Tagged “training-data”Model collapse is the loss of the tails, and keeping the real data bounds itWeak labels elicit what a strong model already knows, so the student beats its teacherAll tags