Training & Learning

Reinforcement Learning from Human Feedback

RLHF

A training method that tunes a model to be helpful by rewarding the responses human reviewers prefer.

RLHF shapes a models behavior using human judgments rather than a fixed set of correct answers. People compare the models outputs and mark which they prefer, those preferences train a reward model, and the main model is then optimized to score well against it. This is a big part of why modern chatbots feel helpful and polite, since it aligns raw capability with what people actually want.

Reference: Wikipedia: Reinforcement Learning from Human Feedback