#RLHF

Також: reinforcement learning from human feedback, dpo, grpo

Reinforcement learning from human feedback - aligning model outputs to human preferences. Covers preference data collection, reward models, DPO and related methods, and alignment evaluation.