rank by recommended for tags saved in tags similar to paper search newest random recommended / saved use "tags", similar uses "paper id" tags one, several (a,b) or all paper id for "similar to paper" last N days empty = any time features auto (concat) tfidf emb concat fusion how papers are compared; auto is tuned model auto (logreg) svm logreg what learns from your papers; auto is tuned regularization C lower = broader matches hide papers I rated yes no only papers you haven't rated