SEAgent: Self-Evolving Computer Use Agent with Autonomous Learning from Experience

Overall framework

Abstract

SEAgent is an agentic self-evolving framework for computer-use agents. It enables agents to learn from interaction with unfamiliar software through curriculum generation, trajectory assessment, experiential learning, and specialist-to-generalist training.

Publication
arXiv preprint
Tong WU 吴桐
Tong WU 吴桐
Assistant Professor @ Fudan

My research interests include 3d vision, long-tailed recognition, and robustness.

Related