RLHF

Reinforcement Learning from Human Feedback for training AI

“RLHF helped the model produce more helpful and harmless responses.”

Origin: Acronym combining reinforcement (Latin re- + fortis) + learning + human (Latin humanus) + feedback