RLHF (Reinforcement Learning from Human Feedback)

A training technique where human raters score model outputs and those scores guide further training via reinforcement learning. RLHF aligns AI coding assistants with developer preferences and reduces harmful or incorrect suggestions.