nostalgebraist comments on OpenAI API base models are not sycophantic, at any size

[nostalgebraist]

It’s possible that the “0 steps RLHF” model is the “Initial Policy” here with HHH prompt context distillation

I wondered about that when I read the original paper, and asked Ethan Perez about it here. He responded:

Good question, there’s no context distillation used in the paper (and none before RLHF)

[janus]

Thanks. That’s pretty odd, then.