Visual Question Answering on GQA (Clean & Robust Accuracy)
60.3Clean AccuracyNone
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| NoneRetained Tokens=576 (Upper Bound)2026.01 | 60.3 | 42.3 | 39.4 | |
| PruMergeRetained Tokens=1922026.01 | 57.2 | 41.5 | 36 | |
| VisionZIPRetained Tokens=1922026.01 | 57 | 40.9 | 36.2 | |
| DivPruneRetained Tokens=1922026.01 | 55.6 | 40.5 | 37.4 | |
| DivPruneRetained Tokens=1282026.01 | 55.6 | 40.4 | 36.5 | |
| PruMergeRetained Tokens=1282026.01 | 55.5 | 41 | 35.3 | |
| FlowCutRetained Tokens=1922026.01 | 55.4 | 40.3 | 35.1 | |
| VisPrunerRetained Tokens=1922026.01 | 55.3 | 40.8 | 36.5 | |
| VisionZIPRetained Tokens=1282026.01 | 55.1 | 40.9 | 35.6 | |
| FlowCutRetained Tokens=1282026.01 | 54.8 | 40.6 | 34.7 | |
| VisPrunerRetained Tokens=1282026.01 | 54.3 | 41.5 | 36.2 | |
| PruMergeRetained Tokens=642026.01 | 53.8 | 41.2 | 34.1 | |
| VisPrunerRetained Tokens=642026.01 | 53.5 | 40.3 | 34.5 | |
| DivPruneRetained Tokens=642026.01 | 52.9 | 40.2 | 36.9 | |
| VisionZIPRetained Tokens=642026.01 | 52.8 | 40.8 | 34.3 | |
| DivPruneRetained Tokens=322026.01 | 50.5 | 40.2 | 35 | |
| PruMergeRetained Tokens=322026.01 | 50.4 | 39.7 | 33.9 | |
| FlowCutRetained Tokens=642026.01 | 50.2 | 38.6 | 34.7 | |
| VisionZIPRetained Tokens=322026.01 | 50.1 | 40 | 33.3 | |
| DivPruneRetained Tokens=162026.01 | 49.3 | 41.2 | 34.1 | |
| VisPrunerRetained Tokens=322026.01 | 49.1 | 37.9 | 33.4 | |
| PruMergeRetained Tokens=162026.01 | 47.6 | 38.5 | 32.4 | |
| VisionZIPRetained Tokens=162026.01 | 47.4 | 37.2 | 32.4 | |
| FlowCutRetained Tokens=322026.01 | 47.2 | 37.2 | 33.9 | |
| VisPrunerRetained Tokens=162026.01 | 45.9 | 38.9 | 32.8 | |
| FlowCutRetained Tokens=162026.01 | 40.7 | 37.8 | 33.6 | |
| NoneRetained Tokens=0 (Blind)2026.01 | 21.7 | 21.7 | 21.7 |