Visual Question Answering on TextVQA (EM)
82.74Exact Match (EM)Baseline
Evaluation Results
| Method | Links | |
|---|---|---|
| BaselineCompression Ratio=0%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 82.74 | |
| IDPrunerCompression Ratio=75%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 80.83 | |
| VisionSelectorCompression Ratio=75%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 80.39 | |
| FastVCompression Ratio=75%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 80.06 | |
| SCOPECompression Ratio=75%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 79.66 | |
| HiPruneCompression Ratio=75%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 76.43 | |
| VisionZipCompression Ratio=75%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 76.21 | |
| DivPruneCompression Ratio=75%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 75.88 | |
| VisPrunerCompression Ratio=75%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 75.17 | |
| VisionSelectorCompression Ratio=90%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 74.52 | |
| IDPrunerCompression Ratio=90%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 74.43 | |
| FastVCompression Ratio=90%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 72.59 | |
| DARTCompression Ratio=75%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 71.68 | |
| SCOPECompression Ratio=90%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 70.61 | |
| DivPruneCompression Ratio=90%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 62.72 | |
| HiPruneCompression Ratio=90%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 62.51 | |
| VisionZipCompression Ratio=90%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 61.06 | |
| VisPrunerCompression Ratio=90%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 59.66 | |
| DARTCompression Ratio=90%, Backbone=Qwen-2.5-7B-Instruct2026.02 | 54.06 | |
| BaselineRetain Tokens=Full2026.02 | 47.78 | |
| IDPrunerRetain Tokens=1282026.02 | 46.39 | |
| IDPrunerRetain Tokens=642026.02 | 44.73 | |
| IDPrunerRetain Tokens=322026.02 | 41.46 |