Hallucination Detection on PopQA Gemini outputs (test)
78.7AUROCCosine Sim.
Evaluation Results
| Method | Links | |
|---|---|---|
| Cosine Sim.Scorer Category=BB2026.06 | 78.7 | |
| Non-Cont.Scorer Category=BB2026.06 | 77.1 | |
| Sonnet judgeScorer Category=Judge2026.06 | 72.8 | |
| Sem. Sets Conf.Scorer Category=BB2026.06 | 72.2 | |
| Semantic Neg.Scorer Category=BB2026.06 | 71.8 | |
| EntailmentScorer Category=BB2026.06 | 71 | |
| GPT-4o judgeScorer Category=Judge2026.06 | 70.2 | |
| P(True)Scorer Category=WB2026.06 | 68.6 | |
| Mean Token Neg.Scorer Category=WB2026.06 | 66.8 | |
| Sequence Prob.Scorer Category=WB2026.06 | 66 | |
| Llama judgeScorer Category=Judge2026.06 | 64.2 | |
| Prob. MarginScorer Category=WB2026.06 | 63.9 | |
| Min Token Prob.Scorer Category=WB2026.06 | 60.6 | |
| Gemini judgeScorer Category=Judge2026.06 | 53.8 | |
| Exact MatchScorer Category=BB2026.06 | 52.6 |