Has anyone noticed a change in the (legacy) GPT4 chat completions quality since release of 4 Turbo?
Prior to last week, we had ChatGPT returning a topic / skill at the start of each response, to enable us to collect stats and do some formatting niceties.
We’re not using function calling yet for reasons to dull to go into yet.
Prior (correct) response:
##GeneralKnowledge
You were looking for blah blah
New (incorrect) response:
GeneralKnowledge
You were looking for blah blah