Reinforcement Learning from Human Feedback (RLHF)

A training method used for finetuning LLMs like GPT, where human evaluators assess and score model outputs to train it to align with human preferences.