RLHF

• RLHF, Reinforcement Learning by Human Feedback。找人提问题，并对模型反馈奖励、惩罚