EffiSkill: Agent Skill Based Automated Code Efficiency Optimization
Published in Under review at EACL 2027, 2026
EffiSkill studies automated code-efficiency optimization with language-model agents. The project builds a reusable skill library from paired slow and optimized programs, then applies those skills to unseen code through execution-free diagnosis, retrieval, plan composition, and candidate generation.
On EffiBench-X, the system improves optimization success rate over the strongest baseline across multiple model and language settings. The current public version is available on arXiv.