Object-level hallucination evaluation on POPE A-OKVQA Adversarial SEEM
57.09AccuracyVideo-LLaVA + VisAlign
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Video-LLaVA + VisAlignenhancement=VisAlign2025.11 | 57.09 | 53.9 | 98.33 | 69.63 | |
| Video-LLaVA2025.11 | 54.1 | 52.14 | 99.6 | 68.45 |