Noise-corrected GRPO: From Noisy Rewards to Unbiased Gradients

Omar El Mansouri, Fathinah Asma Izzati, Mohamed El Amine Seddik +1

Reinforcement learning from human feedback (RLHF) or verifiable rewards (RLVR), the standard paradigm for aligning LLMs or building recent SOTA reasoning models, is highly sensitiv…