Hallucination Detection on SelfAware Gemini outputs (test)
52.8AUROCSonnet judge
Evaluation Results
| Method | Links | |
|---|---|---|
| Sonnet judgeScorer Category=Judge2026.06 | 52.8 | |
| GPT-4o judgeScorer Category=Judge2026.06 | 51.9 | |
| P(True)Scorer Category=WB2026.06 | 51 | |
| Gemini judgeScorer Category=Judge2026.06 | 50.2 | |
| Exact MatchScorer Category=BB2026.06 | 49.9 | |
| Prob. MarginScorer Category=WB2026.06 | 48.6 | |
| Mean Token Neg.Scorer Category=WB2026.06 | 46.9 | |
| Sequence Prob.Scorer Category=WB2026.06 | 46.9 | |
| Llama judgeScorer Category=Judge2026.06 | 44.5 | |
| Cosine Sim.Scorer Category=BB2026.06 | 44.4 | |
| Non-Cont.Scorer Category=BB2026.06 | 42.7 | |
| Semantic Neg.Scorer Category=BB2026.06 | 39.3 | |
| Min Token Prob.Scorer Category=WB2026.06 | 39 | |
| Sem. Sets Conf.Scorer Category=BB2026.06 | 38.7 | |
| EntailmentScorer Category=BB2026.06 | 36.9 |