Reproduce “TestSuiteBench: Coding Agents Struggle to Generate Strong Test Suites from Specifications” 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
Kazuki Egashira, Niels Mündler-Sasahara, Mark Niklas Müller, Martin Vechev
TestSuiteBench evaluates whether coding agents can generate repository-scale test suites from specifications alone. It runs generated tests against reference programs and requirement-targeted mutants, separating validity from mutation score. Across 165 command-line applications, Claude Opus 5 reaches 55.5% mutation score, 96.1% validity, and 91.8% reference pass rate; its valid tests kill just over half the mutants.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.