This roberta-base model classifies scientific review text and output score (decline/borderline/reject).
The model input should be QUESTION_TAG + " " + review_text, where QUESTION_TAG=="MANY_QUESTIONS" if the review_text contains more than one question (?).
