“Training a system with Reinforcement Learning with Human Feedback (RLHF) involves training a reward model, which is an objective function that indicates whether an answer is good or bad.”