Am I the only one to observe a drop in performance for gpt-4-turbo-2024-04-09 over the last days?
Use ase: complex Prompt, processing dialogue transcripts and counting speech acts with output in Json format {‘Speech-act1’: n_occurrences, ‘Speech-act1’: n_occurrences … }
Few days back: used to work ok
Today:
- unable to count occurrences, all counts return 0
- shift back to gpt-3.5-turbo and you get some output, although not the exact figures produced by gpt-4-turbo a couple of days ago