Long Video Understanding on LongTimeScope 300min+
45.1AccuracyVideoMem*
Evaluation results
| Method | Links | |
|---|---|---|
| VideoMem*Size=8B, Input Frames=642025.12 | 45.1 | |
| VideoNSASize=7B, Input Frames=>=642025.12 | 44.4 | |
| Qwen2.5-VLSize=7B, Input Frames=>=642025.12 | 40.7 | |
| VideoLLaMA3Size=7B, Input Frames=>=642025.12 | 39.1 | |
| Qwen3-VL*Size=8B, Input Frames=642025.12 | 36.2 | |
| LLaVA-videoSize=7B, Input Frames=>=642025.12 | 34 | |
| LLaVA-OnevisionSize=7B, Input Frames=>=642025.12 | 30.2 | |
| Video-LLaVASize=7B, Input Frames=>=642025.12 | 17.6 |