@tom_doerr: 智能体深度研究资源精选列表 https://github.com/DavidZWZ/Awesome-Deep-Research…

精心整理的代理式深度研究资源列表 https://github.com/DavidZWZ/Awesome-Deep-Research… — # DavidZWZ/Awesome-Deep-Research 来源: https://github.com/DavidZWZ/Awesome-Deep-Research # 🤖 精选代理式深度研究资源 Awesome arXiv 维护中 欢迎贡献 代码 (https://github.com/DavidZWZ/Awesome-Deep-Research) 欢迎使用 Awesome-Deep-Research! 🚀 本仓库旨在作为您探索前沿“代理式深度研究”（Agentic Deep Research）世界的综合指南。我们精心汇编了一系列资源。无论您是研究人员、开发人员还是爱好者，本仓库都是您探索人工智能与自主代理（Autonomous Agents）迷人交汇点的门户。 若想了解关于信息搜索范式转变的详细分析，请查看我们的立场论文：《从网络搜索到代理式深度研究：通过推理代理激励搜索》(https://arxiv.org/abs/2506.18959) 📄，该论文概述了现有的领域趋势和未来方向。 对于对 RAG 与推理更广泛交叉点感兴趣的研究人员，我们还推荐浏览我们在 Awesome-RAG-Reasoning (https://github.com/DavidZWZ/Awesome-RAG-Reasoning) 🔥🔥🔥 上的综合资源集合。 ## 目录 - 🎯 行业领先的产品和解决方案 - 🔧 开源实现和工具 - 📚 最新研究论文和突破 - 🏆 评估基准和实际应用 - 🤝 贡献与引用 ## 行业领先产品 - Gemini Deep Research (https://gemini.google/overview/deep-research/?hl=en): Google 的高级研究助手，用于深度分析（2024年12月11日） - Deep Research (https://openai.com/index/introducing-deep-research/): OpenAI 的深度研究平台 [API 指南]（2025年2月2日） - Perplexity Deep Research (https://www.perplexity.ai/hub/blog/introducing-perplexity-deep-research): Perplexity 的深度研究分析产品（2025年2月14日） - Grok Agents (https://x.ai/news/grok-3): xAI 由 Grok-3 驱动的自主 DeepSearch 代理（2025年2月19日） - Copilot Researcher (https://www.microsoft.com/en-us/microsoft-365/blog/2025/03/25/introducing-researcher-and-analyst-in-microsoft-365-copilot/): Microsoft 365 Copilot 中的研究员和分析师功能（2025年3月25日） - Research (https://www.anthropic.com/news/research): Anthropic 的研究平台，用于查找和推理信息（2025年4月15日） - Manus (https://manus.im/): 高级研究分析平台（2025年3月6日） - 🦌 DeerFlow (https://deerflow.tech/): 字节跳动的研究分析解决方案（2025年5月9日） - Deep Research (https://chat.qwen.ai/?inputFeature=deep_research): 阿里巴巴通义千问驱动的研究助手（2025年5月14日） - Kimi-Researcher (https://moonshotai.github.io/Kimi-Researcher/): 月之暗面 Kimi 驱动的研究助手（2025年6月20日） - Not Human Search (https://nothumansearch.ai): 专为 AI 代理设计的搜索引擎。可作为 MCP 服务器用于工具发现。 ## 开源实现 - gemini-fullstack-langgraph-quickstart (https://github.com/google-gemini/gemini-fullstack-langgraph-quickstart): Gemini 全栈与 LangGraph 集成。GitHub Repo stars - multi-agent research system (https://github.com/anthropics/anthropic-cookbook/tree/main/patterns/agents/prompts): Anthropic 的多代理研究系统。博客文章 (https://www.anthropic.com/engineering/built-multi-agent-research-system) GitHub Repo stars - gpt-researcher (https://github.com/assafelovic/gpt-researcher): 用于综合研究任务的自主代理。GitHub Repo stars - DeerFlow (https://github.com/bytedance/deer-flow): 字节跳动的开源深度研究框架。GitHub Repo stars - r1-reasoning-rag (https://github.com/deansaco/r1-reasoning-rag): 推理增强型检索增强生成框架。GitHub Repo stars - nanoDeepResearch (https://github.com/liyuan24/nanoDeepResearch): 轻量级深度研究工具包。GitHub Repo stars - deep-research (Aomni) (https://github.com/dzhng/deep-research): Aomni 开发的深度研究助手。GitHub Repo stars - deep-research (u14app) (https://github.com/u14app/deep-research): u14app 开发的深度研究平台。GitHub Repo stars - open-deep-research (https://github.com/btahir/open-deep-research): 开源深度研究框架。GitHub Repo stars - deep-searcher (https://github.com/zilliztech/deep-searcher): 深度搜索和研究工具包。GitHub Repo stars - node-DeepResearch (https://github.com/jina-ai/node-DeepResearch): 用于寻找正确答案的深度研究工具包。GitHub Repo stars - Auto-Deep-Research (https://github.com/HKUDS/Auto-Deep-Research): 自动化深度研究代理。GitHub Repo stars - langgraph-deep-research (https://github.com/foreveryh/langgraph-deep-research): 基于 LangGraph 的深度研究工作流。GitHub Repo stars - DeepResearchAgent (https://github.com/SkyworkAI/DeepResearchAgent): SkyworkAI 开发的深度研究代理。GitHub Repo stars - OpenManus (https://github.com/FoundationAgents/OpenManus): 构建通用 AI 代理的开源框架。GitHub Repo stars - PraisonAI (https://github.com/MervinPraison/PraisonAI): 具备内置深度研究能力的生产级多代理框架。GitHub Repo stars - AtomSearcher (https://github.com/antgroup/Research-Venus): 自动化深度研究代理。GitHub Repo stars ## 最新研究论文 🔥🔥🔥 本节展示了代理式深度研究领域最新且最具影响力的研究论文。每篇论文都代表了在自主研究代理、搜索能力和推理框架构建方面的重要进展。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： - 🤖 深度研究的代理框架 - 🔍 搜索增强的推理模型 - 🌐 用于深度研究的 Web 代理 - 🔄 推理与检索增强生成 - 📊 多模态深度研究 🚀🚀🚀 敬请持续关注该领域的最新突破！ | 标题 | 日期 & 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 | | — | :—: | — | — | — | — | — | — | | CORAL: Towards Autonomous Multi-Agent Evolution for Open-Ended Discovery (https://arxiv.org/abs/2604.01658) | 2026/04/02 (https://github.com/Human-Agent-Society/CORAL) GitHub stars (https://github.com/Human-Agent-Society/CORAL) | Claude (代理主干) | 进化式搜索 (心跳引导) | – | 多代理 | – | 10 个数学/算法/系统基准（包括 Anthropic kernel-engineering） | | Dr. Zero: Self-Evolving Search Agents without Training Data (https://arxiv.org/abs/2601.07055) | 2026/01/11 (https://github.com/facebookresearch/drzero) GitHub stars (https://github.com/facebookresearch/drzero) | Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct | HRPO | Web 搜索 | 多代理 | – | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMultihopQA (2WikiMQA), MuSiQue, Bamboogle | | LEAPS: An LLM-Empowered Adaptive Plugin for Taobao AI Search (https://arxiv.org/abs/2601.05513) | 2026/01/09 | Qwen3-14B | REINFORCE++, GRPO, GSPO | 本地检索 | 单代理 | – | – | | SmartSearch: Process Reward-Guided Query Refinement for Search Agents (https://arxiv.org/abs/2601.04888) | 2026/01/08 (https://github.com/RUC-NLPIR/SmartSearch) GitHub stars (https://github.com/RUC-NLPIR/SmartSearch) | Qwen2.5-3B-Instruct | SFT, DPO, GRPO | Web 搜索 | 单代理 | Asearcher-Base | 2WikiMultihopQA, HotpotQA, Bamboogle, MuSiQue, GAIA, WebWalker | | O-Researcher: An Open Ended Deep Research Model via Multi-Agent Distillation and Agentic RL (https://arxiv.org/abs/2601.03743v1) | 2026/01/07 (https://github.com/OPPO-PersonalAI/O-Researcher) GitHub stars (https://github.com/OPPO-PersonalAI/O-Researcher) | Qwen-2.5-72B-Instruct | GRPO | Web 搜索 | 多代理 | Zhihu-KOL, WideSearch, ELI5 | DeepResearch Bench, DeepResearchGym | | WebAnchor: Anchoring Agent Planning to Stabilize Long-Horizon Web Reasoning (https://arxiv.org/abs/2601.03164) | 2026/01/06 | WebSailor-3B/7B, Tongyi-DR-30B, Qwen-2.5-72B | GRPO | 本地检索 | 单代理 | – | BrowseComp-en, BrowseComp-zh, XBench-DeepSearch, GAIA | | Budget-Aware Tool-Use Enables Effective Agent Scaling (https://arxiv.org/abs/2511.17006) | 2025/11/21 | Gemini-2.5-Flash, Gemini-2.5-Pro, Claude-Sonnet-4 | Prompting | Web 搜索 | 单代理 | – | – | | AutoTool: Efficient Tool Selection for Large Language Model Agents (https://arxiv.org/abs/2511.14650) | 2025/11/18 (https://github.com/jiajingyyyyyy/AutoTool) GitHub stars (https://github.com/jiajingyyyyyy/AutoTool) | Llama4-Scout-17B | Prompting | Web 搜索 | 单代理 | – | AlfWorld, ScienceWorld, ToolQuery-Academia | | Multi-Agent Deep Research: Training Multi-Agent Systems with M-GRPO (https://arxiv.org/abs/2511.13288) | 2025/11/17 (https://github.com/AQ-MedAI/MrlX) GitHub stars (https://github.com/AQ-MedAI/MrlX) | Qwen3-30B-A3B | M-GRPO | Web 搜索 | 多代理 | – | GAIA, XBench-DeepSearch, WebWalkerQA | | Tongyi DeepResearch Technical Report (https://arxiv.org/abs/2510.24701) | 2025/10/28 (https://github.com/Alibaba-NLP/DeepResearch) GitHub stars (https://github.com/Alibaba-NLP/DeepResearch) | Qwen3-30B-A3B-Base | SFT, RL | Web 搜索 | 单代理 | – | HLE, BrowseComp, BrowseComp-ZH, GAIA, XBench-DeepSearch, WebWalkerQA, FRAMES, XBench-DeepSearch-2510 | | TOOLRM: Towards Agentic Tool-Use Reward Modeling (https://arxiv.org/abs/2510.26167) | 2025/10/30 (https://github.com/lirenhao1997/ToolRM) GitHub stars (https://github.com/lirenhao1997/ToolRM) | Qwen3-4B, Qwen3-8B | RL | Web 搜索 | 单代理 | ToolPref-Pairwise-30K | TRBench, ACEBench | | ToolScope: An Agentic Framework for Vision-Guided and Long-Horizon Tool Use (https://arxiv.org/abs/2510.27363) | 2025/10/31 | GPT-4o, Gemini-2.5, Qwen2.5-VL, Llama-3.2-Vision | Prompting | 本地检索 | 多代理 | – | – | | WebSeer: Training Deeper Search Agents through Reinforcement Learning with Self-Reflection (https://arxiv.org/abs/2510.18798) | 2025/10/21 (https://github.com/99hgz/WebSeer) GitHub stars (https://github.com/99hgz/WebSeer) | Qwen-2.5-14B, Qwen-3-14B | RL (冷启动 + RL；自我反思) | Web 搜索 | 单代理 | – | HotpotQA, SimpleQA | | Enterprise Deep Research: Steerable MultiAgent Deep Research for Enterprise Analytics (https://arxiv.org/pdf/2510.17797) | 2025/10/20 (https://github.com/SalesforceAIResearch/enterprise-deep-research) GitHub stars (https://github.com/SalesforceAIResearch/enterprise-deep-research) | – | Prompting | Web 搜索 | 多代理 | – | DeepResearch Bench, DeepConsult | | Stop-RAG: Value-Based Retrieval Control for Iterative RAG (https://arxiv.org/abs/2510.14337) | 2025/10/16 (https://github.com/chosolbee/Stop-RAG) GitHub stars (https://github.com/chosolbee/Stop-RAG) | Llama-3.1-8B-Instruct | 微调 | 本地检索 | 单代理 | MuSiQue, HotpotQA, 2WikiMultihopQA | HotpotQA, MuSiQue, 2WikiMultihopQA | | Towards Agentic Self-Learning LLMs in Search Environment (https://arxiv.org/abs/2510.14253) | 2025/10/16 (https://github.com/forangel2014/Towards-Agentic-Self-Learning) GitHub stars (https://github.com/forangel2014/Towards-Agentic-Self-Learning) | Qwen-2.5-7B-Instruct | RL | Web 搜索 | 多代理 | – | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMultiHopQA, MuSiQue, Bamboogle | | GOAT: A Training Framework for Goal-Oriented Agent with Tools (https://arxiv.org/abs/2510.12218) | 2025/10/14 | Qwen-2-7B, Llama-3-8B-Instruct, Llama-3-70B-Instruct | 微调 | Web 搜索 | 单代理 | – | GOATBench | | ResearStudio: A Human-Intervenable Framework for Building Controllable Deep-Research Agents (https://arxiv.org/abs/2510.12194) | 2025/10/14 (https://github.com/ResearAI/ResearStudio) GitHub stars (https://github.com/ResearAI/ResearStudio) | gpt-4.1, gpt-4.1-mini, o4-mini, Llama-3.3-70B | Prompting | Web 搜索 | 单代理 | – | GAIA | | HiPRAG: Hierarchical Process Rewards for Efficient Agentic Retrieval Augmented Generation (https://arxiv.org/abs/2510.07794) | 2025/10/09 (https://github.com/qualidea1217/HiPRAG) GitHub stars (https://github.com/qualidea1217/HiPRAG) | Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct, Llama-3.2-3B-Instruct | PPO, GRPO | Web 搜索 | 单代理 | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2Wiki, MuSiQue, Bamboogle | | A2SEARCH: Ambiguity-Aware Question Answering with Reinforcement Learning (https://arxiv.org/abs/2510.07958) | 2025/10/09 (https://github.com/zfj1998/A2Search) GitHub stars (https://github.com/zfj1998/A2Search) | Qwen-2.5 系列 | RL | Web 搜索 | 单代理 | NQ | MuSiQue, HotpotQA, 2Wiki, Bamboogle, NQ, TriviaQA, PopQA, AmbigQA | | ReSeek: A Self-Correcting Framework for Search Agents with Instructive Rewards (https://arxiv.org/pdf/2510.00568) | 2025/10/01 (https://tencentbac.github.io/ReSeek/) | Qwen2.5-7B-Instruct, Qwen2.5-3B-Instruct | GRPO | Web 搜索, 本地检索 | 单代理 | NQ, HotpotQA | NQ, TriviaQA, PopQA, HotpotQA, 2WikiMQA, MuSiQue, Bamboogle, FictionalHot | | Process-Supervised Reinforcement Learning for Interactive Multimodal Tool-Use Agents (https://arxiv.org/abs/2509.14480) | 2025/09/17 | Qwen3-8B, Qwen2.5-Omni-7B | RL | 本地检索 | 单代理 | τ-bench, APIGen-MT | τ-bench | | ReSum: Unlocking Long-Horizon Search Intelligence via Context Summarization (https://arxiv.org/abs/2509.13313) | 2025/09/16 (https://github.com/Alibaba-NLP/DeepResearch) GitHub stars (https://github.com/Alibaba-NLP/DeepResearch) | Qwen3-30B-A3B-Thinking | GRPO, SFT | Web 搜索 | 单代理 | SailorFog-QA | BrowseComp-en/zh | | WebWeaver: Structuring Web-Scale Evidence with Dynamic Outlines for Open-Ended Deep Research (https://arxiv.org/abs/2509.13312) | 2025/09/16 (https://github.com/Alibaba-NLP/DeepResearch) GitHub stars (https://github.com/Alibaba-NLP/DeepResearch) | Qwen3-30B-A3B-Instruct | SFT | Web 搜索 | 单代理 | WebWeaver-3k | BrowseComp-en/zh, GAIA, XBench-DeepSearch | | WebResearcher: Unleashing Unbounded Reasoning Capability in Long-Horizon Agents (https://arxiv.org/abs/2509.13309) | 2025/09/16 (https://github.com/Alibaba-NLP/DeepResearch) GitHub stars (https://github.com/Alibaba-NLP/DeepResearch) | Qwen3-30B-A3B | RFT, RL | Web 搜索 | 多代理 | WebFrontier | BrowseComp-en/zh, GAIA, WebWalkerQA, FRAMES, HotpotQA, MuSiQue, 2WikiMultiHopQA | | WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable RL (https://arxiv.org/abs/2509.13305) | 2025/09/16 (https://github.com/Alibaba-NLP/DeepResearch) GitHub stars (https://github.com/Alibaba-NLP/DeepResearch) | Qwen3-30B-A3B | SFT, RL | Web 搜索, 本地检索 | 单代理 | SailorFog-QA-V2 | BrowseComp-EN, BrowseComp-ZH, HLE | | WebExplorer: Explore and Evolve for Training Long-Horizon Web Agents (https://arxiv.org/abs/2509.06501) | 2025/09/08 (https://github.com/hkust-nlp/WebExplorer) GitHub stars (https://github.com/hkust-nlp/WebExplorer) | Qwen3-8B | GRPO, SFT | Web 搜索 | 单代理 | WebExplorer-QA | BrowseComp-en/zh, GAIA, WebWalkerQA, FRAMES, XBench-DeepSearch, HLE | | Atom-Searcher: Enhancing Agentic Deep Research via Fine-Grained Atomic Thought Reward (https://arxiv.org/abs/2508.12800) | 2025/08/18 (https://github.com/antgroup/Research-Venus) GitHub stars (https://github.com/antgroup/Research-Venus) | Qwen2.5-7B | RL(GRPO) | Web 搜索 | 单代理 | NQ, SimpleQA, HotpotQA, 2WikiMultiHopQA, MuSiQue, MultiHopRAG | Bamboogle, NQ, SimpleQA, HotpotQA, 2WikiMultiHopQA, MuSiQue, MultiHopRAG | | MMSearch-R1: Incentivizing LMMs to Search (https://arxiv.org/pdf/2506.20670) | 2025/06/25 (https://github.com/EvolvingLMMs-Lab/multimodal-search-r1) GitHub stars (https://github.c