Disclaimer, I haven't looked at their work to closely.
I think the process they use to select the hyperparameters overfits on the labels in the validation set. The true size of the labelled data includes the 100 from the training set and all labels in the validation set.
I think the process they use to select the hyperparameters overfits on the labels in the validation set. The true size of the labelled data includes the 100 from the training set and all labels in the validation set.