Hallucination Detection on TriviaQA GPT outputs (test)
0.675AUROCP(True)
Evaluation Results
| Method | Links | |
|---|---|---|
| P(True)Scorer Category=WB2026.06 | 0.675 | |
| Non-Cont.Scorer Category=BB2026.06 | 0.653 | |
| Sonnet judgeScorer Category=Judge2026.06 | 0.651 | |
| EntailmentScorer Category=BB2026.06 | 0.64 | |
| Semantic Neg.Scorer Category=BB2026.06 | 0.639 | |
| Sem. Sets Conf.Scorer Category=BB2026.06 | 0.639 | |
| Min Token Prob.Scorer Category=WB2026.06 | 0.617 | |
| Sequence Prob.Scorer Category=WB2026.06 | 0.613 | |
| Mean Token Neg.Scorer Category=WB2026.06 | 0.607 | |
| Exact MatchScorer Category=BB2026.06 | 0.603 | |
| Cosine Sim.Scorer Category=BB2026.06 | 0.599 | |
| Prob. MarginScorer Category=WB2026.06 | 0.595 | |
| Llama judgeScorer Category=Judge2026.06 | 0.589 | |
| Gemini judgeScorer Category=Judge2026.06 | 0.582 | |
| GPT-4o judgeScorer Category=Judge2026.06 | 0.573 |