List your company

Factual Question Answering on TruthfulQA (AUROC, 95% CI)

59.9AUROCB1 entropy
42.01246.65651.355.944Mar 25, 2026

Evaluation results

MethodLinks
B1 entropyCost=Free2026.0359.955.9
SelfCheck (k=5)Cost=6×2026.0358.854.7
Semantic entropy (N=10)Cost=11×2026.0354.851
P(True)Cost=1 call2026.0342.740.8