Concepts
RLHF
An approach that uses human feedback to guide reinforcement learning, often by learning a reward signal from comparisons and optimizing behavior against it. The feedback reflects a particular task, rubric and group of raters. A higher learned reward does not guarantee truthfulness, safety or agreement with every user.
Also called: Reinforcement Learning from Human Feedback
Reviewed
Sources
Member lesson
The definition and sources are public. The complete practical lesson is for members.