Mind linking any examples (or categories) of problems that are definitively not in pre training data but can still be solved by LLMs? Preferably something factual rather than creative, genuinely curious.
Dumb question but anything like this that’s written about on the internet will ultimately end up as training fodder, no?
