Hallucination Detection on FAVA-Annotations (test)
68.19AUROCLLM-Check Attn Score
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LLM-Check Attn ScoreAccess level=White Box2026.02 | 68.19 | 70.53 | — | |
| HALT-LAccess level=Black Box2026.02 | 61.3 | 77.86 | 61.57 | |
| LLM-Check Hidden ScoreAccess level=White Box2026.02 | 57.1 | 65.38 | — | |
| H_overallAccess level=Aggregate Statistics2026.02 | 53.88 | 80 | 40.34 | |
| FAVA ModelAccess level=Black Box2026.02 | 53.29 | 79.9 | — | |
| Delta H_decAccess level=Aggregate Statistics2026.02 | 50.93 | 80 | 44.24 | |
| H_altsAccess level=Aggregate Statistics2026.02 | 48.89 | 80 | 44.01 | |
| PPLAccess level=Aggregate Statistics2026.02 | 46.12 | 80 | 41.77 | |
| LettuceAccess level=Black Box2026.02 | 45.77 | 80.67 | 40.33 |