An alternative framing if the article is that reproving of training examples is /rare/. It works for highly replicated images and, in larger models trained for

An alternative framing if the article is that reproving of training examples is /rare/. It works for highly replicated images and, in larger models trained for longer periods of time, really unusual training examples. Both of these failure modes have pretty obvious mitigations: deduplicate the training data, and trim the outliers.