Hi,
I just created a GPT and I am not sure how to test the responses whether they are consistent for every user. In addition, is there a way to see where the conversation is failing or what’s the quality of overall conversation (like how many thumbs up or down)
thanks.
2 Likes
Hi,
As far as I know, the only visible metric is the number of chats opened. I fully agree that this is a significant limitation when it comes to improving our models. I hope OpenAI provides more visibility regarding the performance of GPTs if they want to quickly develop useful tools based on GPT-4.
1 Like
One option is adding analytics to your GPT via 3rd party platform which can log user messages and your GPT response which you can analyze further
Seems relevant to what you’re asking for