is there any info constitutional_evaluation function??

[yeongsang2]

3.3 Training Loop

[chunhualiao]

I guess evaluate_response_for_RL() could be used as a workaround. This function tries to use OpenAI's text-davinci-003 to return a reward number.

[yeongsang2]

I guess evaluate_response_for_RL() could be used as a workaround. This function tries to use OpenAI's text-davinci-003 to return a reward number.

I think so too, thanks for the reply.