Reproduce “CR-Bench-v2: Evaluating the Real-World Utility of AI Code Review 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
Kristen Pereira, Neelabh Sinha, Rajat Ghosh, Debojyoti Dutta
The paper introduces CR-Bench, which tests whether code-review agents detect known defects and whether their comments are useful. It reconstructs pull requests linked to bugs and scores detection separately from comment usefulness. On the 174-instance verified subset of open-source Python repositories, Gemini Code Assist averages 8.1% KDDR and 78.0% Usefulness Rate; each instance has one labeled defect.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.