TY  - RPRT
TI  - MDPO: Multi-Granularity Direct Preference Optimization for Mathematical Reasoning
AU  - Yunze Lin
PY  - 2025
UR  - https://arxiv.org/abs/2506.15706
ID  - 2506.15706
ER  -