WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918142296457216 |
|---|---|
| author | Li, Kuan Zhang, Zhongwang Yin, Huifeng Ye, Rui Zhao, Yida Zhang, Liwen Ou, Litu Zhang, Dingchu Wu, Xixi Wu, Jialong Wang, Xinyu Qiao, Zile Zhang, Zhen Jiang, Yong Xie, Pengjun Huang, Fei Zhou, Jingren |
| author_facet | Li, Kuan Zhang, Zhongwang Yin, Huifeng Ye, Rui Zhao, Yida Zhang, Liwen Ou, Litu Zhang, Dingchu Wu, Xixi Wu, Jialong Wang, Xinyu Qiao, Zile Zhang, Zhen Jiang, Yong Xie, Pengjun Huang, Fei Zhou, Jingren |
| contents | Transcending human cognitive limitations represents a critical frontier in LLM training. Proprietary agentic systems like DeepResearch have demonstrated superhuman capabilities on extremely complex information-seeking benchmarks such as BrowseComp, a feat previously unattainable. We posit that their success hinges on a sophisticated reasoning pattern absent in open-source models: the ability to systematically reduce extreme uncertainty when navigating vast information landscapes. Based on this insight, we introduce WebSailor, a complete post-training methodology designed to instill this crucial capability. Our approach involves generating novel, high-uncertainty tasks through structured sampling and information obfuscation, RFT cold start, and an efficient agentic RL training algorithm, Duplicating Sampling Policy Optimization (DUPO). With this integrated pipeline, WebSailor significantly outperforms all open-source agents in complex information-seeking tasks, matching proprietary agents' performance and closing the capability gap. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_13305 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learning Li, Kuan Zhang, Zhongwang Yin, Huifeng Ye, Rui Zhao, Yida Zhang, Liwen Ou, Litu Zhang, Dingchu Wu, Xixi Wu, Jialong Wang, Xinyu Qiao, Zile Zhang, Zhen Jiang, Yong Xie, Pengjun Huang, Fei Zhou, Jingren Machine Learning Computation and Language Transcending human cognitive limitations represents a critical frontier in LLM training. Proprietary agentic systems like DeepResearch have demonstrated superhuman capabilities on extremely complex information-seeking benchmarks such as BrowseComp, a feat previously unattainable. We posit that their success hinges on a sophisticated reasoning pattern absent in open-source models: the ability to systematically reduce extreme uncertainty when navigating vast information landscapes. Based on this insight, we introduce WebSailor, a complete post-training methodology designed to instill this crucial capability. Our approach involves generating novel, high-uncertainty tasks through structured sampling and information obfuscation, RFT cold start, and an efficient agentic RL training algorithm, Duplicating Sampling Policy Optimization (DUPO). With this integrated pipeline, WebSailor significantly outperforms all open-source agents in complex information-seeking tasks, matching proprietary agents' performance and closing the capability gap. |
| title | WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learning |
| topic | Machine Learning Computation and Language |
| url | https://arxiv.org/abs/2509.13305 |