Long-form Video Understanding on Video-MME long-form duration
66.1Overall PerformanceColorTrigger
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ColorTriggerModel size=8B, #Frames=128, RGB (%)=37.6 (↓ 62.4%)2026.03 | 66.1 | 76.8 | 66.3 | 55.1 | |
| InternVL-3.5-8B#Frames=128, RGB (%)=1002026.03 | 65.6 | 76.7 | 65.3 | 54.7 | |
| ColorTriggerModel size=8B, #Frames=128, RGB (%)=9.1 (↓ 90.9%)2026.03 | 62.8 | 71.9 | 63 | 53.6 | |
| TimeChat-Online-7B#Frames=1 fps, RGB (%)=1002026.03 | 62.4 | — | — | 48.4 | |
| InternVL-3.5-8B#Frames=128, RGB (%)=0, mode=grayscale2026.03 | 57.3 | 63.4 | 57.9 | 50.6 | |
| Dispider-7B#Frames=1 fps, RGB (%)=1002026.03 | 57.2 | — | — | — | |
| Kangaroo-7B#Frames=64, RGB (%)=1002026.03 | 56 | 66.1 | 55.3 | 46.6 | |
| VideoXL-7B#Frames=128, RGB (%)=1002026.03 | 55.5 | 64 | 53.2 | 49.2 | |
| VideoChat-Online-4B#Frames=2 fps, RGB (%)=1002026.03 | 54.4 | — | — | 47.1 | |
| Video-CCAM-14B#Frames=96, RGB (%)=1002026.03 | 53.2 | 62.2 | 50.6 | 46.7 | |
| LongVA-7B#Frames=128, RGB (%)=1002026.03 | 52.6 | 61.1 | 50.4 | 46.2 | |
| VideoChat2-7B#Frames=16, RGB (%)=1002026.03 | 39.5 | 48.3 | 37 | 33.2 |