I am trying to finetune a pretrained RoBERTa so it recognizes which of the three inputs is the right reason why the input sentence makes no sense, but I have trouble getting
