RLHF vs SFT

Supervised Fine-Tuning (SFT) teaches a model via labeled examples, while Reinforcement Learning from Human Feedback (RLHF) refines it using human preferences for better alignment.