Abstract
Neural theorem provers can be improved through supervised learning on proof repair data that includes compiler feedback and diagnostic explanations, enabling better failure interpretation and correction.
As neural theorem provers become increasingly agentic, the ability to interpret and act on compiler feedback is critical. However, existing Lean datasets consist almost exclusively of correct proofs, offering little supervision for understanding and repairing failures. We study Lean proof repair as a supervised learning problem: given an erroneous proof and compiler feedback, predict both a corrected proof and a natural-language diagnosis grounded in the same feedback. We introduce APRIL (Automated Proof Repair in Lean), a dataset of 260,000 supervised tuples pairing systematically generated proof failures with compiler diagnostics and aligned repair and explanation targets. Training language models on APRIL substantially improves repair accuracy and feedback-conditioned reasoning; in our single-shot repair evaluation setting, a finetuned 4B-parameter model outperforms the strongest open-source baseline. We view diagnostic-conditioned supervision as a complementary training signal for feedback-using provers. Our dataset is available at https://huggingface.co/datasets/uw-math-ai/APRIL{this link}.
Community
Existing Lean datasets contain correct proofs. Models learn error correction with RL, that's expensive. We release a dataset of 260k erroneous Lean proofs, the compiler feedback, error explanation, proof repair reasoning trace, and the corrected proof.
| Model | Baseline | Fine-tuned |
|---|---|---|
| Goedel-Prover-V2-32B | 26.8% | — |
| Goedel-Prover-V2-8B | 15.5% | 34.6% |
| Kimina-Prover-8B | 11.1% | 31.9% |
| Qwen3-4B-Instruct-2507 | 1.1% | 27.4% |
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- LeanTutor: Towards a Verified AI Mathematical Proof Tutor (2026)
- Structured Hints for Sample-Efficient Lean Theorem Proving (2026)
- Decompose-and-Formalise: Recursively Verifiable Natural Language Inference (2026)
- PhysProver: Advancing Automatic Theorem Proving for Physics (2026)
- Pushing the Boundaries of Natural Reasoning: Interleaved Bonus from Formal-Logic Verification (2026)
- Seed-Prover 1.5: Mastering Undergraduate-Level Theorem Proving via Learning from Experience (2025)
- CoReTab: Improving Multimodal Table Understanding with Code-driven Reasoning (2026)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:
@librarian-bot
recommend
Models citing this paper 0
No model linking this paper
Datasets citing this paper 1
Spaces citing this paper 0
No Space linking this paper
Collections including this paper 0
No Collection including this paper