WebSailor: Navigating Super-human Reasoning for Web Agent
This addresses the capability gap in open-source web agents for complex information-seeking tasks, representing a significant but incremental advance.
The paper tackles the problem of open-source LLMs lacking superhuman reasoning for complex web navigation by introducing WebSailor, a post-training methodology that matches proprietary agents' performance on benchmarks like BrowseComp.
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 opensource agents in complex information-seeking tasks, matching proprietary agents' performance and closing the capability gap.