Visual Question Answering on TextVQA (Robust Accuracy Evaluation)
70.3Clean AccuracyNone
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| NoneCompressed Method=None, Retained Tokens (Kmodel)=144, Backbone=Qwen2.5-VL2026.01 | 70.3 | 38 | 33.9 | |
| VisionZIPCompressed Method=VisionZIP, Retained Tokens (Kmodel)=72, Backbone=Qwen2.5-VL2026.01 | 66.2 | 29.6 | 24.2 | |
| VisPrunerCompressed Method=VisPruner, Retained Tokens (Kmodel)=72, Backbone=Qwen2.5-VL2026.01 | 62.8 | 29.4 | 26 | |
| AverageCompressed Method=Average, Retained Tokens (Kmodel)=72, Backbone=Qwen2.5-VL2026.01 | 62.5 | 31 | 28 | |
| DivPruneCompressed Method=DivPrune, Retained Tokens (Kmodel)=72, Backbone=Qwen2.5-VL2026.01 | 58.4 | 34 | 33.8 | |
| VisPrunerRetained Tokens=1922026.01 | 58.2 | 33.9 | 23 | |
| FlowCutRetained Tokens=1922026.01 | 57.8 | 33.6 | 24.3 | |
| DivPruneCompressed Method=DivPrune, Retained Tokens (Kmodel)=36, Backbone=Qwen2.5-VL2026.01 | 57.7 | 28.6 | 24.1 | |
| NoneRetained Tokens=576 (Upper Bound)2026.01 | 57.5 | 34.7 | 26.5 | |
| FlowCutRetained Tokens=1282026.01 | 57.5 | 33.2 | 21.3 | |
| VisionZIPRetained Tokens=1282026.01 | 57 | 33.5 | 21.6 | |
| VisionZIPRetained Tokens=1922026.01 | 56.9 | 34.1 | 24.6 | |
| VisPrunerRetained Tokens=1282026.01 | 56.5 | 34.1 | 21.5 | |
| VisionZIPCompressed Method=VisionZIP, Retained Tokens (Kmodel)=36, Backbone=Qwen2.5-VL2026.01 | 56.1 | 19.2 | 13 | |
| PruMergeRetained Tokens=1922026.01 | 56 | 33.3 | 21.6 | |
| PruMergeRetained Tokens=1282026.01 | 55.1 | 32.9 | 21.9 | |
| VisionZIPRetained Tokens=642026.01 | 54.8 | 32 | 18.3 | |
| FlowCutRetained Tokens=642026.01 | 54.7 | 31.6 | 18.3 | |
| DivPruneRetained Tokens=1922026.01 | 54.6 | 32.7 | 23.5 | |
| VisPrunerRetained Tokens=642026.01 | 54.6 | 31 | 18.9 | |
| PruMergeRetained Tokens=642026.01 | 52.8 | 31.5 | 17.7 | |
| DivPruneRetained Tokens=1282026.01 | 52.7 | 32.2 | 23.2 | |
| AverageCompressed Method=Average, Retained Tokens (Kmodel)=36, Backbone=Qwen2.5-VL2026.01 | 52.1 | 22.5 | 18.2 | |
| DivPruneRetained Tokens=642026.01 | 50.1 | 29.5 | 21.2 | |
| FlowCutRetained Tokens=322026.01 | 50.1 | 29.3 | 17.4 | |
| VisionZIPRetained Tokens=322026.01 | 49.3 | 30.3 | 17.9 | |
| VisPrunerRetained Tokens=322026.01 | 49.2 | 30.7 | 17.8 | |
| PruMergeRetained Tokens=322026.01 | 46.1 | 27.6 | 17.5 | |
| VisionZIPRetained Tokens=162026.01 | 45.8 | 27.2 | 15.8 | |
| DivPruneRetained Tokens=322026.01 | 45.5 | 27.8 | 20.3 | |
| VisPrunerCompressed Method=VisPruner, Retained Tokens (Kmodel)=36, Backbone=Qwen2.5-VL2026.01 | 42.4 | 19.6 | 17.5 | |
| PruMergeRetained Tokens=162026.01 | 39.5 | 25.4 | 15.7 | |
| FlowCutRetained Tokens=162026.01 | 38.7 | 25.7 | 15.1 | |
| DivPruneRetained Tokens=162026.01 | 38.5 | 24.3 | 17 | |
| VisPrunerRetained Tokens=162026.01 | 38.2 | 25.1 | 14.7 | |
| NoneRetained Tokens=0 (Blind)2026.01 | 10.3 | 10.3 | 10.3 | |
| NoneCompressed Method=None, Retained Tokens (Kmodel)=0, Backbone=Qwen2.5-VL2026.01 | 6.8 | 6.8 | 6.8 |