Why not GPT-4
#9
by
Subcode
- opened
Seeing how this is trained off models that are already not performing optimally / as good as GPT-4 the results seem to not be that great.
This should honestly not be trained off models that already perform worse than GPT-4 / score low on benchmarks.