Reproduce “GEPA-Curator: Curated Learning for Self Improving Agents” 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
Zack Yu, Nicholas Lee, Minjae Lee, Lakshya Agarwal, Luca Manolache, Wonjun Kang, Matei Zaharia, Michael W. Mahoney and 3 more
GEPA-Curator improves prompt optimization by selecting task experiences for learning and retrieving a prior success when a task fails. It maintains task histories, skips tasks unlikely to provide useful signal, and gives the optimizer a same-task reference when available. In experiments, it scores 88.6% on GitLab validation; gains vary by setting and do not establish that every failure is noisy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.