Hallucination Detection on SelfAware GPT outputs (test)
0.528AUROCSonnet judge
Evaluation Results
| Method | Links | |
|---|---|---|
| Sonnet judgeScorer Category=Judge2026.06 | 0.528 | |
| Gemini judgeScorer Category=Judge2026.06 | 0.513 | |
| Exact MatchScorer Category=BB2026.06 | 0.499 | |
| Semantic Neg.Scorer Category=BB2026.06 | 0.495 | |
| Sem. Sets Conf.Scorer Category=BB2026.06 | 0.493 | |
| Prob. MarginScorer Category=WB2026.06 | 0.483 | |
| Mean Token Neg.Scorer Category=WB2026.06 | 0.474 | |
| Cosine Sim.Scorer Category=BB2026.06 | 0.471 | |
| Sequence Prob.Scorer Category=WB2026.06 | 0.47 | |
| GPT-4o judgeScorer Category=Judge2026.06 | 0.467 | |
| Min Token Prob.Scorer Category=WB2026.06 | 0.44 | |
| Llama judgeScorer Category=Judge2026.06 | 0.409 | |
| Non-Cont.Scorer Category=BB2026.06 | 0.401 | |
| EntailmentScorer Category=BB2026.06 | 0.355 | |
| P(True)Scorer Category=WB2026.06 | 0.331 |