Visual Question Answering on SeedBench (Accuracy and Avg. Ratio)
66.1AccuracyNüwa+Reroute
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Nüwa+RerouteSource=Ours, Average Token=192, Reroute=true2026.06 | 66.1 | 100.6 | |
| Nüwa+RerouteSource=Ours, Average Token=128, Reroute=true2026.06 | 64.4 | 99.8 | |
| Nüwa+RerouteSource=Ours, Average Token=64, Reroute=true2026.06 | 61.6 | 96.7 | |
| NüwaSource=ICLR’26, Average Token=192, Reroute=false2026.06 | 59.7 | 99.2 | |
| NüwaSource=ICLR’26, Average Token=128, Reroute=false2026.06 | 58.7 | 98.4 | |
| VanillaSource=CVPR’24, Average Token=576, Reroute=false2026.06 | 58.6 | 100 | |
| FastVSource=ECCV’24, Average Token=192, Reroute=false2026.06 | 57.1 | 90.1 | |
| VisionZipSource=CVPR’25, Average Token=192, Reroute=false2026.06 | 56.4 | 97.9 | |
| NüwaSource=ICLR’26, Average Token=64, Reroute=false2026.06 | 56.4 | 95.8 | |
| FastVSource=ECCV’24, Average Token=128, Reroute=false2026.06 | 55.9 | 85.2 | |
| SparseVLMSource=ICML’25, Average Token=192, Reroute=false2026.06 | 55.8 | 95.4 | |
| VisionZipSource=CVPR’25, Average Token=128, Reroute=false2026.06 | 54.9 | 96.9 | |
| PDropSource=CVPR’25, Average Token=192, Reroute=false2026.06 | 54.7 | 95.7 | |
| SparseVLMSource=ICML’25, Average Token=128, Reroute=false2026.06 | 53.4 | 92.9 | |
| PDropSource=CVPR’25, Average Token=128, Reroute=false2026.06 | 53.3 | 93.8 | |
| VisionZipSource=CVPR’25, Average Token=64, Reroute=false2026.06 | 52.2 | 93.2 | |
| FastVSource=ECCV’24, Average Token=64, Reroute=false2026.06 | 51.9 | 77.9 | |
| SparseVLMSource=ICML’25, Average Token=64, Reroute=false2026.06 | 51.1 | 91.5 | |
| PDropSource=CVPR’25, Average Token=64, Reroute=false2026.06 | 40 | 70.3 |