Video Understanding on MVBench (Latency and Efficiency)
67.5MVBench ScoreOuro-Spatial-8B
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Ouro-Spatial-8BModel scale=8B, Number of uniformly sampled frames=322026.06 | 67.5 | — | — | — | — | — | — | — | |
| Qwen3-VL-8BModel scale=8B, Number of uniformly sampled frames=322026.06 | 66.5 | — | — | — | — | — | — | — | |
| Ouro-Spatial-4BModel scale=4B, Number of uniformly sampled frames=322026.06 | 66.4 | — | — | — | — | — | — | — | |
| Qwen3-VL-4BModel scale=4B, Number of uniformly sampled frames=322026.06 | 65.8 | — | — | — | — | — | — | — | |
| FastVIDRetention Ratio R=15%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 60.5 | 11 | 11.7 | — | — | 61.3 | — | 98.5 | |
| LLaVA-Video-7BRetention Ratio R=100%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 60.4 | 94.1 | 100 | — | — | 62.3 | — | 100 | |
| InfoMergeRetention Ratio R=10%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 60 | 6.9 | 7.3 | — | — | 60.5 | — | 97.1 | |
| InfoMergeRetention Ratio R=15%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 59.9 | 11 | 11.7 | — | — | 61.4 | — | 98.6 | |
| HoliTomBefore LLM Retained Ratio=20%2026.05 | 58.7 | 47.5 | 57.5 | 622.3 | 30 | 58.8 | 100.7 | — | |
| FastVIDRetention Ratio R=10%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 58.5 | 6.9 | 7.3 | — | — | 59.4 | — | 95.3 | |
| HoliTomBefore LLM Retained Ratio=25%2026.05 | 58.4 | 49 | 59.3 | 661.3 | 29.9 | 58.8 | 100.7 | — | |
| LLaVA-OV-7BBefore LLM Retained Ratio=100%2026.05 | 58.3 | 82.6 | 100 | 889.9 | 24.4 | 58.4 | 100 | — | |
| HoliTomBefore LLM Retained Ratio=15%2026.05 | 58.1 | 46 | 55.7 | 572.7 | 27.5 | 58.2 | 99.7 | — | |
| VisionZipBefore LLM Retained Ratio=25%2026.05 | 57.9 | 50.5 | 61.1 | 516.6 | 29.4 | 58.2 | 99.7 | — | |
| EarlyTomBefore LLM Retained Ratio=20%2026.05 | 57.8 | 35.1 | 42.4 | 415.3 | 33.4 | 58.1 | 99.3 | — | |
| VisionZipBefore LLM Retained Ratio=20%2026.05 | 57.7 | 48.7 | 58.9 | 495 | 29.8 | 57.7 | 98.8 | — | |
| EarlyTomBefore LLM Retained Ratio=15%2026.05 | 57.5 | 33.6 | 40.7 | 390.6 | 30.4 | 57.3 | 98.1 | — | |
| PruneVidBefore LLM Retained Ratio=25%2026.05 | 57.4 | 50.5 | 61.1 | 703.6 | 29.4 | 57.6 | 98.6 | — | |
| EarlyTomBefore LLM Retained Ratio=25%2026.05 | 57.4 | 36.5 | 44.2 | 426.3 | 32.9 | 58.2 | 99.7 | — | |
| HoliTomBefore LLM Retained Ratio=10%2026.05 | 57.3 | 44.6 | 54 | 556.6 | 29 | 57.9 | 99.1 | — | |
| PruneVidBefore LLM Retained Ratio=20%2026.05 | 57.2 | 49 | 59.3 | 662.1 | 29.5 | 57.1 | 97.8 | — | |
| PruneVidBefore LLM Retained Ratio=15%2026.05 | 56.8 | 47.5 | 57.5 | 574.1 | 27.1 | 57.1 | 97.8 | — | |
| FastVIDBefore LLM Retained Ratio=25%2026.05 | 56.5 | 50.5 | 61.1 | 581.6 | 26.9 | 57.3 | 98.1 | — | |
| VisionZipBefore LLM Retained Ratio=15%2026.05 | 56.5 | 46.9 | 56.8 | 475.9 | 32.1 | 56.7 | 97.1 | — | |
| EarlyTomBefore LLM Retained Ratio=10%2026.05 | 56.5 | 32.2 | 39 | 336.2 | 31.6 | 56.2 | 96.2 | — | |
| FastVIDBefore LLM Retained Ratio=20%2026.05 | 56.3 | 48.7 | 58.9 | 546.6 | 27.6 | 57.3 | 98.1 | — | |
| PruneVidBefore LLM Retained Ratio=10%2026.05 | 56.2 | 45.9 | 55.6 | 592.2 | 28.6 | 56.6 | 96.9 | — | |
| PyramidDropBefore LLM Retained Ratio=100%2026.05 | 56.1 | 51.8 | 62.7 | 813.4 | 28.3 | 56.2 | 96.2 | — | |
| MMG-VidRetention Ratio R=15%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 56.1 | 11 | 11.7 | — | — | 58.6 | — | 94.1 | |
| FastVIDBefore LLM Retained Ratio=15%2026.05 | 56 | 46.9 | 56.8 | 530.8 | 28.7 | 56.8 | 97.3 | — | |
| FastVBefore LLM Retained Ratio=100%2026.05 | 55.9 | 51.1 | 61.9 | 820 | 28.4 | 56.5 | 96.7 | — | |
| FastVIDBefore LLM Retained Ratio=10%2026.05 | 55.9 | 45.2 | 54.7 | 502.1 | 28.3 | 56.5 | 96.7 | — | |
| VisionZipRetention Ratio R=15%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 55.2 | 11 | 11.7 | — | — | 57.8 | — | 92.8 | |
| MMG-VidRetention Ratio R=10%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 54.9 | 6.9 | 7.3 | — | — | 57.2 | — | 91.8 | |
| VisionZipRetention Ratio R=10%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 53.8 | 6.9 | 7.3 | — | — | 56.3 | — | 90.4 | |
| VisionZipBefore LLM Retained Ratio=10%2026.05 | 53.5 | 45.2 | 54.7 | 458.5 | 28.5 | 53.5 | 91.6 | — | |
| DyCokeBefore LLM Retained Ratio=25%2026.05 | 53.1 | 50.5 | 61.1 | 905.6 | 21.1 | 54.1 | 92.6 | — | |
| FastVRetention Ratio R=15%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 50.8 | 16.9 | 18 | — | — | 52.4 | — | 84.2 | |
| FastVRetention Ratio R=10%, Backbone=LLaVA-Video-7B-Qwen22026.06 | 43.2 | 13.1 | 13.9 | — | — | 46.4 | — | 74.5 |