experiential-reinforcement-learning

【效率与协作】Improve RL training efficiency by embedding explicit reflection and revision loops. Models generate initial responses, receive feedback, produce self-reflections describing improvements, revise their attempts, and distill successful corrections into the base policy. Achieves up to 81% improvement on complex tasks through structured behavioral change.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准