Visual Question Answering on VizWiz (Accuracy, Training/Inference Time)
51.66AccuracyVL-Mamba
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| VL-MambaQuery Length=-2026.06 | 51.66 | 7.26 | 0.152 | |
| Q-MambaQuery Length=1282026.06 | 51.03 | 5.52 | 0.095 | |
| Q-MambaQuery Length=2562026.06 | 50.53 | 5.94 | 0.099 | |
| Q-MambaQuery Length=7292026.06 | 49.3 | 7.54 | 0.147 | |
| CobraQuery Length=-2026.06 | 49.22 | 7.62 | 0.129 | |
| Q-MambaQuery Length=5122026.06 | 48.9 | 6.54 | 0.127 | |
| VL-MambaQuery Length=-, Visual scanning order=backward scan only2026.06 | 34.92 | — | — | |
| VL-MambaQuery Length=-, Visual scanning order=forward scan only2026.06 | 29.17 | — | — |