thunlp/KernelGAT

Feverous vs Fever

Opened this issue · 0 comments

Hi, thanks for this great paper.

I'm part of an economics lab at MIT researching AI-human collaboration, where fact-checking is our application of interest. We are interested in using your KGAT fact-checking algorithm. We are interested in using Feverous instead of Fever, since Fever is generally too easy for participants.

We understand that performance won't be as good on Feverous (particularly with NEI claims), but do you think this is feasible / will we get reasonably good output?

Thanks very much for your time!