rlhf

Reinforcement learning from human feedback pipelines, reward modeling, policy optimization, and alignment outcomes.