@emilymbender It's probably Baader-Meinhof, but I've heard "zero-shot" thrown around very casually lately (as though it's a given that out-of-class data will be handled appropriately). Has it come to mean "we are going to throw an LLM at it and claim it does a good job"?