Update README.md

README.md

@@ -9,4 +9,5 @@ metrics:
 - accuracy
 base_model:
 - Qwen/Qwen2.5-7B-Instruct
----
+---
+Through PPO using a reward model trained on merely 5% of the LeanStatement_RL dataset, our model achieved enhanced performance, reaching a **pass@1 compilation success rate** of **95.3%** (465/488) on MiniF2F and **76.0%** (284/374) on ProofNet benchmarks.