Video Question Answering on MM-AU
45.45Accuracy VNo Filtering
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| No FilteringModel=LLaVA-OneVision, Filtering Strategy=None2026.06 | 45.45 | 25.7 | -0.24 | 2,453 | |
| S-scoreModel=LLaVA-OneVision, Filtering Strategy=S-score (Shortcut-Aware Filtering)2026.06 | 37.8 | 10.38 | 7.42 | 836 | |
| LLM-InspectorModel=LLaVA-OneVision, Filtering Strategy=LLM-Inspector2026.06 | 31.79 | 11.47 | 0.32 | 1,252 | |
| B-CleanModel=LLaVA-OneVision, Filtering Strategy=B-Clean2026.06 | 23.54 | -5.02 | -19.91 | 701 |