Glossary
What is RLHF?
RLHF, reinforcement learning from human feedback, is a training step in which people rate an AI model’s answers and the model learns to prefer the better-rated ones. Reviewers compare several replies to the same question and rank them. It is a large part of what makes a model polite and helpful, and what makes it refuse some requests.
Why it matters for a business
Section titled “Why it matters for a business”RLHF sets a model’s tone, how much it hedges, and what it refuses to do. It is a large part of why models from different labs feel different despite similar ability: they were rated by different people against different instructions. It also explains a trap. The hedging a model learned tracks topics the reviewers wanted handled carefully, such as medicine, law and money, not whether a particular answer is right, so a confident tone is no evidence of accuracy.