Research map / Large language models / Agents, code and math
miniF2F-Dafny: LLM-Guided Mathematical Theorem Proving via Auto-Active Verification
Mantas Baksys, Stefan Zetzsche, Olivier Bouissou, Sean B. Holden
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add LLM-generated proof hints (Claude Opus 4.6, pass@4) to Dafny automation | Dafny Verifier · miniF2F-Dafny test | solved (%): 38.9 -> 62.7 | Improved |
| add LLM-generated proof hints (Claude Sonnet 4.5) | Dafny Verifier · miniF2F-Dafny test | solved (%): 38.9 -> 55.7 | Improved |
| add LLM hints from Qwen 3 / GPT-OSS models | Dafny Verifier · miniF2F-Dafny test | pass@4 (%): 38.9 -> 43.4 (GPT-OSS 20B) | Mixed |
| increase attempts from pass@1 to pass@4 with error correction | Claude Opus 4.6 · miniF2F-Dafny test | solved (%): 58.2 -> 62.7 | Improved |
| increase attempts from pass@1 to pass@4 with error correction | Claude Haiku 4.5 · miniF2F-Dafny test | solved (%): 42.2 -> 43.9 | Improved |
| use Dafny automation instead of Lean grind | empty-proof baseline · miniF2F test | solved (%): 32.4 -> 38.9 | Improved |
| use Claude family instead of Qwen/GPT-OSS models | LLM proof hint generator · miniF2F-Dafny test | pass@4 (%): 43.4 -> 62.7 | Improved |
Closest papers by meaning
- APOLLO: Automated LLM and Lean Collaboration for Advanced Formal Reasoning (NeurIPS 2025 · similarity 0.91)
- miniF2F-Lean Revisited: Reviewing Limitations and Charting a Path Forward (NeurIPS 2025 · similarity 0.90)
- ProofAug: Efficient Neural Theorem Proving via Fine-grained Proof Structure Analysis (ICML 2025 · similarity 0.90)
- HERMES: Towards Efficient and Verifiable Mathematical Reasoning in LLMs (ICML 2026 · similarity 0.90)
- LEGO-Prover: Neural Theorem Proving with Growing Libraries (ICLR 2024 · similarity 0.89)
- STP: Self-play LLM Theorem Provers with Iterative Conjecturing and Proving (ICML 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
