Paper 2509.13305
WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learning
- Published
- Sep 2025
- Research lab
- Independent
- Citations
- 60
- GitHub
- Not linked
01 In brief
Summary
WebSailor-V2 is a post-training pipeline for open-source web agents, comprising data construction (SailorFog-QA-V2), SFT, and RL.
The dataset uses a dense knowledge graph with cyclic structures and diverse uncertainty types beyond obfuscation.
Training employs a dual-environment RL framework: a high-fidelity simulator for rapid iteration and a managed real-world environment for stable policy training, integrated in a data-policy feedback loop.
Built on Qwen3-30B-A3B, WebSailor-V2 achieves state-of-the-art results among open-source agents, scoring 35.3 on BrowseComp-EN, 44.1 on BrowseComp-ZH, 30.6 on HLE, and 48.9 on DeepResearch Bench, outperforming larger models like DeepSeek-V3.1 (671B) and rivaling proprietary systems.
The authors emphasize that data quality and environment stability are more critical than the RL algorithm itself, and that SFT cold-start is essential for effective RL.
The agent uses a simple ReAct framework with tools for search, visit, Google Scholar, and Python interpreter.
02 From the paper
Abstract
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.