Fetching the paper…

Step-Controlled DPO: Leveraging Stepwise Error for Enhanced Mathematical Reasoning · Around