Reproduce “AutoCodeSherpa: Symbolic Explanations in AI Coding 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
Sungmin Kang, Haifeng Ruan, Abhik Roychoudhury
AutoCodeSherpa generates executable explanations of bugs to help developers inspect patches. It combines property-based tests describing triggering inputs and visible failures with predicates checking internal program states. In evaluation on Python issues from SWE-bench Verified, 318 of 433 tests had correct input and output conditions, and its patch classifier rejected 105 incorrect patches on a shared 396-issue set; explanations are evidence, not proof.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.