Object Hallucination Detection on MSCOCO Qwen3-VL 3 (test)
86.89AUCOverthinking Score
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Overthinking ScoreClassifier=MLP2026.03 | 86.89 | 53.72 | 71.15 | |
| Overthinking ScoreClassifier=GB2026.03 | 86.65 | 61.69 | 74.43 | |
| MetaTokenClassifier=MLP2026.03 | 86.29 | 47 | 68.03 | |
| MetaTokenClassifier=GB2026.03 | 84.21 | 38.06 | 74.43 | |
| HalLoc2026.03 | 83.85 | 59.5 | 74.75 | |
| MetaTokenClassifier=LR2026.03 | 79.49 | 27.6 | 47.27 | |
| Overthinking ScoreClassifier=LR2026.03 | 77.85 | 37.31 | 55.9 | |
| SVAR2026.03 | 75.56 | 35.69 | 50.2 |