Video Understanding on Average VideoMME, LongVideoBench, MLVU
55.9ScoreVisionZip† + DyTok (7B)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VisionZip† + DyTok (7B)Frames=64, Retain tokens=62722025.12 | 55.9 | 100.7 | |
| VanillaFrames=64, Retain tokens=125442025.12 | 55.5 | 100 | |
| VisionZip†Frames=64, Retain tokens=94082025.12 | 55.5 | 100 | |
| VisionZip† + DyTokFrames=64, Retain tokens=62722025.12 | 55.4 | 99.8 | |
| VisionZip† + DyTokFrames=64, Retain tokens=94082025.12 | 55.3 | 99.6 | |
| VisionZip† + DyTok (7B)Frames=64, Retain tokens=94082025.12 | 55.3 | 99.6 | |
| VisionZip† + DyTok (7B)Frames=64, Retain tokens=22402025.12 | 55.2 | 99.5 | |
| VisionZip†Frames=64, Retain tokens=62722025.12 | 55.1 | 99.3 | |
| VisionZip† + DyTok (7B)Frames=64, Retain tokens=31362025.12 | 55.1 | 99.3 | |
| VisionZip† + DyTokFrames=64, Retain tokens=31362025.12 | 54.6 | 98.4 | |
| VisionZip† + DyTok (7B)Frames=64, Retain tokens=17922025.12 | 54.2 | 97.7 | |
| VisionZip†Frames=64, Retain tokens=31362025.12 | 54 | 97.3 | |
| VisionZip†Frames=64, Retain tokens=22402025.12 | 53.5 | 96.4 | |
| VisionZip† + DyTokFrames=64, Retain tokens=22402025.12 | 53.5 | 96.4 | |
| VisionZipFrames=64, Retain tokens=94082025.12 | 52.9 | 95.3 | |
| VisionZip† + DyTokFrames=64, Retain tokens=17922025.12 | 52.6 | 94.8 | |
| VisionZipFrames=64, Retain tokens=62722025.12 | 52 | 93.7 | |
| VisionZip† + DyTok (7B)Frames=64, Retain tokens=8962025.12 | 52 | 93.7 | |
| VisionZip†Frames=64, Retain tokens=17922025.12 | 51.3 | 92.4 | |
| VisionZip† + DyTokFrames=64, Retain tokens=8962025.12 | 50.3 | 90.6 | |
| VisionZipFrames=64, Retain tokens=31362025.12 | 49.8 | 89.7 | |
| VisionZipFrames=64, Retain tokens=22402025.12 | 46.4 | 83.6 | |
| VisionZip†Frames=64, Retain tokens=8962025.12 | 45.4 | 81.8 | |
| VisionZipFrames=64, Retain tokens=17922025.12 | 43.9 | 79.1 | |
| VisionZipFrames=64, Retain tokens=8962025.12 | 38.7 | 69.7 |