ModelRefs / TruthfulQA — AI Glossary
TruthfulQA — AI Glossary
A benchmark measuring whether models generate truthful answers on questions where humans commonly hold false beliefs.
Overview
TruthfulQA (Lin et al. 2021) contains 817 questions spanning health, law, finance, and conspiracy theories. Human rate for correct answers is 94%; GPT-3 scored 58%. Frontier models reach 85–90% but still fail on specific misconception clusters. Measures both truthfulness and informativeness (avoiding safe non-answers).
Reference details
| Topic | evaluation |
|---|---|
| Last reviewed | 2026-06-24 |
Related terms
Primary source
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to TruthfulQA — AI Glossary.
Frequently asked questions
What is TruthfulQA?
A benchmark measuring whether models generate truthful answers on questions where humans commonly hold false beliefs.
What concepts are related to TruthfulQA?
Closely related concepts include calibration, hallucination, hellaswag.