Visual Question Answering on VQA Text (EM)
79.65Exact MatchBaseline
Evaluation Results
| Method | Links | |
|---|---|---|
| Baseline2026.02 | 79.65 | |
| VisionSelectorRetain Tokens=25%2026.02 | 77.53 | |
| IDPrunerRetain Tokens=25%2026.02 | 77.5 | |
| DivPruneRetain Tokens=25%2026.02 | 75 | |
| IDPrunerRetain Tokens=10%2026.02 | 72.37 |