That will be very common in the corpus, thus pattern matching and retrieval more than 'reasoning'. Not that isn't useful, just expected behavior for an LLM the

That will be very common in the corpus, thus pattern matching and retrieval more than 'reasoning'. Not that isn't useful, just expected behavior for an LLM these days. I would be interested in the output, because it is almost certainly the more common by fundamentally incorrect intro class half truth, that is useful for the typical curriculum, but not near the frontier of current understanding.