[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-DavidZWZ--Awesome-Deep-Research":3,"tool-DavidZWZ--Awesome-Deep-Research":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",151314,2,"2026-04-11T23:32:58",[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。",108322,"2026-04-10T11:39:34",[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},6121,"gemini-cli","google-gemini\u002Fgemini-cli","gemini-cli 是一款由谷歌推出的开源 AI 命令行工具，它将强大的 Gemini 大模型能力直接集成到用户的终端环境中。对于习惯在命令行工作的开发者而言，它提供了一条从输入提示词到获取模型响应的最短路径，无需切换窗口即可享受智能辅助。\n\n这款工具主要解决了开发过程中频繁上下文切换的痛点，让用户能在熟悉的终端界面内直接完成代码理解、生成、调试以及自动化运维任务。无论是查询大型代码库、根据草图生成应用，还是执行复杂的 Git 操作，gemini-cli 都能通过自然语言指令高效处理。\n\n它特别适合广大软件工程师、DevOps 人员及技术研究人员使用。其核心亮点包括支持高达 100 万 token 的超长上下文窗口，具备出色的逻辑推理能力；内置 Google 搜索、文件操作及 Shell 命令执行等实用工具；更独特的是，它支持 MCP（模型上下文协议），允许用户灵活扩展自定义集成，连接如图像生成等外部能力。此外，个人谷歌账号即可享受免费的额度支持，且项目基于 Apache 2.0 协议完全开源，是提升终端工作效率的理想助手。",100752,"2026-04-10T01:20:03",[52,13,15,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":32,"last_commit_at":59,"category_tags":60,"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],{"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":78,"owner_website":79,"owner_url":80,"languages":78,"stars":81,"forks":82,"last_commit_at":83,"license":84,"difficulty_score":85,"env_os":86,"env_gpu":87,"env_ram":87,"env_deps":88,"category_tags":91,"github_topics":93,"view_count":32,"oss_zip_url":78,"oss_zip_packed_at":78,"status":17,"created_at":104,"updated_at":105,"faqs":106,"releases":107},6183,"DavidZWZ\u002FAwesome-Deep-Research","Awesome-Deep-Research","[Up-to-date] Awesome Agentic Deep Research Resources","Awesome-Deep-Research 是一个专注于“代理式深度研究”（Agentic Deep Research）领域的开源资源合集，旨在为探索人工智能与自主智能体交叉前沿的用户提供一站式指南。随着信息检索范式从传统的网页搜索向具备推理能力的智能体搜索转变，该仓库解决了从业者难以系统性获取最新行业动态、开源实现及学术成果的痛点。\n\n它精心整理了包括 Google Gemini、OpenAI、Perplexity 等巨头发布的领先产品，以及相关的开源工具、最新研究论文和评估基准。其独特亮点在于不仅提供了资源列表，还关联了阐述该领域未来方向的位置论文，并推荐了相关的 RAG 与推理技术集合，帮助用户深入理解如何利用智能体进行复杂的信息分析与推理。\n\n无论是希望紧跟技术趋势的研究人员、需要落地实现的开发者，还是对自主智能体充满热情的爱好者，都能从中找到极具价值的参考内容。通过汇聚全球前沿成果，Awesome-Deep-Research 成为了连接理论研究与工程实践的重要桥梁，助力用户高效掌握下一代智能搜索与分析技术的核心脉络。","# 🤖 Awesome Agentic Deep Research Resources\n\n![Awesome](https:\u002F\u002Fcdn.rawgit.com\u002Fsindresorhus\u002Fawesome\u002Fd7305f38d29fed78fa85652e3a63e154dd8e8829\u002Fmedia\u002Fbadge.svg)\n[![arXiv](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-Agentic_Deep_Research-b31b1b.svg)]()\n[![Maintenance](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMaintained%3F-yes-green.svg)]()\n[![Contribution Welcome](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FContributions-welcome-blue)]()\n[![Code](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCode-GithubRepo-c8a4bf.svg)](https:\u002F\u002Fgithub.com\u002FDavidZWZ\u002FAwesome-Deep-Research) \n\n\u003Cp align=\"center\">\n    \u003Cimg src=\"https:\u002F\u002Fi.imgur.com\u002FwaxVImv.png\" alt=\"Oryx Video-ChatGPT\">\n\u003C\u002Fp>\n\nWelcome to Awesome-Deep-Research! 🚀 This repository serves as your comprehensive guide to the cutting-edge world of Agentic Deep Research. We've meticulously curated a collection of resources for you.\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_3cef4aaed6a8.png\" alt=\"DeepResearch Framework\" width=\"80%\" \u002F>\n\u003C\u002Fp>\n\nWhether you're a researcher, developer, or enthusiast, this repository is your gateway to exploring the fascinating intersection of artificial intelligence and autonomous agents. For a detailed analysis of the changing paradigm in information search, check out our position paper: **[From Web Search towards Agentic Deep Research: Incentivizing Search with Reasoning Agents](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.18959)** 📄, which outlines existing domain trends and future directions. For researchers interested in the broader intersection of RAG and Reasoning, we also recommend exploring our comprehensive collection at **[Awesome-RAG-Reasoning](https:\u002F\u002Fgithub.com\u002FDavidZWZ\u002FAwesome-RAG-Reasoning)** 🔥🔥🔥.\n\n## Table of Contents\n- [🎯 Industry-leading products and solutions](#industry-leading-products)\n- [🔧 Open-source implementations and tools](#open-source-implementations)\n- [📚 Latest research papers and breakthroughs](#latest-research-papers)\n- [🏆 Evaluation benchmarks and practical applications](#benchmarks-and-applications)\n- [🤝 Contributing and Citations](#contributing-and-citations)\n\n\n## Industry-Leading Products\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_97183aa10f21.png\" width=\"20\" style=\"vertical-align: -10px;\"\u002F> [Gemini Deep Research](https:\u002F\u002Fgemini.google\u002Foverview\u002Fdeep-research\u002F?hl=en): Google's advanced research assistant for deep analysis (December 11, 2024)\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_bbd16dd1caeb.png\" width=\"20\" style=\"vertical-align: -10px;\"\u002F> [Deep Research](https:\u002F\u002Fopenai.com\u002Findex\u002Fintroducing-deep-research\u002F): OpenAI's deep research platform [[API Guide]](https:\u002F\u002Fcookbook.openai.com\u002Fexamples\u002Fdeep_research_api\u002Fintroduction_to_deep_research_api) (February 2, 2025) \n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_cdee84882928.png\" width=\"16\" style=\"vertical-align: -15px;\"\u002F> [Perplexity Deep Research](https:\u002F\u002Fwww.perplexity.ai\u002Fhub\u002Fblog\u002Fintroducing-perplexity-deep-research): Perplexity's product for in-depth research and analysis (February 14, 2025)\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_f4c38bf79763.png\" width=\"20\" style=\"vertical-align: -10px;\"\u002F> [Grok Agents](https:\u002F\u002Fx.ai\u002Fnews\u002Fgrok-3): xAI's autonomous DeepSearch agents powered by Grok-3 (February 19, 2025)\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_c7fafc73ca50.png\" width=\"16\" style=\"vertical-align: -10px;\"\u002F> [Copilot Researcher](https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fmicrosoft-365\u002Fblog\u002F2025\u002F03\u002F25\u002Fintroducing-researcher-and-analyst-in-microsoft-365-copilot\u002F): Researcher and Analyst in Microsoft 365 Copilot (March 25, 2025)\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_cfa79ca1614b.png\" width=\"20\" style=\"vertical-align: -10px;\"\u002F> [Research](https:\u002F\u002Fwww.anthropic.com\u002Fnews\u002Fresearch): Anthropic's research platform to find and reason with information (April 15, 2025)\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_4dcf6bf0b98b.png\" width=\"20\" style=\"vertical-align: -10px;\"\u002F> [Manus](https:\u002F\u002Fmanus.im\u002F): Advanced research and analysis platform (March 6, 2025)\n- 🦌 [DeerFlow](https:\u002F\u002Fdeerflow.tech\u002F): ByteDance's research and analysis solution (May 9, 2025)\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_e1983aad06be.png\" width=\"16\" style=\"vertical-align: -15px;\"\u002F> [Deep Research](https:\u002F\u002Fchat.qwen.ai\u002F?inputFeature=deep_research): Alibaba's Qwen-powered research assistant (May 14, 2025)\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_7350fb20fac0.png\" width=\"16\" style=\"vertical-align: -10px;\"\u002F> [Kimi-Researcher](https:\u002F\u002Fmoonshotai.github.io\u002FKimi-Researcher\u002F): Moonshot's research assistant powered by Kimi (June 20, 2025)\n\n\n## Open-Source Implementations\n- [gemini-fullstack-langgraph-quickstart](https:\u002F\u002Fgithub.com\u002Fgoogle-gemini\u002Fgemini-fullstack-langgraph-quickstart): Gemini fullstack and LangGraph integration. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgoogle-gemini\u002Fgemini-fullstack-langgraph-quickstart?style=social)\n- [multi-agent research system](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fanthropic-cookbook\u002Ftree\u002Fmain\u002Fpatterns\u002Fagents\u002Fprompts): Multi-agent research system by Anthropic. [Blog post](https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fbuilt-multi-agent-research-system) ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fanthropics\u002Fanthropic-cookbook?style=social)\n- [gpt-researcher](https:\u002F\u002Fgithub.com\u002Fassafelovic\u002Fgpt-researcher): Autonomous agent for comprehensive research tasks. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fassafelovic\u002Fgpt-researcher?style=social)\n- [DeerFlow](https:\u002F\u002Fgithub.com\u002Fbytedance\u002Fdeer-flow): ByteDance's open-source deep research framework. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbytedance\u002Fdeer-flow?style=social)\n- [r1-reasoning-rag](https:\u002F\u002Fgithub.com\u002Fdeansaco\u002Fr1-reasoning-rag): Reasoning-augmented retrieval-augmented generation framework. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdeansaco\u002Fr1-reasoning-rag?style=social)\n- [nanoDeepResearch](https:\u002F\u002Fgithub.com\u002Fliyuan24\u002FnanoDeepResearch): Lightweight deep research toolkit. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fliyuan24\u002FnanoDeepResearch?style=social)\n- [deep-research (Aomni)](https:\u002F\u002Fgithub.com\u002Fdzhng\u002Fdeep-research): Deep research assistant by Aomni. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdzhng\u002Fdeep-research?style=social)\n- [deep-research (u14app)](https:\u002F\u002Fgithub.com\u002Fu14app\u002Fdeep-research): Deep research platform by u14app. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fu14app\u002Fdeep-research?style=social)\n- [open-deep-research](https:\u002F\u002Fgithub.com\u002Fbtahir\u002Fopen-deep-research): Open-source deep research framework. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbtahir\u002Fopen-deep-research?style=social)\n- [deep-searcher](https:\u002F\u002Fgithub.com\u002Fzilliztech\u002Fdeep-searcher): Deep search and research toolkit. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fzilliztech\u002Fdeep-searcher?style=social)\n- [node-DeepResearch](https:\u002F\u002Fgithub.com\u002Fjina-ai\u002Fnode-DeepResearch): Deep research toolkit to find the right answers. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fjina-ai\u002Fnode-DeepResearch?style=social)\n- [Auto-Deep-Research](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FAuto-Deep-Research): Automated deep research agent. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FHKUDS\u002FAuto-Deep-Research?style=social)\n- [langgraph-deep-research](https:\u002F\u002Fgithub.com\u002Fforeveryh\u002Flanggraph-deep-research): Deep research workflows with LangGraph. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fforeveryh\u002Flanggraph-deep-research?style=social)\n- [DeepResearchAgent](https:\u002F\u002Fgithub.com\u002FSkyworkAI\u002FDeepResearchAgent): Deep research agent by SkyworkAI. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSkyworkAI\u002FDeepResearchAgent?style=social)\n- [OpenManus](https:\u002F\u002Fgithub.com\u002FFoundationAgents\u002FOpenManus): An open-source framework for building general AI agents. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FFoundationAgents\u002FOpenManus?style=social)\n- [PraisonAI](https:\u002F\u002Fgithub.com\u002FMervinPraison\u002FPraisonAI): Production-ready multi-agent framework with built-in deep research capabilities. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMervinPraison\u002FPraisonAI?style=social)\n- [AtomSearcher](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus): An Automated deep research agent. ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fantgroup\u002FResearch-Venus?style=social)\n\n## Latest Research Papers\n\n🔥🔥🔥 This section showcases the most recent and impactful research papers in the field of Agentic Deep Research. Each paper represents a significant advancement in the development of autonomous research agents, search capabilities, and reasoning frameworks. The papers are organized chronologically, with the most recent publications at the top. Key areas covered include:\n- 🤖 Agentic frameworks for deep research\n- 🔍 Search-enhanced reasoning models\n- 🌐 Web agents for deep research\n- 🔄 Reasoning and retrieval-augmented generation\n- 📊 Multimodal deep research\n\n🚀🚀🚀 Stay tuned for the hottest breakthroughs in the field!\n\n\n| Title | Date & Code | Base model | Optimization | Search Engine | Agent Architecture | Training Dataset | Evaluation Dataset |\n| --- | :---: | --- | --- | --- | --- | --- | --- |\n| [Dr. Zero: Self-Evolving Search Agents without Training Data](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.07055) | [2026\u002F01\u002F11](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fdrzero) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ffacebookresearch\u002Fdrzero?style=social)](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fdrzero) | Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct | HRPO | Web Search | Multi-Agent | – | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMultihopQA (2WikiMQA), MuSiQue, Bamboogle |\n| [LEAPS: An LLM-Empowered Adaptive Plugin for Taobao AI Search](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.05513) | [2026\u002F01\u002F09]() | Qwen3-14B | REINFORCE++, GRPO, GSPO | Local Retrieval | Single-Agent | – | – |\n| [SmartSearch: Process Reward-Guided Query Refinement for Search Agents](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.04888) | [2026\u002F01\u002F08](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FSmartSearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUC-NLPIR\u002FSmartSearch?style=social)](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FSmartSearch) | Qwen2.5-3B-Instruct | SFT, DPO, GRPO | Web Search | Single-Agent | Asearcher-Base | 2WikiMultihopQA, HotpotQA, Bamboogle, MuSiQue, GAIA, WebWalker |\n| [O-Researcher: An Open Ended Deep Research Model via Multi-Agent Distillation and Agentic RL](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.03743v1) | [2026\u002F01\u002F07](https:\u002F\u002Fgithub.com\u002FOPPO-PersonalAI\u002FO-Researcher) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOPPO-PersonalAI\u002FO-Researcher?style=social)](https:\u002F\u002Fgithub.com\u002FOPPO-PersonalAI\u002FO-Researcher) | Qwen-2.5-72B-Instruct | GRPO | Web Search | Multi-Agent | Zhihu-KOL, WideSearch, ELI5 | DeepResearch Bench, DeepResearchGym |\n| [WebAnchor: Anchoring Agent Planning to Stabilize Long-Horizon Web Reasoning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.03164) | [2026\u002F01\u002F06]() | WebSailor-3B\u002F7B, Tongyi-DR-30B, Qwen-2.5-72B | GRPO | Local Retrieval | Single-Agent | – | BrowseComp-en, BrowseComp-zh, XBench-DeepSearch, GAIA |\n| [Budget-Aware Tool-Use Enables Effective Agent Scaling](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.17006) | [2025\u002F11\u002F21]() | Gemini-2.5-Flash, Gemini-2.5-Pro, Claude-Sonnet-4 | Prompting | Web Search | Single-Agent | – | – |\n| [AutoTool: Efficient Tool Selection for Large Language Model Agents](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.14650) | [2025\u002F11\u002F18](https:\u002F\u002Fgithub.com\u002Fjiajingyyyyyy\u002FAutoTool) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fjiajingyyyyyy\u002FAutoTool?style=social)](https:\u002F\u002Fgithub.com\u002Fjiajingyyyyyy\u002FAutoTool) | Llama4-Scout-17B | Prompting | Web Search | Single-Agent | – | AlfWorld, ScienceWorld, ToolQuery-Academia |\n| [Multi-Agent Deep Research: Training Multi-Agent Systems with M-GRPO](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.13288) | [2025\u002F11\u002F17](https:\u002F\u002Fgithub.com\u002FAQ-MedAI\u002FMrlX) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAQ-MedAI\u002FMrlX?style=social)](https:\u002F\u002Fgithub.com\u002FAQ-MedAI\u002FMrlX) | Qwen3-30B-A3B | M-GRPO | Web Search | Multi-Agent | – | GAIA, XBench-DeepSearch, WebWalkerQA |\n| [Tongyi DeepResearch Technical Report](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.24701) | [2025\u002F10\u002F28](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | Qwen3-30B-A3B-Base | SFT, RL | Web Search | Single-Agent | – | HLE, BrowseComp, BrowseComp-ZH, GAIA, XBench-DeepSearch, WebWalkerQA, FRAMES, XBench-DeepSearch-2510 |\n| [TOOLRM: Towards Agentic Tool-Use Reward Modeling](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.26167) | [2025\u002F10\u002F30](https:\u002F\u002Fgithub.com\u002Flirenhao1997\u002FToolRM) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flirenhao1997\u002FToolRM?style=social)](https:\u002F\u002Fgithub.com\u002Flirenhao1997\u002FToolRM) | Qwen3-4B, Qwen3-8B | RL | Web Search | Single-Agent | ToolPref-Pairwise-30K | TRBench, ACEBench |\n| [ToolScope: An Agentic Framework for Vision-Guided and Long-Horizon Tool Use](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.27363) | [2025\u002F10\u002F31]() | GPT-4o, Gemini-2.5, Qwen2.5-VL, Llama-3.2-Vision | Prompting | Local Retrieval | Multi-Agent | – | – |\n| [WebSeer: Training Deeper Search Agents through Reinforcement Learning with Self-Reflection](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.18798) | [2025\u002F10\u002F21](https:\u002F\u002Fgithub.com\u002F99hgz\u002FWebSeer) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002F99hgz\u002FWebSeer?style=social)](https:\u002F\u002Fgithub.com\u002F99hgz\u002FWebSeer) | Qwen-2.5-14B, Qwen-3-14B | RL (cold start + RL; self-reflection) | Web Search | Single-Agent | – | HotpotQA, SimpleQA |\n| [Enterprise Deep Research: Steerable MultiAgent Deep Research for Enterprise Analytics](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2510.17797) | [2025\u002F10\u002F20](https:\u002F\u002Fgithub.com\u002FSalesforceAIResearch\u002Fenterprise-deep-research) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSalesforceAIResearch\u002Fenterprise-deep-research?style=social)](https:\u002F\u002Fgithub.com\u002FSalesforceAIResearch\u002Fenterprise-deep-research) | – | Prompting | Web Search | Multi-Agent | – | DeepResearch Bench, DeepConsult |\n| [Stop-RAG: Value-Based Retrieval Control for Iterative RAG](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.14337) | [2025\u002F10\u002F16](https:\u002F\u002Fgithub.com\u002Fchosolbee\u002FStop-RAG) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fchosolbee\u002FStop-RAG?style=social)](https:\u002F\u002Fgithub.com\u002Fchosolbee\u002FStop-RAG) | Llama-3.1-8B-Instruct | Fine-tuning | Local Retrieval | Single-Agent | MuSiQue, HotpotQA, 2WikiMultihopQA | HotpotQA, MuSiQue, 2WikiMultihopQA |\n| [Towards Agentic Self-Learning LLMs in Search Environment](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.14253) | [2025\u002F10\u002F16](https:\u002F\u002Fgithub.com\u002Fforangel2014\u002FTowards-Agentic-Self-Learning) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fforangel2014\u002FTowards-Agentic-Self-Learning?style=social)](https:\u002F\u002Fgithub.com\u002Fforangel2014\u002FTowards-Agentic-Self-Learning) | Qwen-2.5-7B-Instruct | RL | Web Search | Multi-Agent | – | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMultiHopQA, MuSiQue, Bamboogle |\n| [GOAT: A Training Framework for Goal-Oriented Agent with Tools](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.12218) | [2025\u002F10\u002F14]() | Qwen-2-7B, Llama-3-8B-Instruct, Llama-3-70B-Instruct | Fine-tuning | Web Search | Single-Agent | – | GOATBench |\n| [ResearStudio: A Human-Intervenable Framework for Building Controllable Deep-Research Agents](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.12194) | [2025\u002F10\u002F14](https:\u002F\u002Fgithub.com\u002FResearAI\u002FResearStudio) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FResearAI\u002FResearStudio?style=social)](https:\u002F\u002Fgithub.com\u002FResearAI\u002FResearStudio) | gpt-4.1, gpt-4.1-mini, o4-mini, Llama-3.3-70B | Prompting | Web Search | Single-Agent | – | GAIA |\n| [HiPRAG: Hierarchical Process Rewards for Efficient Agentic Retrieval Augmented Generation](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.07794) | [2025\u002F10\u002F09](https:\u002F\u002Fgithub.com\u002Fqualidea1217\u002FHiPRAG) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fqualidea1217\u002FHiPRAG?style=social)](https:\u002F\u002Fgithub.com\u002Fqualidea1217\u002FHiPRAG) | Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct, Llama-3.2-3B-Instruct | PPO, GRPO | Web Search | Single-Agent | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2Wiki, MuSiQue, Bamboogle |\n| [A2SEARCH: Ambiguity-Aware Question Answering with Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.07958) | [2025\u002F10\u002F09](https:\u002F\u002Fgithub.com\u002Fzfj1998\u002FA2Search) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fzfj1998\u002FA2Search?style=social)](https:\u002F\u002Fgithub.com\u002Fzfj1998\u002FA2Search) | Qwen-2.5 family | RL | Web Search | Single-Agent | NQ | MuSiQue, HotpotQA, 2Wiki, Bamboogle, NQ, TriviaQA, PopQA, AmbigQA |\n| [ReSeek: A Self-Correcting Framework for Search Agents with Instructive Rewards](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2510.00568) | [2025\u002F10\u002F01](https:\u002F\u002Ftencentbac.github.io\u002FReSeek\u002F) | Qwen2.5-7B-Instruct, Qwen2.5-3B-Instruct | GRPO | Web Search, Local Retrieval | Single-Agent | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMQA, MuSiQue, Bamboogle, FictionalHot |\n| [Process-Supervised Reinforcement Learning for Interactive Multimodal Tool-Use Agents](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.14480) | [2025\u002F09\u002F17]() | Qwen3-8B, Qwen2.5-Omni-7B | RL | Local Retrieval | Single-Agent | τ-bench, APIGen-MT | τ-bench |\n| [ReSum: Unlocking Long-Horizon Search Intelligence via Context Summarization](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.13313) | [2025\u002F09\u002F16](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | Qwen3-30B-A3B-Thinking | GRPO, SFT | Web Search | Single-Agent | SailorFog-QA | BrowseComp-en\u002Fzh |\n| [WebWeaver: Structuring Web-Scale Evidence with Dynamic Outlines for Open-Ended Deep Research](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.13312) | [2025\u002F09\u002F16](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | Qwen3-30B-A3B-Instruct | SFT | Web Search | Single-Agent | WebWeaver-3k | BrowseComp-en\u002Fzh, GAIA, XBench-DeepSearch |\n| [WebResearcher: Unleashing Unbounded Reasoning Capability in Long-Horizon Agents](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.13309) | [2025\u002F09\u002F16](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | Qwen3-30B-A3B | RFT, RL | Web Search | Multi-Agent | WebFrontier | BrowseComp-en\u002Fzh, GAIA, WebWalkerQA, FRAMES, HotpotQA, MuSiQue, 2WikiMultiHopQA |\n| [WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable RL](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.13305) | [2025\u002F09\u002F16](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | Qwen3-30B-A3B | SFT, RL | Web Search, Local Retrieval | Single-Agent | SailorFog-QA-V2 | BrowseComp-EN, BrowseComp-ZH, HLE |\n| [WebExplorer: Explore and Evolve for Training Long-Horizon Web Agents](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.06501) | [2025\u002F09\u002F08](https:\u002F\u002Fgithub.com\u002Fhkust-nlp\u002FWebExplorer) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhkust-nlp\u002FWebExplorer?style=social)](https:\u002F\u002Fgithub.com\u002Fhkust-nlp\u002FWebExplorer) | Qwen3-8B | GRPO, SFT | Web Search | Single-Agent | WebExplorer-QA | BrowseComp-en\u002Fzh, GAIA, WebWalkerQA, FRAMES, XBench-DeepSearch, HLE |\n| [Atom-Searcher: Enhancing Agentic Deep Research via Fine-Grained Atomic Thought Reward](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.12800) | [2025\u002F08\u002F18](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fantgroup\u002FResearch-Venus?style=social)](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus) | Qwen2.5-7B | RL(GRPO) | Web Search | Single-Agent | NQ, SimpleQA, HotpotQA, 2WikiMultiHopQA, MuSiQue, MultiHopRAG | Bamboogle, NQ, SimpleQA, HotpotQA, 2WikiMultiHopQA, MuSiQue, MultiHopRAG |\n| [MMSearch-R1: Incentivizing LMMs to Search](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.20670) | [2025\u002F06\u002F25](https:\u002F\u002Fgithub.com\u002FEvolvingLMMs-Lab\u002Fmultimodal-search-r1) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FEvolvingLMMs-Lab\u002Fmultimodal-search-r1?style=social)](https:\u002F\u002Fgithub.com\u002FRAG-Gym\u002FRAG-Gym) | Qwen2.5-VL-7B | RL(GRPO) | Web Search | Single-Agent | VQA, MetaClip, FVQA, InfoSeek | FVQA-test, InfoSeek, MMSearch, SimpleVQA, LiveVQA |\n| [VideoDeepResearch: Long Video Understanding With Agentic Tool Using](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.10821) | [2025\u002F06\u002F12](https:\u002F\u002Fgithub.com\u002Fyhy-2000\u002FVideoDeepResearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fyhy-2000\u002FVideoDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002Fyhy-2000\u002FVideoDeepResearch) | GPT-4o, Gemini1.5-pro, Qwen2.5-VL-72B-Instruct | Prompting | Local Retrieval | Multi-Agent | – | MLVU, Video-MME, LVBench, LongVideoBench |\n| [Multimodal DeepResearcher: Generating Text-Chart Interleaved Reports From Scratch with Agentic Framework](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.02454) | [2025\u002F06\u002F03]() | Claude3.7-Sonnet, GPT-4o-mini, Qwen3-235B-A22B, Qwen2.5-VL-72B-Instruct | Prompting | Web Search | Multi-Agent | – | Pew Research, Our World in Data, Open Knowledge Foundation |\n| [RAG-Gym: Systematic Optimization of Language Agents for Retrieval-Augmented Generation](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2502.13957v2) | [2025\u002F05\u002F31](https:\u002F\u002Fgithub.com\u002FRAG-Gym\u002FRAG-Gym) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRAG-Gym\u002FRAG-Gym?style=social)](https:\u002F\u002Fgithub.com\u002FRAG-Gym\u002FRAG-Gym) | Llama3.1-8B-Instruct, Qwen2.5-7B-Instruct, GPT-4o-mini | SFT, RL(PPO, DPO) | Local Retrieval | Single-Agent | HotpotQA, MedQA | HotpotQA, 2Wiki, Bamboogle, MedQA |\n| [MaskSearch: A Universal Pre-Training Framework to Enhance Agentic Search Capability](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.20285) | [2025\u002F05\u002F27](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FMaskSearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FMaskSearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FMaskSearch) | Llama3.1-8B, Llama3.2-3B, Llama3.2-1B, Llama3, Qwen2.5-7B, Qwen2.5-3B, Qwen2.5-1.5B, Qwen2.5 | SFT, RL(DAPO) | Local Retrieval | Multi-Agent | HotpotQA | HotpotQA, FanoutQA, Musique, 2WikiMultiHopQA, Bamboogle, FreshQA |\n| [SimpleDeepSearcher: Deep Information Seeking via Web-Powered Reasoning Trajectory Synthesis](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.16834) | [2025\u002F05\u002F25](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FSimpleDeepSearcher?tab=readme-ov-file) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUCAIBox\u002FSimpleDeepSearcher?style=social)](https:\u002F\u002Fgithub.com\u002Fweizhepei\u002FWebAgent-R1) | Qwen2.5-7B-Instruct, Qwen2.5-32B-Instruct, DeepseekDistilled-Qwen2.5-32B, QwQ-32B | SFT | Web Search | Single-Agent | NQ, SimpleQA, HotpotQA, 2WikiMultiHopQA, MuSiQue, MultiHopRAG | Bamboogle, FRAMES, GAIA, NQ, SimpleQA, HotpotQA, 2WikiMultiHopQA, MuSiQue, MultiHopRAG |\n| [WebAgent-R1: Training Web Agents via End-to-End Multi-Turn Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.16421) | [2025\u002F05\u002F22](https:\u002F\u002Fgithub.com\u002Fweizhepei\u002FWebAgent-R1) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fweizhepei\u002FWebAgent-R1?style=social)](https:\u002F\u002Fgithub.com\u002Fweizhepei\u002FWebAgent-R1) | Qwen2.5-3B, Llama3.1-8B | SFT, RL(M-GRPO) | Web Search | Single-Agent | WebArena-Lite, WebArena | WebArena-Lite, WebArena |\n| [R1-Searcher++: Incentivizing the Dynamic Knowledge Acquisition of LLMs via Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.17005) | [2025\u002F05\u002F22](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher-plus) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUCAIBox\u002FR1-Searcher-plus?style=social)](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher-plus) | Qwen2.5-7B-Instruct | SFT, RL | Local Retrieval | Single-Agent | HotpotQA, 2WikiMultiHopQA | HotpotQA, 2WikiMultiHopQA, Musique, Bamboogle |\n| [Process vs. Outcome Reward: Which is Better for Agentic RAG Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.14069) | [2025\u002F05\u002F22](https:\u002F\u002Fgithub.com\u002Fwlzhang2020\u002FReasonRAG) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fwlzhang2020\u002FReasonRAG?style=social)](https:\u002F\u002Fgithub.com\u002Fwlzhang2020\u002FReasonRAG) | Qwen2.5-7B-Instruct | RL(DPO) | Local Retrieval | Single-Agent | PopQA, HotpotQA, 2WikiMultihopQA | PopQA, HotpotQA, 2WikiMultiHopQA, Bamboogle, MuSiQue |\n| [s3 - Efficient Yet Effective Search Agent Training via RL](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.14146) | [2025\u002F05\u002F20](https:\u002F\u002Fgithub.com\u002Fpat-jj\u002Fs3) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpat-jj\u002Fs3?style=social)](https:\u002F\u002Fgithub.com\u002Fpat-jj\u002Fs3) | Qwen2.5-7B-Instruct | RL(PPO) | Local Retrieval | Single-Agent | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2wiki, Musique, MedQA-US, MedMCQA, PubMedQA, BioASQ-Y\u002FN, MMLU-Med |\n| [Demystifying and Enhancing the Efficiency of Large Language Model Based Search Agents](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.12065) | [2025\u002F05\u002F17](https:\u002F\u002Fgithub.com\u002Ftiannuo-yang\u002FSearchAgent-X) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftiannuo-yang\u002FSearchAgent-X?style=social)](https:\u002F\u002Fgithub.com\u002Ftiannuo-yang\u002FSearchAgent-X) | Qwen2.5-14B, Qwen2.5-7B | Prompting | Local Retrieval | Single-Agent | – | Musique, NQ, 2WikiMultiHopQA, HotpotQA, Bamboogle, StrategyQA |\n| [Reinforced Internal-External Knowledge Synergistic Reasoning for Efficient Adaptive Search Agent](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.07596) | [2025\u002F05\u002F12](https:\u002F\u002Fgithub.com\u002Fhzy312\u002Fknowledge-r1) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhzy312\u002Fknowledge-r1?style=social)](https:\u002F\u002Fgithub.com\u002Fhzy312\u002Fknowledge-r1) | Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct | RL(GRPO) | Local Retrieval | Single-Agent | NQ, HotpotQA | PopQA, 2WikiMultihopQA |\n| [ZeroSearch: Incentivize the Search Capability of LLMs without Searching](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.04588) | [2025\u002F05\u002F07](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FZeroSearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FZeroSearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FZeroSearch) | Qwen2.5-3B-Base, Qwen2.5-7B-Base, Qwen2.5-7B-Instruct, Qwen2.5-3B-Instruct, Llama3.2-3B-Instruct, Llama3.2-3B-Base | RL(Reinforce, GRPO, PPO) | Web Search | Single-Agent | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMultiHopQA, Musique, Bamboogle |\n| [Webthinker: Empowering large reasoning models with deep research capability](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.21776) | [2025\u002F04\u002F30](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FWebThinker) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUC-NLPIR\u002FWebThinker?style=social)](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FWebThinker) | GPT-o1, GPT-o3, Deepseek-R1, QwQ-32B, Qwen2.5-32B-Instruct | RL(DPO) | Web Search | Single-Agent | SuperGPQA, WebWalkerQA, OpenThoughts, NaturalReasoning, NuminaMath | GPQA, GAIA, WebWalkerQA, Humanity’s Last Exam |\n| [Pangu Ultra: Pushing the Limits of Dense Large Language Models on Ascend NPUs](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.07866) | [2025\u002F04\u002F11](https:\u002F\u002Fgithub.com\u002Fpangu-tech\u002Fpangu-ultra) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpangu-tech\u002Fpangu-ultra?style=social)](https:\u002F\u002Fgithub.com\u002Fpangu-tech\u002Fpangu-ultra) | Pangu Ultra-135B | SFT, RL | Local Retrieval | Single-Agent | – | – |\n| [Open Deep Search: Democratizing Search with Open-source Reasoning Agents](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.20201) | [2025\u002F03\u002F26](https:\u002F\u002Fgithub.com\u002Fsentient-agi\u002FOpenDeepSearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsentient-agi\u002FOpenDeepSearch?style=social)](https:\u002F\u002Fgithub.com\u002Fsentient-agi\u002FOpenDeepSearch) | Llama3.1-70B, Deepseek-R1 | Prompting | Web Search | Single-Agent | – | SimpleQA, FRAME |\n| [DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real-world Environments](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.03160?) | [2025\u002F03\u002F26](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FDeepResearcher) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGAIR-NLP\u002FDeepResearcher?style=social)](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FDeepResearcher) | Qwen2.5-7B-Instruct | RL(GRPO) | Web Search | Multi-Agent | NQ, TQ, HotpotQA, 2WikiMultiHopQA | MuSiQue, Bamboogle, PopQA, NQ, TQ, HotpotQA, 2WikiMultiHopQA |\n| [ReSearch: Learning to Reason with Search for LLMs via Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.19470) | [2025\u002F03\u002F25](https:\u002F\u002Fgithub.com\u002FAgent-RL\u002FReSearch) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAgent-RL\u002FReSearch?style=social)](https:\u002F\u002Fgithub.com\u002FAgent-RL\u002FReSearch) | Qwen2.5-7B-Instruct, Qwen2.5-32B-Instruct | RL(GRPO) | Web Search | Single-Agent | MuSiQue | HotpotQA, 2WikiMultiHopQA, Musique, Bamboogle |\n| [Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.09516) | [2025\u002F03\u002F12](https:\u002F\u002Fgithub.com\u002FPeterGriffinJin\u002FSearch-R1) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FPeterGriffinJin\u002FSearch-R1?style=social)](https:\u002F\u002Fgithub.com\u002FPeterGriffinJin\u002FSearch-R1) | Qwen2.5-7B-Instruct, Qwen2.5-7B-Base, Qwen2.5-3B-Instruct, Qwen2.5-3B-Base | RL(PPO, GRPO) | Web Search | Single-Agent | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMultiHopQA, Musique, Bamboogle |\n| [Beyond Outlining: Heterogeneous Recursive Planning for Adaptive Long-form Writing with Language Models](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.08275) | [2025\u002F03\u002F11](https:\u002F\u002Fgithub.com\u002Fprincipia-ai\u002FWriteHERE) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fprincipia-ai\u002FWriteHERE?style=social)](https:\u002F\u002Fgithub.com\u002Fprincipia-ai\u002FWriteHERE) | GPT-4o, Claude3.5-Sonnet | Prompting | Web Search | Multi-Agent | – | TELL ME A STORY, WildSeek |\n| [R1-Searcher: Incentivizing the Search Capability in LLMs via Reinforcement Learning](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.05592) | [2025\u002F03\u002F07](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUCAIBox\u002FR1-Searcher?style=social)](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher) | Qwen2.5-7B-Base, Llama3.1-8B-Instruct | SFT, RL(GRPO, Reinforce++) | Web Search, Local Retrieval | Single-Agent | HotpotQA, 2WikiMultiHopQA | HotpotQA, 2WikiMultiHopQA, Musique, Bamboogle |\n| [AutoAgent: A Fully-Automated and Zero-Code Framework for LLM Agents](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2502.05957#page=1.62) | [2025\u002F02\u002F18](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FAutoAgent?tab=readme-ov-file) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FHKUDS\u002FAutoAgent?style=social)](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FAutoAgent?tab=readme-ov-file) | Claude3.5-Sonnet | Prompting | Web Search | Multi-Agent | – | GAIA |\n| [Agentic Reasoning: Reasoning LLMs with Tools for the Deep Research](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.04644) | [2025\u002F02\u002F07](https:\u002F\u002Fgithub.com\u002Ftheworldofagents\u002FAgentic-Reasoning) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftheworldofagents\u002FAgentic-Reasoning?style=social)](https:\u002F\u002Fgithub.com\u002Ftheworldofagents\u002FAgentic-Reasoning) | N\u002FA | Prompting | Web Search | Multi-Agent | – | GPQA |\n| [Search-o1: Agentic Search-Enhanced Large Reasoning Models](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2501.05366) | [2025\u002F01\u002F09](https:\u002F\u002Fgithub.com\u002Fsunnynexus\u002FSearch-o1) [![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsunnynexus\u002FSearch-o1?style=social)](https:\u002F\u002Fgithub.com\u002Fsunnynexus\u002FSearch-o1) | QwQ-32B-Preview | Prompting | Web Search | Single-Agent | – | GPQA, MATH500, AMC2023, AIME2024, LiveCodeBench, NQ, TriviaQA, HotpotQA, 2WikiMultiHopQA, MuSiQue, Bamboogle |\n\n\n## Benchmarks and Applications\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_27c4fb9bebee.png\" alt=\"Benchmarks Plot\" width=\"80%\" \u002F>\n\u003C\u002Fp>\n\n- Humanity's Last Exam [[Paper]](https:\u002F\u002Farxiv.org\u002Fabs\u002F2501.14249) [[Code]](https:\u002F\u002Fgithub.com\u002Fcenterforaisafety\u002Fhle) ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcenterforaisafety\u002Fhle?style=social)\n- BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agents [[Paper]](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.12516) [[Code]](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fsimple-evals) ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopenai\u002Fsimple-evals?style=social)\n- BrowseComp-ZH: Benchmarking Web Browsing Ability of Large Language Models in Chinese ['[Paper]'](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.19314) [[Code]](https:\u002F\u002Fgithub.com\u002FPALIN2018\u002FBrowseComp-ZH) ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FPALIN2018\u002FBrowseComp-ZH?style=social)\n- DeepResearch Bench: A Comprehensive Benchmark for Deep Research Agents [[Paper]](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.11763) [[Code]](https:\u002F\u002Fgithub.com\u002FAyanami0730\u002Fdeep_research_bench) ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAyanami0730\u002Fdeep_research_bench?style=social)\n- MedBrowseComp: Benchmarking Medical Deep Research and Computer Use [[Paper]](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.14963) [[Code]](https:\u002F\u002Fgithub.com\u002Fshan23chen\u002FMedBrowseComp) ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fshan23chen\u002FMedBrowseComp?style=social)\n- Mind2Web 2: Evaluating Agentic Search with Agent-as-a-Judge [[Paper]](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.21506) [[Code]](https:\u002F\u002Fgithub.com\u002FOSU-NLP-Group\u002FMind2Web-2) ![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOSU-NLP-Group\u002FMind2Web-2?style=social)\n\n\n\n## Contributing and Citations\n\n🤝 We welcome contributions to expand this comprehensive collection of Agentic Deep Research resources! \n\n### 📝 How to Contribute\n\n**Adding New Research Papers and Benchmarks:**\n- Submit an issue with the paper details (title, arXiv link, all the categories in our paper table, and GitHub repo if available)\n- Or create a pull request with the paper added to the research papers table or the benchmarks section\n\n**Adding New Open-Source Implementations and New Products:**\n- Submit an issue with the repository details (name, description, release data, GitHub link if available)\n- Or create a pull request with the implementation added to the open-source and products section\n\n\n### 📖 Citation\n\n🔥🔥🔥 If you find this repository useful, please cite our papers:\n\n```bibtex\n@article{zhang2025web,\n  title={From Web Search towards Agentic Deep Research: Incentivizing Search with Reasoning Agents},\n  author={Zhang, Weizhi and Li, Yangning and Bei, Yuanchen and Luo, Junyu and Wan, Guancheng and Yang, Liangwei and Xie, Chenxuan and Yang, Yuyao and Huang, Wei-Chieh and Miao, Chunyu and others},\n  journal={arXiv preprint arXiv:2506.18959},\n  year={2025}\n}\n\n@article{li2025towards,\n  title={Towards Agentic RAG with Deep Reasoning: A Survey of RAG-Reasoning Systems in LLMs},\n  author={Li, Yangning and Zhang, Weizhi and Yang, Yuyao and Huang, Wei-Chieh and Wu, Yaozu and Luo, Junyu and Bei, Yuanchen and Zou, Henry Peng and Luo, Xiao and Zhao, Yusheng and others},\n  journal={arXiv preprint arXiv:2507.09477},\n  year={2025}\n}\n```\n","# 🤖 令人惊叹的代理式深度研究资源\n\n![Awesome](https:\u002F\u002Fcdn.rawgit.com\u002Fsindresorhus\u002Fawesome\u002Fd7305f38d29fed78fa85652e3a63e154dd8e8829\u002Fmedia\u002Fbadge.svg)\n[![arXiv](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-Agentic_Deep_Research-b31b1b.svg)]()\n[![维护中](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMaintained%3F-yes-green.svg)]()\n[![欢迎贡献](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FContributions-welcome-blue)]()\n[![代码](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCode-GithubRepo-c8a4bf.svg)](https:\u002F\u002Fgithub.com\u002FDavidZWZ\u002FAwesome-Deep-Research) \n\n\u003Cp align=\"center\">\n    \u003Cimg src=\"https:\u002F\u002Fi.imgur.com\u002FwaxVImv.png\" alt=\"Oryx Video-ChatGPT\">\n\u003C\u002Fp>\n\n欢迎来到 Awesome-Deep-Research！🚀 本仓库是您通往代理式深度研究前沿世界的全面指南。我们精心整理了一系列资源供您使用。\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_3cef4aaed6a8.png\" alt=\"DeepResearch 框架\" width=\"80%\" \u002F>\n\u003C\u002Fp>\n\n无论您是研究人员、开发者还是爱好者，这个仓库都是您探索人工智能与自主代理之间迷人交汇点的入口。如需深入了解信息搜索范式的转变，请查阅我们的立场论文：**[从网络搜索迈向代理式深度研究：以推理代理激励搜索](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.18959)** 📄，其中概述了当前领域的趋势及未来发展方向。对于关注 RAG 与推理更广泛交叉的研究人员，我们也推荐您浏览我们的综合资源集 **[Awesome-RAG-Reasoning](https:\u002F\u002Fgithub.com\u002FDavidZWZ\u002FAwesome-RAG-Reasoning)** 🔥🔥🔥。\n\n## 目录\n- [🎯 行业领先的产品与解决方案](#industry-leading-products)\n- [🔧 开源实现与工具](#open-source-implementations)\n- [📚 最新研究论文与突破性成果](#latest-research-papers)\n- [🏆 评估基准与实际应用](#benchmarks-and-applications)\n- [🤝 贡献与引用](#contributing-and-citations)\n\n\n## 行业领先的产品\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_97183aa10f21.png\" width=\"20\" style=\"vertical-align: -10px;\"\u002F> [Gemini Deep Research](https:\u002F\u002Fgemini.google\u002Foverview\u002Fdeep-research\u002F?hl=en): Google 的高级研究助手，用于深度分析（2024年12月11日）\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_bbd16dd1caeb.png\" width=\"20\" style=\"vertical-align: -10px;\"\u002F> [Deep Research](https:\u002F\u002Fopenai.com\u002Findex\u002Fintroducing-deep-research\u002F): OpenAI 的深度研究平台 [[API 指南]](https:\u002F\u002Fcookbook.openai.com\u002Fexamples\u002Fdeep_research_api\u002Fintroduction_to_deep_research_api)（2025年2月2日） \n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_cdee84882928.png\" width=\"16\" style=\"vertical-align: -15px;\"\u002F> [Perplexity Deep Research](https:\u002F\u002Fwww.perplexity.ai\u002Fhub\u002Fblog\u002Fintroducing-perplexity-deep-research): Perplexity 的深度研究与分析产品（2025年2月14日）\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_f4c38bf79763.png\" width=\"20\" style=\"vertical-align: -10px;\"\u002F> [Grok Agents](https:\u002F\u002Fx.ai\u002Fnews\u002Fgrok-3): xAI 基于 Grok-3 的自主 DeepSearch 代理（2025年2月19日）\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_c7fafc73ca50.png\" width=\"16\" style=\"vertical-align: -10px;\"\u002F> [Copilot Researcher](https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fmicrosoft-365\u002Fblog\u002F2025\u002F03\u002F25\u002Fintroducing-researcher-and-analyst-in-microsoft-365-copilot\u002F): Microsoft 365 Copilot 中的研究员与分析师功能（2025年3月25日）\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_cfa79ca1614b.png\" width=\"20\" style=\"vertical-align: -10px;\"\u002F> [Research](https:\u002F\u002Fwww.anthropic.com\u002Fnews\u002Fresearch): Anthropic 的研究平台，用于查找并推理信息（2025年4月15日）\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_4dcf6bf0b98b.png\" width=\"20\" style=\"vertical-align: -10px;\"\u002F> [Manus](https:\u002F\u002Fmanus.im\u002F): 高级研究与分析平台（2025年3月6日）\n- 🦌 [DeerFlow](https:\u002F\u002Fdeerflow.tech\u002F): 字节跳动的研究与分析解决方案（2025年5月9日）\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_e1983aad06be.png\" width=\"16\" style=\"vertical-align: -15px;\"\u002F> [Deep Research](https:\u002F\u002Fchat.qwen.ai\u002F?inputFeature=deep_research): 阿里巴巴基于通义千问的研究助手（2025年5月14日）\n- \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_7350fb20fac0.png\" width=\"16\" style=\"vertical-align: -10px;\"\u002F> [Kimi-Researcher](https:\u002F\u002Fmoonshotai.github.io\u002FKimi-Researcher\u002F): Moonshot 基于 Kimi 的研究助手（2025年6月20日）\n\n## 开源实现\n- [gemini-fullstack-langgraph-quickstart](https:\u002F\u002Fgithub.com\u002Fgoogle-gemini\u002Fgemini-fullstack-langgraph-quickstart)：Gemini 全栈与 LangGraph 集成。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fgoogle-gemini\u002Fgemini-fullstack-langgraph-quickstart?style=social)\n- [multi-agent research system](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fanthropic-cookbook\u002Ftree\u002Fmain\u002Fpatterns\u002Fagents\u002Fprompts)：Anthropic 的多智能体研究系统。[博客文章](https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fbuilt-multi-agent-research-system) ![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fanthropics\u002Fanthropic-cookbook?style=social)\n- [gpt-researcher](https:\u002F\u002Fgithub.com\u002Fassafelovic\u002Fgpt-researcher)：用于综合性研究任务的自主智能体。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fassafelovic\u002Fgpt-researcher?style=social)\n- [DeerFlow](https:\u002F\u002Fgithub.com\u002Fbytedance\u002Fdeer-flow)：字节跳动开源的深度研究框架。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbytedance\u002Fdeer-flow?style=social)\n- [r1-reasoning-rag](https:\u002F\u002Fgithub.com\u002Fdeansaco\u002Fr1-reasoning-rag)：推理增强型检索增强生成框架。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdeansaco\u002Fr1-reasoning-rag?style=social)\n- [nanoDeepResearch](https:\u002F\u002Fgithub.com\u002Fliyuan24\u002FnanoDeepResearch)：轻量级深度研究工具包。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fliyuan24\u002FnanoDeepResearch?style=social)\n- [deep-research (Aomni)](https:\u002F\u002Fgithub.com\u002Fdzhng\u002Fdeep-research)：Aomni 的深度研究助手。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fdzhng\u002Fdeep-research?style=social)\n- [deep-research (u14app)](https:\u002F\u002Fgithub.com\u002Fu14app\u002Fdeep-research)：u14app 的深度研究平台。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fu14app\u002Fdeep-research?style=social)\n- [open-deep-research](https:\u002F\u002Fgithub.com\u002Fbtahir\u002Fopen-deep-research)：开源深度研究框架。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbtahir\u002Fopen-deep-research?style=social)\n- [deep-searcher](https:\u002F\u002Fgithub.com\u002Fzilliztech\u002Fdeep-searcher)：深度搜索与研究工具包。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fzilliztech\u002Fdeep-searcher?style=social)\n- [node-DeepResearch](https:\u002F\u002Fgithub.com\u002Fjina-ai\u002Fnode-DeepResearch)：用于寻找正确答案的深度研究工具包。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fjina-ai\u002Fnode-DeepResearch?style=social)\n- [Auto-Deep-Research](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FAuto-Deep-Research)：自动化深度研究智能体。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FHKUDS\u002FAuto-Deep-Research?style=social)\n- [langgraph-deep-research](https:\u002F\u002Fgithub.com\u002Fforeveryh\u002Flanggraph-deep-research)：使用 LangGraph 的深度研究工作流。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fforeveryh\u002Flanggraph-deep-research?style=social)\n- [DeepResearchAgent](https:\u002F\u002Fgithub.com\u002FSkyworkAI\u002FDeepResearchAgent)：SkyworkAI 的深度研究智能体。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSkyworkAI\u002FDeepResearchAgent?style=social)\n- [OpenManus](https:\u002F\u002Fgithub.com\u002FFoundationAgents\u002FOpenManus)：一个用于构建通用 AI 智能体的开源框架。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FFoundationAgents\u002FOpenManus?style=social)\n- [PraisonAI](https:\u002F\u002Fgithub.com\u002FMervinPraison\u002FPraisonAI)：一款具备内置深度研究能力、可直接投入生产的多智能体框架。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FMervinPraison\u002FPraisonAI?style=social)\n- [AtomSearcher](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus)：一款自动化的深度研究智能体。![GitHub 仓库星级](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fantgroup\u002FResearch-Venus?style=social)\n\n## 最新研究论文\n\n🔥🔥🔥 本节展示了代理式深度研究领域中最新且最具影响力的科研论文。每一篇论文都代表着在自主研究智能体、搜索能力及推理框架发展方面的重大进展。论文按时间顺序排列，最新的发表内容位于顶部。涵盖的关键领域包括：\n- 🤖 用于深度研究的代理式框架\n- 🔍 搜索增强型推理模型\n- 🌐 用于深度研究的网络智能体\n- 🔄 推理与检索增强型生成\n- 📊 多模态深度研究\n\n🚀🚀🚀 敬请关注该领域的最新突破！\n\n| 标题 | 日期及代码 | 基础模型 | 优化方法 | 搜索引擎 | 代理架构 | 训练数据集 | 评估数据集 |\n| --- | :---: | --- | --- | --- | --- | --- | --- |\n| [Dr. Zero: 无需训练数据的自进化搜索代理](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.07055) | [2026\u002F01\u002F11](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fdrzero) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ffacebookresearch\u002Fdrzero?style=social)](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fdrzero) | Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct | HRPO | 网络搜索 | 多智能体 | — | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMultihopQA (2WikiMQA), MuSiQue, Bamboogle |\n| [LEAPS: 阿里巴巴AI搜索的LLM赋能自适应插件](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.05513) | [2026\u002F01\u002F09]() | Qwen3-14B | REINFORCE++, GRPO, GSPO | 本地检索 | 单智能体 | — | — |\n| [SmartSearch: 面向搜索代理的过程奖励引导查询精炼](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.04888) | [2026\u002F01\u002F08](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FSmartSearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUC-NLPIR\u002FSmartSearch?style=social)](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FSmartSearch) | Qwen2.5-3B-Instruct | SFT, DPO, GRPO | 网络搜索 | 单智能体 | Asearcher-Base | 2WikiMultihopQA, HotpotQA, Bamboogle, MuSiQue, GAIA, WebWalker |\n| [O-Researcher: 基于多智能体蒸馏与智能体强化学习的开放式深度研究模型](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.03743v1) | [2026\u002F01\u002F07](https:\u002F\u002Fgithub.com\u002FOPPO-PersonalAI\u002FO-Researcher) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOPPO-PersonalAI\u002FO-Researcher?style=social)](https:\u002F\u002Fgithub.com\u002FOPPO-PersonalAI\u002FO-Researcher) | Qwen-2.5-72B-Instruct | GRPO | 网络搜索 | 多智能体 | 知乎KOL、WideSearch、ELI5 | DeepResearch Bench, DeepResearchGym |\n| [WebAnchor: 锚定式智能体规划以稳定长 horizon 网络推理](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.03164) | [2026\u002F01\u002F06]() | WebSailor-3B\u002F7B, Tongyi-DR-30B, Qwen-2.5-72B | GRPO | 本地检索 | 单智能体 | — | BrowseComp-en, BrowseComp-zh, XBench-DeepSearch, GAIA |\n| [预算感知的工具使用实现高效的智能体扩展](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.17006) | [2025\u002F11\u002F21]() | Gemini-2.5-Flash, Gemini-2.5-Pro, Claude-Sonnet-4 | 提示工程 | 网络搜索 | 单智能体 | — | — |\n| [AutoTool: 面向大型语言模型智能体的高效工具选择](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.14650) | [2025\u002F11\u002F18](https:\u002F\u002Fgithub.com\u002Fjiajingyyyyyy\u002FAutoTool) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fjiajingyyyyyy\u002FAutoTool?style=social)](https:\u002F\u002Fgithub.com\u002Fjiajingyyyyyy\u002FAutoTool) | Llama4-Scout-17B | 提示工程 | 网络搜索 | 单智能体 | — | AlfWorld, ScienceWorld, ToolQuery-Academia |\n| [多智能体深度研究：使用M-GRPO训练多智能体系统](https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.13288) | [2025\u002F11\u002F17](https:\u002F\u002Fgithub.com\u002FAQ-MedAI\u002FMrlX) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAQ-MedAI\u002FMrlX?style=social)](https:\u002F\u002Fgithub.com\u002FAQ-MedAI\u002FMrlX) | Qwen3-30B-A3B | M-GRPO | 网络搜索 | 多智能体 | — | GAIA, XBench-DeepSearch, WebWalkerQA |\n| [通义深度研究技术报告](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.24701) | [2025\u002F10\u002F28](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | Qwen3-30B-A3B-Base | SFT, RL | 网络搜索 | 单智能体 | — | HLE, BrowseComp, BrowseComp-ZH, GAIA, XBench-DeepSearch, WebWalkerQA, FRAMES, XBench-DeepSearch-2510 |\n| [TOOLRM: 向智能体工具使用奖励建模迈进](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.26167) | [2025\u002F10\u002F30](https:\u002F\u002Fgithub.com\u002Flirenhao1997\u002FToolRM) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Flirenhao1997\u002FToolRM?style=social)](https:\u002F\u002Fgithub.com\u002Flirenhao1997\u002FToolRM) | Qwen3-4B, Qwen3-8B | RL | 网络搜索 | 单智能体 | ToolPref-Pairwise-30K | TRBench, ACEBench |\n| [ToolScope: 一种用于视觉引导和长 horizon 工具使用的智能体框架](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.27363) | [2025\u002F10\u002F31]() | GPT-4o, Gemini-2.5, Qwen2.5-VL, Llama-3.2-Vision | 提示工程 | 本地检索 | 多智能体 | — | — |\n| [WebSeer: 通过带有自我反思的强化学习训练更深层次的搜索代理](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.18798) | [2025\u002F10\u002F21](https:\u002F\u002Fgithub.com\u002F99hgz\u002FWebSeer) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002F99hgz\u002FWebSeer?style=social)](https:\u002F\u002Fgithub.com\u002F99hgz\u002FWebSeer) | Qwen-2.5-14B, Qwen-3-14B | RL（冷启动+RL；自我反思） | 网络搜索 | 单智能体 | — | HotpotQA, SimpleQA |\n| [企业深度研究：面向企业分析的可引导多智能体深度研究](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2510.17797) | [2025\u002F10\u002F20](https:\u002F\u002Fgithub.com\u002FSalesforceAIResearch\u002Fenterprise-deep-research) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FSalesforceAIResearch\u002Fenterprise-deep-research?style=social)](https:\u002F\u002Fgithub.com\u002FSalesforceAIResearch\u002Fenterprise-deep-research) | — | 提示工程 | 网络搜索 | 多智能体 | — | DeepResearch Bench, DeepConsult |\n| [Stop-RAG: 基于价值的迭代RAG检索控制](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.14337) | [2025\u002F10\u002F16](https:\u002F\u002Fgithub.com\u002Fchosolbee\u002FStop-RAG) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fchosolbee\u002FStop-RAG?style=social)](https:\u002F\u002Fgithub.com\u002Fchosolbee\u002FStop-RAG) | Llama-3.1-8B-Instruct | 微调 | 本地检索 | 单智能体 | MuSiQue, HotpotQA, 2WikiMultihopQA | HotpotQA, MuSiQue, 2WikiMultihopQA |\n| [迈向搜索环境中具有智能体特性的自学习LLM](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.14253) | [2025\u002F10\u002F16](https:\u002F\u002Fgithub.com\u002Fforangel2014\u002FTowards-Agentic-Self-Learning) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fforangel2014\u002FTowards-Agentic-Self-Learning?style=social)](https:\u002F\u002Fgithub.com\u002Fforangel2014\u002FTowards-Agentic-Self-Learning) | Qwen-2.5-7B-Instruct | RL | 网络搜索 | 多智能体 | — | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMultihopQA, MuSiQue, Bamboogle |\n| [GOAT: 面向目标导向型工具使用智能体的训练框架](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.12218) | [2025\u002F10\u002F14]() | Qwen-2-7B, Llama-3-8B-Instruct, Llama-3-70B-Instruct | 微调 | 网络搜索 | 单智能体 | — | GOATBench |\n| [ResearStudio: 一种可供人类干预的可控深度研究智能体构建框架](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.12194) | [2025\u002F10\u002F14](https:\u002F\u002Fgithub.com\u002FResearAI\u002FResearStudio) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FResearAI\u002FResearStudio?style=social)](https:\u002F\u002Fgithub.com\u002FResearAI\u002FResearStudio) | gpt-4.1, gpt-4.1-mini, o4-mini, Llama-3.3-70B | 提示工程 | 网络搜索 | 单智能体 | — | GAIA |\n| [HiPRAG: 用于高效智能体增强检索生成的层次化过程奖励](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.07794) | [2025\u002F10\u002F09](https:\u002F\u002Fgithub.com\u002Fqualidea1217\u002FHiPRAG) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fqualidea1217\u002FHiPRAG?style=social)](https:\u002F\u002Fgithub.com\u002Fqualidea1217\u002FHiPRAG) | Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct, Llama-3.2-3B-Instruct | PPO, GRPO | 网络搜索 | 单智能体 | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2Wiki, MuSiQue, Bamboogle |\n| [A2SEARCH: 基于强化学习的歧义感知问答](https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.07958) | [2025\u002F10\u002F09](https:\u002F\u002Fgithub.com\u002Fzfj1998\u002FA2Search) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fzfj1998\u002FA2Search?style=social)](https:\u002F\u002Fgithub.com\u002Fzfj1998\u002FA2Search) | Qwen-2.5系列 | RL | 网络搜索 | 单智能体 | NQ | MuSiQue, HotpotQA, 2Wiki, Bamboogle, NQ, TriviaQA, PopQA, AmbigQA |\n| [ReSeek: 一种具有指导性奖励的自纠正搜索代理框架](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2510.00568) | [2025\u002F10\u002F01](https:\u002F\u002Ftencentbac.github.io\u002FReSeek\u002F) | Qwen2.5-7B-Instruct, Qwen2.5-3B-Instruct | GRPO | 网络搜索、本地检索 | 单智能体 | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMQA, MuSiQue, Bamboogle, FictionalHot |\n| [面向交互式多模态工具使用智能体的过程监督强化学习](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.14480) | [2025\u002F09\u002F17]() | Qwen3-8B, Qwen2.5-Omni-7B | RL | 本地检索 | 单智能体 | τ-bench, APIGen-MT | τ-bench |\n| [ReSum: 通过上下文摘要解锁长 horizon 搜索智能](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.13313) | [2025\u002F09\u002F16](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | Qwen3-30B-A3B-Thinking | GRPO, SFT | 网络搜索 | 单智能体 | SailorFog-QA | BrowseComp-en\u002Fzh |\n| [WebWeaver: 利用动态提纲构建网络规模证据，支持开放式深度研究](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.13312) | [2025\u002F09\u002F16](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | Qwen3-30B-A3B-Instruct | SFT | 网络搜索 | 单智能体 | WebWeaver-3k | BrowseComp-en\u002Fzh, GAIA, XBench-DeepSearch |\n| [WebResearcher: 在长 horizon 智能体中释放无限推理能力](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.13309) | [2025\u002F09\u002F16](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | Qwen3-30B-A3B | RFT, RL | 网络搜索 | 多智能体 | WebFrontier | BrowseComp-en\u002Fzh, GAIA, WebWalkerQA, FRAMES, HotpotQA, MuSiQue, 2WikiMultihopQA |\n| [WebSailor-V2: 通过合成数据和可扩展RL弥合与专有智能体的鸿沟](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.13305) | [2025\u002F09\u002F16](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FDeepResearch) | Qwen3-30B-A3B | SFT, RL | 网络搜索、本地检索 | 单智能体 | SailorFog-QA-V2 | BrowseComp-EN, BrowseComp-ZH, HLE |\n| [WebExplorer: 探索与进化以训练长 horizon 网络智能体](https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.06501) | [2025\u002F09\u002F08](https:\u002F\u002Fgithub.com\u002Fhkust-nlp\u002FWebExplorer) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhkust-nlp\u002FWebExplorer?style=social)](https:\u002F\u002Fgithub.com\u002Fhkust-nlp\u002FWebExplorer) | Qwen3-8B | GRPO, SFT | 网络搜索 | 单智能体 | WebExplorer-QA | BrowseComp-en\u002Fzh, GAIA, WebWalkerQA, FRAMES, XBench-DeepSearch, HLE |\n| [Atom-Searcher: 通过细粒度原子级思维奖励提升智能体深度研究](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.12800) | [2025\u002F08\u002F18](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fantgroup\u002FResearch-Venus?style=social)](https:\u002F\u002Fgithub.com\u002Fantgroup\u002FResearch-Venus) | Qwen2.5-7B | RL(GRPO) | 网络搜索 | 单智能体 | NQ, SimpleQA, HotpotQA, 2WikiMultihopQA, MuSiQue, MultiHopRAG | Bamboogle, NQ, SimpleQA, HotpotQA, 2WikiMultihopQA, MuSiQue, MultiHopRAG |\n| [MMSearch-R1: 激励LMMs进行搜索](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.20670) | [2025\u002F06\u002F25](https:\u002F\u002Fgithub.com\u002FEvolvingLMMs-Lab\u002Fmultimodal-search-r1) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FEvolvingLMMs-Lab\u002Fmultimodal-search-r1?style=social)](https:\u002F\u002Fgithub.com\u002FRAG-Gym\u002FRAG-Gym) | Qwen2.5-VL-7B | RL(GRPO) | 网络搜索 | 单智能体 | VQA, MetaClip, FVQA, InfoSeek | FVQA-test, InfoSeek, MMSearch, SimpleVQA, LiveVQA |\n| [VideoDeepResearch: 具有智能体工具使用的长视频理解](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.10821) | [2025\u002F06\u002F12](https:\u002F\u002Fgithub.com\u002Fyhy-2000\u002FVideoDeepResearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fyhy-2000\u002FVideoDeepResearch?style=social)](https:\u002F\u002Fgithub.com\u002Fyhy-2000\u002FVideoDeepResearch) | GPT-4o, Gemini1.5-pro, Qwen2.5-VL-72B-Instruct | 提示工程 | 本地检索 | 多智能体 | — | MLVU, Video-MME, LVBench, LongVideoBench |\n| [多模态DeepResearcher: 使用智能体框架从零开始生成文本-图表交错报告](https:\u002F\u002Farxiv.org\u002Fabs\u002F2506.02454) | [2025\u002F06\u002F03]() | Claude3.7-Sonnet, GPT-4o-mini, Qwen3-235B-A22B, Qwen2.5-VL-72B-Instruct | 提示工程 | 网络搜索 | 多智能体 | — | Pew Research, Our World in Data, Open Knowledge Foundation |\n| [RAG-Gym: 针对检索增强生成的语言智能体的系统性优化](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2502.13957v2) | [2025\u002F05\u002F31](https:\u002F\u002Fgithub.com\u002FRAG-Gym\u002FRAG-Gym) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRAG-Gym\u002FRAG-Gym?style=social)](https:\u002F\u002Fgithub.com\u002FRAG-Gym\u002FRAG-Gym) | Llama3.1-8B-Instruct, Qwen2.5-7B-Instruct, GPT-4o-mini | SFT, RL(PPO, DPO) | 本地检索 | 单智能体 | HotpotQA, MedQA | HotpotQA, 2Wiki, Bamboogle, MedQA |\n| [MaskSearch: 一个通用预训练框架，用于增强智能体搜索能力](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.20285) | [2025\u002F05\u002F27](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FMaskSearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FMaskSearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FMaskSearch) | Llama3.1-8B, Llama3.2-3B, Llama3.2-1B, Llama3, Qwen2.5-7B, Qwen2.5-3B, Qwen2.5-1.5B, Qwen2.5 | SFT, RL(DAPO) | 本地检索 | 多智能体 | HotpotQA | HotpotQA, FanoutQA, Musique, 2WikiMultihopQA, Bamboogle, FreshQA |\n| [SimpleDeepSearcher: 通过网络驱动的推理轨迹合成进行深度信息获取](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.16834) | [2025\u002F05\u002F25](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FSimpleDeepSearcher?tab=readme-ov-file) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUCAIBox\u002FSimpleDeepSearcher?style=social)](https:\u002F\u002Fgithub.com\u002Fweizhepei\u002FWebAgent-R1) | Qwen2.5-7B-Instruct, Qwen2.5-32B-Instruct, DeepseekDistilled-Qwen2.5-32B, QwQ-32B | SFT | 网络搜索 | 单智能体 | NQ, SimpleQA, HotpotQA, 2WikiMultihopQA, MuSiQue, MultiHopRAG | Bamboogle, FRAMES, GAIA, NQ, SimpleQA, HotpotQA, 2WikiMultihopQA, MuSiQue, MultiHopRAG |\n| [WebAgent-R1: 通过端到端多轮强化学习训练网络智能体](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.16421) | [2025\u002F05\u002F22](https:\u002F\u002Fgithub.com\u002Fweizhepei\u002FWebAgent-R1) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fweizhepei\u002FWebAgent-R1?style=social)](https:\u002F\u002Fgithub.com\u002Fweizhepei\u002FWebAgent-R1) | Qwen2.5-3B, Llama3.1-8B | SFT, RL(M-GRPO) | 网络搜索 | 单智能体 | WebArena-Lite, WebArena | WebArena-Lite, WebArena |\n| [R1-Searcher++: 通过强化学习激励LLM的动态知识获取](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.17005) | [2025\u002F05\u002F22](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher-plus) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUCAIBox\u002FR1-Searcher-plus?style=social)](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher-plus) | Qwen2.5-7B-Instruct | SFT, RL | 本地检索 | 单智能体 | HotpotQA, 2WikiMultihopQA | HotpotQA, 2WikiMultihopQA, Musique, Bamboogle |\n| [过程奖励 vs. 结果奖励：哪一种更适合智能体RAG强化学习？](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.14069) | [2025\u002F05\u002F22](https:\u002F\u002Fgithub.com\u002Fwlzhang2020\u002FReasonRAG) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fwlzhang2020\u002FReasonRAG?style=social)](https:\u002F\u002Fgithub.com\u002Fwlzhang2020\u002FReasonRAG) | Qwen2.5-7B-Instruct | RL(DPO) | 本地检索 | 单智能体 | PopQA, HotpotQA, 2WikiMultihopQA | PopQA, HotpotQA, 2WikiMultihopQA, Bamboogle, MuSiQue |\n| [s3 - 通过RL高效而有效的搜索智能体训练](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.14146) | [2025\u002F05\u002F20](https:\u002F\u002Fgithub.com\u002Fpat-jj\u002Fs3) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpat-jj\u002Fs3?style=social)](https:\u002F\u002Fgithub.com\u002Fpat-jj\u002Fs3) | Qwen2.5-7B-Instruct | RL(PPO) | 本地检索 | 单智能体 | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2wiki, Musique, MedQA-US, MedMCQA, PubMedQA, BioASQ-Y\u002FN, MMLU-Med |\n| [揭秘并提升基于大型语言模型的搜索智能体效率](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.12065) | [2025\u002F05\u002F17](https:\u002F\u002Fgithub.com\u002Ftiannuo-yang\u002FSearchAgent-X) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftiannuo-yang\u002FSearchAgent-X?style=social)](https:\u002F\u002Fgithub.com\u002Ftiannuo-yang\u002FSearchAgent-X) | Qwen2.5-14B, Qwen2.5-7B | 提示工程 | 本地检索 | 单智能体 | — | Musique, NQ, 2WikiMultihopQA, HotpotQA, Bamboogle, StrategyQA |\n| [强化内外部知识协同推理，以实现高效的自适应搜索智能体](https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.07596) | [2025\u002F05\u002F12](https:\u002F\u002Fgithub.com\u002Fhzy312\u002Fknowledge-r1) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fhzy312\u002Fknowledge-r1?style=social)](https:\u002F\u002Fgithub.com\u002Fhzy312\u002Fknowledge-r1) | Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct | RL(GRPO) | 本地检索 | 单智能体 | NQ, HotpotQA | PopQA, 2WikiMultihopQA |\n| [ZeroSearch: 在不进行搜索的情况下激励LLM的搜索能力](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.04588) | [2025\u002F05\u002F07](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FZeroSearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAlibaba-NLP\u002FZeroSearch?style=social)](https:\u002F\u002Fgithub.com\u002FAlibaba-NLP\u002FZeroSearch) | Qwen2.5-3B-Base, Qwen2.5-7B-Base, Qwen2.5-7B-Instruct, Qwen2.5-3B-Instruct, Llama3.2-3B-Instruct, Llama3.2-3B-Base | RL(Reinforce, GRPO, PPO) | 网络搜索 | 单智能体 | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMultihopQA, Musique, Bamboogle |\n| [Webthinker: 用深度研究能力赋能大型推理模型](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.21776) | [2025\u002F04\u002F30](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FWebThinker) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUC-NLPIR\u002FWebThinker?style=social)](https:\u002F\u002Fgithub.com\u002FRUC-NLPIR\u002FWebThinker) | GPT-o1, GPT-o3, Deepseek-R1, QwQ-32B, Qwen2.5-32B-Instruct | RL(DPO) | 网络搜索 | 单智能体 | SuperGPQA, WebWalkerQA, OpenThoughts, NaturalReasoning, NuminaMath | GPQA, GAIA, WebWalkerQA, Humanity’s Last Exam |\n| [盘古超算：在Ascend NPU上突破密集型大型语言模型的极限](https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.07866) | [2025\u002F04\u002F11](https:\u002F\u002Fgithub.com\u002Fpangu-tech\u002Fpangu-ultra) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fpangu-tech\u002Fpangu-ultra?style=social)](https:\u002F\u002Fgithub.com\u002Fpangu-tech\u002Fpangu-ultra) | 盘古超算-135B | SFT, RL | 本地检索 | 单智能体 | — | — |\n| [开放深度搜索：用开源推理智能体 democratize 搜索](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.20201) | [2025\u002F03\u002F26](https:\u002F\u002Fgithub.com\u002Fsentient-agi\u002FOpenDeepSearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsentient-agi\u002FOpenDeepSearch?style=social)](https:\u002F\u002Fgithub.com\u002Fsentient-agi\u002FOpenDeepSearch) | Llama3.1-70B, Deepseek-R1 | 提示工程 | 网络搜索 | 单智能体 | — | SimpleQA, FRAME |\n| [DeepResearcher: 在真实环境中通过强化学习扩展深度研究](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.03160?) | [2025\u002F03\u002F26](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FDeepResearcher) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FGAIR-NLP\u002FDeepResearcher?style=social)](https:\u002F\u002Fgithub.com\u002FGAIR-NLP\u002FDeepResearcher) | Qwen2.5-7B-Instruct | RL(GRPO) | 网络搜索 | 多智能体 | NQ, TQ, HotpotQA, 2WikiMultihopQA | MuSiQue, Bamboogle, PopQA, NQ, TQ, HotpotQA, 2WikiMultihopQA |\n| [ReSearch: 通过强化学习让LLM学会利用搜索引擎进行推理](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.19470) | [2025\u002F03\u002F25](https:\u002F\u002Fgithub.com\u002FAgent-RL\u002FReSearch) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAgent-RL\u002FReSearch?style=social)](https:\u002F\u002Fgithub.com\u002FAgent-RL\u002FReSearch) | Qwen2.5-7B-Instruct, Qwen2.5-32B-Instruct | RL(GRPO) | 网络搜索 | 单智能体 | MuSiQue | HotpotQA, 2WikiMultihopQA, Musique, Bamboogle |\n| [Search-R1: 通过强化学习训练LLM进行推理并利用搜索引擎](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.09516) | [2025\u002F03\u002F12](https:\u002F\u002Fgithub.com\u002FPeterGriffinJin\u002FSearch-R1) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FPeterGriffinJin\u002FSearch-R1?style=social)](https:\u002F\u002Fgithub.com\u002FPeterGriffinJin\u002FSearch-R1) | Qwen2.5-7B-Instruct, Qwen2.5-7B-Base, Qwen2.5-3B-Instruct, Qwen2.5-3B-Base | RL(PPO, GRPO) | 网络搜索 | 单智能体 | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMultihopQA, Musique, Bamboogle |\n| [超越提纲：用异构递归规划实现语言模型的自适应长篇写作](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.08275) | [2025\u002F03\u002F11](https:\u002F\u002Fgithub.com\u002Fprincipia-ai\u002FWriteHERE) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fprincipia-ai\u002FWriteHERE?style=social)](https:\u002F\u002Fgithub.com\u002Fprincipia-ai\u002FWriteHERE) | GPT-4o, Claude3.5-Sonnet | 提示工程 | 网络搜索 | 多智能体 | — | TELL ME A STORY, WildSeek |\n| [R1-Searcher: 通过强化学习激励LLM的搜索能力](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2503.05592) | [2025\u002F03\u002F07](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FRUCAIBox\u002FR1-Searcher?style=social)](https:\u002F\u002Fgithub.com\u002FRUCAIBox\u002FR1-Searcher) | Qwen2.5-7B-Base, Llama3.1-8B-Instruct | SFT, RL(GRPO, Reinforce++) | 网络搜索、本地检索 | 单智能体 | HotpotQA, 2WikiMultihopQA | HotpotQA, 2WikiMultihopQA, Musique, Bamboogle |\n| [AutoAgent: 一个完全自动化且零代码的LLM智能体框架](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2502.05957#page=1.62) | [2025\u002F02\u002F18](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FAutoAgent?tab=readme-ov-file) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FHKUDS\u002FAutoAgent?style=social)](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FAutoAgent?tab=readme-ov-file) | Claude3.5-Sonnet | 提示工程 | 网络搜索 | 多智能体 | — | GAIA |\n| [智能体推理：使用工具对LLM进行深度研究推理](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.04644) | [2025\u002F02\u002F07](https:\u002F\u002Fgithub.com\u002Ftheworldofagents\u002FAgentic-Reasoning) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Ftheworldofagents\u002FAgentic-Reasoning?style=social)](https:\u002F\u002Fgithub.com\u002Ftheworldofagents\u002FAgentic-Reasoning) | 无 | 提示工程 | 网络搜索 | 多智能体 | — | GPQA |\n| [Search-o1: 具有智能体增强功能的大型推理模型](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2501.05366) | [2025\u002F01\u002F09](https:\u002F\u002Fgithub.com\u002Fsunnynexus\u002FSearch-o1) [![GitHub 星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsunnynexus\u002FSearch-o1?style=social)](https:\u002F\u002Fgithub.com\u002Fsunnynexus\u002FSearch-o1) | QwQ-32B-Preview | 提示工程 | 网络搜索 | 单智能体 | — | GPQA, MATH500, AMC2023, AIME2024, LiveCodeBench, NQ, TriviaQA, HotpotQA, 2WikiMultihopQA, MuSiQue, Bamboogle |\n\n## 基准测试与应用\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_readme_27c4fb9bebee.png\" alt=\"基准测试图表\" width=\"80%\" \u002F>\n\u003C\u002Fp>\n\n- 人类的最后一场考试 [[论文]](https:\u002F\u002Farxiv.org\u002Fabs\u002F2501.14249) [[代码]](https:\u002F\u002Fgithub.com\u002Fcenterforaisafety\u002Fhle) ![GitHub 仓库星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fcenterforaisafety\u002Fhle?style=social)\n- BrowseComp：一个简单但具有挑战性的浏览代理基准测试 [[论文]](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.12516) [[代码]](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fsimple-evals) ![GitHub 仓库星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopenai\u002Fsimple-evals?style=social)\n- BrowseComp-ZH：中文环境下大型语言模型的网页浏览能力基准测试 ['[论文]'](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2504.19314) [[代码]](https:\u002F\u002Fgithub.com\u002FPALIN2018\u002FBrowseComp-ZH) ![GitHub 仓库星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FPALIN2018\u002FBrowseComp-ZH?style=social)\n- DeepResearch 基准：面向深度研究代理的综合性基准测试 [[论文]](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.11763) [[代码]](https:\u002F\u002Fgithub.com\u002FAyanami0730\u002Fdeep_research_bench) ![GitHub 仓库星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FAyanami0730\u002Fdeep_research_bench?style=social)\n- MedBrowseComp：医学领域的深度研究与计算机使用能力基准测试 [[论文]](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2505.14963) [[代码]](https:\u002F\u002Fgithub.com\u002Fshan23chen\u002FMedBrowseComp) ![GitHub 仓库星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fshan23chen\u002FMedBrowseComp?style=social)\n- Mind2Web 2：以代理为评判者的代理式搜索评估 [[论文]](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2506.21506) [[代码]](https:\u002F\u002Fgithub.com\u002FOSU-NLP-Group\u002FMind2Web-2) ![GitHub 仓库星标](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FOSU-NLP-Group\u002FMind2Web-2?style=social)\n\n\n\n## 贡献与引用\n\n🤝 我们欢迎各位贡献者一起扩充这份全面的代理式深度研究资源合集！\n\n### 📝 如何贡献\n\n**添加新的研究论文和基准测试：**\n- 提交一个问题，附上论文详细信息（标题、arXiv 链接、我们论文表格中的所有类别，以及如果有 GitHub 仓库的话，请提供链接）\n- 或者创建一个拉取请求，将论文添加到研究论文表格或基准测试部分\n\n**添加新的开源实现和新产品：**\n- 提交一个问题，附上仓库的详细信息（名称、描述、发布日期、如果有 GitHub 链接请一并提供）\n- 或者创建一个拉取请求，将实现添加到开源和产品部分\n\n\n### 📖 引用\n\n🔥🔥🔥 如果您觉得本仓库对您有所帮助，请引用我们的论文：\n\n```bibtex\n@article{zhang2025web,\n  title={从网络搜索到代理式深度研究：以推理代理激励搜索},\n  author={Zhang, Weizhi and Li, Yangning and Bei, Yuanchen and Luo, Junyu and Wan, Guancheng and Yang, Liangwei and Xie, Chenxuan and Yang, Yuyao and Huang, Wei-Chieh and Miao, Chunyu and others},\n  journal={arXiv 预印本 arXiv:2506.18959},\n  year={2025}\n}\n\n@article{li2025towards,\n  title={迈向具有深度推理能力的代理式 RAG：LLMs 中 RAG-推理系统的综述},\n  author={Li, Yangning and Zhang, Weizhi and Yang, Yuyao and Huang, Wei-Chieh and Wu, Yaozu and Luo, Junyu and Bei, Yuanchen and Zou, Henry Peng and Luo, Xiao and Zhao, Yusheng and others},\n  journal={arXiv 预印本 arXiv:2507.09477},\n  year={2025}\n}\n```","# Awesome-Deep-Research 快速上手指南\n\n`Awesome-Deep-Research` 并非单一的可执行软件，而是一个汇聚了**代理式深度研究（Agentic Deep Research）**领域顶尖产品、开源实现和前沿论文的精选资源库。本指南将帮助你快速定位适合的开发工具并启动项目。\n\n## 环境准备\n\n在开始使用仓库中的开源项目前，请确保你的开发环境满足以下通用要求：\n\n*   **操作系统**: Linux (推荐 Ubuntu 20.04+), macOS, 或 Windows (WSL2)。\n*   **Python 版本**: 大多数项目需要 Python 3.10 或更高版本。\n*   **包管理器**: 推荐使用 `conda` 或 `venv` 进行环境隔离。\n*   **API Keys**: 根据你选择的具体项目，可能需要准备以下密钥：\n    *   LLM API: OpenAI, Anthropic, Google Gemini, 或国内模型 (如 Moonshot\u002FKimi, Alibaba\u002FQwen)。\n    *   搜索 API: Tavily, Serper, Bing Search API 等。\n*   **网络环境**: 部分项目依赖 Hugging Face 或 GitHub 资源，国内开发者建议配置镜像加速。\n\n## 安装步骤\n\n由于该仓库包含多个独立的开源项目，请选择一个适合你的项目进行安装。以下以两个热门项目为例：\n\n### 方案 A：安装 GPT-Researcher (功能全面的自主研究代理)\n\n1.  **克隆仓库**\n    ```bash\n    git clone https:\u002F\u002Fgithub.com\u002Fassafelovic\u002Fgpt-researcher.git\n    cd gpt-researcher\n    ```\n\n2.  **创建虚拟环境并安装依赖**\n    ```bash\n    python -m venv venv\n    source venv\u002Fbin\u002Factivate  # Windows 用户请使用: venv\\Scripts\\activate\n    \n    # 安装核心依赖\n    pip install -r requirements.txt\n    ```\n\n3.  **配置环境变量**\n    在项目根目录创建 `.env` 文件，填入你的 API 密钥：\n    ```bash\n    cp .env.example .env\n    # 编辑 .env 文件，填入 OPENAI_API_KEY, TAVILY_API_KEY 等\n    ```\n\n### 方案 B：安装 DeerFlow (字节跳动开源的深度研究框架)\n\n1.  **克隆仓库**\n    ```bash\n    git clone https:\u002F\u002Fgithub.com\u002Fbytedance\u002Fdeer-flow.git\n    cd deer-flow\n    ```\n\n2.  **安装依赖**\n    ```bash\n    pip install -r requirements.txt\n    ```\n    *注：若下载速度慢，可使用国内镜像源加速：*\n    ```bash\n    pip install -r requirements.txt -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n    ```\n\n3.  **配置模型与搜索接口**\n    参照项目内的 `config.yaml` 或文档说明，配置本地模型路径或云端 API Key。\n\n## 基本使用\n\n安装完成后，即可通过命令行或代码调用深度研究功能。\n\n### 示例 1：运行 GPT-Researcher 进行研究\n\n在终端中直接运行以下命令，启动针对特定主题的深度研究报告生成：\n\n```bash\npython -m gpt_researcher.cli.main --query \"2025 年人工智能代理技术的发展趋势\" --report_type research_report\n```\n\n执行后，系统将自动执行搜索、筛选、综合信息，并在 `outputs` 目录生成 Markdown 或 PDF 格式的报告。\n\n### 示例 2：使用 DeerFlow 进行多轮推理\n\n通过 Python 脚本调用 DeerFlow 的核心逻辑：\n\n```python\nfrom deer_flow import ResearchAgent\n\n# 初始化代理\nagent = ResearchAgent(config_path=\"config.yaml\")\n\n# 执行深度研究任务\nquery = \"分析量子计算在药物研发中的最新突破\"\nresult = agent.run(query)\n\n# 输出结果\nprint(result.summary)\nprint(result.sources)\n```\n\n### 探索更多资源\n\n*   **查看最新论文**: 访问仓库中的 [Latest Research Papers](#latest-research-papers) 章节，获取如 `Dr. Zero`, `O-Researcher` 等最新模型的代码链接。\n*   **尝试在线产品**: 若不想本地部署，可直接体验列表中的行业领先产品，如 [Kimi-Researcher](https:\u002F\u002Fmoonshotai.github.io\u002FKimi-Researcher\u002F) 或 [Deep Research (Qwen)](https:\u002F\u002Fchat.qwen.ai\u002F?inputFeature=deep_research)。","某金融科技公司的量化分析师需要在 48 小时内完成一份关于“全球生成式 AI 在医疗影像诊断领域”的深度竞品与技术趋势报告，以支撑下一季度的投资决策。\n\n### 没有 Awesome-Deep-Research 时\n- **信息搜集碎片化**：分析师需手动在 Google Scholar、arXiv、GitHub 及各大科技博客间反复切换搜索，难以系统性覆盖最新开源项目与商业产品。\n- **技术栈评估耗时**：面对海量零散的 RAG（检索增强生成）与推理代理论文，缺乏权威的基准测试（Benchmarks）参考，难以快速判断哪些算法具备落地价值。\n- **错失前沿动态**：由于无法实时追踪如 OpenAI Deep Research API 或 Grok Agents 等刚刚发布的行业级解决方案，导致报告中的竞品分析滞后于市场现状。\n- **验证成本高昂**：找到的开源代码库质量参差不齐，缺乏经过社区筛选的实现列表，需花费大量时间复现并排除不可用的项目。\n\n### 使用 Awesome-Deep-Research 后\n- **资源一站式聚合**：直接利用该仓库 curated 的清单，瞬间获取从 Google Gemini Deep Research 到最新 arXiv 论文的全链路资源，覆盖学术与工业界双视角。\n- **精准技术选型**：通过仓库中整理的评估基准与应用案例，快速锁定表现最优的 Agentic Deep Research 架构，将技术验证周期从数天缩短至数小时。\n- **同步最前沿情报**：借助其维护的“行业领先产品”列表，即时纳入 Manus、Anthropic Research 等 2025 年最新发布的平台，确保竞品分析具备极高的时效性。\n- **高效落地实施**：直接复用仓库推荐的成熟开源实现工具，避免了重复造轮子，让团队能专注于业务逻辑整合而非基础架构搭建。\n\nAwesome-Deep-Research 将原本需要数周的信息挖掘与验证工作压缩至天级别，成为连接前沿 AI 研究与实际商业落地的关键加速器。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FDavidZWZ_Awesome-Deep-Research_3cef4aae.png","DavidZWZ","Wiz","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FDavidZWZ_276d3756.jpg","CS PhD @ UIC | Student Researcher @ Google |  \r\nEx Research \u002F Applied Scientist Intern @ Roblox, Meta, Amazon.  \r\nWorking on LLMs, Agents, Personalization","Google","Mountain View",null,"https:\u002F\u002Fdavidzwz.github.io\u002F","https:\u002F\u002Fgithub.com\u002FDavidZWZ",698,56,"2026-04-08T23:16:06","MIT",1,"","未说明",{"notes":89,"python":87,"dependencies":90},"该 README 是一个资源列表（Awesome List），汇集了行业领先产品、开源实现工具和研究论文，而非单一的独立软件项目。因此，文中未提供统一的运行环境需求。具体的系统要求（如操作系统、GPU、内存、Python 版本及依赖库）需参考列表中各个具体开源项目（如 gpt-researcher, DeerFlow, OpenManus 等）各自的仓库文档。",[],[14,13,35,92],"其他",[94,95,96,97,98,99,100,101,102,103],"agentic-ai","agentic-rag","deep-research","deep-research-agent","large-language-models","llms","rag","reasoning","reasoning-agent","search-agent","2026-03-27T02:49:30.150509","2026-04-12T18:41:37.031341",[],[]]