Hallucination
Model output that appears confident but is factually incorrect or unsupported.
Why it matters
Side-by-side testing helps spot hallucination patterns for your domain.
Example
A model inventing a citation—compare whether another model flags uncertainty.