RLHF

Reinforcement Learning from Human Feedback - A training technique where models learn from human preferences to become more helpful and safe. Key to modern LLM alignment.