experiential-reinforcement-learning

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.

Independent GitHub snapshot · Automated classification · Original documentation governs