Phase 2 - should we predict both test sets or just unseen?

Hateful Memes
1

If I understood correctly, in Phase 2 we should submit predictions only for the unseen test dataset? Or do we have to submit all ids in the test set (including the ones in the old test set)?

Sorry it’s a dumb question just could not find the info and wanted to be sure.

2

Hi @hvrigazov - That’s correct, in phase 2 you should submit predictions for just the unseen test set. Phase 1 is also continuing to run, so for that one you should submit predictions for just the seen test set. Hope that helps to clarify!