Reproduce “Skill++: Towards Lifelong Agents via Routing-Aware Skill Refinement” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Hanrong Zhang, Minghao Liang, Jiawei Wu, Yankai Chen, Shicheng Fan, Zhiwei Liu, Enze Ma, Junwei Yuan and 5 more

SKILL++ refines incoming skills’ names and descriptions against similar earlier skills, leaving the router and existing entries unchanged. In a growing SRA-Bench library of 731 skills, it improves complete selection of required skills among the router’s top five, measured as FC@5. At the final checkpoint, FC@5 reaches 76.3%, versus 36.4% with original metadata; this is a controlled routing result.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.