Not OP but the screenshot in the article pretty much shows something that it's not at the source.
You'd like to get the "correct" answer straight away, not watch a discussion between two bots.
back
2 comments
Yes, but this is about helping the people who are training the model.
You are missing the point of the model in the first place. By having higher quality RLHF datasets, you get a higher quality final model. CriticGPT is not a product, but a tool to make GPT-4 and future models better.