README
Awesome Deep Research Agent
We maintain a curated collection of papers exploring the path towards Deep Research (DR) Agents, focusing on formulating core concepts and mapping the research landscape.
⌛️ We’re continuously compiling and updating cutting‑edge insights. Feel free to suggest any related work you find valuable!
Build a digital assistant on your screen. Generated by DALL-E-3.
🔥 WELCOME CONTRIBUTE!
🔥 This project is actively maintained, and we welcome your contributions. If you have any suggestions, such as missing papers or information, please feel free to open an issue or submit a pull request.
📰 News
- [2025.09.03] 📄 Updated version of our paper "Deep Research Agents: A Systematic Examination And Roadmap" is now available on ArXiv with expanded analysis and refined research directions!
- [2025.06.18] 🚀 Our comprehensive survey "Deep Research Agents: A Systematic Examination And Roadmap" is officially released on ArXiv - providing systematic insights into the current state and future of deepresearch agents!
🏗️ Our Works Towards DR Agents
✨✨✨ Deep Research Agents: A Systematic Examination And Roadmap
📚 Awesome Papers
Table of Contents- Search Engine Integration
- Tool Use
- Architecture & Workflow
- Tuning Methods
- Industrial Applications
- Benchmarks for DR Agents
Search Engine Integration
📊 Search Engine · API vs Browser Comparison
| Legend | ✔️ Primary focus | 🟫 Secondary/minor focus | — Not present |
|---|
| DR Agent | API | Browser | GAIA | HLE | QA | Base Model |
|---|---|---|---|---|---|---|
| Avatar | 🟫 | — | — | — | Stark | Claude-3-Opus, GPT-4 |
| CoSearch-Agent | ✔️ | — | — | — | — | GPT-3.5-turbo |
| MMAC-Copilot | ✔️ | — | ✔️ | — | — | GPT-3.5, GPT-4 |
| Storm | 🟫 | — | — | — | FreshWiki | GPT-3.5-turbo |
| OpenResearcher | ✔️ | — | — | — | Private QA | DeepSeek-V2-Chat |
| The AI Scientist | ✔️ | — | — | — | MLE-Bench | GPT-4o, o1-mini, o1-preview |
| Gemini DR | ✔️ | ✔️ | — | ✔️ | GPQA | Gemini-2.0-Flash |
| Agent Laboratory | ✔️ | — | — | — | MLE-Bench | GPT-4o, o1-preview |
| Search-o1 | ✔️ | — | — | — | GPQA·NQ·TriviaQA | QwQ-32B-preview |
| WebWalker | — | — | — | — | WebWalkerQA | GPT-4o, Qwen-2.5 |
| Agentic Reasoning | ✔️ | — | — | — | GPQA | DeepSeek-R1, Qwen2.5 |
| AutoAgent | — | ✔️ | ✔️ | — | — | Claude-Sonnet-3.5 |
| Grok DeepSearch | ✔️ | ✔️ | — | — | GPQA | Grok 3 |
| OpenAI DR | — | ✔️ | ✔️ | ✔️ | ✔️ | GPT-o3 |
| Perplexity DR | ✔️ | 🟫 | — | ✔️ | SimoleQA | Flexible |
| Towards an AI Co-Scientist | ✔️ | — | — | — | GPQA | Gemini 2.0 |
| Nouswise | — | — | — | — | — | — |
| AgentRxiv | ✔️ | — | — | — | GPQA·MedQA | GPT-4o-mini |
| Agent-R1 | ✔️ | — | — | — | HotpotQA | Qwen2.5-1.5B-Inst |
| AutoGLM Rumination | — | ✔️ | — | — | GPQA | GLM-Z1-Air |
| Copilot Researcher | — | ✔️ | — | — | — | o3-mini |
| H2O.ai DR | ✔️ | ✔️ | ✔️ | — | — | h2ogpt-oasst1-512-12b |
| Manus | ✔️ | ✔️ | — | — | — | Claude3.5, GPT-4o |
| OpenManus | ✔️ | ✔️ | — | — | — | Claude3.5, GPT-4o |
| OWL | ✔️ | ✔️ | ✔️ | — | — | DeepSeek-R1, Gemini-2.5-Pro, GPT-4o |
| R1-Searcher | 🟫 | — | — | — | 2WikiMultiHopQA, HotpotQA | Llama3.1-8B-Inst, Qwen2.5-7B |
| ReSearch | 🟫 | — | — | — | 2WikiMultiHopQA, HotpotQA | Qwen2.5-7B, Qwen2.5-7B-Inst |
| Search-R1 | 🟫 | — | — | — | 2WikiMultiHopQA, HotpotQA, NQ, TriviaQA | Llama3.2-3B, Qwen2.5-3B/7B |
| DeepResearcher | — | ✔️ | — | — | HotpotQA, NQ, TriviaQA | Qwen2.5-7B-Inst |
| Genspark Super Agent | ✔️ | ✔️ | ✔️ | — | — | Mixture of 9 LLMs |
| WebThinker | ✔️ | — | ✔️ | ✔️ | GPQA, WebWalkerQA | QwQ-32B |
| SWIRL | — | ✔️ | — | — | HotQA, BeerQA | Gemma 2-27B |
| SimpleDeepSearcher | — | ✔️ | ✔️ | — | 2WikiMultiHopQA | Qwen-2.5-7B/32B-In, DeepSeek-D-Qwen-2.5-32B, QwQ-32B |
| Suna AI | ✔️ | ✔️ | — | — | — | GPT-4o, Claude |
| AgenticSeek | — | ✔️ | — | — | — | GPT-4o, DeepSeek-R1, Claude |
| Alita | ✔️ | ✔️ | ✔️ | — | PathVQA | GPT-4o, Claude-Sonnet-4 |
| DeerFlow | ✔️ | — | — | — | — | Doubao-1.5-Pro-32k, DeepSeek-R1, GPT-4o, Qwen |
| PANGU DEEPDIVER | ✔️ | — | — | — | C-SimpleQA, HotpotQA, ProxyQA | Pangu-7B-Reasoner |
| WebDancer | ✔️ | — | ✔️ | — | GAIA, WebWalkerQA | Qwen-2.5, QwQ-32B, DeepSeek-R1, GPT-4o |
| O-agents | ✔️ | — | ✔️ | — | — | GPT-4o, GPT-4.1, Claude-3.7-Sonnet, DeepSeek-R1, Gemini-2.5 |
| Kimi-Researcher | ✔️ | ✔️ | — | ✔️ | SimpleQA | Kimi k1.5/k2 |
| WebSailor | ✔️ | — | ✔️ | — | SimpleQA | Qwen-2.5 |
| Agent-KB | ✔️ | — | ✔️ | — | SWE-bench | GPT-4o, GPT-4.1, Claude-3.7-Sonnet, o3-mini, Qwen-3, DeepSeek-R1 |
| WebShaper | ✔️ | — | ✔️ | — | WebWalkerQA | Qwen-2.5, QwQ-32B |
| Deep Researcher with Test-Time Diffusion | ✔️ | — | ✔️ | ✔️ | — | Gemini-2.5-Pro |
| ChatGPT-Agent | — | — | — | — | — | — |
| AWorld | ✔️ | ✔️ | ✔️ | — | HotpotQA | Gemini-2.5-Pro, GPT-4o |
| Cognitive Kernel-Pro | ✔️ | ✔️ | ✔️ | — | AgentWebQA, WebWalkerQA, Multi-hop URLQA, DocBench, TableBench | Claude-3.7-Sonnet, CK-Pro-8B |
| WebWatcher | ✔️ | — | — | ✔️ | Browsercom-VL, LiveVQA, MMSearch | Qwen-2.5-VL-32B |
| WideSearch | ✔️ | — | — | — | WideSearch | DeepSeek-R1, Doubao-Seed-1.6, Claude Sonnet 4, Gemini-2.5-Pro |
| MiroRL | ✔️ | — | ✔️ | — | — | Qwen3-14B |
Tool Use
📊 Tool Use Capabilities Comparison
| Legend | ✔️ Involved | 🟫 Non Disclosure | — Not present |
|---|
| DR Agent | Code Interp. | Data Analytics | Multimodal | Release |
|---|---|---|---|---|
| CoSearchAgent | — | ✔️ | — | Feb-2024 |
| Storm | ✔️ | — | — | Jul-2024 |
| The AI Scientist | ✔️ | — | — | Aug-2024 |
| Agent Laboratory | ✔️ | — | — | Jan-2025 |
| Agentic Reasoning | ✔️ | — | — | Feb-2025 |
| AutoAgent | ✔️ | — | ✔️ | Feb-2025 |
| Genspark DR | ✔️ | ✔️ | ✔️ | Feb-2025 |
| Grok DeepSearch | ✔️ | ✔️ | ✔️ | Feb-2025 |
| OpenAI DR | ✔️ | ✔️ | ✔️ | Feb-2025 |
| Perplexity DR | ✔️ | ✔️ | ✔️ | Feb-2025 |
| Towards an AI co-scientist | — | ✔️ | ✔️ | Feb-2025 |
| Agent-R1 | ✔️ | — | — | Mar-2025 |
| AutoGLM Romination | ✔️ | — | ✔️ | Mar-2025 |
| Copilot Researcher | ✔️ | ✔️ | 🟫 | Mar-2025 |
| Manus | ✔️ | ✔️ | ✔️ | Mar-2025 |
| OpenManus | ✔️ | ✔️ | — | Mar-2025 |
| OWL | ✔️ | ✔️ | ✔️ | Mar-2025 |
| H2O.ai DR | ✔️ | ✔️ | ✔️ | Mar-2025 |
| Genspark Super Agent | ✔️ | ✔️ | ✔️ | Apr-2025 |
| WebThinker | ✔️ | — | — | Apr-2025 |
| Suna Ai | ✔️ | ✔️ | — | Apr-2025 |
| Tool-Star | ✔️ | ✔️ | — | May-2025 |
| AgenticSeek | ✔️ | ✔️ | — | May-2025 |
| Alita | ✔️ | 🟫 | 🟫 | May-2025 |
| DeerFlow | ✔️ | ✔️ | — | May-2025 |
| O-agents | ✔️ | ✔️ | ✔️ | Jun-2025 |
| Kimi-Researcher | ✔️ | ✔️ | — | Jun-2025 |
| Agent-KB | ✔️ | ✔️ | ✔️ | Jul-2025 |
| AWorld | ✔️ | ✔️ | ✔️ | Jul-2025 |
| Cognitive Kernel-Pro | ✔️ | ✔️ | ✔️ | Aug-2025 |
| WebWatcher | ✔️ | ✔️ | ✔️ | Aug-2025 |
| MiroRL | ✔️ | ✔️ | — | Aug-2025 |
Architecture & Workflow
Static Workflow
- AvaTaR: Optimizing LLM Agents for Tool Usage via Contrastive Reasoning
- Agent Laboratory: Using LLM Agents as Research Assistants
- CoSearchAgent: A Lightweight Collaborative Search Agent with Large Language Models
- AgentRxiv: Towards Collaborative Autonomous Research
- The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery
- Nouswise
Dynamic Single‑Agent Workflow
- OpenResearcher: Unleashing AI for Accelerated Scientific Research
- Search‑o1: Agentic Search‑Enhanced Large Reasoning Models
- Agentic Reasoning: Reasoning LLMs with Tools for the Deep Research
- Agent‑R1: Training Powerful LLM Agents with End‑to‑End Reinforcement Learning
- R1‑Searcher: Incentivizing the Search Capability in LLMs via Reinforcement Learning
- Search‑R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning
- DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real‑world Environments
- WebThinker: Empowering Large Reasoning Models with Deep Research Capability
- Open AI Deep Research
- Google Gemini Deep Research
- Grok Deep Search
- Perplexity Deep Research
- ReSearch: Learning to Reason with Search for LLMs via Reinforcement Learning
- Synthetic Data Generation & Multi-Step RL for Reasoning & Tool Use
- H2O.ai Deep Research
- SimpleDeepSearcher: Deep Information Seeking via Web-Powered Reasoning Trajectory Synthesis
- WebDancer: Towards Autonomous Information Seeking Agency
- Synthetic Data Generation & Multi‑Step RL for Reasoning & Tool Use
- SimpleDeepSearcher: Deep Information Seeking via Web‑Powered Reasoning Trajectory Synthesis
- Suna AI
- Alita: Generalist Agent Enabling Scalable Agentic Reasoning with Minimal Predefinition and Maximal Self‑Evolution
- AgenticSeek
- AutoGLM Rumination
- O-agents
- WebDancer
- WebSailor
- WebShaper
- Kimi-Researcher
- Deep Researcher with Test-Time Diffusion
- WebWatcher
- Chat-Agent
- Cognitive Kernel-Pro
- Tool-Star
- MiroRL
- Memento
Dynamic Multi‑Agent Workflow
- Assisting in Writing Wikipedia-like Articles From Scratch with Large Language Models
- AutoAgent: A Fully-Automated and Zero-Code Framework for LLM Agents
- OpenManus
- Manus
- OWL
- PANGU DEEPDIVER: ADAPTIVE SEARCH INTENSITY SCALING VIA OPEN-WEB REINFORCEMENT LEARNING
- H2O.ai Deep Research
- Copilot Researche
- Genspark Super Agent
- DeerFlow
- MMAC-Copilot
- Agent KB
- Towards an AI co-scientist
- WebWalker
- AWorld
- Atom-Searcher
Tuning Methods
📊 Tuning Methods Comparison
| Legend | ✔️ Implemented | 🟫 Details Unknown | — Not present |
|---|
| DR Agent | SFT | RL | Base Model | Data | Reward Design |
|---|---|---|---|---|---|
| Gemini DR | 🟫 | 🟫 | Gemini-2.0-Flash | — | 🟫 |
| WebWalker | — | — | GPT-4o, Qwen-2.5 (7–72B) | WebWalkerQA | — |
| Grok DeepSearch | — | 🟫 | Grok 3 | — | 🟫 |
| OpenAI DR | — | 🟫 | GPT-o3 | — | 🟫 |
| Agentic Reasoning | ✔️ | — | DeepSeek-R1, Qwen2.5 | GPQA | Rule-Outcome |
| AutoAgent | — | ✔️ | Claude-Sonnet-3.5 | — | — |
| Towards an AI co-scientist | — | — | Gemini 2.0 | — | — |
| Agent-R1 | — | PPO · Reinforce++ · GRPO | Qwen2.5-1.5B-Inst | HotpotQA | Rule-Outcome |
| AutoGLM Rumination | 🟫 | 🟫 | GLM-Z1-Air | — | 🟫 |
| H2O.ai DR | ✔️ | 🟫 | h2ogpt-oasst1-512-12b | — | 🟫 |
| Copilot Researcher | 🟫 | 🟫 | o3-mini | — | — |
| ReSearch | — | GRPO | Qwen2.5-7B-Inst · Qwen2.5-32B-Inst | 2WikiMultiHopQA | Rule-Outcome |
| R1-Searcher | ✔️ | Reinforce++ · GRPO | Qwen2.5-7B-Inst / LLaMA-3.1-8B-Inst | 2WikiMultiHopQA · HotpotQA | Rule-Outcome |
| Search-R1 | ✔️ | PPO · GRPO | Qwen2.5-3B/7B · LLaMA3.2-3B-Inst | NQ · HotpotQA | Rule-Outcome |
| Nouswise | 🟫 | 🟫 | Nouswise | — | 🟫 |
| DeepResearcher | — | GRPO | Qwen2.5-7B-Inst | NQ · HotpotQA | Rule-Outcome |
| Genspark Super Agent | — | 🟫 | Mixture of Agents | — | 🟫 |
| WebThinker | ✔️ | Iterative Online DPO | QwQ-32B | Expert Dataset | Rule-Outcome |
| SWIRL | — | Offline-RL | Gemma 2-27B | HotPotQA | — |
| SimpleDeepSearcher | ✔️ | PPO | Qwen-2.5-7B-In · Qwen-2.5-32B-In · Deepseek-Distilled-Qwen-32B · QwQ-32B | NQ · HotpotQA · 2WikiMultiHopQA · Musique · SimpleQA · MultiHop-RAG | Process-based reward |
| PANGU DEEPDIVER | ✔️ | GRPO | Pangu-7B-Reasoner | WebPuzzle | Rule-Outcome |
| Tool-Star | ✔️ | GRPO | Qwen-2.5 | NuminaMath · HotpotQA · 2WikiMultiHopQA | Rule-Outcome |
| WebDancer | ✔️ | DAPO | Qwen-2.5-7B/32B · QwQ-32B · DeepSeek-R1 · GPT-4o | CRAWLQA · E2HQA | Rule-Outcome |
| O-agents | — | — | GPT-4o · GPT-4.1 · Claude-3.7-Sonnet · DeepSeek-R1 · Gemini-2.5 | — | — |
| Kimi-Researcher | — | REINFORCE | Kimi k1.5/k2 | — | Rule-Outcome |
| WebSailor | ✔️ | DUPO | Qwen-2.5-3B/7B/32B/72B | SailorFog-QA | Rule-Outcome |
| Agent-KB | — | — | GPT-4o · GPT-4.1 · Claude-3.7-Sonnet · o3-mini · Qwen-3 32B · DeepSeek-R1 | — | — |
| WebShaper | ✔️ | GRPO | Qwen-2.5-3B/7B/32B/72B · QwQ-32B | WebShaper | Rule-Outcome |
| Cognitive Kernel-Pro | ✔️ | — | Claude-3.7-Sonnet · CK-Pro-8B | OpenWebVoyager · Multi-hop URLQA · AgentWebQA · WebWalkerQA · DocBench · TableBench | — |
| WebWatcher | — | GRPO | Qwen-2.5-VL-32B | BrowseComp-VL · Long-tail VQA · Hard VQA | Rule-Outcome |
| MiroRL | ✔️ | GRPO | Qwen3-14B | MiroRL-GenQA | Rule-Outcome |
| Atom-Searcher | ✔️ | GRPO | Qwen2.5-7B-Inst | 2WikiMultiHopQA · HotpotQA | Atomic Thought Reward (ATR) |
Benchmarks for DR Agents
📊 QA Benchmarks (Hotpot / 2Wiki / NQ / TQ / GPQA)
| DR Agent | Base Model | Hotpot | 2Wiki | NQ | TQ | GPQA | Release |
|---|---|---|---|---|---|---|---|
| Search-o1 | QwQ-32B-preview | 57.3 | 71.4 | 49.7 | 74.1 | 57.9 | Jan-2025 |
| Agentic Reasoning | DeepSeek-R1, Qwen2.5 | — | — | — | — | 67.0 | Feb-2025 |
| Grok DeepSearch | Grok3 | — | — | — | — | 84.6 | Feb-2025 |
| AgentRxiv | GPT-4o-mini | — | — | — | — | 41.0 | Mar-2025 |
| R1-Searcher | Qwen2.5-7B-Base | 71.9 | 63.8 | — | — | — | Mar-2025 |
| ReSearch | Qwen2.5-32B-Inst | 67.7 | 50.0 | — | — | — | Mar-2025 |
| Search-R1 | Qwen2.5-7B-Inst | 34.5 | 36.9 | 40.9 | 55.2 | — | Mar-2025 |
| DeepResearcher | Qwen2.5-7B-Inst | 64.3 | 66.6 | 61.9 | 85.0 | — | Apr-2025 |
| WebThinker | QwQ-32B | — | — | — | — | 68.7 | Apr-2025 |
| SimpleDeepSearch | QwQ-32B | 73.5 | — | — | — | — | Apr-2025 |
| SWIRL | Gemma 2-27B | 72.0 | — | — | — | — | Apr-2025 |
| Tool-Star | Qwen2.5-3B | 51.9 | 40.0 | — | — | — | May-2025 |
📊 GAIA (Test and Val) Benchmarks
| DR Agent | Base Model | GAIA L-1 | L-2 | L-3 | Ave. | Release | Split |
|---|---|---|---|---|---|---|---|
| MMAC-Copilot | GPT-3.5, GPT-4 | 45.16 | 20.75 | 6.12 | 25.91 | Mar-2024 | Test |
| H2O.ai DR | Claude-3.7-Sonnet | 89.25 | 79.87 | 61.22 | 79.73 | Mar-2025 | Test |
| Alita | Claude-Sonnet-4, GPT-4o | 92.47 | 71.70 | 55.10 | 75.42 | May-2025 | Test |
| Agent-KB | GPT-4.1, Claude-3.7 | 84.91 | 74.42 | 57.69 | 75.15 | Jul-2025 | Test |
| O-agents | Claude-3.7 | 83.02 | 74.42 | 53.85 | 73.93 | Jun-2025 | Test |
| WebDancer | QwQ-32B | 61.5 | 50.0 | 25.0 | 51.5 | May-2025 | Test |
| WebShaper | Qwen-2.5-72B | 69.2 | 63.4 | 16.6 | 60.1 | Jul-2025 | Test |
| Deep Researcher w/ Test-Time Diffusion | Gemini-2.5-Pro | — | — | — | 69.1 | Jul-2025 | Test |
| Cognitive Kernel-Pro | Claude-3.7-Sonnet | 83.02 | 68.60 | 53.85 | 70.91 | Aug-2025 | Test |
| AutoAgent | Claude-Sonnet-3.5 | 71.7 | 53.5 | 26.9 | 55.2 | Feb-2025 | Dev |
| OpenAI DR | GPT-o3-customized | 78.7 | 73.2 | 58.0 | 67.4 | Feb-2025 | Dev |
| Manus | Claude 3.5, GPT-4o | 86.5 | 70.1 | 57.7 | 71.4 | Mar-2025 | Dev |
| OWL | Claude-3.7-Sonnet | 84.9 | 68.6 | 42.3 | 69.7 | Mar-2025 | Dev |
| H2O.ai DR | h2ogpt-oasst1-512-12b | 67.92 | 67.44 | 42.31 | 63.64 | Mar-2025 | Dev |
| Genspark Super Agent | Claude 3 Opus | 87.8 | 72.7 | 58.8 | 73.1 | Apr-2025 | Dev |
| WebThinker | QwQ-32B | 53.8 | 44.2 | 16.7 | 44.7 | Apr-2025 | Dev |
| SimpleDeepSearch | QwQ-32B | 50.5 | 45.8 | 13.8 | 43.9 | Apr-2025 | Dev |
| Alita | Claude-Sonnet-4, GPT-4o | 75.15 | — | 87.27 | — | May-2025 | Dev |
📄 Citation
If you find this work helpful, please cite our paper:
@article{huang2025deep,
title={Deep Research Agents: A Systematic Examination And Roadmap},
author={Huang, Yuxuan and Chen, Yihang and Zhang, Haozheng and Li, Kang and Fang, Meng and Yang, Linyi and Li, Xiaoguang and Shang, Lifeng and Xu, Songcen and Hao, Jianye and others},
journal={arXiv preprint arXiv:2506.18096},
year={2025}
}
Contributors
Similar repos
README
Awesome Deep Research Agent
We maintain a curated collection of papers exploring the path towards Deep Research (DR) Agents, focusing on formulating core concepts and mapping the research landscape.
⌛️ We’re continuously compiling and updating cutting‑edge insights. Feel free to suggest any related work you find valuable!
Build a digital assistant on your screen. Generated by DALL-E-3.
🔥 WELCOME CONTRIBUTE!
🔥 This project is actively maintained, and we welcome your contributions. If you have any suggestions, such as missing papers or information, please feel free to open an issue or submit a pull request.
📰 News
- [2025.09.03] 📄 Updated version of our paper "Deep Research Agents: A Systematic Examination And Roadmap" is now available on ArXiv with expanded analysis and refined research directions!
- [2025.06.18] 🚀 Our comprehensive survey "Deep Research Agents: A Systematic Examination And Roadmap" is officially released on ArXiv - providing systematic insights into the current state and future of deepresearch agents!
🏗️ Our Works Towards DR Agents
✨✨✨ Deep Research Agents: A Systematic Examination And Roadmap
📚 Awesome Papers
Table of Contents- Search Engine Integration
- Tool Use
- Architecture & Workflow
- Tuning Methods
- Industrial Applications
- Benchmarks for DR Agents
Search Engine Integration
📊 Search Engine · API vs Browser Comparison
| Legend | ✔️ Primary focus | 🟫 Secondary/minor focus | — Not present |
|---|
| DR Agent | API | Browser | GAIA | HLE | QA | Base Model |
|---|---|---|---|---|---|---|
| Avatar | 🟫 | — | — | — | Stark | Claude-3-Opus, GPT-4 |
| CoSearch-Agent | ✔️ | — | — | — | — | GPT-3.5-turbo |
| MMAC-Copilot | ✔️ | — | ✔️ | — | — | GPT-3.5, GPT-4 |
| Storm | 🟫 | — | — | — | FreshWiki | GPT-3.5-turbo |
| OpenResearcher | ✔️ | — | — | — | Private QA | DeepSeek-V2-Chat |
| The AI Scientist | ✔️ | — | — | — | MLE-Bench | GPT-4o, o1-mini, o1-preview |
| Gemini DR | ✔️ | ✔️ | — | ✔️ | GPQA | Gemini-2.0-Flash |
| Agent Laboratory | ✔️ | — | — | — | MLE-Bench | GPT-4o, o1-preview |
| Search-o1 | ✔️ | — | — | — | GPQA·NQ·TriviaQA | QwQ-32B-preview |
| WebWalker | — | — | — | — | WebWalkerQA | GPT-4o, Qwen-2.5 |
| Agentic Reasoning | ✔️ | — | — | — | GPQA | DeepSeek-R1, Qwen2.5 |
| AutoAgent | — | ✔️ | ✔️ | — | — | Claude-Sonnet-3.5 |
| Grok DeepSearch | ✔️ | ✔️ | — | — | GPQA | Grok 3 |
| OpenAI DR | — | ✔️ | ✔️ | ✔️ | ✔️ | GPT-o3 |
| Perplexity DR | ✔️ | 🟫 | — | ✔️ | SimoleQA | Flexible |
| Towards an AI Co-Scientist | ✔️ | — | — | — | GPQA | Gemini 2.0 |
| Nouswise | — | — | — | — | — | — |
| AgentRxiv | ✔️ | — | — | — | GPQA·MedQA | GPT-4o-mini |
| Agent-R1 | ✔️ | — | — | — | HotpotQA | Qwen2.5-1.5B-Inst |
| AutoGLM Rumination | — | ✔️ | — | — | GPQA | GLM-Z1-Air |
| Copilot Researcher | — | ✔️ | — | — | — | o3-mini |
| H2O.ai DR | ✔️ | ✔️ | ✔️ | — | — | h2ogpt-oasst1-512-12b |
| Manus | ✔️ | ✔️ | — | — | — | Claude3.5, GPT-4o |
| OpenManus | ✔️ | ✔️ | — | — | — | Claude3.5, GPT-4o |
| OWL | ✔️ | ✔️ | ✔️ | — | — | DeepSeek-R1, Gemini-2.5-Pro, GPT-4o |
| R1-Searcher | 🟫 | — | — | — | 2WikiMultiHopQA, HotpotQA | Llama3.1-8B-Inst, Qwen2.5-7B |
| ReSearch | 🟫 | — | — | — | 2WikiMultiHopQA, HotpotQA | Qwen2.5-7B, Qwen2.5-7B-Inst |
| Search-R1 | 🟫 | — | — | — | 2WikiMultiHopQA, HotpotQA, NQ, TriviaQA | Llama3.2-3B, Qwen2.5-3B/7B |
| DeepResearcher | — | ✔️ | — | — | HotpotQA, NQ, TriviaQA | Qwen2.5-7B-Inst |
| Genspark Super Agent | ✔️ | ✔️ | ✔️ | — | — | Mixture of 9 LLMs |
| WebThinker | ✔️ | — | ✔️ | ✔️ | GPQA, WebWalkerQA | QwQ-32B |
| SWIRL | — | ✔️ | — | — | HotQA, BeerQA | Gemma 2-27B |
| SimpleDeepSearcher | — | ✔️ | ✔️ | — | 2WikiMultiHopQA | Qwen-2.5-7B/32B-In, DeepSeek-D-Qwen-2.5-32B, QwQ-32B |
| Suna AI | ✔️ | ✔️ | — | — | — | GPT-4o, Claude |
| AgenticSeek | — | ✔️ | — | — | — | GPT-4o, DeepSeek-R1, Claude |
| Alita | ✔️ | ✔️ | ✔️ | — | PathVQA | GPT-4o, Claude-Sonnet-4 |
| DeerFlow | ✔️ | — | — | — | — | Doubao-1.5-Pro-32k, DeepSeek-R1, GPT-4o, Qwen |
| PANGU DEEPDIVER | ✔️ | — | — | — | C-SimpleQA, HotpotQA, ProxyQA | Pangu-7B-Reasoner |
| WebDancer | ✔️ | — | ✔️ | — | GAIA, WebWalkerQA | Qwen-2.5, QwQ-32B, DeepSeek-R1, GPT-4o |
| O-agents | ✔️ | — | ✔️ | — | — | GPT-4o, GPT-4.1, Claude-3.7-Sonnet, DeepSeek-R1, Gemini-2.5 |
| Kimi-Researcher | ✔️ | ✔️ | — | ✔️ | SimpleQA | Kimi k1.5/k2 |
| WebSailor | ✔️ | — | ✔️ | — | SimpleQA | Qwen-2.5 |
| Agent-KB | ✔️ | — | ✔️ | — | SWE-bench | GPT-4o, GPT-4.1, Claude-3.7-Sonnet, o3-mini, Qwen-3, DeepSeek-R1 |
| WebShaper | ✔️ | — | ✔️ | — | WebWalkerQA | Qwen-2.5, QwQ-32B |
| Deep Researcher with Test-Time Diffusion | ✔️ | — | ✔️ | ✔️ | — | Gemini-2.5-Pro |
| ChatGPT-Agent | — | — | — | — | — | — |
| AWorld | ✔️ | ✔️ | ✔️ | — | HotpotQA | Gemini-2.5-Pro, GPT-4o |
| Cognitive Kernel-Pro | ✔️ | ✔️ | ✔️ | — | AgentWebQA, WebWalkerQA, Multi-hop URLQA, DocBench, TableBench | Claude-3.7-Sonnet, CK-Pro-8B |
| WebWatcher | ✔️ | — | — | ✔️ | Browsercom-VL, LiveVQA, MMSearch | Qwen-2.5-VL-32B |
| WideSearch | ✔️ | — | — | — | WideSearch | DeepSeek-R1, Doubao-Seed-1.6, Claude Sonnet 4, Gemini-2.5-Pro |
| MiroRL | ✔️ | — | ✔️ | — | — | Qwen3-14B |
Tool Use
📊 Tool Use Capabilities Comparison
| Legend | ✔️ Involved | 🟫 Non Disclosure | — Not present |
|---|
| DR Agent | Code Interp. | Data Analytics | Multimodal | Release |
|---|---|---|---|---|
| CoSearchAgent | — | ✔️ | — | Feb-2024 |
| Storm | ✔️ | — | — | Jul-2024 |
| The AI Scientist | ✔️ | — | — | Aug-2024 |
| Agent Laboratory | ✔️ | — | — | Jan-2025 |
| Agentic Reasoning | ✔️ | — | — | Feb-2025 |
| AutoAgent | ✔️ | — | ✔️ | Feb-2025 |
| Genspark DR | ✔️ | ✔️ | ✔️ | Feb-2025 |
| Grok DeepSearch | ✔️ | ✔️ | ✔️ | Feb-2025 |
| OpenAI DR | ✔️ | ✔️ | ✔️ | Feb-2025 |
| Perplexity DR | ✔️ | ✔️ | ✔️ | Feb-2025 |
| Towards an AI co-scientist | — | ✔️ | ✔️ | Feb-2025 |
| Agent-R1 | ✔️ | — | — | Mar-2025 |
| AutoGLM Romination | ✔️ | — | ✔️ | Mar-2025 |
| Copilot Researcher | ✔️ | ✔️ | 🟫 | Mar-2025 |
| Manus | ✔️ | ✔️ | ✔️ | Mar-2025 |
| OpenManus | ✔️ | ✔️ | — | Mar-2025 |
| OWL | ✔️ | ✔️ | ✔️ | Mar-2025 |
| H2O.ai DR | ✔️ | ✔️ | ✔️ | Mar-2025 |
| Genspark Super Agent | ✔️ | ✔️ | ✔️ | Apr-2025 |
| WebThinker | ✔️ | — | — | Apr-2025 |
| Suna Ai | ✔️ | ✔️ | — | Apr-2025 |
| Tool-Star | ✔️ | ✔️ | — | May-2025 |
| AgenticSeek | ✔️ | ✔️ | — | May-2025 |
| Alita | ✔️ | 🟫 | 🟫 | May-2025 |
| DeerFlow | ✔️ | ✔️ | — | May-2025 |
| O-agents | ✔️ | ✔️ | ✔️ | Jun-2025 |
| Kimi-Researcher | ✔️ | ✔️ | — | Jun-2025 |
| Agent-KB | ✔️ | ✔️ | ✔️ | Jul-2025 |
| AWorld | ✔️ | ✔️ | ✔️ | Jul-2025 |
| Cognitive Kernel-Pro | ✔️ | ✔️ | ✔️ | Aug-2025 |
| WebWatcher | ✔️ | ✔️ | ✔️ | Aug-2025 |
| MiroRL | ✔️ | ✔️ | — | Aug-2025 |
Architecture & Workflow
Static Workflow
- AvaTaR: Optimizing LLM Agents for Tool Usage via Contrastive Reasoning
- Agent Laboratory: Using LLM Agents as Research Assistants
- CoSearchAgent: A Lightweight Collaborative Search Agent with Large Language Models
- AgentRxiv: Towards Collaborative Autonomous Research
- The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery
- Nouswise
Dynamic Single‑Agent Workflow
- OpenResearcher: Unleashing AI for Accelerated Scientific Research
- Search‑o1: Agentic Search‑Enhanced Large Reasoning Models
- Agentic Reasoning: Reasoning LLMs with Tools for the Deep Research
- Agent‑R1: Training Powerful LLM Agents with End‑to‑End Reinforcement Learning
- R1‑Searcher: Incentivizing the Search Capability in LLMs via Reinforcement Learning
- Search‑R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning
- DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real‑world Environments
- WebThinker: Empowering Large Reasoning Models with Deep Research Capability
- Open AI Deep Research
- Google Gemini Deep Research
- Grok Deep Search
- Perplexity Deep Research
- ReSearch: Learning to Reason with Search for LLMs via Reinforcement Learning
- Synthetic Data Generation & Multi-Step RL for Reasoning & Tool Use
- H2O.ai Deep Research
- SimpleDeepSearcher: Deep Information Seeking via Web-Powered Reasoning Trajectory Synthesis
- WebDancer: Towards Autonomous Information Seeking Agency
- Synthetic Data Generation & Multi‑Step RL for Reasoning & Tool Use
- SimpleDeepSearcher: Deep Information Seeking via Web‑Powered Reasoning Trajectory Synthesis
- Suna AI
- Alita: Generalist Agent Enabling Scalable Agentic Reasoning with Minimal Predefinition and Maximal Self‑Evolution
- AgenticSeek
- AutoGLM Rumination
- O-agents
- WebDancer
- WebSailor
- WebShaper
- Kimi-Researcher
- Deep Researcher with Test-Time Diffusion
- WebWatcher
- Chat-Agent
- Cognitive Kernel-Pro
- Tool-Star
- MiroRL
- Memento
Dynamic Multi‑Agent Workflow
- Assisting in Writing Wikipedia-like Articles From Scratch with Large Language Models
- AutoAgent: A Fully-Automated and Zero-Code Framework for LLM Agents
- OpenManus
- Manus
- OWL
- PANGU DEEPDIVER: ADAPTIVE SEARCH INTENSITY SCALING VIA OPEN-WEB REINFORCEMENT LEARNING
- H2O.ai Deep Research
- Copilot Researche
- Genspark Super Agent
- DeerFlow
- MMAC-Copilot
- Agent KB
- Towards an AI co-scientist
- WebWalker
- AWorld
- Atom-Searcher
Tuning Methods
📊 Tuning Methods Comparison
| Legend | ✔️ Implemented | 🟫 Details Unknown | — Not present |
|---|
| DR Agent | SFT | RL | Base Model | Data | Reward Design |
|---|---|---|---|---|---|
| Gemini DR | 🟫 | 🟫 | Gemini-2.0-Flash | — | 🟫 |
| WebWalker | — | — | GPT-4o, Qwen-2.5 (7–72B) | WebWalkerQA | — |
| Grok DeepSearch | — | 🟫 | Grok 3 | — | 🟫 |
| OpenAI DR | — | 🟫 | GPT-o3 | — | 🟫 |
| Agentic Reasoning | ✔️ | — | DeepSeek-R1, Qwen2.5 | GPQA | Rule-Outcome |
| AutoAgent | — | ✔️ | Claude-Sonnet-3.5 | — | — |
| Towards an AI co-scientist | — | — | Gemini 2.0 | — | — |
| Agent-R1 | — | PPO · Reinforce++ · GRPO | Qwen2.5-1.5B-Inst | HotpotQA | Rule-Outcome |
| AutoGLM Rumination | 🟫 | 🟫 | GLM-Z1-Air | — | 🟫 |
| H2O.ai DR | ✔️ | 🟫 | h2ogpt-oasst1-512-12b | — | 🟫 |
| Copilot Researcher | 🟫 | 🟫 | o3-mini | — | — |
| ReSearch | — | GRPO | Qwen2.5-7B-Inst · Qwen2.5-32B-Inst | 2WikiMultiHopQA | Rule-Outcome |
| R1-Searcher | ✔️ | Reinforce++ · GRPO | Qwen2.5-7B-Inst / LLaMA-3.1-8B-Inst | 2WikiMultiHopQA · HotpotQA | Rule-Outcome |
| Search-R1 | ✔️ | PPO · GRPO | Qwen2.5-3B/7B · LLaMA3.2-3B-Inst | NQ · HotpotQA | Rule-Outcome |
| Nouswise | 🟫 | 🟫 | Nouswise | — | 🟫 |
| DeepResearcher | — | GRPO | Qwen2.5-7B-Inst | NQ · HotpotQA | Rule-Outcome |
| Genspark Super Agent | — | 🟫 | Mixture of Agents | — | 🟫 |
| WebThinker | ✔️ | Iterative Online DPO | QwQ-32B | Expert Dataset | Rule-Outcome |
| SWIRL | — | Offline-RL | Gemma 2-27B | HotPotQA | — |
| SimpleDeepSearcher | ✔️ | PPO | Qwen-2.5-7B-In · Qwen-2.5-32B-In · Deepseek-Distilled-Qwen-32B · QwQ-32B | NQ · HotpotQA · 2WikiMultiHopQA · Musique · SimpleQA · MultiHop-RAG | Process-based reward |
| PANGU DEEPDIVER | ✔️ | GRPO | Pangu-7B-Reasoner | WebPuzzle | Rule-Outcome |
| Tool-Star | ✔️ | GRPO | Qwen-2.5 | NuminaMath · HotpotQA · 2WikiMultiHopQA | Rule-Outcome |
| WebDancer | ✔️ | DAPO | Qwen-2.5-7B/32B · QwQ-32B · DeepSeek-R1 · GPT-4o | CRAWLQA · E2HQA | Rule-Outcome |
| O-agents | — | — | GPT-4o · GPT-4.1 · Claude-3.7-Sonnet · DeepSeek-R1 · Gemini-2.5 | — | — |
| Kimi-Researcher | — | REINFORCE | Kimi k1.5/k2 | — | Rule-Outcome |
| WebSailor | ✔️ | DUPO | Qwen-2.5-3B/7B/32B/72B | SailorFog-QA | Rule-Outcome |
| Agent-KB | — | — | GPT-4o · GPT-4.1 · Claude-3.7-Sonnet · o3-mini · Qwen-3 32B · DeepSeek-R1 | — | — |
| WebShaper | ✔️ | GRPO | Qwen-2.5-3B/7B/32B/72B · QwQ-32B | WebShaper | Rule-Outcome |
| Cognitive Kernel-Pro | ✔️ | — | Claude-3.7-Sonnet · CK-Pro-8B | OpenWebVoyager · Multi-hop URLQA · AgentWebQA · WebWalkerQA · DocBench · TableBench | — |
| WebWatcher | — | GRPO | Qwen-2.5-VL-32B | BrowseComp-VL · Long-tail VQA · Hard VQA | Rule-Outcome |
| MiroRL | ✔️ | GRPO | Qwen3-14B | MiroRL-GenQA | Rule-Outcome |
| Atom-Searcher | ✔️ | GRPO | Qwen2.5-7B-Inst | 2WikiMultiHopQA · HotpotQA | Atomic Thought Reward (ATR) |
Benchmarks for DR Agents
📊 QA Benchmarks (Hotpot / 2Wiki / NQ / TQ / GPQA)
| DR Agent | Base Model | Hotpot | 2Wiki | NQ | TQ | GPQA | Release |
|---|---|---|---|---|---|---|---|
| Search-o1 | QwQ-32B-preview | 57.3 | 71.4 | 49.7 | 74.1 | 57.9 | Jan-2025 |
| Agentic Reasoning | DeepSeek-R1, Qwen2.5 | — | — | — | — | 67.0 | Feb-2025 |
| Grok DeepSearch | Grok3 | — | — | — | — | 84.6 | Feb-2025 |
| AgentRxiv | GPT-4o-mini | — | — | — | — | 41.0 | Mar-2025 |
| R1-Searcher | Qwen2.5-7B-Base | 71.9 | 63.8 | — | — | — | Mar-2025 |
| ReSearch | Qwen2.5-32B-Inst | 67.7 | 50.0 | — | — | — | Mar-2025 |
| Search-R1 | Qwen2.5-7B-Inst | 34.5 | 36.9 | 40.9 | 55.2 | — | Mar-2025 |
| DeepResearcher | Qwen2.5-7B-Inst | 64.3 | 66.6 | 61.9 | 85.0 | — | Apr-2025 |
| WebThinker | QwQ-32B | — | — | — | — | 68.7 | Apr-2025 |
| SimpleDeepSearch | QwQ-32B | 73.5 | — | — | — | — | Apr-2025 |
| SWIRL | Gemma 2-27B | 72.0 | — | — | — | — | Apr-2025 |
| Tool-Star | Qwen2.5-3B | 51.9 | 40.0 | — | — | — | May-2025 |
📊 GAIA (Test and Val) Benchmarks
| DR Agent | Base Model | GAIA L-1 | L-2 | L-3 | Ave. | Release | Split |
|---|---|---|---|---|---|---|---|
| MMAC-Copilot | GPT-3.5, GPT-4 | 45.16 | 20.75 | 6.12 | 25.91 | Mar-2024 | Test |
| H2O.ai DR | Claude-3.7-Sonnet | 89.25 | 79.87 | 61.22 | 79.73 | Mar-2025 | Test |
| Alita | Claude-Sonnet-4, GPT-4o | 92.47 | 71.70 | 55.10 | 75.42 | May-2025 | Test |
| Agent-KB | GPT-4.1, Claude-3.7 | 84.91 | 74.42 | 57.69 | 75.15 | Jul-2025 | Test |
| O-agents | Claude-3.7 | 83.02 | 74.42 | 53.85 | 73.93 | Jun-2025 | Test |
| WebDancer | QwQ-32B | 61.5 | 50.0 | 25.0 | 51.5 | May-2025 | Test |
| WebShaper | Qwen-2.5-72B | 69.2 | 63.4 | 16.6 | 60.1 | Jul-2025 | Test |
| Deep Researcher w/ Test-Time Diffusion | Gemini-2.5-Pro | — | — | — | 69.1 | Jul-2025 | Test |
| Cognitive Kernel-Pro | Claude-3.7-Sonnet | 83.02 | 68.60 | 53.85 | 70.91 | Aug-2025 | Test |
| AutoAgent | Claude-Sonnet-3.5 | 71.7 | 53.5 | 26.9 | 55.2 | Feb-2025 | Dev |
| OpenAI DR | GPT-o3-customized | 78.7 | 73.2 | 58.0 | 67.4 | Feb-2025 | Dev |
| Manus | Claude 3.5, GPT-4o | 86.5 | 70.1 | 57.7 | 71.4 | Mar-2025 | Dev |
| OWL | Claude-3.7-Sonnet | 84.9 | 68.6 | 42.3 | 69.7 | Mar-2025 | Dev |
| H2O.ai DR | h2ogpt-oasst1-512-12b | 67.92 | 67.44 | 42.31 | 63.64 | Mar-2025 | Dev |
| Genspark Super Agent | Claude 3 Opus | 87.8 | 72.7 | 58.8 | 73.1 | Apr-2025 | Dev |
| WebThinker | QwQ-32B | 53.8 | 44.2 | 16.7 | 44.7 | Apr-2025 | Dev |
| SimpleDeepSearch | QwQ-32B | 50.5 | 45.8 | 13.8 | 43.9 | Apr-2025 | Dev |
| Alita | Claude-Sonnet-4, GPT-4o | 75.15 | — | 87.27 | — | May-2025 | Dev |
📄 Citation
If you find this work helpful, please cite our paper:
@article{huang2025deep,
title={Deep Research Agents: A Systematic Examination And Roadmap},
author={Huang, Yuxuan and Chen, Yihang and Zhang, Haozheng and Li, Kang and Fang, Meng and Yang, Linyi and Li, Xiaoguang and Shang, Lifeng and Xu, Songcen and Hao, Jianye and others},
journal={arXiv preprint arXiv:2506.18096},
year={2025}
}