Comment by muldvarp

I think the training data contains more than enough information for an LLM to learn what kind of proof is considered beautiful or elegant by humans.

Of course it also contains more than enough information to learn what kind of question is interesting to humans.