VeriSoftBench / README.md
maxRyeery's picture
Update README.md
741c2a1 verified
metadata
license: mit

VeriSoftBench

VeriSoftBench is a benchmark for evaluating neural theorem provers on software verification tasks in Lean 4.

The dataset contains 500 theorem-proving tasks drawn from 23 real-world Lean 4 repositories spanning compiler verification, type system formalization, applied verification (zero-knowledge proofs, smart contracts), semantic frameworks, and more.

📄 Paper (arXiv): https://arxiv.org/html/2602.18307v1
💻 Full benchmark + pipeline + setup: https://github.com/utopia-group/VeriSoftBench


Dataset Contents

This Hugging Face release contains the dataset of the benchmark tasks only. For the full end-to-end evaluation pipeline, please refer to the Github repository:

👉 https://github.com/utopia-group/VeriSoftBench

Each task in verisoftbench.jsonl contains:

  • Theorem name, statement, and source location
  • Filtered dependencies (library defs, repo defs, local context, lemmas)
  • Ground truth proof
  • Metadata (category, difficulty metrics, Aristotle subset membership)

Citation

@misc{xin2026verisoftbenchrepositoryscaleformalverification,
      title={VeriSoftBench: Repository-Scale Formal Verification Benchmarks for Lean}, 
      author={Yutong Xin and Qiaochu Chen and Greg Durrett and Işil Dillig},
      year={2026},
      eprint={2602.18307},
      archivePrefix={arXiv},
      primaryClass={cs.SE},
      url={https://arxiv.org/abs/2602.18307}, 
}