Audio-visual understanding on Audio-visual understanding evaluation suite (test)
46.7WorldSense ScoreFull tokens
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Full tokensR (Rv-Ra)=100 (100-100), TFLOPs=111.0, Model=Qwen2.5-Omni-7B2026.05 | 46.7 | 64 | 34.1 | 65.3 | 33.3 | 48.7 | |
| SEATSR (Rv-Ra)=35 (30-65), TFLOPs=36.7, Model=Qwen2.5-Omni-7B2026.05 | 46.2 | 62.1 | 35 | 66.8 | 36.2 | 49.3 | |
| DivPrune-omR (Rv-Ra)=35 (30-65), TFLOPs=37.4, Model=Qwen2.5-Omni-7B2026.05 | 45.4 | 59.7 | 34.7 | 66.3 | 35.4 | 48.3 | |
| SEATSR (Rv-Ra)=25 (20-55), TFLOPs=26.5, Model=Qwen2.5-Omni-7B2026.05 | 45.3 | 60.9 | 34.7 | 66.5 | 35.7 | 48.6 | |
| OmniZipR (Rv-Ra)=35 (30-65), TFLOPs=38.2, Model=Qwen2.5-Omni-7B2026.05 | 45.2 | 60.4 | 34.5 | 66.3 | 34.7 | 48.2 | |
| VisionZip-omR (Rv-Ra)=35 (30-65), TFLOPs=37.4, Model=Qwen2.5-Omni-7B2026.05 | 45.2 | 60.2 | 34.6 | 66.3 | 35.4 | 48.3 | |
| DivPruneR (Rv-Ra)=35 (24-100), TFLOPs=37.4, Model=Qwen2.5-Omni-7B2026.05 | 44.8 | 60.7 | 33.9 | 66.1 | 35.7 | 48.2 | |
| RandomR (Rv-Ra)=35 (30-65), TFLOPs=37.4, Model=Qwen2.5-Omni-7B2026.05 | 44.6 | 59.7 | 34.1 | 65 | 34.9 | 47.7 | |
| FastVIDR (Rv-Ra)=35 (24-100), TFLOPs=37.1, Model=Qwen2.5-Omni-7B2026.05 | 44.6 | 59.8 | 34.6 | 65.4 | 35.8 | 48.1 | |
| VisionZipR (Rv-Ra)=35 (24-100), TFLOPs=37.4, Model=Qwen2.5-Omni-7B2026.05 | 44.5 | 60.6 | 33.6 | 66.2 | 35.3 | 48 | |
| DivPrune-omR (Rv-Ra)=25 (20-55), TFLOPs=27.6, Model=Qwen2.5-Omni-7B2026.05 | 44.5 | 59.2 | 34.5 | 66.2 | 35.4 | 48 | |
| DyCokeR (Rv-Ra)=35 (24-100), TFLOPs=37.7, Model=Qwen2.5-Omni-7B2026.05 | 44.4 | 59.9 | 34.6 | 66 | 34.7 | 47.9 | |
| VisionZip-omR (Rv-Ra)=25 (20-55), TFLOPs=27.6, Model=Qwen2.5-Omni-7B2026.05 | 44.4 | 58.3 | 33.7 | 65.4 | 34.7 | 47.3 | |
| SEATSR (Rv-Ra)=15 (10-45), TFLOPs=17.3, Model=Qwen2.5-Omni-7B2026.05 | 44.1 | 58.7 | 33.7 | 66 | 34.6 | 47.4 | |
| FastVR (Rv-Ra)=35 (24-100), TFLOPs=38.3, Model=Qwen2.5-Omni-7B2026.05 | 43.7 | 59.6 | 34.7 | 65.2 | 34.8 | 47.6 | |
| SEATSR (Rv-Ra)=10 (6-35), TFLOPs=12.2, Model=Qwen2.5-Omni-7B2026.05 | 43.5 | 57.8 | 33.6 | 64.6 | 35.1 | 46.9 | |
| FastVIDR (Rv-Ra)=25 (12-100), TFLOPs=27.0, Model=Qwen2.5-Omni-7B2026.05 | 43.4 | 58.5 | 33.8 | 64.3 | 34.8 | 47 | |
| FastV-omR (Rv-Ra)=35 (30-65), TFLOPs=38.3, Model=Qwen2.5-Omni-7B2026.05 | 43.3 | 58.7 | 34.4 | 65.6 | 34.3 | 47.2 | |
| DivPruneR (Rv-Ra)=25 (12-100), TFLOPs=27.6, Model=Qwen2.5-Omni-7B2026.05 | 43.2 | 59.7 | 34.3 | 65.3 | 35.9 | 47.7 | |
| RandomR (Rv-Ra)=25 (20-55), TFLOPs=27.6, Model=Qwen2.5-Omni-7B2026.05 | 43 | 57.3 | 33.5 | 64.2 | 34.6 | 46.5 | |
| VisionZip-omR (Rv-Ra)=15 (10-45), TFLOPs=17.6, Model=Qwen2.5-Omni-7B2026.05 | 43 | 57.1 | 33.6 | 63.1 | 33.8 | 46.1 | |
| OmniZipR (Rv-Ra)=25 (25-25), TFLOPs=23.9, Model=Qwen2.5-Omni-7B2026.05 | 42.7 | 55.5 | 32.2 | 66 | 34 | 46.1 | |
| FastVR (Rv-Ra)=25 (12-100), TFLOPs=28.3, Model=Qwen2.5-Omni-7B2026.05 | 42.4 | 56.5 | 33.4 | 62.8 | 33.8 | 45.8 | |
| VisionZipR (Rv-Ra)=25 (12-100), TFLOPs=27.6, Model=Qwen2.5-Omni-7B2026.05 | 42.4 | 59.7 | 34.4 | 64.1 | 34.5 | 47 | |
| DivPrune-omR (Rv-Ra)=15 (10-45), TFLOPs=17.6, Model=Qwen2.5-Omni-7B2026.05 | 42.4 | 57.6 | 33 | 64.4 | 34.4 | 46.4 | |
| VisionZip-omR (Rv-Ra)=10 (6-35), TFLOPs=12.8, Model=Qwen2.5-Omni-7B2026.05 | 41.8 | 52.8 | 33.4 | 60.7 | 34.4 | 44.6 | |
| FastV-omR (Rv-Ra)=25 (20-55), TFLOPs=28.3, Model=Qwen2.5-Omni-7B2026.05 | 41.6 | 56.1 | 33 | 63.4 | 32.7 | 45.4 | |
| DivPrune-omR (Rv-Ra)=10 (6-35), TFLOPs=12.8, Model=Qwen2.5-Omni-7B2026.05 | 41.3 | 55.1 | 33.5 | 63.4 | 34.8 | 45.6 | |
| RandomR (Rv-Ra)=15 (10-45), TFLOPs=18.0, Model=Qwen2.5-Omni-7B2026.05 | 40.4 | 55.7 | 32.7 | 63.4 | 32.8 | 45 | |
| RandomR (Rv-Ra)=10 (6-35), TFLOPs=12.8, Model=Qwen2.5-Omni-7B2026.05 | 39.2 | 52 | 31.6 | 60.4 | 32.9 | 43.2 | |
| FastV-omR (Rv-Ra)=15 (10-45), TFLOPs=24.3, Model=Qwen2.5-Omni-7B2026.05 | 38.4 | 51.8 | 31.2 | 58.8 | 31.2 | 42.3 |