AI glossary

What is RLHF?

Reinforcement learning from human feedback: training a model toward answers people prefer, using human rankings of outputs.

RLHF and related methods such as DPO (direct preference optimization) are why chat models are helpful and follow instructions. Some approaches use AI feedback guided by written principles.

▶Practice it hands-onHow LLMs Are Built (and Why They Hallucinate)

Related terms

Pretraining Fine-Tuning

← All 43 AI terms