Temperature setting

#4
by SdrstrmKR - opened

Thank you for the dataset and an interesting paper!

I was wondering about a quote in your paper:"When generating responses, we set the sampling temperature to 1." (p.4)
Would setting a lower temperature prevent some of the hallucinations? Where you able to try different temperature settings?

Thanks again!
/KR

Sign up or log in to comment