Comment by sandGorgon

but isnt that a direct contradiction of this particular paper anyways - that chatgpt anyways outperforms human annotation.

so permit me to act as devil's advocate to your statement - prove that (in context of this paper), your hypothesis is still correct.