Research map / Large language models / Agents, code and math
Learning to Prove Theorems via Interacting with Proof Assistants
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Combine ASTactic with hammer | ASTactic · CoqGym test set | success rate: 24.8 -> 30.0 | Helped |
| Combine ASTactic with auto | ASTactic · CoqGym test set | success rate: 12.2 -> 12.8 | Helped |
| Use ASTactic instead of Coq's built-in easy tactic | Coq theorem proving · CoqGym test set | success rate: 4.9 -> 12.2 | Helped |
| Increase beam width from 1 to 20 (ablation) | ASTactic · CoqGym test set | success rate: 1.0 -> 12.2 | Helped |
| Increase beam width from 1 to 20 (ablation) | ASTactic · CoqGym test set | average runtime: 0.2 -> 3.3 | Hurt |
| Increase beam width from 5 to 20 (ablation) | ASTactic · CoqGym test set | success rate: 6.5 -> 12.2 | Helped |
| Increase beam width from 10 to 20 (ablation) | ASTactic · CoqGym test set | success rate: 10.8 -> 12.2 | Helped |
| Increase beam width from 15 to 20 (ablation) | ASTactic · CoqGym test set | success rate: 12.0 -> 12.2 | Helped |
| Increase beam width from 20 to 25 (ablation) | ASTactic · CoqGym test set | success rate: 12.2 -> 11.7 | Hurt |
| Increase beam width from 20 to 25 (ablation) | ASTactic · CoqGym test set | average runtime: 3.3 -> 3.9 | Hurt |
Closest papers by meaning
- Lean-STaR: Learning to Interleave Thinking and Proving (ICLR 2025 · similarity 0.90)
- ProofAug: Efficient Neural Theorem Proving via Fine-grained Proof Structure Analysis (ICML 2025 · similarity 0.89)
- Reviving DSP for Advanced Theorem Proving in the Era of Reasoning Models (NeurIPS 2025 · similarity 0.89)
- REFACTOR: Learning to Extract Theorems from Proofs (ICLR 2024 · similarity 0.89)
- ImProver: Agent-Based Automated Proof Optimization (ICLR 2025 · similarity 0.88)
- DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search (ICLR 2025 · similarity 0.88)
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.