [Search-R1](https:\u002F\u002Fgithub.com\u002FPeterGriffinJin\u002FSearch-R1) | PPO\u002FGRPO | Single | Outcome | Multi | Search | All | Search |\n| [R1-Searcher](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher) | PPO\u002FDPO | Single | Both | Multi | Search | All | Yes |\n| [C-3PO](https:\u002F\u002Fgithub.com\u002FChen-GX\u002FC-3PO) | PPO | Multi | Outcome | Multi | Search | Model | Yes |\n| [DeepRetrieval](https:\u002F\u002Fgithub.com\u002Fpat-jj\u002FDeepRetrieval) | GRPO | Single | Outcome | Multi | Query Generation\u002FIR | Rule | Yes (Search) |\n| [SSRL](https:\u002F\u002Fgithub.com\u002FTsinghuaC3I\u002FSSRL) | GRPO | Single | Outcome | Multi | Self-Search | Rule | Yes (Self-search) |\n| [Research-Venus](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus) | GRPO | Single | Both | Multi | Deep Research | Model (atomic thought) | Yes (Search) |\n| [DeepResearch](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | RL-based | Single | Outcome | Multi | Deep Research | Model | Yes (Search, Browse) |\n| [DeepDive](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FDeepDive) | GRPO | Single | Outcome | Multi | KG-augmented Search | Rule | Yes (KG + Search) |\n\n\u003C\u002Fdetails>\n\n## 🌐 Web & GUI Agent\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [MobileAgent](https:\u002F\u002Fgithub.com\u002FX-PLUG\u002FMobileAgent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FX-PLUG\u002FMobileAgent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | X-PLUG (TongyiQwen) | [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.11543) | veRL |\n| [InfiGUI-G1](https:\u002F\u002Fgithub.com\u002FInfiXAI\u002FInfiGUI-G1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FInfiXAI\u002FInfiGUI-G1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | InfiX AI | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.05731) | veRL |\n| [UI-AGILE](https:\u002F\u002Fgithub.com\u002FKDEGroup\u002FUI-AGILE) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FKDEGroup\u002FUI-AGILE?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | Xiamen University | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.22025) | Custom |\n| [gui-rcpo](https:\u002F\u002Fgithub.com\u002Fzju-real\u002Fgui-rcpo) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fzju-real\u002Fgui-rcpo?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | Zhejiang University | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.05615) | Custom |\n| [Grounding-R1](https:\u002F\u002Fgithub.com\u002FYan98\u002FGrounding-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FYan98\u002FGrounding-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Salesforce | [blog](https:\u002F\u002Fhuggingface.co\u002Fblog\u002FHelloKKMe\u002Fgrounding-r1) | trl |\n| [AgentCPM-GUI](https:\u002F\u002Fgithub.com\u002FOpenBMB\u002FAgentCPM-GUI) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenBMB\u002FAgentCPM-GUI?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | OpenBMB\u002FTsinghua\u002FRUC | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.01391) | Huggingface |\n| [TTI](https:\u002F\u002Fgithub.com\u002Ftest-time-interaction\u002FTTI) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftest-time-interaction\u002FTTI?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | CMU | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.07976) | Custom |\n| [SE-GUI](https:\u002F\u002Fgithub.com\u002FYXB-NKU\u002FSE-GUI) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FYXB-NKU\u002FSE-GUI?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Nankai University\u002Fvivo | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.12370) | trl |\n| [ARPO](https:\u002F\u002Fgithub.com\u002Fdvlab-research\u002FARPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdvlab-research\u002FARPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | CUHK\u002FHKUST | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.16282) | veRL |\n| [GUI-G1](https:\u002F\u002Fgithub.com\u002FYuqi-Zhou\u002FGUI-G1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FYuqi-Zhou\u002FGUI-G1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | RUC | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.15810) | TRL |\n| [WebAgent-R1](https:\u002F\u002Fgithub.com\u002Fweizhepei\u002FWebAgent-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fweizhepei\u002FWebAgent-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Amazon\u002FUVA | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.16421) | Custom |\n| [GUI-R1](https:\u002F\u002Fgithub.com\u002Fritzz-ai\u002FGUI-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fritzz-ai\u002FGUI-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | CAS\u002FNUS | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.10458) | veRL |\n| [UI-R1](https:\u002F\u002Fgithub.com\u002Flll6gg\u002FUI-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flll6gg\u002FUI-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | vivo\u002FCUHK | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.21620) | TRL |\n| [CollabUIAgents](https:\u002F\u002Fgithub.com\u002FTHUNLP-MT\u002FCollabUIAgents) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUNLP-MT\u002FCollabUIAgents?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Tsinghua\u002FAlibaba\u002FHKUST | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.14496) | Custom |\n| [WebAgent](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FWebAgent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FWebAgent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | Alibaba | [paper1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2501.07572), [paper2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.22648) | LLaMA-Factory |\n| [UI-TARS](https:\u002F\u002Fgithub.com\u002Fbytedance\u002FUI-TARS) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbytedance\u002FUI-TARS?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | ByteDance Seed | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.02544) | Custom |\n| [DigiQ](https:\u002F\u002Fgithub.com\u002FDigiRL-agent\u002Fdigiq) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FDigiRL-agent\u002Fdigiq?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | UC Berkeley\u002FCMU\u002FAmazon | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.15760) | Custom |\n| [ZeroGUI](https:\u002F\u002Fgithub.com\u002FOpenGVLab\u002FZeroGUI) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenGVLab\u002FZeroGUI?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Shanghai AI Lab | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.23762) | Custom |\n| [InfiGUI-R1](https:\u002F\u002Fgithub.com\u002FReallm-Labs\u002FInfiGUI-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FReallm-Labs\u002FInfiGUI-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | Zhejiang University | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.14239) | Custom |\n| [GUI-Agent-RL](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FGUI-Agent-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FGUI-Agent-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Microsoft | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.18906) | Custom |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [MobileAgent](https:\u002F\u002Fgithub.com\u002FX-PLUG\u002FMobileAgent) | semi-online RL | Single | Both | Multi | MobileGUI\u002FAutomation | Rule | Yes |\n| [InfiGUI-G1](https:\u002F\u002Fgithub.com\u002FInfiXAI\u002FInfiGUI-G1) | AEPO | Single | Outcome | Single | GUI\u002FGrounding | Rule | No |\n| [UI-AGILE](https:\u002F\u002Fgithub.com\u002FKDEGroup\u002FUI-AGILE) | GRPO | Single | Outcome | Single | GUI Grounding | Rule (continuous) | No |\n| [gui-rcpo](https:\u002F\u002Fgithub.com\u002Fzju-real\u002Fgui-rcpo) | RCPO | Single | Outcome | Single | GUI Grounding | Rule (self-supervised) | No |\n| [Grounding-R1](https:\u002F\u002Fgithub.com\u002FYan98\u002FGrounding-R1) | GRPO | Single | Outcome | Multi | GUI Grounding | Model | Yes |\n| [AgentCPM-GUI](https:\u002F\u002Fgithub.com\u002FOpenBMB\u002FAgentCPM-GUI) | GRPO | Single | Outcome | Multi | Mobile GUI | Model | Yes |\n| [TTI](https:\u002F\u002Fgithub.com\u002Ftest-time-interaction\u002FTTI) | REINFORCE\u002FBC | Single | Outcome | Multi | Web | External | Web Browsing |\n| [SE-GUI](https:\u002F\u002Fgithub.com\u002FYXB-NKU\u002FSE-GUI) | GRPO | Single | Both | Single | GUI Grounding | Rule | Yes |\n| [ARPO](https:\u002F\u002Fgithub.com\u002Fdvlab-research\u002FARPO) | GRPO | Single | Outcome | Multi | GUI | External | Computer Use |\n| [GUI-G1](https:\u002F\u002Fgithub.com\u002FYuqi-Zhou\u002FGUI-G1) | GRPO | Single | Outcome | Single | GUI | Rule\u002FExternal | No |\n| [WebAgent-R1](https:\u002F\u002Fgithub.com\u002Fweizhepei\u002FWebAgent-R1) | M-GRPO | Single | Outcome | Multi | Web Navigation (WebArena-Lite) | Rule (task success) | Yes (Web browsing) |\n| [GUI-R1](https:\u002F\u002Fgithub.com\u002Fritzz-ai\u002FGUI-R1) | GRPO | Single | Outcome | Multi | GUI | Rule | No |\n| [UI-R1](https:\u002F\u002Fgithub.com\u002Flll6gg\u002FUI-R1) | GRPO | Single | Process | Both | GUI | Rule | Computer\u002FPhone Use |\n| [CollabUIAgents](https:\u002F\u002Fgithub.com\u002FTHUNLP-MT\u002FCollabUIAgents) | DPO (credit re-assignment) | Multi | Process | Multi | GUI (Mobile + Web) | Model (LLM) | Yes (GUI interaction) |\n| [WebAgent](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FWebAgent) | DAPO | Multi | Process | Multi | Web | Model | Yes |\n| [UI-TARS](https:\u002F\u002Fgithub.com\u002Fbytedance\u002FUI-TARS) | Multi-turn RL | Single | Both | Multi | GUI (Cross-platform) | Model | Yes (GUI actions) |\n| [DigiQ](https:\u002F\u002Fgithub.com\u002FDigiRL-agent\u002Fdigiq) | Value-based offline RL | Single | Outcome | Multi | Android Device Control | Model (Q-function) | Yes |\n| [ZeroGUI](https:\u002F\u002Fgithub.com\u002FOpenGVLab\u002FZeroGUI) | Online RL | Single | Outcome | Multi | GUI Agent | Rule | Yes (GUI actions) |\n| [InfiGUI-R1](https:\u002F\u002Fgithub.com\u002FReallm-Labs\u002FInfiGUI-R1) | RL + sub-goal guidance | Single | Both | Multi | GUI Reasoning | Rule | Yes |\n| [GUI-Agent-RL](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FGUI-Agent-RL) | Value-based RL (VEM) | Single | Outcome | Multi | GUI (Web Shopping) | Model | Yes |\n\n\u003C\u002Fdetails>\n\n## 🔨 Tool-Use Agent\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [MATPO](https:\u002F\u002Fgithub.com\u002Fmzf666\u002FMATPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmzf666\u002FMATPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | MiroMind AI | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.04678) | Custom |\n| [MiroRL](https:\u002F\u002Fgithub.com\u002FMiroMindAI\u002FMiroRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMiroMindAI\u002FMiroRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | MiroMindAI | [HF Repo](https:\u002F\u002Fhuggingface.co\u002Fmiromind-ai) | veRL |\n| [verl-tool](https:\u002F\u002Fgithub.com\u002FTIGER-AI-Lab\u002Fverl-tool) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTIGER-AI-Lab\u002Fverl-tool?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | TIGER-Lab | [X](https:\u002F\u002Fx.com\u002FDongfuJiang\u002Fstatus\u002F1929198238017720379) | veRL |\n| [Multi-Turn-RL-Agent](https:\u002F\u002Fgithub.com\u002FSiliangZeng\u002FMulti-Turn-RL-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSiliangZeng\u002FMulti-Turn-RL-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | University of Minnesota | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.11821) | Custom |\n| [Tool-N1](https:\u002F\u002Fgithub.com\u002FNVlabs\u002FTool-N1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNVlabs\u002FTool-N1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | NVIDIA | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.00024) | veRL |\n| [Tool-Star](https:\u002F\u002Fgithub.com\u002Fdongguanting\u002FTool-Star) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdongguanting\u002FTool-Star?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | RUC | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.16410) | LLaMA-Factory |\n| [RL-Factory](https:\u002F\u002Fgithub.com\u002FSimple-Efficient\u002FRL-Factory) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSimple-Efficient\u002FRL-Factory?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Simple-Efficient | [model](https:\u002F\u002Fhuggingface.co\u002FSimple-Efficient\u002FRLFactory-Qwen3-8B-GRPO) | veRL |\n| [ReTool](https:\u002F\u002Fgithub.com\u002FReTool-RL\u002FReTool) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FReTool-RL\u002FReTool?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | ByteDance | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.11536) | veRL |\n| [AWorld](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FAWorld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FinclusionAI\u002FAWorld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | Ant Group (inclusionAI) | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.20404) | veRL |\n| [Agent-R1](https:\u002F\u002Fgithub.com\u002F0russwest0\u002FAgent-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002F0russwest0\u002FAgent-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | USTC | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.14460) | veRL |\n| [ReCall](https:\u002F\u002Fgithub.com\u002FAgent-RL\u002FReCall) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAgent-RL\u002FReCall?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | BaiChuan | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.19470) | veRL |\n| [ToolRL](https:\u002F\u002Fgithub.com\u002Fqiancheng0\u002FToolRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fqiancheng0\u002FToolRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | UIUC | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.13958) | veRL |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [MATPO](https:\u002F\u002Fgithub.com\u002Fmzf666\u002FMATPO) | GRPO (multi-agent) | Multi | Outcome | Multi | Tool-use\u002FSearch | Rule | Yes (MCP: Serper, Web scraping) |\n| [MiroRL](https:\u002F\u002Fgithub.com\u002FMiroMindAI\u002FMiroRL) | GRPO | Single | Both | Multi | Reasoning\u002FPlanning\u002FToolUse | Rule-based | MCP |\n| [verl-tool](https:\u002F\u002Fgithub.com\u002FTIGER-AI-Lab\u002Fverl-tool) | PPO\u002FGRPO | Single | Both | Both | Math\u002FCode | Rule\u002FExternal | Yes |\n| [Multi-Turn-RL-Agent](https:\u002F\u002Fgithub.com\u002FSiliangZeng\u002FMulti-Turn-RL-Agent) | GRPO | Single | Both | Multi | Tool-use\u002FMath | Rule\u002FExternal | Yes |\n| [Tool-N1](https:\u002F\u002Fgithub.com\u002FNVlabs\u002FTool-N1) | PPO | Single | Outcome | Multi | Math\u002FDialogue | All | Yes |\n| [Tool-Star](https:\u002F\u002Fgithub.com\u002Fdongguanting\u002FTool-Star) | PPO\u002FDPO\u002FORPO\u002FSimPO\u002FKTO | Single | Outcome | Multi | Multi-modal\u002FTool Use\u002FDialogue | Model\u002FExternal | Yes |\n| [RL-Factory](https:\u002F\u002Fgithub.com\u002FSimple-Efficient\u002FRL-Factory) | GRPO | Multi | Both | Multi | Tool-use\u002FNL2SQL | All | MCP |\n| [ReTool](https:\u002F\u002Fgithub.com\u002FReTool-RL\u002FReTool) | PPO | Single | Outcome | Multi | Math | External | Code |\n| [AWorld](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FAWorld) | GRPO | Both | Outcome | Multi | Search\u002FWeb\u002FCode | External\u002FRule | Yes |\n| [Agent-R1](https:\u002F\u002Fgithub.com\u002F0russwest0\u002FAgent-R1) | PPO\u002FGRPO | Single | Both | Multi | Tool-use\u002FQA | Model | Yes |\n| [ReCall](https:\u002F\u002Fgithub.com\u002FAgent-RL\u002FReCall) | PPO\u002FGRPO\u002FRLOO\u002FREINFORCE++\u002FReMax | Single | Outcome | Multi | Tool-use\u002FMath\u002FQA | All | Yes |\n| [ToolRL](https:\u002F\u002Fgithub.com\u002Fqiancheng0\u002FToolRL) | GRPO\u002FPPO | Single | Outcome | Multi | Tool Learning | Rule\u002FExternal | Yes |\n\n\u003C\u002Fdetails>\n\n## 💻 Code & SWE Agent\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [CUDA-Agent](https:\u002F\u002Fgithub.com\u002FBytedTsinghua-SIA\u002FCUDA-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FBytedTsinghua-SIA\u002FCUDA-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | ByteDance\u002FTsinghua | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.24286) | Custom |\n| [LLM-in-Sandbox](https:\u002F\u002Fgithub.com\u002Fllm-in-sandbox\u002Fllm-in-sandbox) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fllm-in-sandbox\u002Fllm-in-sandbox?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | RUC\u002FMSRA\u002FTHU | [Paper](https:\u002F\u002Fhuggingface.co\u002Fpapers\u002F2601.16206) | rllm (w\u002F veRL) |\n| [PPP-Agent](https:\u002F\u002Fgithub.com\u002Fsunnweiwei\u002FPPP-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsunnweiwei\u002FPPP-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | CMU\u002FOpenHands | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.02208) | veRL |\n| [RepoDeepSearch](https:\u002F\u002Fgithub.com\u002FMizersy\u002FRepoDeepSearch) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMizersy\u002FRepoDeepSearch?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | PKU, Bytedance, BIT | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.03012) | veRL |\n| [CUDA-L1](https:\u002F\u002Fgithub.com\u002Fdeepreinforce-ai\u002FCUDA-L1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdeepreinforce-ai\u002FCUDA-L1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | DeepReinforce AI | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.14111) | Custom |\n| [MedAgentGym](https:\u002F\u002Fgithub.com\u002Fwshi83\u002FMedAgentGym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fwshi83\u002FMedAgentGym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Emory\u002FGeorgia Tech | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.04405) | Hugginface |\n| [CURE](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FCURE) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGen-Verse\u002FCURE?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | University of Chicago \u003Cbr> Princeton\u002FByteDance | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.03136) | Huggingface |\n| [Time-R1](https:\u002F\u002Fgithub.com\u002Fulab-uiuc\u002FTime-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fulab-uiuc\u002FTime-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | UIUC | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.13508) | veRL |\n| [ML-Agent](https:\u002F\u002Fgithub.com\u002FMASWorks\u002FML-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMASWorks\u002FML-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | MASWorks | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.23723) | Custom |\n| [SkyRL](https:\u002F\u002Fgithub.com\u002FNovaSky-AI\u002FSkyRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNovaSky-AI\u002FSkyRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | NovaSky | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.16108) | veRL |\n| [digitalhuman](https:\u002F\u002Fgithub.com\u002FTencent\u002Fdigitalhuman) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTencent\u002Fdigitalhuman?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | Tencent | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.03112) | veRL |\n| [sweet_rl](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fsweet_rl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ffacebookresearch\u002Fsweet_rl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | Meta\u002FUCB | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.15478) | OpenRLHF |\n| [swe-rl](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fswe-rl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ffacebookresearch\u002Fswe-rl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Meta\u002FUIUC\u002FCMU | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.18449) | Custom |\n| [rllm](https:\u002F\u002Fgithub.com\u002Fagentica-project\u002Frllm) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fagentica-project\u002Frllm?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | Berkeley Sky Computing Lab \u003Cbr> BAIR \u002F Together AI | [Notion Blog](https:\u002F\u002Fpretty-radio-b75.notion.site\u002FrLLM-A-Framework-for-Post-Training-Language-Agents-21b81902c146819db63cd98a54ba5f31) | veRL |\n| [open-r1](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Fopen-r1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhuggingface\u002Fopen-r1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | HuggingFace | -- | TRL |\n| [R1-Code-Interpreter](https:\u002F\u002Fgithub.com\u002Fyongchao98\u002FR1-Code-Interpreter) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fyongchao98\u002FR1-Code-Interpreter?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | MIT | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.21668) | Custom |\n| [CTRL](https:\u002F\u002Fgithub.com\u002FHKUNLP\u002Fcritic-rl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FHKUNLP\u002Fcritic-rl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | HKU\u002FByteDance | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.03492) | Custom |\n| [DeepAnalyze](https:\u002F\u002Fgithub.com\u002Fruc-datalab\u002FDeepAnalyze) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fruc-datalab\u002FDeepAnalyze?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | RUC\u002FTsinghua | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.16872) | Custom |\n| [AceCoder](https:\u002F\u002Fgithub.com\u002FTIGER-AI-Lab\u002FAceCoder) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTIGER-AI-Lab\u002FAceCoder?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Waterloo (TIGER-Lab) | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.01718) | Custom |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [CUDA-Agent](https:\u002F\u002Fgithub.com\u002FBytedTsinghua-SIA\u002FCUDA-Agent) | Agentic RL (staged) | Single | Outcome | Multi | CUDA Kernel Generation | Rule (correctness + performance) | Yes (compile\u002Fverify\u002Fprofile) |\n| [LLM-in-Sandbox](https:\u002F\u002Fgithub.com\u002Fllm-in-sandbox\u002Fllm-in-sandbox) | GRPO++ | Single | Outcome | Multi | Code\u002FSWE + General (Math\u002FSci\u002FBio) | Rule | Yes (Code Sandbox w\u002F Terminal, File, Internet) |\n| [PPP-Agent](https:\u002F\u002Fgithub.com\u002Fsunnweiwei\u002FPPP-Agent) | PPP-RL | Single | Both | Multi | SWE\u002FResearch | Rule+Model | Search, Ask, Browse |\n| [RepoDeepSearch](https:\u002F\u002Fgithub.com\u002FMizersy\u002FRepoDeepSearch) | GRPO | Single | Both | Multi | Search\u002FRepair | Rule\u002FExternal | Yes |\n| [CUDA-L1](https:\u002F\u002Fgithub.com\u002Fdeepreinforce-ai\u002FCUDA-L1) | Contrastive RL | Single | Outcome | Single | CUDA Optimization | Rule (performance) | No |\n| [MedAgentGym](https:\u002F\u002Fgithub.com\u002Fwshi83\u002FMedAgentGym) | SFT\u002FDPO\u002FPPO\u002FGRPO | Single | Outcome | Multi | Medical\u002FCode | External | Yes |\n| [CURE](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FCURE) | PPO | Single | Outcome | Single | Code | External | No |\n| [Time-R1](https:\u002F\u002Fgithub.com\u002Fulab-uiuc\u002FTime-R1) | PPO\u002FGRPO\u002FDPO | Multi | Outcome | Multi | Temporal | All | Code |\n| [ML-Agent](https:\u002F\u002Fgithub.com\u002FMASWorks\u002FML-Agent) | Custom | Single | Process | Multi | Code | All | Yes |\n| [SkyRL](https:\u002F\u002Fgithub.com\u002FNovaSky-AI\u002FSkyRL) | PPO\u002FGRPO | Single | Outcome | Multi | Math\u002FCode | All | Code |\n| [digitalhuman](https:\u002F\u002Fgithub.com\u002FTencent\u002Fdigitalhuman) | PPO\u002FGRPO\u002FReMax\u002FRLOO | Multi | Outcome | Multi | Empathy\u002FMath\u002FCode\u002FMultimodalQA | Rule\u002FModel\u002FExternal | Yes |\n| [sweet_rl](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fsweet_rl) | DPO | Multi | Process | Multi | Design\u002FCode | Model | Web Browsing |\n| [swe-rl](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fswe-rl) | RL-based | Single | Outcome | Single | SWE (SWE-bench) | Rule (similarity) | No |\n| [rllm](https:\u002F\u002Fgithub.com\u002Fagentica-project\u002Frllm) | PPO\u002FGRPO | Single | Outcome | Multi | Code Edit | External | Yes |\n| [open-r1](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Fopen-r1) | GRPO | Single | Outcome | Single | Math\u002FCode | All | Yes |\n| [R1-Code-Interpreter](https:\u002F\u002Fgithub.com\u002Fyongchao98\u002FR1-Code-Interpreter) | GRPO | Single | Outcome | Multi | Code Interpretation | Rule\u002FExternal | Yes (Code exec) |\n| [CTRL](https:\u002F\u002Fgithub.com\u002FHKUNLP\u002Fcritic-rl) | RL (critique-revision) | Single | Process | Multi | Code Refinement | Model | Yes (Code exec) |\n| [DeepAnalyze](https:\u002F\u002Fgithub.com\u002Fruc-datalab\u002FDeepAnalyze) | Curriculum RL | Single | Outcome | Multi | Data Science | Rule\u002FExternal | Yes (Code exec) |\n| [AceCoder](https:\u002F\u002Fgithub.com\u002FTIGER-AI-Lab\u002FAceCoder) | GRPO | Single | Outcome | Single | Code Generation | External (test cases) | Yes |\n\n\u003C\u002Fdetails>\n\n## 🤔 Reasoning Agent\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [Agent0](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FAgent0) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Faiming-lab\u002FAgent0?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | UNC‑Chapel Hill \u002F Salesforce Research \u002F Stanford University | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.16043) | veRL |\n| [KG-R1](https:\u002F\u002Fgithub.com\u002FJinyeop3110\u002FKG-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FJinyeop3110\u002FKG-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | UIUC\u002FGoogle | [Paper1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.09516), [Paper2](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.15117) | veRL |\n| [AgentFlow](https:\u002F\u002Fgithub.com\u002Flupantech\u002FAgentFlow) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flupantech\u002FAgentFlow?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.09 | Stanford University | [arXiv](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.05592) | veRL |\n| [ARPO](https:\u002F\u002Fgithub.com\u002Fdongguanting\u002FARPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdongguanting\u002FARPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | RUC, Kuaishou | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.19849) | veRL |\n| [terminal-bench-rl](https:\u002F\u002Fgithub.com\u002FDanau5tin\u002Fterminal-bench-rl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FDanau5tin\u002Fterminal-bench-rl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | Individual (Danau5tin) | N\u002FA | rLLM |\n| [MOTIF](https:\u002F\u002Fgithub.com\u002Fpurbeshmitra\u002FMOTIF) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpurbeshmitra\u002FMOTIF?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | University of Maryland | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.02851) | trl |\n| [cmriat\u002Fl0](https:\u002F\u002Fgithub.com\u002Fcmriat\u002Fl0) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcmriat\u002Fl0?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | CMRIAT | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.23667) | veRL |\n| [agent-distillation](https:\u002F\u002Fgithub.com\u002FNardien\u002Fagent-distillation) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNardien\u002Fagent-distillation?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | KAIST | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.17612) | Custom |\n| [EasyR1](https:\u002F\u002Fgithub.com\u002Fhiyouga\u002FEasyR1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhiyouga\u002FEasyR1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | Individual | [repo1](https:\u002F\u002Fgithub.com\u002Fhiyouga\u002FEasyR1)\u002F[paper2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2409.19256) | veRL |\n| [AutoCoA](https:\u002F\u002Fgithub.com\u002FADaM-BJTU\u002FAutoCoA) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FADaM-BJTU\u002FAutoCoA?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | BJTU | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.06580) | veRL |\n| [ToRL](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FToRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGAIR-NLP\u002FToRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | SJTU | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.23383) | veRL |\n| [ReMA](https:\u002F\u002Fgithub.com\u002Fziyuwan\u002FReMA-public) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fziyuwan\u002FReMA-public?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | SJTU, UCL | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.09501) | veRL |\n| [Agentic-Reasoning](https:\u002F\u002Fgithub.com\u002Ftheworldofagents\u002FAgentic-Reasoning) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftheworldofagents\u002FAgentic-Reasoning?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Oxford | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2502.04644) | Custom |\n| [SimpleTIR](https:\u002F\u002Fgithub.com\u002Fltzheng\u002FSimpleTIR) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fltzheng\u002FSimpleTIR?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | NTU, Bytedance | [Notion Blog](https:\u002F\u002Fsimpletir.notion.site\u002Freport) | veRL |\n| [openrlhf_async_pipline](https:\u002F\u002Fgithub.com\u002Fyyht\u002Fopenrlhf_async_pipline) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fyyht\u002Fopenrlhf_async_pipline?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.5 | OpenRLHF | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2405.11143) | OpenRLHF |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [Agent0](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FAgent0) | ADPO | Multi | Process | Multi | Math\u002FVisual | Model\u002FVerifier | Yes |\n| [KG-R1](https:\u002F\u002Fgithub.com\u002FJinyeop3110\u002FKG-R1) | GRPO\u002FPPO | Single | Both | Multi | KGQA | Rule\u002FModel | KG Retrieval |\n| [AgentFlow](https:\u002F\u002Fgithub.com\u002Flupantech\u002FAgentFlow) | Flow-GRPO | Single | Outcome | Multi | Search\u002FMath\u002FQA | Model\u002FExternal | Yes |\n| [ARPO](https:\u002F\u002Fgithub.com\u002Fdongguanting\u002FARPO) | GRPO | Single | Outcome | Multi | Math\u002FCoding | Model\u002FRule | Yes |\n| [terminal-bench-rl](https:\u002F\u002Fgithub.com\u002FDanau5tin\u002Fterminal-bench-rl) | GRPO | Single | Outcome | Multi | Coding\u002FTerminal | Model+External Verifier | Yes |\n| [MOTIF](https:\u002F\u002Fgithub.com\u002Fpurbeshmitra\u002FMOTIF) | GRPO | Single | Outcome | Multi | QA | Rule | No |\n| [cmriat\u002Fl0](https:\u002F\u002Fgithub.com\u002Fcmriat\u002Fl0) | PPO | Multi | Process | Multi | QA | All | Yes |\n| [agent-distillation](https:\u002F\u002Fgithub.com\u002FNardien\u002Fagent-distillation) | PPO | Single | Process | Multi | QA\u002FMath | External | Yes |\n| [EasyR1](https:\u002F\u002Fgithub.com\u002Fhiyouga\u002FEasyR1) | GRPO | Single | Process | Multi | Vision-Language | Model | Yes |\n| [AutoCoA](https:\u002F\u002Fgithub.com\u002FADaM-BJTU\u002FAutoCoA) | GRPO | Multi | Outcome | Multi | Reasoning\u002FMath\u002FQA | All | Yes |\n| [ToRL](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FToRL) | GRPO | Single | Outcome | Single | Math | Rule\u002FExternal | Yes |\n| [ReMA](https:\u002F\u002Fgithub.com\u002Fziyuwan\u002FReMA-public) | PPO | Multi | Outcome | Multi | Math | Rule | No |\n| [Agentic-Reasoning](https:\u002F\u002Fgithub.com\u002Ftheworldofagents\u002FAgentic-Reasoning) | Custom | Single | Process | Multi | QA\u002FMath | External | Web Browsing |\n| [SimpleTIR](https:\u002F\u002Fgithub.com\u002Fltzheng\u002FSimpleTIR) | PPO\u002FGRPO (with extensions) | Single | Outcome | Multi | Math, Coding | All | Yes |\n| [openrlhf_async_pipline](https:\u002F\u002Fgithub.com\u002Fyyht\u002Fopenrlhf_async_pipline) | PPO\u002FREINFORCE++\u002FDPO\u002FRLOO | Single | Outcome | Multi | Dialogue\u002FReasoning\u002FQA | All | No |\n\n\u003C\u002Fdetails>\n\n## 👥 Multi-Agent RL\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [PettingLLMs](https:\u002F\u002Fgithub.com\u002Fpettingllms-ai\u002FPettingLLMs) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpettingllms-ai\u002FPettingLLMs?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | Intel \u002F UCSD | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.11062) | Custom |\n| [MASPRM](https:\u002F\u002Fgithub.com\u002Fmilad1378yz\u002FMASPRM) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmilad1378yz\u002FMASPRM?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | UBC \u002F Huawei | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.24803) | Custom |\n| [ARIA](https:\u002F\u002Fgithub.com\u002Frhyang2021\u002FARIA) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Frhyang2021\u002FARIA?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Fudan University | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.00539) | Custom |\n| [AMPO](https:\u002F\u002Fgithub.com\u002FMozerWang\u002FAMPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMozerWang\u002FAMPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Tongyi Lab, Alibaba | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.02156) | veRL |\n| [MAPoRL](https:\u002F\u002Fgithub.com\u002Fchanwoo-park-official\u002FMAPoRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fchanwoo-park-official\u002FMAPoRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | Academic | -- | Custom |\n| [FlowReasoner](https:\u002F\u002Fgithub.com\u002Fsail-sg\u002FFlowReasoner) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsail-sg\u002FFlowReasoner?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | Sea AI Lab \u002F NUS | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.15257) | Custom |\n| [DrMAS](https:\u002F\u002Fgithub.com\u002FlangfengQ\u002FDrMAS) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FlangfengQ\u002FDrMAS?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | NTU | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.08847) | Custom |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [PettingLLMs](https:\u002F\u002Fgithub.com\u002Fpettingllms-ai\u002FPettingLLMs) | AT-GRPO | Multi | Both | Multi | Game\u002FCode\u002FMath\u002FPlanning | Rule (verifiable) | No |\n| [MASPRM](https:\u002F\u002Fgithub.com\u002Fmilad1378yz\u002FMASPRM) | PRM (trained from MCTS rollouts) | Multi | Process | Multi | Reasoning (GSM8K\u002FMATH\u002FMMLU) | Learned PRM | No |\n| [ARIA](https:\u002F\u002Fgithub.com\u002Frhyang2021\u002FARIA) | REINFORCE | Both | Process | Multi | Negotiation\u002FBargaining | Other | No |\n| [AMPO](https:\u002F\u002Fgithub.com\u002FMozerWang\u002FAMPO) | BC\u002FAMPO(GRPO improvement) | Multi | Outcome | Multi | Social Interaction | Model-based | No |\n| [MAPoRL](https:\u002F\u002Fgithub.com\u002Fchanwoo-park-official\u002FMAPoRL) | PPO | Multi | Outcome | Multi | Collaborative LLM Tasks | Rule | No |\n| [FlowReasoner](https:\u002F\u002Fgithub.com\u002Fsail-sg\u002FFlowReasoner) | GRPO | Multi | Outcome | Multi | Multi-agent Workflow Design | Rule | Yes |\n| [DrMAS](https:\u002F\u002Fgithub.com\u002FlangfengQ\u002FDrMAS) | GRPO (agent-wise) | Multi | Outcome | Multi | Multi-agent LLM Systems | Rule | No |\n\n\u003C\u002Fdetails>\n\n## 🧠 Memory\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [MEM1](https:\u002F\u002Fgithub.com\u002FMIT-MI\u002FMEM1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMIT-MI\u002FMEM1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | MIT | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.15841) | veRL (based on Search-R1) |\n| [Memento](https:\u002F\u002Fgithub.com\u002FAgent-on-the-Fly\u002FMemento) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAgent-on-the-Fly\u002FMemento?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | UCL, Huawei | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.16153) | Custom |\n| [MemAgent](https:\u002F\u002Fgithub.com\u002FBytedTsinghua-SIA\u002FMemAgent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FBytedTsinghua-SIA\u002FMemAgent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Bytedance, Tsinghua-SIA | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.02259) | veRL |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [MEM1](https:\u002F\u002Fgithub.com\u002FMIT-MI\u002FMEM1) | PPO\u002FGRPO | Single | Outcome | Multi | WebShop\u002FGSM8K\u002FQA | Rule\u002FModel | Yes |\n| [Memento](https:\u002F\u002Fgithub.com\u002FAgent-on-the-Fly\u002FMemento) | soft Q-Learning | Single | Outcome | Multi | Research\u002FQA\u002FCode\u002FWeb | External\u002FRule | Yes |\n| [MemAgent](https:\u002F\u002Fgithub.com\u002FBytedTsinghua-SIA\u002FMemAgent) | PPO, GRPO, DPO | Multi | Outcome | Multi | Long-context QA | Rule\u002FModel\u002FExternal | Yes |\n\n\u003C\u002Fdetails>\n\n## 🦾 Embodied\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [Embodied-R1](https:\u002F\u002Fgithub.com\u002Fpickxiguapi\u002FEmbodied-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpickxiguapi\u002FEmbodied-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Tianjing University | [Paper](http:\u002F\u002Farxiv.org\u002Fabs\u002F2508.13998) | veRL |\n| [STeCa](https:\u002F\u002Fgithub.com\u002FWangHanLinHenry\u002FSTeCa) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FWangHanLinHenry\u002FSTeCa?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | The Hong Kong Polytechnic University | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.14276) | FastChat\u002FTRL |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [Embodied-R1](https:\u002F\u002Fgithub.com\u002Fpickxiguapi\u002FEmbodied-R1) | GRPO | Single | Outcome | Single | Grounding\u002FWaypoint | Rule | No |\n| [STeCa](https:\u002F\u002Fgithub.com\u002FWangHanLinHenry\u002FSTeCa) | DPO (RFT) | Single | Both | Multi | Embodied\u002FHousehold | Rule\u002FMC | Environment Actions |\n\n\u003C\u002Fdetails>\n\n## 🏷️ Domain-Specific\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework | Domain |\n| :----: | :----: | :----: |  :----: | :----: | :----: | :----: |\n| [MedSAM-Agent](https:\u002F\u002Fgithub.com\u002FCUHK-AIM-Group\u002FMedSAM-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FCUHK-AIM-Group\u002FMedSAM-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | CUHK\u002FTencent | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.03320) | Custom | Medical |\n| [OS-R1](https:\u002F\u002Fgithub.com\u002FLHY-24\u002FOS-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FLHY-24\u002FOS-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | ISCAS | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.12551) | Custom | OS\u002FSystems |\n| [MMedAgent-RL](https:\u002F\u002Fgithub.com\u002FJanerhYang\u002FMMedAgent-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FJanerhYang\u002FMMedAgent-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | Unknown | [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.00555) | Unknown | Medical |\n| [DoctorAgent-RL](https:\u002F\u002Fgithub.com\u002FJarvisUSTC\u002FDoctorAgent-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FJarvisUSTC\u002FDoctorAgent-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | UCAS\u002FCAS\u002FUSTC | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.19630) | RAGEN | Medical |\n| [Biomni](https:\u002F\u002Fgithub.com\u002Fsnap-stanford\u002FBiomni) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsnap-stanford\u002FBiomni?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | Stanford University (SNAP) | [Paper](https:\u002F\u002Fwww.biorxiv.org\u002Fcontent\u002F10.1101\u002F2025.05.30.656746v1) | Custom | Biomedical |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [MedSAM-Agent](https:\u002F\u002Fgithub.com\u002FCUHK-AIM-Group\u002FMedSAM-Agent) | GRPO (via veRL) | Single | Both | Multi | Medical Image Segmentation | Model (clinical fidelity) | Yes (SAM\u002FMedSAM2) |\n| [OS-R1](https:\u002F\u002Fgithub.com\u002FLHY-24\u002FOS-R1) | GRPO (via veRL) | Single | Outcome | Multi | Linux Kernel Tuning | Rule | Yes (LightRAG, kernel config) |\n| [MMedAgent-RL](https:\u002F\u002Fgithub.com\u002FJanerhYang\u002FMMedAgent-RL) | Unknown | Multi | Unknown | Unknown | Unknown | Unknown | Unknown |\n| [DoctorAgent-RL](https:\u002F\u002Fgithub.com\u002FJarvisUSTC\u002FDoctorAgent-RL) | GRPO | Multi | Both | Multi | Consultation\u002FDiagnosis | Model\u002FRule | No |\n| [Biomni](https:\u002F\u002Fgithub.com\u002Fsnap-stanford\u002FBiomni) | TBD | Single | TBD | Single | scRNAseq\u002FCRISPR\u002FADMET\u002FKnowledge | TBD | Yes |\n\n\u003C\u002Fdetails>\n\n## 🎯 Reward & Training Methodology\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | Focus |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [ToolPRMBench](https:\u002F\u002Fgithub.com\u002FDavid-Li0406\u002FToolPRMBench) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FDavid-Li0406\u002FToolPRMBench?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | Arizona State University | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.12294) | PRM Benchmark for Tool-Use |\n| [RLVR-World](https:\u002F\u002Fgithub.com\u002Fthuml\u002FRLVR-World) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fthuml\u002FRLVR-World?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | THU ML Group | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.13934) | RLVR for World Models |\n| [AgentPRM](https:\u002F\u002Fgithub.com\u002Fsanjibanc\u002Fagent_prm) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsanjibanc\u002Fagent_prm?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Cornell | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.10325) | Process Reward for Agents |\n| [Agentic-Reward-Modeling](https:\u002F\u002Fgithub.com\u002FTHU-KEG\u002FAgentic-Reward-Modeling) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHU-KEG\u002FAgentic-Reward-Modeling?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | THU-KEG | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.19328) | Agentic Reward Agent |\n| [AgentRM](https:\u002F\u002Fgithub.com\u002Fthunlp\u002FAgentRM) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fthunlp\u002FAgentRM?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | THUNLP\u002FTsinghua | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.18407) | Generalizable Agent RM |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [ToolPRMBench](https:\u002F\u002Fgithub.com\u002FDavid-Li0406\u002FToolPRMBench) | N\u002FA (Benchmark) | Single | Process | Multi | Tool-Use | Rule\u002FModel | Yes |\n| [RLVR-World](https:\u002F\u002Fgithub.com\u002Fthuml\u002FRLVR-World) | RLVR | Single | Outcome | Multi | World Modeling (Language\u002FVideo) | Model (verifiable) | No |\n| [AgentPRM](https:\u002F\u002Fgithub.com\u002Fsanjibanc\u002Fagent_prm) | PPO\u002FDPO + PRM | Single | Process | Multi | ALFWorld\u002FGeneral | Model (PRM) | Yes |\n| [Agentic-Reward-Modeling](https:\u002F\u002Fgithub.com\u002FTHU-KEG\u002FAgentic-Reward-Modeling) | DPO\u002FBest-of-N | Single | Outcome | Single | General Instruction | Model (Reward Agent) | Yes (Verification) |\n| [AgentRM](https:\u002F\u002Fgithub.com\u002Fthunlp\u002FAgentRM) | MCTS\u002FRM-guided | Single | Outcome | Multi | 9 Agent Tasks | Model (regression PRM) | Yes |\n\n\u003C\u002Fdetails>\n\n## 🛡️ Safety\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [SafeSearch](https:\u002F\u002Fgithub.com\u002Famazon-science\u002FSafeSearch) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Famazon-science\u002FSafeSearch?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | Amazon Science | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.17017) | veRL |\n| [curiosity_redteam](https:\u002F\u002Fgithub.com\u002FImprobable-AI\u002Fcuriosity_redteam) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FImprobable-AI\u002Fcuriosity_redteam?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.2 | MIT | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2402.19464) | Custom |\n| [RLbreaker](https:\u002F\u002Fgithub.com\u002FXuanChen-xc\u002FRLbreaker) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FXuanChen-xc\u002FRLbreaker?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.6 | Purdue | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2406.08705) | Custom |\n| [xJailbreak](https:\u002F\u002Fgithub.com\u002FAegis1863\u002FxJailbreak) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAegis1863\u002FxJailbreak?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | Academic | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2501.16727) | Custom |\n| [Auto-RT](https:\u002F\u002Fgithub.com\u002Ficip-cas\u002FAuto-RT) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ficip-cas\u002FAuto-RT?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | ICIP-CAS | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2501.01830) | Custom |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [SafeSearch](https:\u002F\u002Fgithub.com\u002Famazon-science\u002FSafeSearch) | PPO (GAE\u002FGRPO) | Single | Both | Multi | Safe QA\u002FSearch | Rule + Model | Search |\n| [curiosity_redteam](https:\u002F\u002Fgithub.com\u002FImprobable-AI\u002Fcuriosity_redteam) | RL + Curiosity | Single | Outcome | Multi | Red Teaming | Model | Yes (iterative query) |\n| [RLbreaker](https:\u002F\u002Fgithub.com\u002FXuanChen-xc\u002FRLbreaker) | Custom PPO | Single | Outcome | Multi | Jailbreaking | Model | Yes (mutator selection) |\n| [xJailbreak](https:\u002F\u002Fgithub.com\u002FAegis1863\u002FxJailbreak) | RL | Single | Outcome | Multi | Jailbreaking | Model (embedding) | Yes (iterative) |\n| [Auto-RT](https:\u002F\u002Fgithub.com\u002Ficip-cas\u002FAuto-RT) | PPO | Single | Outcome | Multi | Red Teaming | Model | Yes (strategy exploration) |\n\n\u003C\u002Fdetails>\n\n## 👁️ VLM Agent\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [multimodal-search-r1](https:\u002F\u002Fgithub.com\u002FEvolvingLMMs-Lab\u002Fmultimodal-search-r1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FEvolvingLMMs-Lab\u002Fmultimodal-search-r1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | ByteDance\u002FNTU | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.20670) | Custom |\n| [DeepEyesV2](https:\u002F\u002Fgithub.com\u002FVisual-Agent\u002FDeepEyesV2) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FVisual-Agent\u002FDeepEyesV2?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | Xiaohongshu | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.05271) | Custom |\n| [VDeepEyes](https:\u002F\u002Fgithub.com\u002FVisual-Agent\u002FDeepEyes) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FVisual-Agent\u002FDeepEyes?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Xiaohongshu\u002FXJTU | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.14362) | veRL |\n| [CoSo](https:\u002F\u002Fgithub.com\u002FlangfengQ\u002FCoSo) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FlangfengQ\u002FCoSo?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | NTU\u002FAlibaba | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.03792) | Custom |\n| [RL4VLM](https:\u002F\u002Fgithub.com\u002FRL4VLM\u002FRL4VLM) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRL4VLM\u002FRL4VLM?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.5 | UC Berkeley | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2405.10292) | Custom |\n| [VSC-RL](https:\u002F\u002Fgithub.com\u002Fai-agents-2030\u002FVSC_RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fai-agents-2030\u002FVSC_RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Liverpool\u002FHuawei\u002FTianjin\u002FUCL | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.07949) | Custom |\n| [AlphaDrive](https:\u002F\u002Fgithub.com\u002Fhustvl\u002FAlphaDrive) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhustvl\u002FAlphaDrive?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | HUST\u002FHorizon Robotics | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.07608) | Custom |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [multimodal-search-r1](https:\u002F\u002Fgithub.com\u002FEvolvingLMMs-Lab\u002Fmultimodal-search-r1) | GRPO | Single | Outcome | Multi | Multimodal Search | Rule | Yes (Search) |\n| [DeepEyesV2](https:\u002F\u002Fgithub.com\u002FVisual-Agent\u002FDeepEyesV2) | Outcome RL | Single | Outcome | Multi | Multimodal Reasoning | Rule | Yes (Code exec, Web search) |\n| [VDeepEyes](https:\u002F\u002Fgithub.com\u002FVisual-Agent\u002FDeepEyes) | PPO\u002FGRPO | Multi | Process | Multi | VQA | All | Yes |\n| [CoSo](https:\u002F\u002Fgithub.com\u002FlangfengQ\u002FCoSo) | Soft RL (counterfactual) | Single | Outcome | Multi | Android\u002FCard\u002FEmbodied | Rule | Yes |\n| [RL4VLM](https:\u002F\u002Fgithub.com\u002FRL4VLM\u002FRL4VLM) | PPO | Single | Outcome | Multi | GymCards\u002FALFWorld | Rule | Yes |\n| [VSC-RL](https:\u002F\u002Fgithub.com\u002Fai-agents-2030\u002FVSC_RL) | Variational RL | Single | Outcome | Multi | Mobile Device Control | Rule | Yes |\n| [AlphaDrive](https:\u002F\u002Fgithub.com\u002Fhustvl\u002FAlphaDrive) | GRPO | Single | Outcome | Multi | Autonomous Driving | Rule (4 planning rewards) | No |\n\n\u003C\u002Fdetails>\n\n## 🔄 Self-Evolution\n\n> ⚠️ **Note**: The definition of \"Self-Evolution\" in the context of RL for LLM agents is still evolving and not yet well-established. This category currently collects works whose paper titles explicitly contain \"self-evolving\" or \"self-evolution\", where the agent improves itself through RL-driven feedback loops.\n\n\n| Github Repo | 🌟 Stars | Date | Org | Paper Link | RL Framework |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [AgentEvolver](https:\u002F\u002Fgithub.com\u002Fmodelscope\u002FAgentEvolver) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmodelscope\u002FAgentEvolver?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | Alibaba\u002FTongyi Lab | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.10395) | Custom |\n| [SEAgent](https:\u002F\u002Fgithub.com\u002FSunzeY\u002FSEAgent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSunzeY\u002FSEAgent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | Shanghai AI Lab \u002F CUHK | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.04700) | Custom |\n| [MemSkill](https:\u002F\u002Fgithub.com\u002FViktorAxelsen\u002FMemSkill) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FViktorAxelsen\u002FMemSkill?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | NTU\u002FUIUC\u002FUIC\u002FTsinghua | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.02474) | Custom |\n| [MemRL](https:\u002F\u002Fgithub.com\u002FMemTensor\u002FMemRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMemTensor\u002FMemRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | SJTU\u002FXidian\u002FNUS\u002FUSTC\u002FMemTensor | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.03192) | Custom |\n| [RAGEN](https:\u002F\u002Fgithub.com\u002FRAGEN-AI\u002FRAGEN) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRAGEN-AI\u002FRAGEN?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | RAGEN-AI | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.20073) | veRL |\n| [WebRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FWebRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002FWebRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.11 | Tsinghua\u002FZhipu AI | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2411.02337) | Custom |\n\n\u003Cdetails>\n\u003Csummary>📋 Click to view technical details\u003C\u002Fsummary>\n\n| Github Repo | RL Algorithm | Single\u002FMulti Agent | Outcome\u002FProcess Reward | Single\u002FMulti Turn | Task | Reward Type | Tool usage |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [AgentEvolver](https:\u002F\u002Fgithub.com\u002Fmodelscope\u002FAgentEvolver) | ADCA-GRPO | Single | Outcome | Multi | Social Game\u002FTool-use | Rule | Yes |\n| [SEAgent](https:\u002F\u002Fgithub.com\u002FSunzeY\u002FSEAgent) | GRPO | Single | Outcome | Multi | Computer Use (OSWorld) | Model | Yes (Screenshot-based) |\n| [MemSkill](https:\u002F\u002Fgithub.com\u002FViktorAxelsen\u002FMemSkill) | PPO | Single | Process | Multi | QA\u002FALFWorld | Model (learned skills) | Yes |\n| [MemRL](https:\u002F\u002Fgithub.com\u002FMemTensor\u002FMemRL) | RL-based (Q-value) | Single | Process | Multi | HLE\u002FBigCodeBench\u002FALFWorld | Model (retrieval) | Yes |\n| [RAGEN](https:\u002F\u002Fgithub.com\u002FRAGEN-AI\u002FRAGEN) | PPO\u002FGRPO (StarPO) | Single | Both | Multi | TextGame | All | Yes |\n| [WebRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FWebRL) | Actor-Critic RL + ORM | Single | Outcome | Multi | Web Navigation (WebArena) | Model (ORM) | Yes (Web browsing) |\n\n\u003C\u002Fdetails>\n\n## ⛰️ Environment\n\n| Github Repo | 🌟 Stars | Date | Org | Task |\n| :----: | :----: | :----: |  :----: | :----: |\n| [OpenSandbox](https:\u002F\u002Fgithub.com\u002Falibaba\u002FOpenSandbox) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Falibaba\u002FOpenSandbox?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | Alibaba | Code\u002FGUI\u002FAgent Eval |\n| [OpenEnv](https:\u002F\u002Fgithub.com\u002Fmeta-pytorch\u002FOpenEnv) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmeta-pytorch\u002FOpenEnv?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | Meta (PyTorch) | Chess\u002FArcade\u002FFinance |\n| [NeMo-Gym](https:\u002F\u002Fgithub.com\u002FNVIDIA-NeMo\u002FGym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNVIDIA-NeMo\u002FGym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | NVIDIA | Multi-step\u002FMulti-turn |\n| [open-trajectory-gym](https:\u002F\u002Fgithub.com\u002Fwestonbrown\u002Fopen-trajectory-gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fwestonbrown\u002Fopen-trajectory-gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | Individual | CTF\u002FSecurity |\n| [R2E-Gym](https:\u002F\u002Fgithub.com\u002FR2E-Gym\u002FR2E-Gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FR2E-Gym\u002FR2E-Gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | UC Berkeley\u002FANU | SWE |\n| [LoCoBench-Agent](https:\u002F\u002Fgithub.com\u002FSalesforceAIResearch\u002FLoCoBench-Agent) | ![](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSalesforceAIResearch\u002FLoCoBench-Agent.svg?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700) | 2025.11 | Salesforce AI Research | SWE |\n| [Simia-Agent-Training](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FSimia-Agent-Training) | ![](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FSimia-Agent-Training.svg?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700) | 2025.10 | Microsoft | ToolUse\u002FAPI |\n| [PaperArena](https:\u002F\u002Fgithub.com\u002FMelmaphother\u002FPaperArena) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMelmaphother\u002FPaperArena?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | University of Science and Technology of China | ScientificLiteratureQA |\n| [enterprise-deep-research](https:\u002F\u002Fgithub.com\u002FSalesforceAIResearch\u002Fenterprise-deep-research) | ![](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSalesforceAIResearch\u002Fenterprise-deep-research.svg?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700) | 2025.9 | Salesforce AI Research | DeepResearch |\n| [CompassVerifier](https:\u002F\u002Fgithub.com\u002Fopen-compass\u002FCompassVerifier) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopen-compass\u002FCompassVerifier?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | Shanghai AI Lab | Reasoning |\n| [SWE-smith](https:\u002F\u002Fgithub.com\u002FSWE-bench\u002FSWE-smith) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSWE-bench\u002FSWE-smith?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | Princeton\u002FStanford\u002FSWE-bench | SWE |\n| [SWE-Gym](https:\u002F\u002Fgithub.com\u002FSWE-Gym\u002FSWE-Gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSWE-Gym\u002FSWE-Gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.12 | UC Berkeley\u002FUIUC\u002FCMU\u002FApple | SWE |\n| [Mind2Web-2](https:\u002F\u002Fgithub.com\u002FOSU-NLP-Group\u002FMind2Web-2) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOSU-NLP-Group\u002FMind2Web-2?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Ohio State University | Web |\n| [gem](https:\u002F\u002Fgithub.com\u002Faxon-rl\u002Fgem) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Faxon-rl\u002Fgem?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Sea AI Lab | Math\u002FCode\u002FGame\u002FQA |\n| [MLE-Dojo](https:\u002F\u002Fgithub.com\u002FMLE-Dojo\u002FMLE-Dojo) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMLE-Dojo\u002FMLE-Dojo?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | GIT, Stanford | MLE |\n| [atropos](https:\u002F\u002Fgithub.com\u002FNousResearch\u002Fatropos) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNousResearch\u002Fatropos?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | Nous Research | Game\u002FCode\u002FTool |\n| [InternBootcamp](https:\u002F\u002Fgithub.com\u002FInternLM\u002FInternBootcamp) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FInternLM\u002FInternBootcamp?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | InternBootcamp | Coding\u002FQA\u002FGame |\n| [loong](https:\u002F\u002Fgithub.com\u002Fcamel-ai\u002Floong) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcamel-ai\u002Floong?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | CAMEL-AI.org | RLVR |\n| [DataSciBench](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FDataSciBench) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002FDataSciBench?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Tsinghua | data analysis |\n| [reasoning-gym](https:\u002F\u002Fgithub.com\u002Fopen-thought\u002Freasoning-gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopen-thought\u002Freasoning-gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | open-thought | Math\u002FGame |\n| [llmgym](https:\u002F\u002Fgithub.com\u002Ftensorzero\u002Fllmgym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftensorzero\u002Fllmgym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | tensorzero | TextGame\u002FTool |\n| [debug-gym](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fdebug-gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fdebug-gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.11 | Microsoft Research | Debugging\u002FGame\u002FCode |\n| [gym-llm](https:\u002F\u002Fgithub.com\u002Frsanchezmo\u002Fgym-llm) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Frsanchezmo\u002Fgym-llm?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.8 | Rodrigo Sánchez Molina | Control\u002FGame |\n| [AgentGym](https:\u002F\u002Fgithub.com\u002FWooooDyy\u002FAgentGym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FWooooDyy\u002FAgentGym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.6 | Fudan | Web\u002FGame |\n| [tau-bench](https:\u002F\u002Fgithub.com\u002Fsierra-research\u002Ftau-bench) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsierra-research\u002Ftau-bench?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.6 | Sierra | Tool |\n| [appworld](https:\u002F\u002Fgithub.com\u002FStonyBrookNLP\u002Fappworld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FStonyBrookNLP\u002Fappworld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.6 | Stony Brook University | Phone Use |\n| [android_world](https:\u002F\u002Fgithub.com\u002Fgoogle-research\u002Fandroid_world) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgoogle-research\u002Fandroid_world?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.5 | Google Research | Phone Use |\n| [TheAgentCompany](https:\u002F\u002Fgithub.com\u002FTheAgentCompany\u002FTheAgentCompany) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTheAgentCompany\u002FTheAgentCompany?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.3 | CMU, Duke | Coding |\n| [LlamaGym](https:\u002F\u002Fgithub.com\u002FKhoomeiK\u002FLlamaGym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FKhoomeiK\u002FLlamaGym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.3 | Rohan Pandey | Game|\n| [visualwebarena](https:\u002F\u002Fgithub.com\u002Fweb-arena-x\u002Fvisualwebarena)   | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fweb-arena-x\u002Fvisualwebarena?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.1 | CMU | Web |\n| [LMRL-Gym](https:\u002F\u002Fgithub.com\u002Fabdulhaim\u002FLMRL-Gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fabdulhaim\u002FLMRL-Gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2023.12 | UC Berkeley | Game |\n| [OSWorld](https:\u002F\u002Fgithub.com\u002Fxlang-ai\u002FOSWorld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fxlang-ai\u002FOSWorld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2023.10 | HKU, CMU, Salesforce, Waterloo | Computer Use |\n| [webarena](https:\u002F\u002Fgithub.com\u002Fweb-arena-x\u002Fwebarena) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fweb-arena-x\u002Fwebarena?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2023.7 | CMU | Web |\n| [AgentBench](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FAgentBench) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002FAgentBench?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2023.7 | Tsinghua University | Game\u002FWeb\u002FQA\u002FTool |\n| [WebShop](https:\u002F\u002Fgithub.com\u002Fprinceton-nlp\u002FWebShop) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fprinceton-nlp\u002FWebShop?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2022.7 | Princeton-NLP | Web |\n| [ScienceWorld](https:\u002F\u002Fgithub.com\u002Fallenai\u002FScienceWorld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fallenai\u002FScienceWorld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2022.3 | AllenAI | TextGame\u002FScienceQA |\n| [alfworld](https:\u002F\u002Fgithub.com\u002Falfworld\u002Falfworld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Falfworld\u002Falfworld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2020.10 | Microsoft, CMU, UW | Embodied |\n| [factorio-learning-environment](https:\u002F\u002Fgithub.com\u002FJackHopkins\u002Ffactorio-learning-environment) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FJackHopkins\u002Ffactorio-learning-environment?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2021.6 | JackHopkins | Game |\n| [jericho](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fjericho) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fjericho?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2018.10 | Microsoft, GIT | TextGame |\n| [TextWorld](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FTextWorld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FTextWorld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2018.6 | Microsoft Research | TextGame |\n\n## Under Review\u002FWaiting for Open Source\n- [JoyAgents-R1: Joint Evolution Dynamics for Versatile Multi-LLM Agents with Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.19846)\n- [Shop-R1: Rewarding LLMs to Simulate Human Behavior in Online Shopping via Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.17842)\n- [Training Long-Context, Multi-Turn Software Engineering Agents with Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.03501)\n- [Acting Less is Reasoning More! Teaching Model to Act Efficiently](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.14870)\n- [Agentic Reasoning and Tool Integration for LLMs via Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.01441)\n- [ComputerRL: Scaling End-to-End Online Reinforcement Learning for Computer Use Agents](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.14040)\n- [Atom-Searcher: Enhancing Agentic Deep Research via Fine-Grained Atomic Thought Reward](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus)\n- [MUA-RL: MULTI-TURN USER-INTERACTING AGENTREINFORCEMENT LEARNING FOR AGENTIC TOOL USE](https:\u002F\u002Fgithub.com\u002Fzzwkk\u002FMUA-RL)\n- [Understanding Tool-Integrated Reasoning](https:\u002F\u002Fzhongwenxu.notion.site\u002FUnderstanding-Tool-Integrated-Reasoning-2551c4e140e3805489fadcc802a1ea83)\n- [Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.19828)\n- [Encouraging Good Processes Without the Need for Good Answers: Reinforcement Learning for LLM Agent Planning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.19598)\n- [SFR-DeepResearch: Towards Effective Reinforcement Learning for Autonomously Reasoning Single Agents](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.06283)\n- [WebExplorer: Explore and Evolve for Training Long-Horizon Web Agents](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.06501)\n- [EnvX: Agentize Everything with Agentic AI](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.08088)\n- [UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.02544)\n- [UI-Venus Technical Report: Building High-performance UI Agents with RFT](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.10833)\n- [Agent2 : An Agent-Generates-Agent Framework for Reinforcement Learning Automation](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.13368)\n- [Tool-R1: Sample-Efficient Reinforcement Learning for Agentic Tool Use](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.12867v1)\n- [Adversarial Reinforcement Learning for Large Language Model Agent Safety](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.05442)\n- [Learning to Refine: An Agentic RL Approach for Iterative SPARQL Query Construction](https:\u002F\u002Fwww.arxiv.org\u002Fabs\u002F2511.11770)\n- [InfoFlow: Reinforcing Search Agent Via Reward Density Optimization](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.26575)\n\n## Star History\n\n[![Star History Chart](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fthinkwee_AgentsMeetRL_readme_02ad232660f9.png)](https:\u002F\u002Fwww.star-history.com\u002F#thinkwee\u002FagentsMeetRL&Date)\n\n\n## Citation\n\nIf you find this repository useful, please consider citing it:\n\n```bibtex\n@misc{agentsMeetRL,\n  title={When LLM Agents Meet Reinforcement Learning: A Comprehensive Survey},\n  author={AgentsMeetRL Contributors},\n  year={2025},\n  url={https:\u002F\u002Fgithub.com\u002Fthinkwee\u002FagentsMeetRL}\n}\n```\n\n---\n\n\u003Cdiv align=\"center\">\n  \u003Cp>Made with ❤️ by the AgentsMeetRL community\u003C\u002Fp>\n\u003C\u002Fdiv>\n","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fthinkwee_AgentsMeetRL_readme_020ab9c761ce.png\" alt=\"NOVER Logo\" width=\"500\">\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n\n![基础框架](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F基础框架-18-BFA2DB?style=for-the-badge)\n![通用](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F通用-16-4E6813?style=for-the-badge)\n![搜索与RAG](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F搜索_&_RAG-25-845C40?style=for-the-badge)\n![Web与GUI](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FWeb_&_GUI-20-A259FF?style=for-the-badge)\n\u003Cbr>\n![工具](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F工具-12-D89F7B?style=for-the-badge)\n![代码与SWE](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F代码_&_SWE-19-A47B67?style=for-the-badge)\n![推理](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F推理-15-FF69B4?style=for-the-badge)\n![多智能体](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F多--智能体-7-1F4CAD?style=for-the-badge)\n\u003Cbr>\n![记忆](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F记忆-3-007a88?style=for-the-badge)\n![具身](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F具身-2-C0C5CE?style=for-the-badge)\n![领域特定](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F领域--特定-5-ffc884?style=for-the-badge)\n![奖励与训练](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F奖励_&_训练-5-9B59B6?style=for-the-badge)\n\u003Cbr>\n![安全](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F安全-5-E74C3C?style=for-the-badge)\n![VLM智能体](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FVLM_智能体-7-2ECC71?style=for-the-badge)\n![自我进化](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F自我--进化-6-F39C12?style=for-the-badge)\n![环境](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F环境-40-FA5A4C?style=for-the-badge)\n\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n\n[![交互式仪表盘](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F📊_交互式_仪表盘-访问_网站-蓝色?style=for-the-badge)](https:\u002F\u002Fthinkwee.top\u002Famr\u002F)\n\n\u003C\u002Fdiv>\n\n# 当LLM智能体遇上强化学习\n\n**AgentsMeetRL** 是一个精彩的列表，汇总了使用强化学习训练 LLM 智能体的 **开源仓库**：\n - 🤖 判断一个项目是否为智能体项目的标准是：它必须具备以下至少一项：多轮交互或工具使用（因此，TIR 项目和工具集成推理也被纳入本仓库）。\n - ⚠️ 本项目基于对使用 LLM 编码智能体的开源仓库进行的代码分析，其中可能包含不准确的情况。尽管已人工审核，但仍可能存在遗漏。如果您发现任何错误，请随时通过 issue 或 PR 告知我们——我们非常欢迎！\n - 🚀 我们特别关注各个项目所依赖的强化学习框架、RL 算法、奖励机制以及环境，以便大家参考这些优秀的开源项目是如何做出技术选择的。请查看每个表格下方的 [点击查看技术细节]。\n - 📅 最后更新日期：2026年3月24日\n - 🤗 欢迎随时提交您自己的项目——我们非常期待您的贡献！\n\n分类体系：\n - **基础框架**：用于 LLM 智能体的通用 RL 训练框架（例如 veRL、OpenRLHF、trl）\n - **通用\u002F多任务**：在多个任务或环境中进行训练和评估的智能体系统\n - **搜索与RAG**：利用检索工具增强 LLM 推理能力的搜索增强型推理智能体\n - **Web与GUI**：与网页浏览器、移动\u002F桌面 GUI 或操作系统交互的智能体\n - **工具使用**：经过训练以调用外部工具（API、代码执行器、MCP 等）的智能体\n - **代码与SWE**：软件工程和代码生成智能体\n - **推理**：具备工具集成或多轮推理能力的智能体（数学、问答、视觉等）\n - **多智能体RL**：通过强化学习实现的多智能体协作、谈判或信用分配\n - **记忆**：能够学习管理、检索或演化记忆的智能体\n - **具身**：在具身化\u002F物理仿真环境中运行的智能体\n - **领域特定**：针对特定领域的 RL 智能体（如医疗、操作系统调优等）\n - **奖励与训练**：用于智能体的进程\u002F结果奖励模型及训练方法\n - **安全**：用于智能体安全对齐、对抗性红队测试以及防越狱\u002F攻防的强化学习\n - **VLM智能体**：通过强化学习训练的视觉-语言模型智能体，用于多模态交互\n - **自我进化**：通过 RL 反馈循环实现自我进化的智能体（⚠️ 此定义仍在社区中不断发展）\n - **环境**：用于智能体训练\u002F评估的基准、模拟环境和沙盒环境\n\n部分枚举：\n - 奖励类型枚举：\n   - 外部验证器：例如编译器或数学求解器\n   - 基于规则：例如具有精确匹配评分的 LaTeX 解析器\n   - 基于模型：例如经过训练的验证 LLM 或奖励 LLM\n   - 自定义\n\n---\n\n## 更新\n- 📢 **2026年3月更新**：将分类体系由12类重组为16类。新增约70个仓库，涵盖2025年9月至2026年3月期间的内容。新增类别包括多智能体RL、奖励与训练、安全、VLM智能体、自我进化以及领域特定。原GUI和Web合并为Web与GUI，TextGame和Biomedical作为独立类别已被取消。总仓库数量从约134个增加到205个。\n\n## 🔧 基础框架\n\n| GitHub 仓库 | 🌟 星数 | 日期 | 组织 | 论文链接 |\n| :----: | :----: | :----: |  :----: | :----: |\n| [Open-AgentRL](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FOpen-AgentRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGen-Verse\u002FOpen-AgentRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | Gen-Verse | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.02488) |\n| [OpenClaw-RL](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FOpenClaw-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGen-Verse\u002FOpenClaw-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | Gen-Verse | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2603.10165) |\n| [Claw-R1](https:\u002F\u002Fgithub.com\u002FAgentR1\u002FClaw-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAgentR1\u002FClaw-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | 中国科学技术大学 | -- |\n| [prime-rl](https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fprime-rl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FPrimeIntellect-ai\u002Fprime-rl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Prime Intellect | -- |\n| [NeMo-RL](https:\u002F\u002Fgithub.com\u002FNVIDIA-NeMo\u002FRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNVIDIA-NeMo\u002FRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | 英伟达 | -- |\n| [RLinf](https:\u002F\u002Fgithub.com\u002FRLinf\u002FRLinf) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRLinf\u002FRLinf?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | 清华大学\u002FInfinigence AI\u002F北京大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.15965) |\n| [siiRL](https:\u002F\u002Fgithub.com\u002Fsii-research\u002FsiiRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsii-research\u002FsiiRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | 上海创新研究院 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.13833) |\n| [slime](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fslime) | ![](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002Fslime?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700) | 2025.6 | 清华大学 (THUDM) | [博客](https:\u002F\u002Flmsys.org\u002Fblog\u002F2025-07-09-slime\u002F) |\n| [agent-lightning](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fagent-lightning) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fagent-lightning?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 微软研究院 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.03680) |\n| [AReaL](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FAReaL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FinclusionAI\u002FAReaL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 蚂蚁集团\u002F清华大学 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.24298) |\n| [ROLL](https:\u002F\u002Fgithub.com\u002Falibaba\u002FROLL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Falibaba\u002FROLL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 阿里巴巴 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.06122) |\n| [MARTI](https:\u002F\u002Fgithub.com\u002FTsinghuaC3I\u002FMARTI) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTsinghuaC3I\u002FMARTI?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 清华大学 | -- |\n| [RL2](https:\u002F\u002Fgithub.com\u002FChenmienTan\u002FRL2) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FChenmienTan\u002FRL2?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | Accio | – |\n| [verifiers](https:\u002F\u002Fgithub.com\u002Fwillccbb\u002Fverifiers) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fwillccbb\u002Fverifiers?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | 个人 | -- |\n| [oat](https:\u002F\u002Fgithub.com\u002Fsail-sg\u002Foat) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsail-sg\u002Foat?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.11 | 新加坡国立大学\u002FSea AI | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2411.01493) |\n| [veRL](https:\u002F\u002Fgithub.com\u002Fvolcengine\u002Fverl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fvolcengine\u002Fverl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.10 | 字节跳动 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2409.19256) |\n| [OpenRLHF](https:\u002F\u002Fgithub.com\u002FOpenRLHF\u002FOpenRLHF) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenRLHF\u002FOpenRLHF?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2023.7 | OpenRLHF | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2405.11143) |\n| [trl](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftrl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhuggingface\u002Ftrl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2019.11 | HuggingFace | -- |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单智能体\u002F多智能体 | 结果奖励\u002F过程奖励 | 单轮\u002F多轮 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [Open-AgentRL](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FOpen-AgentRL) | GRPO-TCR | 单智能体 | 两者均有 | 多轮 | 推理\u002FGUI\u002F编码 | 模型（PRM） | 是（SandboxFusion） |\n| [OpenClaw-RL](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FOpenClaw-RL) | GRPO\u002FOPD | 单\u002F多智能体 | 两者均有 | 多轮 | 终端\u002FGUI\u002FSWE\u002F工具调用 | 模型\u002F外部 | 是 |\n| [Claw-R1](https:\u002F\u002Fgithub.com\u002FAgentR1\u002FClaw-R1) | 通用强化学习框架 | 多智能体 | 两者均有 | 多轮 | 通用智能体 | 全部 | 是（与框架无关） |\n| [prime-rl](https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fprime-rl) | GRPO\u002FPPO | 多智能体 | 结果奖励 | 多轮 | 数学\u002F代码\u002F搜索 | 模型\u002F外部 | 是 |\n| [NeMo-RL](https:\u002F\u002Fgithub.com\u002FNVIDIA-NeMo\u002FRL) | GRPO\u002FDAPO\u002FGDPO\u002FDPO | 单智能体 | 结果奖励 | 多轮 | 数学\u002F推理\u002F代码 | 规则\u002F外部 | 否 |\n| [RLinf](https:\u002F\u002Fgithub.com\u002FRLinf\u002FRLinf) | PPO\u002FGRPO\u002FDAPO\u002FSAC\u002FREINFORCE++\u002FCrossQ\u002FRLPD | 单\u002F多智能体 | 两者均有 | 多轮 | 机器人技术\u002F数学\u002F代码\u002FQA\u002FVQA | 全部（规则\u002F模型\u002F外部） | 是 |\n| [siiRL](https:\u002F\u002Fgithub.com\u002Fsii-research\u002FsiiRL) | PPO\u002FGRPO\u002FCPGD\u002FMARFT | 多智能体 | 两者均有 | 多轮 | LLM\u002FVLM\u002FLLM-MAS 后训练 | 模型\u002F规则 | 计划中 |\n| [slime](https:\u002F\u002Fgithub.com\u002FTHUDM\u002Fslime) | GRPO\u002FGSPO\u002FREINFORCE++ | 单智能体 | 两者均有 | 双向 | 数学\u002F代码 | 外部验证器 | 是 |\n| [agent-lightning](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fagent-lightning) | PPO\u002F自定义\u002F自动提示优化 | 多智能体 | 结果奖励 | 多轮 | 计算器\u002FSQL | 模型\u002F外部\u002F规则 | 是 |\n| [AReaL](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FAReaL) | PPO | 单\u002F多智能体 | 结果奖励 | 双向 | 数学\u002F代码 | 外部 | 是 |\n| [ROLL](https:\u002F\u002Fgithub.com\u002Falibaba\u002FROLL) | PPO\u002FGRPO\u002FReinforce++\u002FTOPR\u002FRAFT++ | 多智能体 | 两者均有 | 多轮 | 数学\u002FQA\u002F代码\u002F对齐 | 全部 | 是 |\n| [MARTI](https:\u002F\u002Fgithub.com\u002FTsinghuaC3I\u002FMARTI) | PPO\u002FGRPO\u002FREINFORCE++\u002FTTRL | 多智能体 | 两者均有 | 多轮 | 数学 | 全部 | 是 |\n| [RL2](https:\u002F\u002Fgithub.com\u002FChenmienTan\u002FRL2) | Dr. GRPO\u002FPPO\u002FDPO | 单智能体 | 两者均有 | 双向 | QA\u002F对话 | 规则\u002F模型\u002F外部 | 是 |\n| [verifiers](https:\u002F\u002Fgithub.com\u002Fwillccbb\u002Fverifiers) | GRPO | 多智能体 | 结果奖励 | 双向 | 推理\u002F数学\u002F代码 | 全部 | 代码 |\n| [oat](https:\u002F\u002Fgithub.com\u002Fsail-sg\u002Foat) | PPO\u002FGRPO | 单智能体 | 结果奖励 | 多轮 | 数学\u002F对齐 | 外部 | 否 |\n| [veRL](https:\u002F\u002Fgithub.com\u002Fvolcengine\u002Fverl) | PPO\u002FGRPO | 单智能体 | 结果奖励 | 双向 | 数学\u002FQA\u002F推理\u002F搜索 | 全部 | 是 |\n| [OpenRLHF](https:\u002F\u002Fgithub.com\u002FOpenRLHF\u002FOpenRLHF) | PPO\u002FREINFORCE++\u002FGRPO\u002FDPO\u002FIPO\u002FKTO\u002FRLOO | 多智能体 | 两者均有 | 双向 | 对话\u002F聊天\u002F补全 | 规则\u002F模型\u002F外部 | 是 |\n| [trl](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftrl) | PPO\u002FGRPO\u002FDPO | 单智能体 | 两者均有 | 单轮 | QA | 自定义 | 否 |\n\n\u003C\u002Fdetails>\n\n\n\n## 💪 通用\u002F多任务\n\n| GitHub 仓库 | 🌟 星数 | 发布日期 | 机构 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [MetaClaw](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FMetaClaw) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Faiming-lab\u002FMetaClaw?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | 北卡罗来纳大学教堂山分校（AIMING 实验室） | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2603.17187) | 自定义 |\n| [SkillRL](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FSkillRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Faiming-lab\u002FSkillRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | 北卡罗来纳大学教堂山分校（AIMING 实验室） | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.08234) | 自定义 |\n| [LLM-in-Sandbox](https:\u002F\u002Fgithub.com\u002Fllm-in-sandbox\u002Fllm-in-sandbox) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fllm-in-sandbox\u002Fllm-in-sandbox?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | 中国人民大学\u002F微软亚洲研究院\u002F清华大学 | [论文](https:\u002F\u002Fhuggingface.co\u002Fpapers\u002F2601.16206) | rllm（结合 veRL） |\n| [youtu-agent](https:\u002F\u002Fgithub.com\u002FTencentCloudADP\u002Fyoutu-agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTencentCloudADP\u002Fyoutu-agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.12 | 腾讯优图实验室 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2512.24615) | 自定义 |\n| [DEPO](https:\u002F\u002Fgithub.com\u002FOpenCausaLab\u002FDEPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenCausaLab\u002FDEPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | 香港科技大学\u002F上海交通大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.15392) | LLaMA-Factory |\n| [SPEAR](https:\u002F\u002Fgithub.com\u002FTencentYoutuResearch\u002FSPEAR) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTencentYoutuResearch\u002FSPEAR?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | 腾讯优图实验室 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.22601) | veRL\u002Fverl-agent |\n| [DeepAgent](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FDeepAgent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUC-NLPIR\u002FDeepAgent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | 中国人民大学\u002F小红书 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.21618) | 自定义 |\n| [AgentRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FAgentRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002FAgentRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | 清华大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.04206) | veRL |\n| [AgentGym-RL](https:\u002F\u002Fgithub.com\u002FWooooDyy\u002FAgentGym-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FWooooDyy\u002FAgentGym-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | 复旦大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.08755) | veRL |\n| [Agent_Foundation_Models](https:\u002F\u002Fgithub.com\u002FOPPO-PersonalAI\u002FAgent_Foundation_Models) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOPPO-PersonalAI\u002FAgent_Foundation_Models?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | OPPO 个人 AI 实验室 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.13167) | veRL |\n| [Trinity-RFT](https:\u002F\u002Fgithub.com\u002Fmodelscope\u002FTrinity-RFT) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmodelscope\u002FTrinity-RFT?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 阿里巴巴 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.17826) | veRL |\n| [SPA-RL-Agent](https:\u002F\u002Fgithub.com\u002FWangHanLinHenry\u002FSPA-RL-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FWangHanLinHenry\u002FSPA-RL-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 香港理工大学 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.20732) | TRL |\n| [verl-agent](https:\u002F\u002Fgithub.com\u002FlangfengQ\u002Fverl-agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FlangfengQ\u002Fverl-agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 新加坡南洋理工大学\u002FSkywork | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.10978) | veRL |\n| [VAGEN](https:\u002F\u002Fgithub.com\u002FRAGEN-AI\u002FVAGEN) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRAGEN-AI\u002FVAGEN?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | RAGEN-AI | [论文](https:\u002F\u002Fwww.notion.so\u002FVAGEN-Training-VLM-Agents-with-Multi-Turn-Reinforcement-Learning-1bfde13afb6e80b792f6d80c7c2fcad0) | veRL |\n| [ART](https:\u002F\u002Fgithub.com\u002FOpenPipe\u002FART) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenPipe\u002FART?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | OpenPipe | [论文](https:\u002F\u002Fgithub.com\u002FOpenPipe\u002FART#-citation) | TRL |\n| [OpenManus-RL](https:\u002F\u002Fgithub.com\u002FOpenManus\u002FOpenManus-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenManus\u002FOpenManus-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | 伊利诺伊大学厄巴纳-香槟分校\u002FMetaGPT | —— | 自定义 |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单智能体\u002F多智能体 | 结果奖励\u002F过程奖励 | 单轮\u002F多轮 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [MetaClaw](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FMetaClaw) | GRPO (LoRA) | 单智能体 | 过程奖励 | 多轮 | 通用代理任务 | 模型（PRM）奖励 | 是（技能增强） |\n| [SkillRL](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FSkillRL) | GRPO | 单智能体 | 结果奖励 | 多轮 | ALFWorld\u002FWebShop\u002F搜索 | 规则奖励 | 是（网页搜索、动作） |\n| [LLM-in-Sandbox](https:\u002F\u002Fgithub.com\u002Fllm-in-sandbox\u002Fllm-in-sandbox) | GRPO++ | 单智能体 | 结果奖励 | 多轮 | 数学\u002F物理\u002F化学\u002F生物医学\u002F长上下文\u002FIF\u002FSWE | 规则奖励 | 是（代码沙箱，含终端、文件、互联网） |\n| [youtu-agent](https:\u002F\u002Fgithub.com\u002FTencentCloudADP\u002Fyoutu-agent) | 无训练 GRPO | 单智能体 | 结果奖励 | 多轮 | 深度研究\u002F数据分析\u002F工具使用 | 模型\u002F外部奖励 | 是（网页搜索、代码、文件） |\n| [DEPO](https:\u002F\u002Fgithub.com\u002FOpenCausaLab\u002FDEPO) | KTO + 效率损失 | 单智能体 | 结果与过程奖励 | 多轮 | BabyAI\u002FWebShop 等代理任务 | 规则奖励 | 是 |\n| [SPEAR](https:\u002F\u002Fgithub.com\u002FTencentYoutuResearch\u002FSPEAR) | GRPO\u002FGiGPO + SIL | 单智能体 | 结果与过程奖励 | 多轮 | 数学\u002F代理任务 | 规则\u002F外部奖励 | 是（搜索、沙箱、浏览器） |\n| [DeepAgent](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FDeepAgent) | ToolPO | 单智能体 | 结果奖励 | 多轮 | ToolBench\u002FALFWorld\u002FWebShop\u002FGAIA\u002FHLE | 模型奖励 | 是（16,000+ RapidAPIs） |\n| [AgentRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FAgentRL) | GRPO\u002FREINFORCE++\u002FRLOO\u002FReMax\u002FGAE | 单智能体 | 结果奖励 | 多轮 | 代理任务 | 外部奖励 | 是 |\n| [AgentGym-RL](https:\u002F\u002Fgithub.com\u002FWooooDyy\u002FAgentGym-RL) | PPO\u002FGRPO\u002FRLOO\u002FREINFORCE++ | 单智能体 | 结果奖励 | 多轮 | 网页\u002F搜索\u002F游戏\u002F具身智能\u002F科学 | 规则\u002F模型\u002F外部奖励 | 是（网页、搜索、环境 API） |\n| [Agent_Foundation_Models](https:\u002F\u002Fgithub.com\u002FOPPO-PersonalAI\u002FAgent_Foundation_Models) | DAPO\u002FPPO | 单智能体 | 结果奖励 | 单轮 | QA\u002F代码\u002F数学 | 规则\u002F外部奖励 | 是 |\n| [Trinity-RFT](https:\u002F\u002Fgithub.com\u002Fmodelscope\u002FTrinity-RFT) | PPO\u002FGRPO | 单智能体 | 结果奖励 | 结果与过程奖励 | 数学\u002F文本游戏\u002F网页 | 所有奖励类型 | 是 |\n| [SPA-RL-Agent](https:\u002F\u002Fgithub.com\u002FWangHanLinHenry\u002FSPA-RL-Agent) | PPO | 单智能体 | 过程奖励 | 多轮 | 导航\u002F网页\u002F文本游戏 | 模型奖励 | 否 |\n| [verl-agent](https:\u002F\u002Fgithub.com\u002FlangfengQ\u002Fverl-agent) | PPO\u002FGRPO\u002FGiGPO\u002FDAPO\u002FRLOO\u002FREINFORCE++ | 多智能体 | 结果与过程奖励 | 多轮 | 手机使用\u002F数学\u002F代码\u002F网页\u002F文本游戏 | 所有奖励类型 | 是 |\n| [VAGEN](https:\u002F\u002Fgithub.com\u002FRAGEN-AI\u002FVAGEN) | PPO\u002FGRPO | 单智能体 | 结果与过程奖励 | 多轮 | 文本游戏\u002F导航 | 所有奖励类型 | 是 |\n| [ART](https:\u002F\u002Fgithub.com\u002FOpenPipe\u002FART) | GRPO | 多智能体 | 结果与过程奖励 | 多轮 | 文本游戏 | 所有奖励类型 | 是 |\n| [OpenManus-RL](https:\u002F\u002Fgithub.com\u002FOpenManus\u002FOpenManus-RL) | PPO\u002FDPO\u002FGRPO | 多智能体 | 结果奖励 | 多轮 | 文本游戏 | 所有奖励类型 | 是 |\n\n\u003C\u002Fdetails>\n\n\n\n## 🔍 搜索与 RAG 代理\n\n| GitHub 仓库 | 🌟 星数 | 发布日期 | 机构 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [ProRAG](https:\u002F\u002Fgithub.com\u002Flilinwz\u002FProRAG) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flilinwz\u002FProRAG?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | 人大 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.21912) | 自定义 |\n| [MemSearcher](https:\u002F\u002Fgithub.com\u002Ficip-cas\u002FMemSearcher) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ficip-cas\u002FMemSearcher?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | 中科院 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.02805) | 自定义 |\n| [ReSeek](https:\u002F\u002Fgithub.com\u002FTencentBAC\u002FReSeek) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTencentBAC\u002FReSeek?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | 腾讯PCG BAC\u002F清华大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.00568) | veRL |\n| [AutoGraph-R1](https:\u002F\u002Fgithub.com\u002FHKUST-KnowComp\u002FAutoGraph-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FHKUST-KnowComp\u002FAutoGraph-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | 香港科技大学KnowComp | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.15339) | 自定义 |\n| [Tree-GRPO](https:\u002F\u002Fgithub.com\u002FAMAP-ML\u002FTree-GRPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAMAP-ML\u002FTree-GRPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | 高德地图 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.21240) | veRL |\n| [ASearcher](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FASearcher) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FinclusionAI\u002FASearcher?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | 蚂蚁集团研究强化学习实验室 \u003Cbr> 清华大学 & UW | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.07976) | RealHF\u002FAReaL |\n| [Graph-R1](https:\u002F\u002Fgithub.com\u002FLHRLAB\u002FGraph-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FLHRLAB\u002FGraph-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | 北邮\u002FNTU\u002FNUS | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.21892) | veRL |\n| [Kimi-Researcher](https:\u002F\u002Fgithub.com\u002Fmoonshotai\u002FKimi-Researcher) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmoonshotai\u002FKimi-Researcher?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Moonshot AI | [博客](https:\u002F\u002Fmoonshotai.github.io\u002FKimi-Researcher\u002F) | 自定义 |\n| [R-Search](https:\u002F\u002Fgithub.com\u002FQingFei1\u002FR-Search) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FQingFei1\u002FR-Search?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 个人 | -- | veRL |\n| [R1-Searcher-plus](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher-plus) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUCAIBox\u002FR1-Searcher-plus?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 人大 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.17005) | 自定义 |\n| [StepSearch](https:\u002F\u002Fgithub.com\u002FZillwang\u002FStepSearch) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FZillwang\u002FStepSearch?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 商汤科技 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.15107) | veRL |\n| [AutoRefine](https:\u002F\u002Fgithub.com\u002Fsyr-cn\u002FAutoRefine) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsyr-cn\u002FAutoRefine?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 中国科学技术大学 | [论文](https:\u002F\u002Fwww.arxiv.org\u002Fpdf\u002F2505.11277) | veRL |\n| [ZeroSearch](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FZeroSearch) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FZeroSearch?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 阿里巴巴 |[论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.04588) | veRL |\n| [ReasonRAG](https:\u002F\u002Fgithub.com\u002Fwlzhang2020\u002FReasonRAG) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fwlzhang2020\u002FReasonRAG?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 香港城市大学 \u002F 华为 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.14069) | 自定义 |\n| [Agentic-RAG-R1](https:\u002F\u002Fgithub.com\u002Fjiangxinke\u002FAgentic-RAG-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fjiangxinke\u002FAgentic-RAG-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.12 | 北京大学 | -- | 自定义 |\n| [WebThinker](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FWebThinker) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUC-NLPIR\u002FWebThinker?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | 人大 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.21776) | 自定义 |\n| [DeepResearcher](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FDeepResearcher) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGAIR-NLP\u002FDeepResearcher?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | 上海交通大学 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.03160) | veRL |\n| [Search-R1](https:\u002F\u002Fgithub.com\u002FPeterGriffinJin\u002FSearch-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FPeterGriffinJin\u002FSearch-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | UIUC\u002FGoogle | [论文1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.09516), [论文2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.15117) | veRL |\n| [R1-Searcher](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUCAIBox\u002FR1-Searcher?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | 人大 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.05592) | OpenRLHF |\n| [C-3PO](https:\u002F\u002Fgithub.com\u002FChen-GX\u002FC-3PO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FChen-GX\u002FC-3PO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 阿里巴巴 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2502.06205) | OpenRLHF |\n| [DeepRetrieval](https:\u002F\u002Fgithub.com\u002Fpat-jj\u002FDeepRetrieval) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpat-jj\u002FDeepRetrieval?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | UIUC | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.00223) | veRL |\n| [SSRL](https:\u002F\u002Fgithub.com\u002FTsinghuaC3I\u002FSSRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTsinghuaC3I\u002FSSRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | 清华大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.10874) | 自定义 |\n| [Research-Venus](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fantgroup\u002FResearch-Venus?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | 蚂蚁集团 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.12800) | 自定义 |\n| [DeepResearch](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | 阿里巴巴\u002F通义实验室 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.24701) | 自定义 |\n| [DeepDive](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FDeepDive) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002FDeepDive?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | 清华大学\u002FTHUDM | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.10446) | 自定义 |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单\u002F多智能体 | 结果\u002F过程奖励 | 单\u002F多轮 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [ProRAG](https:\u002F\u002Fgithub.com\u002Flilinwz\u002FProRAG) | GRPO + DGA（双粒度优势） | 单 | 两者 | 多 | 多跳 RAG | 模型（通过 MCTS 的 PRM） | 是（检索） |\n| [MemSearcher](https:\u002F\u002Fgithub.com\u002Ficip-cas\u002FMemSearcher) | 多上下文 GRPO | 单 | 结果 | 多 | 搜索\u002FQA + 记忆 | 规则\u002F模型 | 是（网络搜索 + 记忆） |\n| [ReSeek](https:\u002F\u002Fgithub.com\u002FTencentBAC\u002FReSeek) | GRPO\u002FPPO | 单 | 两者 | 多 | QA\u002F搜索 | 规则 | 搜索\u002FJUDGE |\n| [AutoGraph-R1](https:\u002F\u002Fgithub.com\u002FHKUST-KnowComp\u002FAutoGraph-R1) | GRPO（通过 VeRL） | 单 | 结果 | 多 | 面向 QA 的知识图谱构建 | 规则 | 是（图谱检索） |\n| [Tree-GRPO](https:\u002F\u002Fgithub.com\u002FAMAP-ML\u002FTree-GRPO) | GRPO\u002FTree-GRPO | 单 | 结果 | 多 | 搜索 | 规则 | 搜索 |\n| [ASearcher](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FASearcher) | PPO\u002FGRPO + 解耦 PPO | 单 | 结果 | 多 | 数学\u002F代码\u002F搜索问答 | 外部\u002F规则 | 是 |\n| [Graph-R1](https:\u002F\u002Fgithub.com\u002FLHRLAB\u002FGraph-R1) | GRPO\u002FREINFORCE++\u002FPPO | 单 | 结果 | 多 | 知识图谱问答 | 规则（EM\u002FF1） | 是（图谱检索） |\n| [Kimi-Researcher](https:\u002F\u002Fgithub.com\u002Fmoonshotai\u002FKimi-Researcher) | REINFORCE | 单 | 结果 | 多 | 研究 | 结果 | 搜索、浏览、编码 |\n| [R-Search](https:\u002F\u002Fgithub.com\u002FQingFei1\u002FR-Search) | PPO\u002FGRPO | 单 | 两者 | 多 | QA\u002F搜索 | 全部 | 是 |\n| [R1-Searcher-plus](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher-plus) | 自定义 | 单 | 结果 | 多 | 搜索 | 模型 | 搜索 |\n| [StepSearch](https:\u002F\u002Fgithub.com\u002FZillwang\u002FStepSearch) | PPO | 单 | 迁移 | 多 | QA | 模型 | 搜索 |\n| [AutoRefine](https:\u002F\u002Fgithub.com\u002Fsyr-cn\u002FAutoRefine) | PPO\u002FGRPO | 多 | 两者 | 多 | RAG QA | 规则 | 搜索 |\n| [ZeroSearch](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FZeroSearch) | PPO\u002FGRPO\u002FREINFORCE | 单 | 结果 | 多 | QA\u002F搜索 | 规则 | 是 |\n| [ReasonRAG](https:\u002F\u002Fgithub.com\u002Fwlzhang2020\u002FReasonRAG) | DPO + 基于 MCTS 的 PRM | 单 | 过程 | 多 | 多跳 QA | 模型（PRM） | 是（维基百科搜索） |\n| [Agentic-RAG-R1](https:\u002F\u002Fgithub.com\u002Fjiangxinke\u002FAgentic-RAG-R1) | GRPO | 单 | 结果 | 多 | 知识密集型 QA | 规则\u002F模型 | 是（维基百科\u002F文档搜索） |\n| [WebThinker](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FWebThinker) | DPO | 单 | 结果 | 多 | 推理\u002FQA\u002F研究 | 模型\u002F外部 | 网络浏览 |\n| [DeepResearcher](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FDeepResearcher) | PPO\u002FGRPO | 多 | 结果 | 多 | 研究 | 全部 | 是 |\n| [Search-R1](https:\u002F\u002Fgithub.com\u002FPeterGriffinJin\u002FSearch-R1) | PPO\u002FGRPO | 单 | 结果 | 多 | 搜索 | 全部 | 搜索 |\n| [R1-Searcher](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher) | PPO\u002FDPO | 单 | 两者 | 多 | 搜索 | 全部 | 是 |\n| [C-3PO](https:\u002F\u002Fgithub.com\u002FChen-GX\u002FC-3PO) | PPO | 多 | 结果 | 多 | 搜索 | 模型 | 是 |\n| [DeepRetrieval](https:\u002F\u002Fgithub.com\u002Fpat-jj\u002FDeepRetrieval) | GRPO | 单 | 结果 | 多 | 查询生成\u002F信息检索 | 规则 | 是（搜索） |\n| [SSRL](https:\u002F\u002Fgithub.com\u002FTsinghuaC3I\u002FSSRL) | GRPO | 单 | 结果 | 多 | 自我搜索 | 规则 | 是（自我搜索） |\n| [Research-Venus](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus) | GRPO | 单 | 两者 | 多 | 深度研究 | 模型（原子思维） | 是（搜索） |\n| [DeepResearch](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | 基于强化学习 | 单 | 结果 | 多 | 深度研究 | 模型 | 是（搜索、浏览） |\n| [DeepDive](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FDeepDive) | GRPO | 单 | 结果 | 多 | 知识图谱增强的搜索 | 规则 | 是（知识图谱 + 搜索） |\n\n\u003C\u002Fdetails>\n\n\n\n## 🌐 网络与 GUI 代理\n\n| GitHub 仓库 | 🌟 星数 | 日期 | 组织 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [MobileAgent](https:\u002F\u002Fgithub.com\u002FX-PLUG\u002FMobileAgent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FX-PLUG\u002FMobileAgent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | X-PLUG (通义千问) | [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.11543) | veRL |\n| [InfiGUI-G1](https:\u002F\u002Fgithub.com\u002FInfiXAI\u002FInfiGUI-G1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FInfiXAI\u002FInfiGUI-G1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | InfiX AI | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.05731) | veRL |\n| [UI-AGILE](https:\u002F\u002Fgithub.com\u002FKDEGroup\u002FUI-AGILE) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FKDEGroup\u002FUI-AGILE?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | 厦门大学 | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.22025) | 自定义 |\n| [gui-rcpo](https:\u002F\u002Fgithub.com\u002Fzju-real\u002Fgui-rcpo) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fzju-real\u002Fgui-rcpo?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | 浙江大学 | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.05615) | 自定义 |\n| [Grounding-R1](https:\u002F\u002Fgithub.com\u002FYan98\u002FGrounding-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FYan98\u002FGrounding-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | Salesforce | [blog](https:\u002F\u002Fhuggingface.co\u002Fblog\u002FHelloKKMe\u002Fgrounding-r1) | trl |\n| [AgentCPM-GUI](https:\u002F\u002Fgithub.com\u002FOpenBMB\u002FAgentCPM-GUI) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenBMB\u002FAgentCPM-GUI?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | OpenBMB\u002F清华大学\u002F中国人民大学 | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.01391) | Huggingface |\n| [TTI](https:\u002F\u002Fgithub.com\u002Ftest-time-interaction\u002FTTI) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftest-time-interaction\u002FTTI?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 卡内基梅隆大学 | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.07976) | 自定义 |\n| [SE-GUI](https:\u002F\u002Fgithub.com\u002FYXB-NKU\u002FSE-GUI) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FYXB-NKU\u002FSE-GUI?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 南开大学\u002Fvivo | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.12370) | trl |\n| [ARPO](https:\u002F\u002Fgithub.com\u002Fdvlab-research\u002FARPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdvlab-research\u002FARPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 香港中文大学\u002F香港科技大学 | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.16282) | veRL |\n| [GUI-G1](https:\u002F\u002Fgithub.com\u002FYuqi-Zhou\u002FGUI-G1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FYuqi-Zhou\u002FGUI-G1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 中国人民大学 | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.15810) | TRL |\n| [WebAgent-R1](https:\u002F\u002Fgithub.com\u002Fweizhepei\u002FWebAgent-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fweizhepei\u002FWebAgent-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 亚马逊\u002F弗吉尼亚大学 | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.16421) | 自定义 |\n| [GUI-R1](https:\u002F\u002Fgithub.com\u002Fritzz-ai\u002FGUI-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fritzz-ai\u002FGUI-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | 中国科学院\u002FNUS | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.10458) | veRL |\n| [UI-R1](https:\u002F\u002Fgithub.com\u002Flll6gg\u002FUI-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flll6gg\u002FUI-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | vivo\u002F香港中文大学 | [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.21620) | TRL |\n| [CollabUIAgents](https:\u002F\u002Fgithub.com\u002FTHUNLP-MT\u002FCollabUIAgents) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUNLP-MT\u002FCollabUIAgents?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 清华大学\u002F阿里巴巴\u002F香港科技大学 | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.14496) | 自定义 |\n| [WebAgent](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FWebAgent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FWebAgent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | 阿里巴巴 | [paper1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2501.07572), [paper2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.22648) | LLaMA-Factory |\n| [UI-TARS](https:\u002F\u002Fgithub.com\u002Fbytedance\u002FUI-TARS) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbytedance\u002FUI-TARS?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | 字节跳动 Seed | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.02544) | 自定义 |\n| [DigiQ](https:\u002F\u002Fgithub.com\u002FDigiRL-agent\u002Fdigiq) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FDigiRL-agent\u002Fdigiq?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 加州大学伯克利分校\u002F卡内基梅隆大学\u002F亚马逊 | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.15760) | 自定义 |\n| [ZeroGUI](https:\u002F\u002Fgithub.com\u002FOpenGVLab\u002FZeroGUI) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOpenGVLab\u002FZeroGUI?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 上海人工智能实验室 | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.23762) | 自定义 |\n| [InfiGUI-R1](https:\u002F\u002Fgithub.com\u002FReallm-Labs\u002FInfiGUI-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FReallm-Labs\u002FInfiGUI-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | 浙江大学 | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.14239) | 自定义 |\n| [GUI-Agent-RL](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FGUI-Agent-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FGUI-Agent-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 微软 | [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.18906) | 自定义 |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单智能体\u002F多智能体 | 结果奖励\u002F过程奖励 | 单轮\u002F多轮 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [MobileAgent](https:\u002F\u002Fgithub.com\u002FX-PLUG\u002FMobileAgent) | 半在线强化学习 | 单智能体 | 同时使用 | 多轮 | 移动 GUI\u002F自动化 | 规则 | 是 |\n| [InfiGUI-G1](https:\u002F\u002Fgithub.com\u002FInfiXAI\u002FInfiGUI-G1) | AEPO | 单智能体 | 结果奖励 | 单轮 | GUI\u002F接地 | 规则 | 否 |\n| [UI-AGILE](https:\u002F\u002Fgithub.com\u002FKDEGroup\u002FUI-AGILE) | GRPO | 单智能体 | 结果奖励 | 单轮 | GUI 接地 | 规则（连续） | 否 |\n| [gui-rcpo](https:\u002F\u002Fgithub.com\u002Fzju-real\u002Fgui-rcpo) | RCPO | 单智能体 | 结果奖励 | 单轮 | GUI 接地 | 规则（自监督） | 否 |\n| [Grounding-R1](https:\u002F\u002Fgithub.com\u002FYan98\u002FGrounding-R1) | GRPO | 单智能体 | 结果奖励 | 多轮 | GUI 接地 | 模型 | 是 |\n| [AgentCPM-GUI](https:\u002F\u002Fgithub.com\u002FOpenBMB\u002FAgentCPM-GUI) | GRPO | 单智能体 | 结果奖励 | 多轮 | 移动 GUI | 模型 | 是 |\n| [TTI](https:\u002F\u002Fgithub.com\u002Ftest-time-interaction\u002FTTI) | REINFORCE\u002FBC | 单智能体 | 结果奖励 | 多轮 | 网页 | 外部工具 | 网页浏览 |\n| [SE-GUI](https:\u002F\u002Fgithub.com\u002FYXB-NKU\u002FSE-GUI) | GRPO | 单智能体 | 同时使用 | 单轮 | GUI 接地 | 规则 | 是 |\n| [ARPO](https:\u002F\u002Fgithub.com\u002Fdvlab-research\u002FARPO) | GRPO | 单智能体 | 结果奖励 | 多轮 | GUI | 外部工具 | 计算机操作 |\n| [GUI-G1](https:\u002F\u002Fgithub.com\u002FYuqi-Zhou\u002FGUI-G1) | GRPO | 单智能体 | 结果奖励 | 单轮 | GUI | 规则\u002F外部工具 | 否 |\n| [WebAgent-R1](https:\u002F\u002Fgithub.com\u002Fweizhepei\u002FWebAgent-R1) | M-GRPO | 单智能体 | 结果奖励 | 多轮 | 网页导航（WebArena-Lite） | 规则（任务成功） | 是（网页浏览） |\n| [GUI-R1](https:\u002F\u002Fgithub.com\u002Fritzz-ai\u002FGUI-R1) | GRPO | 单智能体 | 结果奖励 | 多轮 | GUI | 规则 | 否 |\n| [UI-R1](https:\u002F\u002Fgithub.com\u002Flll6gg\u002FUI-R1) | GRPO | 单智能体 | 过程奖励和结果奖励 | 同时使用 | GUI | 规则 | 计算机\u002F手机使用 |\n| [CollabUIAgents](https:\u002F\u002Fgithub.com\u002FTHUNLP-MT\u002FCollabUIAgents) | DPO（信用再分配） | 多智能体 | 过程奖励 | 多轮 | GUI（移动 + 网页） | 模型（LLM） | 是（GUI 交互） |\n| [WebAgent](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FWebAgent) | DAPO | 多智能体 | 过程奖励 | 多轮 | 网页 | 模型 | 是 |\n| [UI-TARS](https:\u002F\u002Fgithub.com\u002Fbytedance\u002FUI-TARS) | 多轮强化学习 | 单智能体 | 同时使用 | 多轮 | GUI（跨平台） | 模型 | 是（GUI 操作） |\n| [DigiQ](https:\u002F\u002Fgithub.com\u002FDigiRL-agent\u002Fdigiq) | 基于价值的离线强化学习 | 单智能体 | 结果奖励 | 多轮 | 安卓设备控制 | 模型（Q 函数） | 是 |\n| [ZeroGUI](https:\u002F\u002Fgithub.com\u002FOpenGVLab\u002FZeroGUI) | 在线强化学习 | 单智能体 | 结果奖励 | 多轮 | GUI 智能体 | 规则 | 是（GUI 操作） |\n| [InfiGUI-R1](https:\u002F\u002Fgithub.com\u002FReallm-Labs\u002FInfiGUI-R1) | 强化学习 + 子目标引导 | 单智能体 | 同时使用 | 多轮 | GUI 推理 | 规则 | 是 |\n| [GUI-Agent-RL](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FGUI-Agent-RL) | 基于价值的强化学习（VEM） | 单智能体 | 结果奖励 | 多轮 | GUI（网上购物） | 模型 | 是 |\n\n\u003C\u002Fdetails>\n\n## 🔨 工具使用智能体\n\n\n| GitHub 仓库 | 🌟 星数 | 发布日期 | 组织 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [MATPO](https:\u002F\u002Fgithub.com\u002Fmzf666\u002FMATPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmzf666\u002FMATPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | MiroMind AI | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.04678) | 自定义 |\n| [MiroRL](https:\u002F\u002Fgithub.com\u002FMiroMindAI\u002FMiroRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMiroMindAI\u002FMiroRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | MiroMindAI | [HF 仓库](https:\u002F\u002Fhuggingface.co\u002Fmiromind-ai) | veRL |\n| [verl-tool](https:\u002F\u002Fgithub.com\u002FTIGER-AI-Lab\u002Fverl-tool) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTIGER-AI-Lab\u002Fverl-tool?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | TIGER-Lab | [X](https:\u002F\u002Fx.com\u002FDongfuJiang\u002Fstatus\u002F1929198238017720379) | veRL |\n| [Multi-Turn-RL-Agent](https:\u002F\u002Fgithub.com\u002FSiliangZeng\u002FMulti-Turn-RL-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSiliangZeng\u002FMulti-Turn-RL-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 明尼苏达大学 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.11821) | 自定义 |\n| [Tool-N1](https:\u002F\u002Fgithub.com\u002FNVlabs\u002FTool-N1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNVlabs\u002FTool-N1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | NVIDIA | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.00024) | veRL |\n| [Tool-Star](https:\u002F\u002Fgithub.com\u002Fdongguanting\u002FTool-Star) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdongguanting\u002FTool-Star?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 人大 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.16410) | LLaMA-Factory |\n| [RL-Factory](https:\u002F\u002Fgithub.com\u002FSimple-Efficient\u002FRL-Factory) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSimple-Efficient\u002FRL-Factory?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | Simple-Efficient | [模型](https:\u002F\u002Fhuggingface.co\u002FSimple-Efficient\u002FRLFactory-Qwen3-8B-GRPO) | veRL |\n| [ReTool](https:\u002F\u002Fgithub.com\u002FReTool-RL\u002FReTool) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FReTool-RL\u002FReTool?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | 字节跳动 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.11536) | veRL |\n| [AWorld](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FAWorld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FinclusionAI\u002FAWorld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | 蚂蚁集团 (inclusionAI) | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.20404) | veRL |\n| [Agent-R1](https:\u002F\u002Fgithub.com\u002F0russwest0\u002FAgent-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002F0russwest0\u002FAgent-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | 中国科学技术大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.14460) | veRL |\n| [ReCall](https:\u002F\u002Fgithub.com\u002FAgent-RL\u002FReCall) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAgent-RL\u002FReCall?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | 百川 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.19470) | veRL |\n| [ToolRL](https:\u002F\u002Fgithub.com\u002Fqiancheng0\u002FToolRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fqiancheng0\u002FToolRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | UIUC | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.13958) | veRL |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单\u002F多智能体 | 结果\u002F过程奖励 | 单\u002F多回合 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [MATPO](https:\u002F\u002Fgithub.com\u002Fmzf666\u002FMATPO) | GRPO (多智能体) | 多 | 结果 | 多 | 工具使用\u002F搜索 | 规则 | 是（MCP：Serper，网页抓取） |\n| [MiroRL](https:\u002F\u002Fgithub.com\u002FMiroMindAI\u002FMiroRL) | GRPO | 单 | 两者 | 多 | 推理\u002F规划\u002F工具使用 | 基于规则 | MCP |\n| [verl-tool](https:\u002F\u002Fgithub.com\u002FTIGER-AI-Lab\u002Fverl-tool) | PPO\u002FGRPO | 单 | 两者 | 两者 | 数学\u002F代码 | 规则\u002F外部 | 是 |\n| [Multi-Turn-RL-Agent](https:\u002F\u002Fgithub.com\u002FSiliangZeng\u002FMulti-Turn-RL-Agent) | GRPO | 单 | 两者 | 多 | 工具使用\u002F数学 | 规则\u002F外部 | 是 |\n| [Tool-N1](https:\u002F\u002Fgithub.com\u002FNVlabs\u002FTool-N1) | PPO | 单 | 结果 | 多 | 数学\u002F对话 | 全部 | 是 |\n| [Tool-Star](https:\u002F\u002Fgithub.com\u002Fdongguanting\u002FTool-Star) | PPO\u002FDPO\u002FORPO\u002FSimPO\u002FKTO | 单 | 结果 | 多 | 多模态\u002F工具使用\u002F对话 | 模型\u002F外部 | 是 |\n| [RL-Factory](https:\u002F\u002Fgithub.com\u002FSimple-Efficient\u002FRL-Factory) | GRPO | 多 | 两者 | 多 | 工具使用\u002FNL2SQL | 全部 | MCP |\n| [ReTool](https:\u002F\u002Fgithub.com\u002FReTool-RL\u002FReTool) | PPO | 单 | 结果 | 多 | 数学 | 外部 | 代码 |\n| [AWorld](https:\u002F\u002Fgithub.com\u002FinclusionAI\u002FAWorld) | GRPO | 两者 | 结果 | 多 | 搜索\u002F网络\u002F代码 | 外部\u002F规则 | 是 |\n| [Agent-R1](https:\u002F\u002Fgithub.com\u002F0russwest0\u002FAgent-R1) | PPO\u002FGRPO | 单 | 两者 | 多 | 工具使用\u002FQA | 模型 | 是 |\n| [ReCall](https:\u002F\u002Fgithub.com\u002FAgent-RL\u002FReCall) | PPO\u002FGRPO\u002FRLOO\u002FREINFORCE++\u002FReMax | 单 | 结果 | 多 | 工具使用\u002F数学\u002FQA | 全部 | 是 |\n| [ToolRL](https:\u002F\u002Fgithub.com\u002Fqiancheng0\u002FToolRL) | GRPO\u002FPPO | 单 | 结果 | 多 | 工具学习 | 规则\u002F外部 | 是 |\n\n\u003C\u002Fdetails>\n\n## 💻 代码与软件工程智能体\n\n| GitHub 仓库 | 🌟 星数 | 发布日期 | 机构 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [CUDA-Agent](https:\u002F\u002Fgithub.com\u002FBytedTsinghua-SIA\u002FCUDA-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FBytedTsinghua-SIA\u002FCUDA-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | 字节跳动\u002F清华大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.24286) | 自定义 |\n| [LLM-in-Sandbox](https:\u002F\u002Fgithub.com\u002Fllm-in-sandbox\u002Fllm-in-sandbox) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fllm-in-sandbox\u002Fllm-in-sandbox?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | 人大\u002FMSRA\u002F清华 | [论文](https:\u002F\u002Fhuggingface.co\u002Fpapers\u002F2601.16206) | rllm (w\u002F veRL) |\n| [PPP-Agent](https:\u002F\u002Fgithub.com\u002Fsunnweiwei\u002FPPP-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsunnweiwei\u002FPPP-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | 卡内基梅隆大学\u002FOpenHands | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.02208) | veRL |\n| [RepoDeepSearch](https:\u002F\u002Fgithub.com\u002FMizersy\u002FRepoDeepSearch) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMizersy\u002FRepoDeepSearch?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | 北大、字节跳动、北理工 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.03012) | veRL |\n| [CUDA-L1](https:\u002F\u002Fgithub.com\u002Fdeepreinforce-ai\u002FCUDA-L1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdeepreinforce-ai\u002FCUDA-L1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | DeepReinforce AI | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.14111) | 自定义 |\n| [MedAgentGym](https:\u002F\u002Fgithub.com\u002Fwshi83\u002FMedAgentGym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fwshi83\u002FMedAgentGym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 埃默里大学\u002F佐治亚理工学院 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.04405) | Hugginface |\n| [CURE](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FCURE) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGen-Verse\u002FCURE?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 芝加哥大学 \u003Cbr> 普林斯顿大学\u002F字节跳动 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.03136) | Huggingface |\n| [Time-R1](https:\u002F\u002Fgithub.com\u002Fulab-uiuc\u002FTime-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fulab-uiuc\u002FTime-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | UIUC | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.13508) | veRL |\n| [ML-Agent](https:\u002F\u002Fgithub.com\u002FMASWorks\u002FML-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMASWorks\u002FML-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | MASWorks | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.23723) | 自定义 |\n| [SkyRL](https:\u002F\u002Fgithub.com\u002FNovaSky-AI\u002FSkyRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNovaSky-AI\u002FSkyRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | NovaSky | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.16108) | veRL |\n| [digitalhuman](https:\u002F\u002Fgithub.com\u002FTencent\u002Fdigitalhuman) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTencent\u002Fdigitalhuman?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | 腾讯 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.03112) | veRL |\n| [sweet_rl](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fsweet_rl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ffacebookresearch\u002Fsweet_rl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | Meta\u002FUCB | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.15478) | OpenRLHF |\n| [swe-rl](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fswe-rl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ffacebookresearch\u002Fswe-rl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | Meta\u002FUIUC\u002FCMU | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.18449) | 自定义 |\n| [rllm](https:\u002F\u002Fgithub.com\u002Fagentica-project\u002Frllm) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fagentica-project\u002Frllm?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | 伯克利天空计算实验室 \u003Cbr> BAIR \u002F Together AI | [Notion 博客](https:\u002F\u002Fpretty-radio-b75.notion.site\u002FrLLM-A-Framework-for-Post-Training-Language-Agents-21b81902c146819db63cd98a54ba5f31) | veRL |\n| [open-r1](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Fopen-r1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhuggingface\u002Fopen-r1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | HuggingFace | -- | TRL |\n| [R1-Code-Interpreter](https:\u002F\u002Fgithub.com\u002Fyongchao98\u002FR1-Code-Interpreter) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fyongchao98\u002FR1-Code-Interpreter?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | MIT | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.21668) | 自定义 |\n| [CTRL](https:\u002F\u002Fgithub.com\u002FHKUNLP\u002Fcritic-rl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FHKUNLP\u002Fcritic-rl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 香港大学\u002F字节跳动 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.03492) | 自定义 |\n| [DeepAnalyze](https:\u002F\u002Fgithub.com\u002Fruc-datalab\u002FDeepAnalyze) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fruc-datalab\u002FDeepAnalyze?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | 人大\u002F清华 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.16872) | 自定义 |\n| [AceCoder](https:\u002F\u002Fgithub.com\u002FTIGER-AI-Lab\u002FAceCoder) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTIGER-AI-Lab\u002FAceCoder?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 渥太华大学 (TIGER-Lab) | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.01718) | 自定义 |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单\u002F多智能体 | 结果\u002F过程奖励 | 单轮\u002F多轮 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [CUDA-Agent](https:\u002F\u002Fgithub.com\u002FBytedTsinghua-SIA\u002FCUDA-Agent) | 智能体强化学习（分阶段） | 单智能体 | 结果奖励 | 多轮 | CUDA 核函数生成 | 规则奖励（正确性 + 性能） | 是（编译\u002F验证\u002F性能分析） |\n| [LLM-in-Sandbox](https:\u002F\u002Fgithub.com\u002Fllm-in-sandbox\u002Fllm-in-sandbox) | GRPO++ | 单智能体 | 结果奖励 | 多轮 | 编码\u002FSWE + 通用任务（数学\u002F科学\u002F生物） | 规则奖励 | 是（带终端、文件和互联网的代码沙箱） |\n| [PPP-Agent](https:\u002F\u002Fgithub.com\u002Fsunnweiwei\u002FPPP-Agent) | PPP-RL | 单智能体 | 结果和过程奖励 | 多轮 | SWE\u002F科研 | 规则+模型奖励 | 搜索、提问、浏览 |\n| [RepoDeepSearch](https:\u002F\u002Fgithub.com\u002FMizersy\u002FRepoDeepSearch) | GRPO | 单智能体 | 结果和过程奖励 | 多轮 | 搜索\u002F修复 | 规则\u002F外部奖励 | 是 |\n| [CUDA-L1](https:\u002F\u002Fgithub.com\u002Fdeepreinforce-ai\u002FCUDA-L1) | 对比强化学习 | 单智能体 | 结果奖励 | 单轮 | CUDA 优化 | 规则奖励（性能） | 否 |\n| [MedAgentGym](https:\u002F\u002Fgithub.com\u002Fwshi83\u002FMedAgentGym) | SFT\u002FDPO\u002FPPO\u002FGRPO | 单智能体 | 结果奖励 | 多轮 | 医疗\u002F编码 | 外部奖励 | 是 |\n| [CURE](https:\u002F\u002Fgithub.com\u002FGen-Verse\u002FCURE) | PPO | 单智能体 | 结果奖励 | 单轮 | 编码 | 外部奖励 | 否 |\n| [Time-R1](https:\u002F\u002Fgithub.com\u002Fulab-uiuc\u002FTime-R1) | PPO\u002FGRPO\u002FDPO | 多智能体 | 结果奖励 | 多轮 | 时序相关任务 | 全部 | 代码 |\n| [ML-Agent](https:\u002F\u002Fgithub.com\u002FMASWorks\u002FML-Agent) | 自定义 | 单智能体 | 迁移奖励 | 多轮 | 编码 | 全部 | 是 |\n| [SkyRL](https:\u002F\u002Fgithub.com\u002FNovaSky-AI\u002FSkyRL) | PPO\u002FGRPO | 单智能体 | 结果奖励 | 多轮 | 数学\u002F编码 | 全部 | 代码 |\n| [digitalhuman](https:\u002F\u002Fgithub.com\u002FTencent\u002Fdigitalhuman) | PPO\u002FGRPO\u002FReMax\u002FRLOO | 多智能体 | 结果奖励 | 多轮 | 同理心\u002F数学\u002F编码\u002F多模态问答 | 规则\u002F模型\u002F外部奖励 | 是 |\n| [sweet_rl](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fsweet_rl) | DPO | 多智能体 | 过程奖励 | 多轮 | 设计\u002F编码 | 模型奖励 | 网页浏览 |\n| [swe-rl](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fswe-rl) | 基于强化学习 | 单智能体 | 结果奖励 | 单轮 | SWE（SWE-bench） | 规则奖励（相似性） | 否 |\n| [rllm](https:\u002F\u002Fgithub.com\u002Fagentica-project\u002Frllm) | PPO\u002FGRPO | 单智能体 | 结果奖励 | 多轮 | 代码编辑 | 外部奖励 | 是 |\n| [open-r1](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Fopen-r1) | GRPO | 单智能体 | 结果奖励 | 单轮 | 数学\u002F编码 | 全部 | 是 |\n| [R1-Code-Interpreter](https:\u002F\u002Fgithub.com\u002Fyongchao98\u002FR1-Code-Interpreter) | GRPO | 单智能体 | 结果奖励 | 多轮 | 代码解释 | 规则\u002F外部奖励 | 是（代码执行） |\n| [CTRL](https:\u002F\u002Fgithub.com\u002FHKUNLP\u002Fcritic-rl) | 强化学习（批评-修正） | 单智能体 | 过程奖励 | 多轮 | 代码优化 | 模型奖励 | 是（代码执行） |\n| [DeepAnalyze](https:\u002F\u002Fgithub.com\u002Fruc-datalab\u002FDeepAnalyze) | 课程制强化学习 | 单智能体 | 结果奖励 | 多轮 | 数据科学 | 规则\u002F外部奖励 | 是（代码执行） |\n| [AceCoder](https:\u002F\u002Fgithub.com\u002FTIGER-AI-Lab\u002FAceCoder) | GRPO | 单智能体 | 结果奖励 | 单轮 | 代码生成 | 外部奖励（测试用例） | 是 |\n\n\u003C\u002Fdetails>\n\n## 🤔 推理智能体\n\n\n| GitHub 仓库 | 🌟 星数 | 发布日期 | 机构 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [Agent0](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FAgent0) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Faiming-lab\u002FAgent0?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | 北卡罗来纳大学教堂山分校 \u002F Salesforce Research \u002F 斯坦福大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.16043) | veRL |\n| [KG-R1](https:\u002F\u002Fgithub.com\u002FJinyeop3110\u002FKG-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FJinyeop3110\u002FKG-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | 伊利诺伊大学厄巴纳-香槟分校\u002F谷歌 | [论文1](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.09516), [论文2](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.15117) | veRL |\n| [AgentFlow](https:\u002F\u002Fgithub.com\u002Flupantech\u002FAgentFlow) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flupantech\u002FAgentFlow?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.09 | 斯坦福大学 | [arXiv](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.05592) | veRL |\n| [ARPO](https:\u002F\u002Fgithub.com\u002Fdongguanting\u002FARPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdongguanting\u002FARPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | 中国人民大学、快手 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.19849) | veRL |\n| [terminal-bench-rl](https:\u002F\u002Fgithub.com\u002FDanau5tin\u002Fterminal-bench-rl) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FDanau5tin\u002Fterminal-bench-rl?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | 个人（Danau5tin） | 无 | rLLM |\n| [MOTIF](https:\u002F\u002Fgithub.com\u002Fpurbeshmitra\u002FMOTIF) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpurbeshmitra\u002FMOTIF?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 马里兰大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.02851) | trl |\n| [cmriat\u002Fl0](https:\u002F\u002Fgithub.com\u002Fcmriat\u002Fl0) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcmriat\u002Fl0?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | CMRIAT | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.23667) | veRL |\n| [agent-distillation](https:\u002F\u002Fgithub.com\u002FNardien\u002Fagent-distillation) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNardien\u002Fagent-distillation?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | KAIST | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.17612) | 自定义 |\n| [EasyR1](https:\u002F\u002Fgithub.com\u002Fhiyouga\u002FEasyR1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhiyouga\u002FEasyR1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | 个人 | [repo1](https:\u002F\u002Fgithub.com\u002Fhiyouga\u002FEasyR1)\u002F[paper2](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2409.19256) | veRL |\n| [AutoCoA](https:\u002F\u002Fgithub.com\u002FADaM-BJTU\u002FAutoCoA) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FADaM-BJTU\u002FAutoCoA?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | 北京交通大学 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.06580) | veRL |\n| [ToRL](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FToRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGAIR-NLP\u002FToRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | 上海交通大学 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.23383) | veRL |\n| [ReMA](https:\u002F\u002Fgithub.com\u002Fziyuwan\u002FReMA-public) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fziyuwan\u002FReMA-public?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | 上海交通大学、伦敦大学学院 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.09501) | veRL |\n| [Agentic-Reasoning](https:\u002F\u002Fgithub.com\u002Ftheworldofagents\u002FAgentic-Reasoning) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftheworldofagents\u002FAgentic-Reasoning?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 牛津大学 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2502.04644) | 自定义 |\n| [SimpleTIR](https:\u002F\u002Fgithub.com\u002Fltzheng\u002FSimpleTIR) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fltzheng\u002FSimpleTIR?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 新加坡国立大学、字节跳动 | [Notion 博客](https:\u002F\u002Fsimpletir.notion.site\u002Freport) | veRL |\n| [openrlhf_async_pipline](https:\u002F\u002Fgithub.com\u002Fyyht\u002Fopenrlhf_async_pipline) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fyyht\u002Fopenrlhf_async_pipline?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.5 | OpenRLHF | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2405.11143) | OpenRLHF |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单智能体\u002F多智能体 | 结果奖励\u002F过程奖励 | 单轮\u002F多轮 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [Agent0](https:\u002F\u002Fgithub.com\u002Faiming-lab\u002FAgent0) | ADPO | 多智能体 | 过程奖励 | 多轮 | 数学\u002F视觉 | 模型\u002F验证器 | 是 |\n| [KG-R1](https:\u002F\u002Fgithub.com\u002FJinyeop3110\u002FKG-R1) | GRPO\u002FPPO | 单智能体 | 双重奖励 | 多轮 | 知识图谱问答 | 规则\u002F模型 | 知识图谱检索 |\n| [AgentFlow](https:\u002F\u002Fgithub.com\u002Flupantech\u002FAgentFlow) | Flow-GRPO | 单智能体 | 结果奖励 | 多轮 | 搜索\u002F数学\u002F问答 | 模型\u002F外部工具 | 是 |\n| [ARPO](https:\u002F\u002Fgithub.com\u002Fdongguanting\u002FARPO) | GRPO | 单智能体 | 结果奖励 | 多轮 | 数学\u002F编程 | 模型\u002F规则 | 是 |\n| [terminal-bench-rl](https:\u002F\u002Fgithub.com\u002FDanau5tin\u002Fterminal-bench-rl) | GRPO | 单智能体 | 结果奖励 | 多轮 | 编程\u002F终端 | 模型+外部验证器 | 是 |\n| [MOTIF](https:\u002F\u002Fgithub.com\u002Fpurbeshmitra\u002FMOTIF) | GRPO | 单智能体 | 结果奖励 | 多轮 | 问答 | 规则 | 否 |\n| [cmriat\u002Fl0](https:\u002F\u002Fgithub.com\u002Fcmriat\u002Fl0) | PPO | 多智能体 | 过程奖励 | 多轮 | 问答 | 全部 | 是 |\n| [agent-distillation](https:\u002F\u002Fgithub.com\u002FNardien\u002Fagent-distillation) | PPO | 单智能体 | 过程奖励 | 多轮 | 问答\u002F数学 | 外部工具 | 是 |\n| [EasyR1](https:\u002F\u002Fgithub.com\u002Fhiyouga\u002FEasyR1) | GRPO | 单智能体 | 过程奖励 | 多轮 | 视觉-语言 | 模型 | 是 |\n| [AutoCoA](https:\u002F\u002Fgithub.com\u002FADaM-BJTU\u002FAutoCoA) | GRPO | 多智能体 | 结果奖励 | 多轮 | 推理\u002F数学\u002F问答 | 全部 | 是 |\n| [ToRL](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FToRL) | GRPO | 单智能体 | 结果奖励 | 单轮 | 数学 | 规则\u002F外部工具 | 是 |\n| [ReMA](https:\u002F\u002Fgithub.com\u002Fziyuwan\u002FReMA-public) | PPO | 多智能体 | 结果奖励 | 多轮 | 数学 | 规则 | 否 |\n| [Agentic-Reasoning](https:\u002F\u002Fgithub.com\u002Ftheworldofagents\u002FAgentic-Reasoning) | 自定义 | 单智能体 | 过程奖励 | 多轮 | 问答\u002F数学 | 外部工具 | 网页浏览 |\n| [SimpleTIR](https:\u002F\u002Fgithub.com\u002Fltzheng\u002FSimpleTIR) | PPO\u002FGRPO（带扩展） | 单智能体 | 结果奖励 | 多轮 | 数学、编程 | 全部 | 是 |\n| [openrlhf_async_pipline](https:\u002F\u002Fgithub.com\u002Fyyht\u002Fopenrlhf_async_pipline) | PPO\u002FREINFORCE++\u002FDPO\u002FRLOO | 单智能体 | 结果奖励 | 多轮 | 对话\u002F推理\u002F问答 | 全部 | 否 |\n\n\u003C\u002Fdetails>\n\n## 👥 多智能体强化学习\n\n\n| GitHub 仓库 | 🌟 星数 | 日期 | 机构 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [PettingLLMs](https:\u002F\u002Fgithub.com\u002Fpettingllms-ai\u002FPettingLLMs) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpettingllms-ai\u002FPettingLLMs?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | Intel \u002F UCSD | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.11062) | 自定义 |\n| [MASPRM](https:\u002F\u002Fgithub.com\u002Fmilad1378yz\u002FMASPRM) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmilad1378yz\u002FMASPRM?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.10 | UBC \u002F Huawei | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.24803) | 自定义 |\n| [ARIA](https0:\u002F\u002Fgithub.com\u002Frhyang2021\u002FARIA) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Frhyang2021\u002FARIA?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 复旦大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.00539) | 自定义 |\n| [AMPO](https:\u002F\u002Fgithub.com\u002FMozerWang\u002FAMPO) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMozerWang\u002FAMPO?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 阿里巴巴通义实验室 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.02156) | veRL |\n| [MAPoRL](https:\u002F\u002Fgithub.com\u002Fchanwoo-park-official\u002FMAPoRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fchanwoo-park-official\u002FMAPoRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | 学术界 | —— | 自定义 |\n| [FlowReasoner](https:\u002F\u002Fgithub.com\u002Fsail-sg\u002FFlowReasoner) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsail-sg\u002FFlowReasoner?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | 海 AI 实验室 \u002F 新加坡国立大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.15257) | 自定义 |\n| [DrMAS](https:\u002F\u002Fgithub.com\u002FlangfengQ\u002FDrMAS) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FlangfengQ\u002FDrMAS?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | 南洋理工大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.08847) | 自定义 |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单\u002F多智能体 | 结果\u002F过程奖励 | 单\u002F多回合 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [PettingLLMs](https:\u002F\u002Fgithub.com\u002Fpettingllms-ai\u002FPettingLLMs) | AT-GRPO | 多 | 两者 | 多 | 游戏\u002F代码\u002F数学\u002F规划 | 规则（可验证） | 否 |\n| [MASPRM](https:\u002F\u002Fgithub.com\u002Fmilad1378yz\u002FMASPRM) | PRM（由 MCTS 滚出训练） | 多 | 过程 | 多 | 推理（GSM8K\u002FMATH\u002FMMLU） | 学习型 PRM | 否 |\n| [ARIA](https:\u002F\u002Fgithub.com\u002Frhyang2021\u002FARIA) | REINFORCE | 两者 | 过程 | 多 | 谈判\u002F讨价还价 | 其他 | 否 |\n| [AMPO](https:\u002F\u002Fgithub.com\u002FMozerWang\u002FAMPO) | BC\u002FAMPO（GRPO 改进） | 多 | 结果 | 多 | 社交互动 | 基于模型 | 否 |\n| [MAPoRL](https:\u002F\u002Fgithub.com\u002Fchanwoo-park-official\u002FMAPoRL) | PPO | 多 | 结果 | 多 | LLM 协作任务 | 规则 | 否 |\n| [FlowReasoner](https:\u002F\u002Fgithub.com\u002Fsail-sg\u002FFlowReasoner) | GRPO | 多 | 结果 | 多 | 多智能体工作流设计 | 规则 | 是 |\n| [DrMAS](https:\u002F\u002Fgithub.com\u002FlangfengQ\u002FDrMAS) | GRPO（按智能体） | 多 | 结果 | 多 | 多智能体 LLM 系统 | 规则 | 否 |\n\n\u003C\u002Fdetails>\n\n## 🧠 记忆\n\n\n| GitHub 仓库 | 🌟 星数 | 日期 | 机构 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [MEM1](https0:\u002F\u002Fgithub.com\u002FMIT-MI\u002FMEM1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMIT-MI\u002FMEM1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | MIT | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.15841) | veRL（基于 Search-R1） |\n| [Memento](https:\u002F\u002Fgithub.com\u002FAgent-on-the-Fly\u002FMemento) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAgent-on-the-Fly\u002FMemento?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | UCL、华为 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.16153) | 自定义 |\n| [MemAgent](https:\u002F\u002Fgithub.com\u002FBytedTsinghua-SIA\u002FMemAgent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FBytedTsinghua-SIA\u002FMemAgent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 字节跳动、清华 SIA | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.02259) | veRL |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单\u002F多智能体 | 结果\u002F过程奖励 | 单\u002F多回合 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [MEM1](https:\u002F\u002Fgithub.com\u002FMIT-MI\u002FMEM1) | PPO\u002FGRPO | 单 | 结果 | 多 | 网店\u002FGSM8K\u002FQA | 规则\u002F模型 | 是 |\n| [Memento](https:\u002F\u002Fgithub.com\u002FAgent-on-the-Fly\u002FMemento) | 软 Q-Learning | 单 | 结果 | 多 | 研究\u002FQA\u002F代码\u002F网络 | 外部\u002F规则 | 是 |\n| [MemAgent](https:\u002F\u002Fgithub.com\u002FBytedTsinghua-SIA\u002FMemAgent) | PPO、GRPO、DPO | 多 | 结果 | 多 | 长上下文 QA | 规则\u002F模型\u002F外部 | 是 |\n\n\u003C\u002Fdetails>\n\n## 🦾 具身\n\n\n| GitHub 仓库 | 🌟 星数 | 日期 | 机构 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [Embodied-R1](https:\u002F\u002Fgithub.com\u002Fpickxiguapi\u002FEmbodied-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpickxiguapi\u002FEmbodied-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 天津大学 | [论文](http:\u002F\u002Farxiv.org\u002Fabs\u002F2508.13998) | veRL |\n| [STeCa](https:\u002F\u002Fgithub.com\u002FWangHanLinHenry\u002FSTeCa) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FWangHanLinHenry\u002FSTeCa?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 香港理工大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.14276) | FastChat\u002FTRL |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单\u002F多智能体 | 结果\u002F过程奖励 | 单\u002F多回合 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [Embodied-R1](https:\u002F\u002Fgithub.com\u002Fpickxiguapi\u002FEmbodied-R1) | GRPO | 单 | 结果 | 单 | 定位\u002F航点 | 规则 | 否 |\n| [STeCa](https:\u002F\u002Fgithub.com\u002FWangHanLinHenry\u002FSTeCa) | DPO（RFT） | 单 | 两者 | 多 | 具身\u002F家务 | 规则\u002FMC | 环境动作 |\n\n\u003C\u002Fdetails>\n\n## 🏷️ 领域专用\n\n\n| GitHub 仓库 | 🌟 星数 | 日期 | 机构 | 论文链接 | 强化学习框架 | 领域 |\n| :----: | :----: | :----: |  :----: | :----: | :----: | :----: |\n| [MedSAM-Agent](https:\u002F\u002Fgithub.com\u002FCUHK-AIM-Group\u002FMedSAM-Agent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FCUHK-AIM-Group\u002FMedSAM-Agent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | 香港中文大学\u002F腾讯 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.03320) | 自定义 | 医疗 |\n| [OS-R1](https:\u002F\u002Fgithub.com\u002FLHY-24\u002FOS-R1) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FLHY-24\u002FOS-R1?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | 中国科学院计算技术研究所 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.12551) | 自定义 | 操作系统\u002F系统 |\n| [MMedAgent-RL](https:\u002F\u002Fgithub.com\u002FJanerhYang\u002FMMedAgent-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FJanerhYang\u002FMMedAgent-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | 未知 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.00555) | 未知 | 医疗 |\n| [DoctorAgent-RL](https:\u002F\u002Fgithub.com\u002FJarvisUSTC\u002FDoctorAgent-RL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FJarvisUSTC\u002FDoctorAgent-RL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 中国科学院大学\u002F中国科学院\u002F中国科学技术大学 | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.19630) | RAGEN | 医疗 |\n| [Biomni](https:\u002F\u002Fgithub.com\u002Fsnap-stanford\u002FBiomni) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsnap-stanford\u002FBiomni?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | 斯坦福大学（SNAP） | [论文](https:\u002F\u002Fwww.biorxiv.org\u002Fcontent\u002F10.1101\u002F2025.05.30.656746v1) | 自定义 | 生物医学 |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单智能体\u002F多智能体 | 结果奖励\u002F过程奖励 | 单轮\u002F多轮 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [MedSAM-Agent](https:\u002F\u002Fgithub.com\u002FCUHK-AIM-Group\u002FMedSAM-Agent) | GRPO（通过 veRL） | 单 | 两者 | 多 | 医学图像分割 | 模型（临床保真度） | 是（SAM\u002FMedSAM2） |\n| [OS-R1](https:\u002F\u002Fgithub.com\u002FLHY-24\u002FOS-R1) | GRPO（通过 veRL） | 单 | 结果 | 多 | Linux 内核调优 | 规则 | 是（LightRAG、内核配置） |\n| [MMedAgent-RL](https:\u002F\u002Fgithub.com\u002FJanerhYang\u002FMMedAgent-RL) | 未知 | 多 | 未知 | 未知 | 未知 | 未知 | 未知 |\n| [DoctorAgent-RL](https:\u002F\u002Fgithub.com\u002FJarvisUSTC\u002FDoctorAgent-RL) | GRPO | 多 | 两者 | 多 | 问诊\u002F诊断 | 模型\u002F规则 | 否 |\n| [Biomni](https:\u002F\u002Fgithub.com\u002Fsnap-stanford\u002FBiomni) | 待定 | 单 | 待定 | 单 | scRNAseq\u002FCRISPR\u002FADMET\u002F知识 | 待定 | 是 |\n\n\u003C\u002Fdetails>\n\n## 🎯 奖励与训练方法\n\n\n| GitHub 仓库 | 🌟 星数 | 日期 | 机构 | 论文链接 | 重点 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [ToolPRMBench](https:\u002F\u002Fgithub.com\u002FDavid-Li0406\u002FToolPRMBench) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FDavid-Li0406\u002FToolPRMBench?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | 亚利桑那州立大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.12294) | 工具使用 PRM 基准测试 |\n| [RLVR-World](https:\u002F\u002Fgithub.com\u002Fthuml\u002FRLVR-World) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fthuml\u002FRLVR-World?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 清华大学机器学习组 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.13934) | 用于世界模型的 RLVR |\n| [AgentPRM](https:\u002F\u002Fgithub.com\u002Fsanjibanc\u002Fagent_prm) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsanjibanc\u002Fagent_prm?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 康奈尔大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.10325) | 针对智能体的过程奖励 |\n| [Agentic-Reward-Modeling](https:\u002F\u002Fgithub.com\u002FTHU-KEG\u002FAgentic-Reward-Modeling) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHU-KEG\u002FAgentic-Reward-Modeling?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 清华大学 KEG 小组 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.19328) | 聚合式奖励代理 |\n| [AgentRM](https:\u002F\u002Fgithub.com\u002Fthunlp\u002FAgentRM) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fthunlp\u002FAgentRM?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 清华大学 THUNLP | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.18407) | 可泛化的智能体 RM |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单智能体\u002F多智能体 | 结果奖励\u002F过程奖励 | 单轮\u002F多轮 | 任务 | 奖励类型 | 工具使用 |\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [ToolPRMBench](https:\u002F\u002Fgithub.com\u002FDavid-Li0406\u002FToolPRMBench) | 无（基准测试） | 单 | 过程 | 多 | 工具使用 | 规则\u002F模型 | 是 |\n| [RLVR-World](https:\u002F\u002Fgithub.com\u002Fthuml\u002FRLVR-World) | RLVR | 单 | 结果 | 多 | 世界建模（语言\u002F视频） | 模型（可验证） | 否 |\n| [AgentPRM](https:\u002F\u002Fgithub.com\u002Fsanjibanc\u002Fagent_prm) | PPO\u002FDPO + PRM | 单 | 过程 | 多 | ALFWorld\u002F通用 | 模型（PRM） | 是 |\n| [Agentic-Reward-Modeling](https:\u002F\u002Fgithub.com\u002FTHU-KEG\u002FAgentic-Reward-Modeling) | DPO\u002FBest-of-N | 单 | 结果 | 单 | 通用指令 | 模型（奖励代理） | 是（验证） |\n| [AgentRM](https:\u002F\u002Fgithub.com\u002Fthunlp\u002FAgentRM) | MCTS\u002FRM 引导 | 单 | 结果 | 多 | 9 个智能体任务 | 模型（回归 PRM） | 是 |\n\n\u003C\u002Fdetails>\n\n## 🛡️ 安全\n\n\n| GitHub 仓库 | 🌟 星数 | 日期 | 组织 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [SafeSearch](https:\u002F\u002Fgithub.com\u002Famazon-science\u002FSafeSearch) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Famazon-science\u002FSafeSearch?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | Amazon Science | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.17017) | veRL |\n| [curiosity_redteam](https:\u002F\u002Fgithub.com\u002FImprobable-AI\u002Fcuriosity_redteam) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FImprobable-AI\u002Fcuriosity_redteam?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.2 | MIT | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2402.19464) | 自定义 |\n| [RLbreaker](https:\u002F\u002Fgithub.com\u002FXuanChen-xc\u002FRLbreaker) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FXuanChen-xc\u002FRLbreaker?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.6 | 普渡大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2406.08705) | 自定义 |\n| [xJailbreak](https:\u002F\u002Fgithub.com\u002FAegis1863\u002FxJailbreak) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAegis1863\u002FxJailbreak?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | 学术界 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2501.16727) | 自定义 |\n| [Auto-RT](https:\u002F\u002Fgithub.com\u002Ficip-cas\u002FAuto-RT) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ficip-cas\u002FAuto-RT?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | ICIP-CAS | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2501.01830) | 自定义 |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单\u002F多智能体 | 结果\u002F过程奖励 | 单\u002F多回合 | 任务 |  Beliebte Suchanfragen:\n\n## 🔄 自我进化\n\n> ⚠️ **注**: 在针对大语言模型智能体的强化学习背景下，“自我进化”的定义仍在发展中，尚未完全确立。本类别目前收录了论文标题中明确包含“self-evolving”或“self-evolution”的相关工作，这些工作中的智能体通过强化学习驱动的反馈循环实现自我改进。\n\n\n| GitHub 仓库 | 🌟 星数 | 发布日期 | 机构 | 论文链接 | 强化学习框架 |\n| :----: | :----: | :----: |  :----: | :----: | :----: |\n| [AgentEvolver](https:\u002F\u002Fgithub.com\u002Fmodelscope\u002FAgentEvolver) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmodelscope\u002FAgentEvolver?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.11 | 阿里巴巴\u002F通义实验室 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.10395) | 自定义 |\n| [SEAgent](https:\u002F\u002Fgithub.com\u002FSunzeY\u002FSEAgent) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSunzeY\u002FSEAgent?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.8 | 上海人工智能实验室 \u002F 香港中文大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.04700) | 自定义 |\n| [MemSkill](https:\u002F\u002Fgithub.com\u002FViktorAxelsen\u002FMemSkill) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FViktorAxelsen\u002FMemSkill?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.2 | 南洋理工大学\u002F伊利诺伊大学厄巴纳-香槟分校\u002F芝加哥大学\u002F清华大学 | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.02474) | 自定义 |\n| [MemRL](https:\u002F\u002Fgithub.com\u002FMemTensor\u002FMemRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMemTensor\u002FMemRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | 上海交通大学\u002F西安电子科技大学\u002F新加坡国立大学\u002F中国科学技术大学\u002FMemTensor | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.03192) | 自定义 |\n| [RAGEN](https:\u002F\u002Fgithub.com\u002FRAGEN-AI\u002FRAGEN) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRAGEN-AI\u002FRAGEN?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | RAGEN-AI | [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.20073) | veRL |\n| [WebRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FWebRL) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002FWebRL?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.11 | 清华大学\u002F智谱AI | [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2411.02337) | 自定义 |\n\n\u003Cdetails>\n\u003Csummary>📋 点击查看技术细节\u003C\u002Fsummary>\n\n| GitHub 仓库 | 强化学习算法 | 单智能体\u002F多智能体 | 结果奖励\u002F过程奖励 | 单轮\u002F多轮 | 任务 |  Beliebte Artikel\n| :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: |\n| [AgentEvolver](https:\u002F\u002Fgithub.com\u002Fmodelscope\u002FAgentEvolver) | ADCA-GRPO | 单智能体 | 结果奖励 | 多轮 | 社交游戏\u002F工具使用 | 规则 | 是 |\n| [SEAgent](https:\u002F\u002Fgithub.com\u002FSunzeY\u002FSEAgent) | GRPO | 单智能体 | 结果奖励 | 多轮 | 计算机使用 (OSWorld) | 模型 | 是 (基于截图) |\n| [MemSkill](https:\u002F\u002Fgithub.com\u002FViktorAxelsen\u002FMemSkill) | PPO | 单智能体 | 过程奖励 | 多轮 | QA\u002FALFWorld | 模型 (学习到的技能) | 是 |\n| [MemRL](https:\u002F\u002Fgithub.com\u002FMemTensor\u002FMemRL) | 基于强化学习 (Q值) | 单智能体 | 过程奖励 | 多轮 | HLE\u002FBigCodeBench\u002FALFWorld | 模型 (检索) | 是 |\n| [RAGEN](https:\u002F\u002Fgithub.com\u002FRAGEN-AI\u002FRAGEN) | PPO\u002FGRPO (StarPO) | 单智能体 | 结果奖励和过程奖励 | 多轮 | 文本游戏 | 全部 | 是 |\n| [WebRL](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FWebRL) | 行动者-评论家强化学习 + ORM | 单智能体 | 结果奖励 | 多轮 | 网页导航 (WebArena) | 模型 (ORM) | 是 (网页浏览) |\n\n\u003C\u002Fdetails>\n\n## ⛰️ 环境\n\n| GitHub 仓库 | 🌟 星数 | 发布日期 | 组织 | 任务 |\n| :----: | :----: | :----: |  :----: | :----: |\n| [OpenSandbox](https:\u002F\u002Fgithub.com\u002Falibaba\u002FOpenSandbox) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Falibaba\u002FOpenSandbox?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | 阿里巴巴 | 代码\u002FGUI\u002F智能体评估 |\n| [OpenEnv](https:\u002F\u002Fgithub.com\u002Fmeta-pytorch\u002FOpenEnv) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmeta-pytorch\u002FOpenEnv?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | Meta (PyTorch) | 国际象棋\u002F街机\u002F金融 |\n| [NeMo-Gym](https:\u002F\u002Fgithub.com\u002FNVIDIA-NeMo\u002FGym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNVIDIA-NeMo\u002FGym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.1 | NVIDIA | 多步\u002F多轮 |\n| [open-trajectory-gym](https:\u002F\u002Fgithub.com\u002Fwestonbrown\u002Fopen-trajectory-gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fwestonbrown\u002Fopen-trajectory-gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2026.3 | 个人 | CTF\u002F安全 |\n| [R2E-Gym](https:\u002F\u002Fgithub.com\u002FR2E-Gym\u002FR2E-Gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FR2E-Gym\u002FR2E-Gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | UC Berkeley\u002FANU | 软件工程 |\n| [LoCoBench-Agent](https:\u002F\u002Fgithub.com\u002FSalesforceAIResearch\u002FLoCoBench-Agent) | ![](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSalesforceAIResearch\u002FLoCoBench-Agent.svg?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700) | 2025.11 | Salesforce AI Research | 软件工程 |\n| [Simia-Agent-Training](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FSimia-Agent-Training) | ![](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FSimia-Agent-Training.svg?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700) | 2025.10 | 微软 | 工具使用\u002FAPI |\n| [PaperArena](https:\u002F\u002Fgithub.com\u002FMelmaphother\u002FPaperArena) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMelmaphother\u002FPaperArena?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.9 | 中国科学技术大学 | 科学文献问答 |\n| [enterprise-deep-research](https:\u002F\u002Fgithub.com\u002FSalesforceAIResearch\u002Fenterprise-deep-research) | ![](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSalesforceAIResearch\u002Fenterprise-deep-research.svg?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700) | 2025.9 | Salesforce AI Research | 深度研究 |\n| [CompassVerifier](https:\u002F\u002Fgithub.com\u002Fopen-compass\u002FCompassVerifier) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopen-compass\u002FCompassVerifier?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.7 | 上海人工智能实验室 | 推理 |\n| [SWE-smith](https:\u002F\u002Fgithub.com\u002FSWE-bench\u002FSWE-smith) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSWE-bench\u002FSWE-smith?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | 普林斯顿\u002F斯坦福\u002FSWE-bench | 软件工程 |\n| [SWE-Gym](https:\u002F\u002Fgithub.com\u002FSWE-Gym\u002FSWE-Gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSWE-Gym\u002FSWE-Gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.12 | UC Berkeley\u002FUIUC\u002FCMU\u002F苹果 | 软件工程 |\n| [Mind2Web-2](https:\u002F\u002Fgithub.com\u002FOSU-NLP-Group\u002FMind2Web-2) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOSU-NLP-Group\u002FMind2Web-2?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.6 | 俄亥俄州立大学 | 网页 |\n| [gem](https:\u002F\u002Fgithub.com\u002Faxon-rl\u002Fgem) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Faxon-rl\u002Fgem?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | 海洋人工智能实验室 | 数学\u002F代码\u002F游戏\u002F问答 |\n| [MLE-Dojo](https:\u002F\u002Fgithub.com\u002FMLE-Dojo\u002FMLE-Dojo) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMLE-Dojo\u002FMLE-Dojo?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.5 | GIT, 斯坦福 | 机器学习工程 |\n| [atropos](https:\u002F\u002Fgithub.com\u002FNousResearch\u002Fatropos) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FNousResearch\u002Fatropos?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | Nous Research | 游戏\u002F代码\u002F工具 |\n| [InternBootcamp](https:\u002F\u002Fgithub.com\u002FInternLM\u002FInternBootcamp) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FInternLM\u002FInternBootcamp?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.4 | InternBootcamp | 编程\u002F问答\u002F游戏 |\n| [loong](https:\u002F\u002Fgithub.com\u002Fcamel-ai\u002Floong) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcamel-ai\u002Floong?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.3 | CAMEL-AI.org | RLVR |\n| [DataSciBench](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FDataSciBench) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002FDataSciBench?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.2 | 清华大学 | 数据分析 |\n| [reasoning-gym](https:\u002F\u002Fgithub.com\u002Fopen-thought\u002Freasoning-gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopen-thought\u002Freasoning-gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | open-thought | 数学\u002F游戏 |\n| [llmgym](https:\u002F\u002Fgithub.com\u002Ftensorzero\u002Fllmgym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftensorzero\u002Fllmgym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2025.1 | tensorzero | 文本游戏\u002F工具 |\n| [debug-gym](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fdebug-gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fdebug-gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.11 | 微软研究院 | 调试\u002F游戏\u002F代码 |\n| [gym-llm](https:\u002F\u002Fgithub.com\u002Frsanchezmo\u002Fgym-llm) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Frsanchezmo\u002Fgym-llm?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.8 | Rodrigo Sánchez Molina | 控制\u002F游戏 |\n| [AgentGym](https:\u002F\u002Fgithub.com\u002FWooooDyy\u002FAgentGym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FWooooDyy\u002FAgentGym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.6 | 复旦大学 | 网页\u002F游戏 |\n| [tau-bench](https:\u002F\u002Fgithub.com\u002Fsierra-research\u002Ftau-bench) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsierra-research\u002Ftau-bench?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.6 | Sierra | 工具 |\n| [appworld](https:\u002F\u002Fgithub.com\u002FStonyBrookNLP\u002Fappworld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FStonyBrookNLP\u002Fappworld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.6 | 石溪大学 | 手机使用 |\n| [android_world](https:\u002F\u002Fgithub.com\u002Fgoogle-research\u002Fandroid_world) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgoogle-research\u002Fandroid_world?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.5 | Google 研究院 | 手机使用 |\n| [TheAgentCompany](https:\u002F\u002Fgithub.com\u002FTheAgentCompany\u002FTheAgentCompany) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTheAgentCompany\u002FTheAgentCompany?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.3 | CMU, 杜克大学 | 编程 |\n| [LlamaGym](https:\u002F\u002Fgithub.com\u002FKhoomeiK\u002FLlamaGym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FKhoomeiK\u002FLlamaGym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.3 | Rohan Pandey | 游戏 |\n| [visualwebarena](https:\u002F\u002Fgithub.com\u002Fweb-arena-x\u002Fvisualwebarena)   | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fweb-arena-x\u002Fvisualwebarena?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2024.1 | CMU | 网页 |\n| [LMRL-Gym](https:\u002F\u002Fgithub.com\u002Fabdulhaim\u002FLMRL-Gym) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fabdulhaim\u002FLMRL-Gym?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2023.12 | UC Berkeley | 游戏 |\n| [OSWorld](https:\u002F\u002Fgithub.com\u002Fxlang-ai\u002FOSWorld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fxlang-ai\u002FOSWorld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2023.10 | 香港大学、CMU、Salesforce、滑铁卢 | 计算机使用 |\n| [webarena](https:\u002F\u002Fgithub.com\u002Fweb-arena-x\u002Fwebarena) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fweb-arena-x\u002Fwebarena?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2023.7 | CMU | 网页 |\n| [AgentBench](https:\u002F\u002Fgithub.com\u002FTHUDM\u002FAgentBench) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FTHUDM\u002FAgentBench?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2023.7 | 清华大学 | 游戏\u002F网页\u002F问答\u002F工具 |\n| [WebShop](https:\u002F\u002Fgithub.com\u002Fprinceton-nlp\u002FWebShop) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fprinceton-nlp\u002FWebShop?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2022.7 | Princeton-NLP | 网页 |\n| [ScienceWorld](https:\u002F\u002Fgithub.com\u002Fallenai\u002FScienceWorld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fallenai\u002FScienceWorld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2022.3 | AllenAI | 文本游戏\u002F科学问答 |\n| [alfworld](https:\u002F\u002Fgithub.com\u002Falfworld\u002Falfworld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Falfworld\u002Falfworld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2020.10 | 微软、CMU、华盛顿大学 | 身体化 |\n| [factorio-learning-environment](https:\u002F\u002Fgithub.com\u002FJackHopkins\u002Ffactorio-learning-environment) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FJackHopkins\u002Ffactorio-learning-environment?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2021.6 | JackHopkins | 游戏 |\n| [jericho](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fjericho) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002Fjericho?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2018.10 | 微软、GIT | 文本游戏 |\n| [TextWorld](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002FTextWorld) | \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fmicrosoft\u002FTextWorld?style=for-the-badge&logo=github&logoColor=white&labelColor=181717&color=ffd700\" alt=\"Stars\"> | 2018.6 | 微软研究院 | 文本游戏 |\n\n## 审核中\u002F等待开源\n- [JoyAgents-R1：基于强化学习的多功能多大模型智能体联合进化动力学](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.19846)\n- [Shop-R1：通过强化学习奖励大模型模拟在线购物中的人类行为](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.17842)\n- [利用强化学习训练长上下文、多轮次的软件工程智能体](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.03501)\n- [少行动，多推理！教导模型高效行动](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.14870)\n- [基于强化学习的大模型代理式推理与工具集成](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.01441)\n- [ComputerRL：面向计算机使用智能体的端到端在线强化学习规模化](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.14040)\n- [Atom-Searcher：通过细粒度原子级思维奖励增强代理式深度研究](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus)\n- [MUA-RL：用于代理式工具使用的多轮用户交互智能体强化学习](https:\u002F\u002Fgithub.com\u002Fzzwkk\u002FMUA-RL)\n- [理解工具集成式推理](https:\u002F\u002Fzhongwenxu.notion.site\u002FUnderstanding-Tool-Integrated-Reasoning-2551c4e140e3805489fadcc802a1ea83)\n- [Memory-R1：通过强化学习提升大语言模型智能体的记忆管理与利用能力](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.19828)\n- [鼓励良好过程，无需良好答案：大模型智能体规划的强化学习](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.19598)\n- [SFR-DeepResearch：迈向自主推理单智能体的有效强化学习](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.06283)\n- [WebExplorer：探索与进化，用于训练长 horizon 的网页智能体](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.06501)\n- [EnvX：用代理式 AI 实现万物智能化](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.08088)\n- [UI-TARS-2 技术报告：利用多轮次强化学习推进 GUI 智能体发展](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.02544)\n- [UI-Venus 技术报告：借助 RFT 构建高性能 GUI 智能体](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.10833)\n- [Agent2：一种用于强化学习自动化的智能体生成智能体框架](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.13368)\n- [Tool-R1：针对代理式工具使用的样本高效强化学习](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.12867v1)\n- [面向大语言模型智能体安全的对抗性强化学习](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.05442)\n- [学习精炼：一种用于迭代构建 SPARQL 查询的代理式强化学习方法](https:\u002F\u002Fwww.arxiv.org\u002Fabs\u002F2511.11770)\n- [InfoFlow：通过奖励密度优化强化搜索智能体](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.26575)\n\n## 星标历史\n\n[![星标历史图表](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fthinkwee_AgentsMeetRL_readme_02ad232660f9.png)](https:\u002F\u002Fwww.star-history.com\u002F#thinkwee\u002FagentsMeetRL&Date)\n\n\n## 引用\n\n如果您觉得本仓库有用，请考虑引用：\n\n```bibtex\n@misc{agentsMeetRL,\n  title={当大模型智能体遇到强化学习：全面综述},\n  author={AgentsMeetRL 贡献者},\n  year={2025},\n  url={https:\u002F\u002Fgithub.com\u002Fthinkwee\u002FagentsMeetRL}\n}\n```\n\n---\n\n\u003Cdiv align=\"center\">\n  \u003Cp>由 AgentsMeetRL 社区用心制作\u003C\u002Fp>\n\u003C\u002Fdiv>","# AgentsMeetRL 快速上手指南\n\n**AgentsMeetRL** 并非一个单一的 Python 库，而是一个汇总了**使用强化学习（RL）训练大语言模型（LLM）智能体**的开源项目清单。它涵盖了从基础框架（如 veRL, OpenRLHF）到具体应用场景（如代码生成、Web 操作、多智能体协作）的各类仓库。\n\n本指南将指导你如何利用该清单选择合适的基础框架，并以目前社区最流行的 **OpenRLHF** 或 **veRL** 为例，快速搭建一个 LLM 智能体强化学习环境。\n\n## 1. 环境准备\n\n在开始之前，请确保你的开发环境满足以下要求。由于强化学习训练对算力要求较高，建议使用配备 NVIDIA GPU 的 Linux 服务器。\n\n*   **操作系统**: Linux (Ubuntu 20.04\u002F22.04 推荐)\n*   **硬件要求**:\n    *   GPU: 至少 1 张 NVIDIA A100\u002FH100 或同等算力显卡 (显存建议 24GB+)\n    *   CPU: 8 核以上\n    *   内存: 32GB+\n*   **软件依赖**:\n    *   Python: 3.9 - 3.11\n    *   CUDA: 11.8 或 12.1+\n    *   Git\n*   **前置知识**: 熟悉 PyTorch 和 Hugging Face Transformers 基础用法。\n\n> 💡 **国内开发者提示**: 建议配置国内镜像源以加速依赖下载。\n> *   Pip 源：`https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple`\n> *   Hugging Face 镜像：`export HF_ENDPOINT=https:\u002F\u002Fhf-mirror.com`\n\n## 2. 安装步骤\n\n由于 AgentsMeetRL 包含多个独立项目，你需要根据需求选择其中一个“基础框架（Base Framework）”进行安装。以下以清单中星数较高且通用的 **OpenRLHF** 为例（其他如 `veRL`, `trl` 安装逻辑类似）。\n\n### 步骤 2.1: 克隆项目\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FOpenRLHF\u002FOpenRLHF.git\ncd OpenRLHF\n```\n\n### 步骤 2.2: 创建虚拟环境并安装依赖\n推荐使用 Conda 管理环境：\n\n```bash\nconda create -n openrlhf python=3.10 -y\nconda activate openrlhf\n\n# 使用清华源加速安装\npip install -r requirements.txt -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n\n# 安装 OpenRLHF 包本身\npip install -e . -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n\n### 步骤 2.3: 验证安装\n检查是否成功导入核心模块：\n```bash\npython -c \"import openrlhf; print('OpenRLHF installed successfully!')\"\n```\n\n## 3. 基本使用\n\n以下示例展示如何使用 OpenRLHF 框架，通过 PPO 算法对一个简单的数学推理智能体进行强化学习微调。此流程对应 AgentsMeetRL 分类中的 **Base Framework** 和 **Reasoning** 类别。\n\n### 步骤 3.1: 准备数据\n创建一个简单的 JSONL 格式提示文件 `prompt.jsonl`，每行包含一个用于训练的 prompt：\n```json\n{\"prompt\": \"Calculate 23 * 45.\"}\n{\"prompt\": \"What is the square root of 144?\"}\n```\n\n### 步骤 3.2: 启动训练\n使用 `openrlhf` 命令行工具启动 PPO 训练。以下命令假设使用本地单卡环境，并加载一个小型预训练模型（如 Qwen-1.5-0.5B 或 TinyLlama）作为演示。\n\n```bash\nopenrlhf train_ppo \\\n  --pretrain_models Qwen\u002FQwen1.5-0.5B-Chat \\\n  --reward_pretrain_models Qwen\u002FQwen1.5-0.5B-Chat \\\n  --save_path .\u002Fcheckpoint\u002Fllama-0.5b-ppo \\\n  --micro_train_batch_size 2 \\\n  --train_batch_size 8 \\\n  --micro_rollout_batch_size 4 \\\n  --rollout_batch_size 8 \\\n  --max_samples 100 \\\n  --max_epochs 1 \\\n  --prompt_max_len 1024 \\\n  --generate_max_len 1024 \\\n  --zero_stage 2 \\\n  --bf16 \\\n  --actor_learning_rate 5e-7 \\\n  --critic_learning_rate 9e-6 \\\n  --init_kl_coef 0.01 \\\n  --prompt_data prompt.jsonl \\\n  --input_key prompt \\\n  --apply_chat_template \\\n  --normalize_reward \\\n  --adam_offload \\\n  --flash_attn \\\n  --gradient_checkpointing\n```\n\n### 步骤 3.3: 查看结果\n训练完成后，模型权重将保存在 `.\u002Fcheckpoint\u002Fllama-0.5b-ppo` 目录下。你可以使用 Hugging Face `transformers` 加载该模型进行测试：\n\n```python\nfrom transformers import AutoTokenizer, AutoModelForCausalLM\nimport torch\n\nmodel_path = \".\u002Fcheckpoint\u002Fllama-0.5b-ppo\"\ntokenizer = AutoTokenizer.from_pretrained(model_path)\nmodel = AutoModelForCausalLM.from_pretrained(model_path, device_map=\"auto\")\n\ninput_text = \"Calculate 123 + 456.\"\ninputs = tokenizer(input_text, return_tensors=\"pt\").to(model.device)\noutputs = model.generate(**inputs, max_new_tokens=50)\nprint(tokenizer.decode(outputs[0], skip_special_tokens=True))\n```\n\n---\n**进阶探索**:\n访问 [AgentsMeetRL 交互式看板](https:\u002F\u002Fthinkwee.top\u002Famr\u002F) 浏览更多特定领域的项目（如 **Web & GUI**, **Code & SWE**, **Multi-Agent RL**），点击对应项目的 \"Click to view technical details\" 查看其具体的 Reward 类型和环境配置，重复上述安装步骤即可复用其技术栈。","某 AI 初创团队正致力于开发一款能自主操作浏览器完成复杂数据抓取与表单填写的智能助手，急需引入强化学习（RL）来提升代理在动态网页环境中的决策能力。\n\n### 没有 AgentsMeetRL 时\n- **技术选型迷茫**：面对 GitHub 上数百个零散的 RL 项目，团队难以快速区分哪些是专为\"Web & GUI\"场景设计，哪些仅适用于纯文本推理，导致大量时间浪费在无效代码阅读上。\n- **架构重复造轮子**：由于缺乏对现有“基础框架”（如 veRL、OpenRLHF）的系统梳理，开发人员误以为需要从头搭建训练基础设施，延误了核心算法的研发进度。\n- **奖励机制设计困难**：在定义代理操作浏览器的成功标准时，找不到成熟的\"Reward & Training\"案例参考，导致模型训练收敛缓慢且容易出现死循环。\n- **安全隐患被忽视**：团队专注于功能实现，却因未查阅\"Safety\"分类下的对抗性测试项目，导致代理在面对恶意网页弹窗时缺乏防御机制。\n\n### 使用 AgentsMeetRL 后\n- **精准定位资源**：通过 AgentsMeetRL 的分类标签，团队直接锁定了 20 个专注于\"Web & GUI\"和\"Tool-Use\"的开源项目，半天内就完成了技术栈调研。\n- **复用成熟框架**：依据列表中推荐的通用 RL 训练框架，团队直接集成了经过验证的代码库，将原本需两周的基础设施搭建工作缩短至两天。\n- **优化奖励模型**：参考列表中\"Reward & Training\"类别的成功实践，团队快速设计了基于页面状态变化的稀疏奖励函数，显著提升了代理的操作成功率。\n- **构建安全防线**：利用 AgentsMeetRL 提供的安全对齐项目，团队为代理添加了防注入和异常拦截模块，确保其在真实网络环境中的鲁棒性。\n\nAgentsMeetRL 将分散的强化学习智能体资源转化为结构化的技术地图，帮助开发者从“盲目摸索”转向“站在巨人肩膀上创新”。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fthinkwee_AgentsMeetRL_020ab9c7.png","thinkwee","Wei Liu","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fthinkwee_b7ecdce4.png","Too Stupid to Give Up Thinking","King's College London","London",null,"thinkwee2767","https:\u002F\u002Fthinkwee.top\u002Fabout","https:\u002F\u002Fgithub.com\u002Fthinkwee",[83],{"name":84,"color":85,"percentage":86},"HTML","#e34c26",100,922,41,"2026-04-07T07:09:14",1,"","未说明",{"notes":94,"python":92,"dependencies":95},"AgentsMeetRL 本身是一个开源项目汇总列表（Awesome List），用于整理和分类使用强化学习训练 LLM Agent 的仓库，而非一个可直接运行的单一软件工具。因此，README 中未提供具体的操作系统、GPU、内存、Python 版本或依赖库安装需求。用户需根据列表中具体引用的子项目（如 OpenRLHF, veRL, trl 等）查阅其各自的文档以获取运行环境要求。",[],[13,35,14],[98,99,100,101,102,103,104,105,106,107,108,109,110],"agent","agentic-ai","agentic-coding","agentic-workflow","awesome-list","large-language-model","llm","multiagent","reinforcement","rlhf","rlvr","tool-learning","llm-age","2026-03-27T02:49:30.150509","2026-04-08T01:08:46.619626",[],[]]