Reproduce “CI-Repair-Bench: A Repository-Aware Benchmark for Automated Patch Validation via CI Workflows” 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
Rabeya Khatun Muna, Md Nakhla Rafi, Tse-Hsun, Chen
CI-Repair-Bench evaluates automated repair by applying patches to failing commits and rerunning a standardized target workflow, requiring every job to pass. It contains 567 failures from 103 Python repositories, labeled across 12 categories. In its reference pipeline, GPT-5-mini achieves 18.9% Pass@1, with formatting easier than dependency, configuration, and environment failures.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.