I Grpo Self Feedback Reasoning

Improve LLM reasoning through iterative refinement where the model refines its best previous attempts. Two-stage training: exploratory draft generation, then conditioned refinement using GRPO. Dynamic conditioning signals evolve with policy, enabling state-of-the-art math reasoning on AIME (85.62%).

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit f331c2374f

Frequently asked questions

npx get-skill add adu2021/i-grpo-self-feedback-reasoning