Comment by vvrm

Another problem with the title: the article is about DPO, which doesn’t do reinforcement learning. So not RLHF. I guess RLHF has more of a name recognition than DPO.