I’m trying to call ChatCompletion (gpt-4-0125-preview) using an array of N objects as an input (+ instructions and prompt).
I expect gpt to return an array with N different responses, but the model is not consistent enough to always return those N responses. In fact, now I hardly ever get the expected amount of responses.
I built the call in the previous version (gpt-4-1106-preview), it has been used for more than a month and it has always worked as expected. I strongly believe that this is a bug, but I need the fixed encoding of the latest version. Is there something I can do?
Thanks in advance!
1 Like