Skill∞ encodes the skill optimizer itself as a skill.
Skill optimization improves itself to beat handcrafted agent skills
Tao et al. show that a meta agent which rewrites its own optimization procedure lifts pass rates on 83 agentic tasks, with the main gain coming from the self-improvement loop.
Top University
Wangcheng Tao · Zhanglu Yan · Han Wu · Weng-Fai Wong
National University of Singapore · City University of Hong Kong
Research Digest··2 min read
Tao et al. present Skill∞, a framework that makes skill optimization recursively self-improving: a meta agent follows a meta-skill to optimize task skills, and the same procedure optimizes the meta-skill itself.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.