RLHF (Reinforcement Learning from Human Feedback)

Concept

A training method for AI models where a reward model is trained from human feedback. It requires collecting a significant amount of human labels.

Mentioned in 1 video