← Back to Glossary
RLHF / RLAIF
Reinforcement Learning from Human / AI Feedback
Training models using reward scores from human graders (RLHF) or AI evaluators (RLAIF) to ensure outputs are helpful, honest, and safe.
Why it exists
Supervised fine-tuning alone struggles with subjective quality and safety. Reinforcement learning aligns model outputs with human preference rankings.
How it works
Human or AI evaluators rank candidate outputs. A reward model is trained on these preferences, and algorithms like PPO or DPO optimize the LLM.
Related terms
Where you'll use this
roadmap