AI Glossary

Hallucination

Also called confabulation

What does “Hallucination” mean?

Hallucination: When an AI model states something false or unsupported as if it were true, such as invented facts, quotes, citations or features. It happens because models generate plausible-sounding text rather than looking facts up. Grounding, retrieval and asking the model to admit uncertainty can reduce it, but checking important claims is still necessary.

Hallucination visual guide: When an AI model states something false or unsupported as if it were true, such as invented facts, quotes, citations or features. It happens because models generate plausible-sounding text rather than looking facts up. Grounding, retrieval and asking the model to admit uncertainty can reduce it, but checking important claims is still necessary.
Visual guide Open full size (new tab)

What is an example of “Hallucination”?

Lawyers have been sanctioned after filing briefs that cited court cases an AI tool had made up.

Keep learning

See also

GroundingTermRetrieval-augmented generationTermKnowledge cutoffTermSycophancyTerm

Same topic

Chain of thoughtTermIn-context learningTermInference costTermFrequency and presence penaltiesTermFew-shot promptingTerm