Video Understanding on VideoMME (Short, Medium, Long, Overall)
73Accuracy (Short Segment)EchoPrune
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EchoPruneNumber of Frames (F)=80, Compression Ratio=5x, Token Budget=20%2026.05 | 73 | 62.3 | 53.4 | 62.9 | |
| EchoPruneNumber of Frames (F)=160, Compression Ratio=10x, Token Budget=10%2026.05 | 72.7 | 63.2 | 53.3 | 63.1 | |
| EchoPruneNumber of Frames (F)=64, Compression Ratio=4x, Token Budget=25%2026.05 | 72.4 | 60.1 | 51.6 | 61.4 | |
| EchoPruneNumber of Frames (F)=320, Compression Ratio=20x, Token Budget=5% (ultra)2026.05 | 72.2 | 61.4 | 54.8 | 62.8 | |
| Qwen2.5VL-7BNumber of Frames (F)=162026.05 | 66.4 | 56.4 | 48.2 | 57 |