Picture for Hongming Wu

Hongming Wu

M-GRPO: Stabilizing Self-Supervised Reinforcement Learning for Large Language Models with Momentum-Anchored Policy Optimization

Add code
Dec 15, 2025
Viaarxiv icon