File size: 839 Bytes
3438e5a
 
 
 
 
 
d56c73c
 
 
 
 
 
 
 
 
3438e5a
 
aef8a47
1ca83c1
aef8a47
 
 
98430bc
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
---
license: apache-2.0
datasets:
- SJTULean/LeanStatement_SFT
language:
- en
result:
  dataset:
    type: Formalizer Lean
    name: miniF2F
  metrics:
    - name: pass@1
      type: pass@1
      value: 0.9406
      verified: false
base_model:
- Qwen/Qwen2.5-7B-Instruct
---
# LeanFormalizer_SFT

<!-- Provide a quick summary of what the model is/does. -->

Based on Qwen2.5-7b and trained on our LeanStatement_SFT dataset, our model achieves **state-of-the-art** performance in formal mathematics verification, with a **pass@1 compilation success rate** of 94.1% (459/488) on MiniF2F and 75.4% (282/374) on ProofNet benchmarks as of December 2024(while our PPO model achieves 95.3% on MiniF2F and 76.0% on Proofnet). These results demonstrate the effectiveness of scaling laws in mathematical formalization using the Lean theorem prover.