Visual Question Answering on POPE (Accuracy, Avg. Ratio)
86.4Accuracy (POPE)Nüwa
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| NüwaSource=ICLR’26, Average Token=192, Reroute=false2026.06 | 86.4 | 99.2 | |
| Nüwa+RerouteSource=Ours, Average Token=192, Reroute=true2026.06 | 86 | 100.6 | |
| VanillaSource=CVPR’24, Average Token=576, Reroute=false2026.06 | 85.9 | 100 | |
| NüwaSource=ICLR’26, Average Token=128, Reroute=false2026.06 | 85.5 | 98.4 | |
| VisionZipSource=CVPR’25, Average Token=192, Reroute=false2026.06 | 85.3 | 97.9 | |
| Nüwa+RerouteSource=Ours, Average Token=128, Reroute=true2026.06 | 85.3 | 99.8 | |
| SparseVLMSource=ICML’25, Average Token=192, Reroute=false2026.06 | 83.6 | 95.4 | |
| VisionZipSource=CVPR’25, Average Token=128, Reroute=false2026.06 | 83.2 | 96.9 | |
| NüwaSource=ICLR’26, Average Token=64, Reroute=false2026.06 | 83 | 95.8 | |
| Nüwa+RerouteSource=Ours, Average Token=64, Reroute=true2026.06 | 82.6 | 96.7 | |
| PDropSource=CVPR’25, Average Token=192, Reroute=false2026.06 | 82.3 | 95.7 | |
| PDropSource=CVPR’25, Average Token=128, Reroute=false2026.06 | 82.3 | 93.8 | |
| PruMergeSource=ICCV’25, Average Token=128, Reroute=false2026.06 | 81.5 | 94.7 | |
| SparseVLMSource=ICML’25, Average Token=128, Reroute=false2026.06 | 80.5 | 92.9 | |
| SparseVLMSource=ICML’25, Average Token=64, Reroute=false2026.06 | 77.5 | 91.5 | |
| VisionZipSource=CVPR’25, Average Token=64, Reroute=false2026.06 | 77 | 93.2 | |
| PruMergeSource=ICCV’25, Average Token=64, Reroute=false2026.06 | 75.7 | 92.1 | |
| FastVSource=ECCV’24, Average Token=192, Reroute=false2026.06 | 64.8 | 90.1 | |
| FastVSource=ECCV’24, Average Token=128, Reroute=false2026.06 | 59.6 | 85.2 | |
| FastVSource=ECCV’24, Average Token=64, Reroute=false2026.06 | 59.6 | 77.9 | |
| PDropSource=CVPR’25, Average Token=64, Reroute=false2026.06 | 55.9 | 70.3 |