CODESKILL: Learning Self-Evolving Skills for Coding Agents
CODESKILL is a proposed framework aimed at enhancing coding agents' abilities through self-evolving skills. It utilizes a learnable management policy to extract and maintain procedural skills from coding-agent trajectories. Experiments indicate that CODESKILL significantly improves task performance while keeping the skill bank size stable.
- ▪CODESKILL reformulates skill extraction and maintenance as a learnable management policy.
- ▪The framework improves the average pass rate by 9.69 over the no-skill baseline.
- ▪CODESKILL maintains a compact skill bank during iterative construction.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.25430 (cs) [Submitted on 25 May 2026] Title:CODESKILL: Learning Self-Evolving Skills for Coding Agents Authors:Yanzhou Li, Yiran Zhang, Xiaoyu Zhang, Xiaoxia Liu, Yang Liu View a PDF of the paper titled CODESKILL: Learning Self-Evolving Skills for Coding Agents, by Yanzhou Li and 4 other authors View PDF HTML (experimental) Abstract:Coding agents produce rich trajectories while solving software-engineering tasks. To enable agent self-evolution, these trajectories can be distilled into reusable procedural skills that compactly encode experience to guide future behavior.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
