SafetyEst. 2021

TruthfulQA

TruthfulQA measures whether language models generate truthful answers to questions. It includes 817 questions spanning 38 categories where humans might give false answers due to misconceptions, superstitions, or conspiracy theories.

Metrics

Truthfulness (%) on 817 questions

Created By

Stephanie Lin et al.

Top Model Scores

RankModelScoreDate
1Claude Opus 4.682.4%2026-02
2GPT-5.280.1%2026-03
3Gemini 3 Ultra78.6%2026-01
4Grok 476.3%2026-02
5Llama 4 405B74.9%2026-01