用語

RLHF

定義は英語です。

Reinforcement learning from human feedback: a training step that uses people’s preferences between outputs.