Question about prompt (especially about reward model prompt)

[tunantu]

Hi authors, when I tested the reward model you given, I found the reward scores are always below 0. May I get your reward prompt when you trained this model and applied it?

Also, does the prompt in LLaVA you used is identical to the original LLaVA prompt?

Thanks.