[tulip4attoo]

I don't see the reward model that they are using. I think that is an important factor when doing rlhf.

[aldarisbm]

What does this impact? how aligned to the base model does the output model stays?