Reproduce “LeanLean: Benchmarking Repository-Scale Lean Proof Compression” 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
Kári Rögnvaldsson, Niels Mündler-Sasahara, Jasper Dekoninck, Martin Vechev
The authors introduce LeanLean, a benchmark testing whether agents can reduce Lean code across 64 repositories while preserving protected results. Agents simplify proofs and change declaration dependencies; scores average relative token reductions per repository, with failures receiving zero. Opus 5 achieves the highest score, 48.30%, but only 21% on the sixteen largest repositories; token savings do not establish improved maintainability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.