Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Goedel-Prover-V2: Scaling Formal Theorem Proving with Scaffolded Data Synthesis and Self-Correction

About

We introduce Goedel-Prover-V2, a series of open-source language models that set a new state-of-the-art in automated theorem proving. Built on the standard expert iteration and reinforcement learning pipeline, our approach incorporates three key innovations: (1) Scaffolded data synthesis: We generate synthetic tasks of increasing difficulty to train the model to master increasingly complex theorems; (2) Verifier-guided self-correction: We enable the model to iteratively revise its proofs by leveraging feedback from the Lean compiler; (3) Model averaging: We merge model checkpoints to mitigate the decrease in model output diversity in later stages of training. Our small model, Goedel-Prover-V2-8B, reaches 84.6% pass@32 on MiniF2F and outperforms DeepSeek-Prover-V2-671B under the same metric, despite being 80X smaller. Our flagship model, Goedel-Prover-V2-32B, achieves 88.1% on MiniF2F at pass@32 in standard mode and 90.4% in self-correction mode, outperforming prior SOTA by a large margin. Additionally, our flagship model solves 86 problems on PutnamBench at pass@184, securing the first place among open-source models on the leaderboard, surpassing DeepSeek-Prover-V2-671B's record of solving 47 problems by pass@1024 with a significantly smaller model size and compute budget. At the time of its release (July-August 2025), Goedel-Prover-V2 achieves the strongest overall performance among all open-source theorem provers. It also ranks among the top-performing models--including closed-source systems with publicly reported performance--under a constrained test-time compute budget. Our models, code, and data are released at https://github.com/Goedel-LM/Goedel-Prover-V2.

Yong Lin, Shange Tang, Bohan Lyu, Ziran Yang, Jui-Hui Chung, Haoyu Zhao, Lai Jiang, Yihan Geng, Jiawei Ge, Jingruo Sun, Jiayun Wu, Jiri Gesi, Ximing Lu, David Acuna, Kaiyu Yang, Hongzhou Lin, Yejin Choi, Danqi Chen, Sanjeev Arora, Chi Jin• 2025

Related benchmarks

TaskDatasetResultRank
Formal Theorem ProvingMiniF2F (test)
Pass@192.6
128
Automated Theorem ProvingMiniF2F (test)
Success Rate92.2
93
Theorem ProvingPutnamBench Lean
Solved Rate86
23
Theorem ProvingPutnam-Bench
Pass@3213
21
Theorem ProvingProver-Bench
Pass@3253.2
19
Theorem ProvingMathOlympiad-Bench
Pass@3220.3
16
Theorem ProvingProofNet (test)
Pass@3222
15
Formal Theorem ProvingPutnamBench
Solve Rate13
14
Auto-formalizationFormalMath-Lite
Pass@898.1
13
Auto-formalizationMiniF2F (test)
Pass@898.4
13
Showing 10 of 26 rows

Other info

Follow for update