LLM = Large *Language* Model.

They are designed to do language prediction, not logical deduction.

Attempting to extract truth or accuracy or facts or intelligence from a probabilistic algorithm is an exercise in frustration.