← AI tech topics
Hallucination & confabulation
Hallucination is the failure mode that made LLM testing a discipline: the model states something false with the same fluency and confidence as something true. It is not a bug to be patched but a consequence of how generative models work — they optimize for plausible continuations, not verified facts. Mitigations (RAG, citations, abstention training) reduce the rate without eliminating it, so testing has to measure it: the linked findings document fabricated citations, invented APIs and confident wrong answers across vendors.
Findings (5)
- Confusion about knowledge cutoff and current dateHallucinationLow
- Fabricated citations and referencesHallucinationHigh
- Fabrication instead of admitting uncertaintyHallucinationHigh
- Poor uncertainty calibration / overconfidenceHallucinationMedium
- Reasoning model knowingly fabricates unverifiable referencesHallucinationHigh
Methods
References
- Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs — arXiv
- Dated Data: Tracing Knowledge Cutoffs in Large Language Models — arXiv
- Gorilla: Large Language Model Connected with Massive APIs — arXiv
- Incident 541: ChatGPT Produced False Court Case Law Presented by Legal Counsel in Court — AI Incident Database
- Self-contradictory Hallucinations of Large Language Models: Evaluation, Detection and Mitigation — arXiv
- Survey of Hallucination in Natural Language Generation — ACM Computing Surveys
- TruthfulQA: Measuring How Models Mimic Human Falsehoods — arXiv
Cite this
Qlarify Labs. (2026). Hallucination & confabulation. Retrieved from https://labs.qlarify.fi/topics/hallucination