SkillWeaver: Web Agents can Self-Improve by Discovering and Honing Skills
This addresses the challenge of procedural knowledge abstraction and skill refinement for web agents, with incremental advancements in skill composition and transfer.
The paper tackles the problem of autonomous web agents lacking self-improvement capabilities by introducing SkillWeaver, a framework that enables agents to autonomously discover, practice, and distill skills into reusable APIs, resulting in relative success rate improvements of 31.8% on WebArena and 39.8% on real-world websites.
To survive and thrive in complex environments, humans have evolved sophisticated self-improvement mechanisms through environment exploration, hierarchical abstraction of experiences into reuseable skills, and collaborative construction of an ever-growing skill repertoire. Despite recent advancements, autonomous web agents still lack crucial self-improvement capabilities, struggling with procedural knowledge abstraction, refining skills, and skill composition. In this work, we introduce SkillWeaver, a skill-centric framework enabling agents to self-improve by autonomously synthesizing reusable skills as APIs. Given a new website, the agent autonomously discovers skills, executes them for practice, and distills practice experiences into robust APIs. Iterative exploration continually expands a library of lightweight, plug-and-play APIs, significantly enhancing the agent's capabilities. Experiments on WebArena and real-world websites demonstrate the efficacy of SkillWeaver, achieving relative success rate improvements of 31.8% and 39.8%, respectively. Additionally, APIs synthesized by strong agents substantially enhance weaker agents through transferable skills, yielding improvements of up to 54.3% on WebArena. These results demonstrate the effectiveness of honing diverse website interactions into APIs, which can be seamlessly shared among various web agents.