Elamur Research⌕
☰
home / papers / 2025 / WebSailor: Navigating Super-human Reasoning for Web Agent

WebSailor: Navigating Super-human Reasoning for Web Agent

Paper · 2025 · Kuan Li, Zhongwang Zhang, Huifeng Yin et al.

Summary

Open-source LLMs and web agents score close to zero on demanding browsing benchmarks such as BrowseComp-en/zh, while proprietary systems like DeepResearch handle them well. The authors argue that open models lack one skill: steadily narrowing extreme uncertainty when no fixed solution route exists. They call these "Level 3" tasks, as opposed to simple lookups or standard multi-hop questions. WebSailor is a complete post-training recipe for ReAct agents that use search and visit tools. First, the SailorFog-QA dataset is built. Knowledge graphs are grown stochastically from Wikidata seed entities using real web content, subgraphs are sampled, and the resulting questions are deliberately blurred, for example by making dates vague. Second, successful tool-use traces from QwQ and DeepSeek-R1 are kept, but their wordy reasoning is rewritten into short, action-focused thoughts. A small RFT cold start is followed by RL with DUPO, a sampling scheme designed to make agentic RL faster. The resulting models outperform every open-source agent and approach proprietary systems, which substantially narrows the gap between open and closed web agents.

Key points

  • Task taxonomy: information-seeking tasks fall into three levels. Level 3 combines high uncertainty with no predefined reasoning path, and existing training data rarely covers it.
  • SailorFog-QA difficulty: the data comes from random-walk-like graph expansion, subgraph sampling and obfuscation. Some questions push even o3 to make up to about 40 tool calls.
  • Training pipeline: the RFT cold start uses only a little over 2k high-quality examples. It is followed by Duplicating Sampling Policy Optimization (DUPO), which applies dynamic sampling both before and during RL.
  • Agent setup: the agent follows the ReAct framework with two tools. Search returns 10 results per query, and visit returns a webpage summary focused on a stated goal.
  • Results: WebSailor models come in 3B, 7B, 32B and 72B sizes and lead all open-source agents on BrowseComp-en/zh. They also outperform browsing-enabled Grok-3 and Doubao, and they transfer well to simpler benchmarks (GAIA, XBench-DeepSearch, SimpleQA). Code is available at github.com/Alibaba-NLP/WebAgent.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 19,503 worksArtificial Intelligence · 18,805 worksArts and Humanities · 17,959 worksLanguage and Linguistics · 17,789 worksInformation Systems · 1,888 works

🧭 Topics 1Research subjects and languages the work is about.Reasoning · 1,948 works

Cite

Cite as Kuan Li, Zhongwang Zhang, Huifeng Yin, Liwen Zhang, Litu Ou, Jialong Wu, Wenbiao Yin, Baixuan Li, Zhengwei Tao, Xinyu Wang, Weizhou Shen, Junkai Zhang, Dingchu Zhang, Xixi Wu, Yong Jiang, Ming Yan, Pengjun Xie, Fei Huang and Jingren Zhou. 2025. WebSailor: Navigating Super-human Reasoning for Web Agent. arXiv:2507.02592.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2507.02592 · arXiv 2507.02592

@misc{li2025websailor,
  title = {{WebSailor: Navigating Super-human Reasoning for Web Agent}},
  author = {Kuan Li, Zhongwang Zhang, Huifeng Yin, Liwen Zhang, Litu Ou, Jialong Wu, Wenbiao Yin, Baixuan Li, Zhengwei Tao, Xinyu Wang, Weizhou Shen, Junkai Zhang, Dingchu Zhang, Xixi Wu, Yong Jiang, Ming Yan, Pengjun Xie, Fei Huang and Jingren Zhou},
  year = {2025},
  doi = {10.48550/arxiv.2507.02592},
  eprint = {2507.02592},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2507.02592}
}

References

60 references, 32 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Dario Amodei, Chris Olah, Jacob Steinhardt, Paul Christiano, John Schulman, and Dan Mané. Concrete problems in ai safety. arXiv preprint arXiv:1606.06565, 2016.
  2. Baian Chen, Chang Shu, Ehsan Shareghi, Nigel Collier, Karthik Narasimhan, and Shunyu Yao. Fireact: Toward language agent fine-tuning. arXiv preprint arXiv:2310.05915, 2023.
  3. Hardy Chen, Haoqin Tu, Fali Wang, Hui Liu, Xianfeng Tang, Xinya Du, Yuyin Zhou, and Cihang Xie. Sft or rl? an early investigation into training r1-like reasoning large vision-language models. arXiv preprint arXiv:2504.11468, 2025.
  4. Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde De Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, et al. Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374, 2021.
  5. Tianzhe Chu, Yuexiang Zhai, Jihan Yang, Shengbang Tong, Saining Xie, Dale Schuurmans, Quoc V Le, Sergey Levine, and Yi Ma. Sft memorizes, rl generalizes: A comparative study of foundation model post-training. arXiv preprint arXiv:2501.17161, 2025.
  6. ByteDance Doubao. Doubao, 2025. URL http://www.doubao.com/. ↗
  7. Google. Measuring compositional generalization, 2020. URL https://research.google/blog/measuri ng-compositional-generalization/. ↗
  8. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning. arXiv preprint arXiv:2501.12948, 2025.
  9. Xanh Ho, Anh-Khoa Duong Nguyen, Saku Sugawara, and Akiko Aizawa. Constructing a multi-hop qa dataset for comprehensive evaluation of reasoning steps, 2020. URL https://arxiv.org/abs/2011.0 1060. ↗
  10. Jingcheng Hu, Yinmin Zhang, Qi Han, Daxin Jiang, Xiangyu Zhang, and Heung-Yeung Shum. Open-reasoner-zero: An open source approach to scaling up reinforcement learning on the base model. arXiv preprint arXiv:2503.24290, 2025.
  11. Yuheng Huang, Jiayang Song, Zhijie Wang, Shengming Zhao, Huaming Chen, Felix Juefei-Xu, and Lei Ma. Look before you leap: An exploratory study of uncertainty measurement for large language models. arXiv preprint arXiv:2307.10236, 2023. — [arXiv:2307.10236](https://arxiv.org/abs/2307.10236) find
  12. Bowen Jin, Hansi Zeng, Zhenrui Yue, Jinsung Yoon, Sercan Arik, Dong Wang, Hamed Zamani, and Jiawei Han. Search-r1: Training llms to reason and leverage search engines with reinforcement learning. arXiv preprint arXiv:2503.09516, 2025.
  13. Jina.ai. Jina, 2025. URL https://jina.ai/. ↗
  14. Mandar Joshi, Eunsol Choi, Daniel S Weld, and Luke Zettlemoyer. Triviaqa: A large scale distantly supervised challenge dataset for reading comprehension. arXiv preprint arXiv:1705.03551, 2017.
  15. Kyle Jurado, Sydney C Ludvigson, and Serena Ng. Measuring uncertainty. American Economic Review, 105(3):1177–1216, 2015. find
  16. Sanyam Kapoor, Nate Gruver, Manley Roberts, Katherine Collins, Arka Pal, Umang Bhatt, Adrian Weller, Samuel Dooley, Micah Goldblum, and Andrew Gordon Wilson. Large language models must be taught to know what they don’t know. arXiv preprint arXiv:2406.08391, 2024. find
  17. Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, et al. Natural questions: a benchmark for question answering research. Transactions of the Association for Computational Linguistics, 7:4…
  18. Kuan Li, Liwen Zhang, Yong Jiang, Pengjun Xie, Fei Huang, Shuai Wang, and Minhao Cheng. Lara: Benchmarking retrieval-augmented generation and long-context llms–no silver bullet for lc or rag routing. arXiv preprint arXiv:2502.09977, 2025a. — [arXiv:2502.09977](https://arxiv.org/abs/2502.09977) find
  19. Xiaoxi Li, Guanting Dong, Jiajie Jin, Yuyao Zhang, Yujia Zhou, Yutao Zhu, Peitian Zhang, and Zhicheng Dou. Search-o1: Agentic search-enhanced large reasoning models. arXiv preprint arXiv:2501.05366, 2025b.
  20. Xiaoxi Li, Jiajie Jin, Guanting Dong, Hongjin Qian, Yutao Zhu, Yongkang Wu, Ji-Rong Wen, and Zhicheng Dou. Webthinker: Empowering large reasoning models with deep research capability. CoRR, abs/2504.21776, 2025c. doi: 10.48550/ARXIV.2504.21776. URL https://doi.org/10.48550/a rXiv.2504.21776.
  21. Yuxuan Liu, Tianchi Yang, Shaohan Huang, Zihan Zhang, Haizhen Huang, Furu Wei, Weiwei Deng, Feng Sun, and Qi Zhang. Calibrating llm-based evaluator. In Nicoletta Calzolari, Min-Yen Kan, Véronique Hoste, Alessandro Lenci, Sakriani Sakti, and Nianwen Xue (eds.), Proceedings of the 2024 Joint Internati…
  22. ELRA and ICCL, 2024. URL https://aclanthology.org/2024.lrec-main.237. ↗
  23. Grégoire Mialon, Clémentine Fourrier, Thomas Wolf, Yann LeCun, and Thomas Scialom. Gaia: a benchmark for general ai assistants. In The Twelfth International Conference on Learning Representations, 2023.
  24. OpenAI. Hello GPT-4o, 2024. URL https://openai.com/index/hello-gpt-4o/. ↗
  25. OpenAI. Deep research system card, 2025a. URL https://cdn.openai.com/deep-research-system-c ard.pdf. ↗
  26. OpenAI. Introducing openai gpt-4.1, 2025b. URL https://openai.com/index/gpt-4-1/. ↗
  27. OpenAI. Introducing openai o3 and o4-mini, 2025c. URL https://openai.com/index/introducing-o 3-and-o4-mini/. ↗
  28. OpenAI. Introducing simpleqa, 2025d. URL https://openai.com/index/introducing-simpleqa/. ↗
  29. Qwen Team. QwQ-32B: Embracing the power of reinforcement learning, March 2025. URL https://qwenlm.github.io/blog/qwq-32b/. ↗
  30. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Y Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300, 2024.
  31. Guangming Sheng, Chi Zhang, Zilingfeng Ye, Xibin Wu, Wang Zhang, Ru Zhang, Yanghua Peng, Haibin Lin, and Chuan Wu. Hybridflow: A flexible and efficient rlhf framework. In Proceedings of the Twentieth European Conference on Computer Systems, pp. 1279–1297, 2025.
  32. Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper, and Bryan Catanzaro. Megatron-lm: Training multi-billion parameter language models using model parallelism. arXiv preprint arXiv:1909.08053, 2019. find
  33. Huatong Song, Jinhao Jiang, Yingqian Min, Jie Chen, Zhipeng Chen, Wayne Xin Zhao, Lei Fang, and Ji-Rong Wen. R1-searcher: Incentivizing the search capability in llms via reinforcement learning. arXiv preprint arXiv:2503.05592, 2025.
  34. Chuanneng Sun, Songjun Huang, and Dario Pompili. Llm-based multi-agent reinforcement learning: Current and future directions. arXiv preprint arXiv:2405.11106, 2024. — [arXiv:2405.11106](https://arxiv.org/abs/2405.11106) find
  35. Yiyou Sun, Georgia Zhou, Hao Wang, Dacheng Li, Nouha Dziri, and Dawn Song. Climbing the ladder of reasoning: What llms can-and still can’t-solve after sft? arXiv preprint arXiv:2504.11741, 2025. find
  36. Gokul Swamy, Sanjiban Choudhury, Wen Sun, Zhiwei Steven Wu, and J Andrew Bagnell. All roads lead to likelihood: The value of reinforcement learning in fine-tuning. arXiv preprint arXiv:2503.01067, 2025. — [arXiv:2503.01067](https://arxiv.org/abs/2503.01067) find
  37. Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. Musique: Multihop questions via single-hop question composition. Transactions of the Association for Computational Linguistics, 10:539–554, 2022. — ♫ MuSiQue: Multihop Questions via Single-hop Question Composition
  38. Minzheng Wang, Longze Chen, Fu Cheng, Shengyi Liao, Xinghua Zhang, Bingli Wu, Haiyang Yu, Nan Xu, Lei Zhang, Run Luo, Yunshui Li, Min Yang, Fei Huang, and Yongbin Li. Leave no document behind: Benchmarking long-context llms with extended multi-doc QA. In Yaser Al-Onaizan, Mohit Bansal, and Yun-Nung…
  39. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824–24837, 2022.
  40. Jason Wei, Nguyen Karina, Hyung Won Chung, Yunxin Joy Jiao, Spencer Papay, Amelia Glaese, John Schulman, and William Fedus. Measuring short-form factuality in large language models. arXiv preprint arXiv:2411.04368, 2024.
  41. Jason Wei, Zhiqing Sun, Spencer Papay, Scott McKinney, Jeffrey Han, Isa Fulford, Hyung Won Chung, Alex Tachard Passos, William Fedus, and Amelia Glaese. Browsecomp: A simple yet challenging benchmark for browsing agents. arXiv preprint arXiv:2504.12516, 2025.
  42. Thaddäus Wiedemer, Prasanna Mayilvahanan, Matthias Bethge, and Wieland Brendel. Compositional generalization from first principles. Advances in Neural Information Processing Systems, 36:6941–6960, 2023. find
  43. Tom D Wilson. Models in information behaviour research. Journal of documentation, 55(3):249–270, 1999. find
  44. Jialong Wu, Baixuan Li, Runnan Fang, Wenbiao Yin, Liwen Zhang, Zhengwei Tao, Dingchu Zhang, Zekun Xi, Yong Jiang, Pengjun Xie, et al. Webdancer: Towards autonomous information seeking agency. arXiv preprint arXiv:2505.22648, 2025a.
  45. Jialong Wu, Wenbiao Yin, Yong Jiang, Zhenglin Wang, Zekun Xi, Runnan Fang, Linhai Zhang, Yulan He, Deyu Zhou, Pengjun Xie, and Fei Huang. Webwalker: Benchmarking llms in web traversal, 2025b. URL https://arxiv.org/abs/2501.07572. x.ai. Grok 3 beta — the age of reasoning agents, 2025. URL https://x.a… ↗
  46. Xbench-Team. Xbench-deepsearch, 2025. URL https://xbench.org/agi/aisearch. ↗
  47. Zhi-Qin John Xu, Yaoyu Zhang, Tao Luo, Yanyang Xiao, and Zheng Ma. Frequency principle: Fourier analysis sheds light on deep neural networks. arXiv preprint arXiv:1901.06523, 2019. — [arXiv:1901.06523](https://arxiv.org/abs/1901.06523) find
  48. Zhi-Qin John Xu, Yaoyu Zhang, and Tao Luo. Overview frequency principle/spectral bias in deep learning. Communications on Applied Mathematics and Computation, pp. 1–38, 2024. find
  49. Zhi-Qin John Xu, Yaoyu Zhang, and Zhangchen Zhou. An overview of condensation phenomenon in deep learning. arXiv preprint arXiv:2504.09484, 2025. — [arXiv:2504.09484](https://arxiv.org/abs/2504.09484) find
  50. An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, et al. Qwen2.5 technical report. arXiv preprint arXiv:2412.15115, 2024. find
  51. Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William W Cohen, Ruslan Salakhutdinov, and Christopher D Manning. Hotpotqa: A dataset for diverse, explainable multi-hop question answering. arXiv preprint arXiv:1809.09600, 2018.
  52. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. React: Synergizing reasoning and acting in language models. In International Conference on Learning Representations (ICLR), 2023.
  53. Yixin Ye, Zhen Huang, Yang Xiao, Ethan Chern, Shijie Xia, and Pengfei Liu. Limo: Less is more for reasoning. arXiv preprint arXiv:2502.03387, 2025.
  54. Huifeng Yin, Yu Zhao, Minghao Wu, Xuanfan Ni, Bo Zeng, Hao Wang, Tianqi Shi, Liangying Shao, Chenyang Lyu, Longyue Wang, et al. Towards widening the distillation bottleneck for reasoning models. arXiv e-prints, pp. arXiv–2503, 2025. find
  55. Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Tiantian Fan, Gaohong Liu, Lingjun Liu, Xin Liu, et al. Dapo: An open-source llm reinforcement learning system at scale. arXiv preprint arXiv:2503.14476, 2025.
  56. Yang Yue, Zhiqi Chen, Rui Lu, Andrew Zhao, Zhaokai Wang, Shiji Song, and Gao Huang. Does reinforcement learning really incentivize reasoning capacity in llms beyond the base model? arXiv preprint arXiv:2504.13837, 2025.
  57. Dingchu Zhang, Yida Zhao, Jialong Wu, Baixuan Li, Wenbiao Yin, Liwen Zhang, Yong Jiang, Yufeng Li, Kewei Tu, Pengjun Xie, and Fei Huang. Evolvesearch: An iterative self-evolving search agent, 2025a. URL https://arxiv.org/abs/2505.22501. ↗
  58. Shaokun Zhang, Yi Dong, Jieyu Zhang, Jan Kautz, Bryan Catanzaro, Andrew Tao, Qingyun Wu, Zhiding Yu, and Guilin Liu. Nemotron-research-tool-n1: Tool-using language models with reinforced reasoning. arXiv preprint arXiv:2505.00024, 2025b. — [arXiv:2505.00024](https://arxiv.org/abs/2505.00024) find
  59. Yuxiang Zheng, Dayuan Fu, Xiangkun Hu, Xiaojie Cai, Lyumanshan Ye, Pengrui Lu, and Pengfei Liu. Deepresearcher: Scaling deep research via reinforcement learning in real-world environments, 2025. URL https://arxiv.org/abs/2504.03160. ↗
  60. Peilin Zhou, Bruce Leon, Xiang Ying, Can Zhang, Yifan Shao, Qichen Ye, Dading Chong, Zhiling Jin, Chenxuan Xie, Meng Cao, et al. Browsecomp-zh: Benchmarking web browsing ability of large language models in chinese. arXiv preprint arXiv:2504.19314, 2025. — [arXiv:2504.19314](https://arxiv.org/abs/2504.19314) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/li-2025-websailor-navigating-super-human-reasoning/