Comment by emp17344

Maybe LLMs always fail to generalize outside their data set, and it’s just less noticeable with written language.