Reproduce “Bug Finding With Agentic Symbolic Execution” 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
George Granberry, Wolfgang Ahrendt, Andy Gordon, Moa Johansson
The authors build an AI-agent pipeline that supplies contracts and a database-aware environment so CrossHair can symbolically test Python web applications, then validates candidate bugs with concrete tests against the unmodified app. Across 5,406 contracted functions in 11 applications, it confirmed 1,050 bugs; the inspection baseline found more overall, while missing 312 pipeline findings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.