Evaluation on booster dataset

[alexxx-yyer]

How do you evaluate on the Booster dataset?
As far as I know, ground truth is only available for the training split.
Do you evaluate on the training set, or is there another evaluation protocol?

[Windsrain]

Yes, we evaluate it on the training set, because our work aims at zero-shot.