Video Grounding on video-SAL2
18.9Video-SAL2 MetricFull Token
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Full TokenBackbone=Video-SALMONN2+ 7B, Comp.↑ (%)=02026.05 | 18.9 | 100 | |
| ContextGuardBackbone=Video-SALMONN2+ 7B, Comp.↑ (%)=552026.05 | 19.9 | 100.5 | |
| OmniZipBackbone=Video-SALMONN2+ 7B, Comp.↑ (%)=492026.05 | 22.4 | 95.7 | |
| Full TokenBackbone=Video-SALMONN2+ 3B, Comp.↑ (%)=02026.05 | 22.7 | 100 | |
| FastVBackbone=Video-SALMONN2+ 7B, Comp.↑ (%)=502026.05 | 24 | 92.8 | |
| ContextGuardBackbone=Video-SALMONN2+ 3B, Comp.↑ (%)=562026.05 | 24.2 | 96.9 | |
| RandomBackbone=Video-SALMONN2+ 7B, Comp.↑ (%)=502026.05 | 25.7 | 90.7 | |
| OmniZipBackbone=Video-SALMONN2+ 3B, Comp.↑ (%)=492026.05 | 26.3 | 93.7 | |
| FastVBackbone=Video-SALMONN2+ 3B, Comp.↑ (%)=502026.05 | 27.4 | 90.2 | |
| RandomBackbone=Video-SALMONN2+ 3B, Comp.↑ (%)=502026.05 | 30.3 | 90.6 | |
| Full TokenBackbone=Qwen2.5-Omni 7B, Comp.↑ (%)=02026.05 | 48.1 | 100 | |
| ContextGuardBackbone=Qwen2.5-Omni 7B, Comp.↑ (%)=552026.05 | 48.1 | 100 | |
| FastVBackbone=Qwen2.5-Omni 7B, Comp.↑ (%)=502026.05 | 49.8 | 98.1 | |
| RandomBackbone=Qwen2.5-Omni 7B, Comp.↑ (%)=502026.05 | 50.4 | 95.4 | |
| ContextGuardBackbone=Qwen2.5-Omni 3B, Comp.↑ (%)=612026.05 | 52.3 | 100.5 | |
| OmniZipBackbone=Qwen2.5-Omni 3B, Comp.↑ (%)=542026.05 | 52.4 | 98.9 | |
| OmniZipBackbone=Qwen2.5-Omni 7B, Comp.↑ (%)=542026.05 | 52.8 | 97.3 | |
| Full TokenBackbone=Qwen2.5-Omni 3B, Comp.↑ (%)=02026.05 | 53.5 | 100 | |
| FastVBackbone=Qwen2.5-Omni 3B, Comp.↑ (%)=502026.05 | 54.6 | 98.2 | |
| RandomBackbone=Qwen2.5-Omni 3B, Comp.↑ (%)=502026.05 | 56.1 | 95.7 |