I have found a few articles on the forum regarding this topic, but wanted to confirm absolutely. Using 3.5-turbo and comparing with web/chat 3.5 we are getting very different results. Changing the max token size and/or temperature hasn't helped and we've also tried davancii 003 and not chat as we are preforming summarization tasks on text data. Is it true that the web chat is "fine tuned" and vastly different from api results? Any suggestions? Thank you for confirming!
#API / Web Chat resulting in different results
granite raptor
halcyon helm
I have found a few articles on the forum regarding this topic, but wanted to con...
I think it's just a case of different settings, so different temperature, system message etc. Could be wrong though...