Reinforcement Learning with Self-Feedback (RLSF)

A post-training method that uses the model's own internal confidence scores or self-generated critiques as reward signals, elimi