List your company

Hallucination Detection on PopQA GPT outputs (test)

0.802AUROCMean Token Neg.
0.561760.624130.68650.74887Jun 1, 2026

Evaluation results

MethodLinks
Mean Token Neg.Scorer Category=WB2026.060.802
P(True)Scorer Category=WB2026.060.795
Prob. MarginScorer Category=WB2026.060.784
Cosine Sim.Scorer Category=BB2026.060.774
Sequence Prob.Scorer Category=WB2026.060.754
Llama judgeScorer Category=Judge2026.060.744
Sonnet judgeScorer Category=Judge2026.060.714
Non-Cont.Scorer Category=BB2026.060.684
Min Token Prob.Scorer Category=WB2026.060.679
GPT-4o judgeScorer Category=Judge2026.060.643
EntailmentScorer Category=BB2026.060.607
Sem. Sets Conf.Scorer Category=BB2026.060.602
Semantic Neg.Scorer Category=BB2026.060.589
Exact MatchScorer Category=BB2026.060.575
Gemini judgeScorer Category=Judge2026.060.571