RLHF

Reinforcement learning from human feedback. The training stage that turns a base model into a helpful assistant.