Object Hallucination Detection on MSCOCO Gemma 3 (test)
85.59AUCOverthinking Score
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Overthinking ScoreClassifier=GB2026.03 | 85.59 | 59.89 | 74.54 | |
| Overthinking ScoreClassifier=MLP2026.03 | 85.38 | 56.84 | 72.07 | |
| HalLoc2026.03 | 79.27 | 49.96 | 67.11 | |
| MetaTokenClassifier=MLP2026.03 | 78.4 | 33.63 | 60.03 | |
| MetaTokenClassifier=GB2026.03 | 77.23 | 34.75 | 67.15 | |
| Overthinking ScoreClassifier=LR2026.03 | 76.68 | 31.2 | 67.53 | |
| SVAR2026.03 | 74.11 | 32.75 | 47.84 | |
| MetaTokenClassifier=LR2026.03 | 73.91 | 28.78 | 48.27 |