That's not how it works… They aren't able to literally regurgitate everything they've read, no matter how you prompt them. That would obviously violate the pige

That's not how it works… They aren't able to literally regurgitate everything they've read, no matter how you prompt them. That would obviously violate the pigeonhole principle. LLMs are, of course, a lossy compression format, and figuring out just how lossy the format is, and the degree of lossiness depends on the frequency of the given string in the training data. It's clearly worthwhile to investigate how exactly it depends.