Video Understanding on LongVideoBench 1-60min
56.8AccuracyCaCoVID
Evaluation Results
| Method | Links | |
|---|---|---|
| CaCoVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 56.8 | |
| CaCoVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 56.5 | |
| LLaVA-OneVision-7BBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=100%2026.02 | 56.4 | |
| DivPruneBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 56.4 | |
| CaCoVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 56.2 | |
| PruneVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 56.1 | |
| VisionZipBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 56 | |
| FrameFusionBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 55.7 | |
| VisionZipBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 55.7 | |
| DivPruneBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 55.6 | |
| CaCoVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 55.2 | |
| FrameFusionBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 55.1 | |
| DivPruneBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 55.1 | |
| PruneVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 55.1 | |
| PruneVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 54.9 | |
| PruneVIDBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 54.4 | |
| Qwen2.5-VL-3BBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=100%2026.02 | 54.3 | |
| VisionZipBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 54 | |
| FastVBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=25%2026.02 | 53.5 | |
| DivPruneBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 53.4 | |
| DivPruneBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 53.3 | |
| CaCoVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 53.3 | |
| FastVBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=20%2026.02 | 53 | |
| PruneVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 53 | |
| FrameFusionBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 52.9 | |
| CaCoVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 52.8 | |
| FrameFusionBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 52.4 | |
| CaCoVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 52.4 | |
| DivPruneBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 52.3 | |
| PruneVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 52.1 | |
| PruneVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 52.1 | |
| VisionZipBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 52 | |
| VisionZipBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 51.9 | |
| VisionZipBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 51.7 | |
| DivPruneBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 51.5 | |
| VisionZipBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=10%2026.02 | 51.5 | |
| CaCoVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=10%2026.02 | 51.5 | |
| PruneVIDBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=10%2026.02 | 51.3 | |
| FrameFusionBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 50.7 | |
| DivPruneBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=10%2026.02 | 50.5 | |
| FrameFusionBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 49.9 | |
| FastVBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=15%2026.02 | 49.8 | |
| FastVBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=25%2026.02 | 49.5 | |
| FrameFusionBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 49.1 | |
| VisionZipBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 48.9 | |
| FastVBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=20%2026.02 | 48.2 | |
| FastVBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=15%2026.02 | 47.8 | |
| FastVBase Model=LLaVA-OneVision-7B, Equivalent Retention Ratio=10%2026.02 | 47.6 | |
| FastVBase Model=Qwen2.5-VL-3B, Equivalent Retention Ratio=10%2026.02 | 45 |