Video Understanding on MLVU 3-120min (test)
47.7AccuracyLLaVA-OneVision-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaVA-OneVision-7BBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=100%2026.02 | 47.7 | |
| DivPruneBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 46.4 | |
| CaCoVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 46.3 | |
| CaCoVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 46.2 | |
| PruneVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 45.7 | |
| Qwen2.5-VL-3BBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=100%2026.02 | 45.6 | |
| DivPruneBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 45.6 | |
| PruneVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 45.4 | |
| CaCoVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 45.3 | |
| CaCoVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 44.9 | |
| DivPruneBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 44.9 | |
| FrameFusionBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 44.8 | |
| CaCoVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 44.8 | |
| VisionZipBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 44.7 | |
| CaCoVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 44.6 | |
| PruneVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 44.3 | |
| FrameFusionBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 44.2 | |
| FrameFusionBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 44.2 | |
| CaCoVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=10%2026.02 | 44.2 | |
| DivPruneBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 44.1 | |
| DivPruneBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 43.9 | |
| DivPruneBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 43.9 | |
| VisionZipBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 43.8 | |
| VisionZipBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 43.8 | |
| PruneVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 43.7 | |
| VisionZipBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 43.5 | |
| VisionZipBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 43.5 | |
| FrameFusionBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 43.4 | |
| CaCoVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 43.2 | |
| FastVBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 43 | |
| PruneVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 43 | |
| PruneVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 42.7 | |
| VisionZipBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=10%2026.02 | 42.4 | |
| PruneVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 42.1 | |
| FastVBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 42 | |
| DivPruneBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 42 | |
| VisionZipBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 41.9 | |
| FrameFusionBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 41.8 | |
| FastVBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 41.7 | |
| FastVBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 41.7 | |
| VisionZipBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 41.5 | |
| FrameFusionBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 41.5 | |
| FastVBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 41.3 | |
| DivPruneBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=10%2026.02 | 40.8 | |
| PruneVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=10%2026.02 | 40.4 | |
| FrameFusionBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 39.8 | |
| FastVBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 37.3 | |
| FastVBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 35.2 | |
| FastVBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=10%2026.02 | 32 |