Visual Question Answering on Text VQA (Accuracy, Avg., Retain %)
81.56AccuracyBase
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| BaseRatio=100%, Model Backbone=Qwen3-VL-4B2026.05 | 81.56 | 81.95 | 100 | |
| IVC-PruneRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 81.56 | 76.81 | 93.74 | |
| IVC-PruneRatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 80.96 | 75.09 | 91.64 | |
| ERASERatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 80.41 | 78.98 | 96.38 | |
| ERASERatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 80.31 | 77.1 | 94.09 | |
| ERASERatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 79.65 | 73.25 | 89.39 | |
| DARTRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 76.96 | 69.49 | 84.8 | |
| IVC-PruneRatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 76.23 | 66.15 | 80.73 | |
| DARTRatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 74.13 | 63.72 | 77.75 | |
| CDPrunerRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 73.16 | 60.84 | 74.24 | |
| PruneSIDRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 69.87 | 66.97 | 81.72 | |
| CDPrunerRatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 69.65 | 55.15 | 67.3 | |
| PruneSIDRatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 66.98 | 63.72 | 77.75 | |
| DARTRatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 64 | 50.73 | 61.91 | |
| CDPrunerRatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 59.96 | 43.71 | 53.34 | |
| PruneSIDRatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 56.36 | 51.54 | 62.9 |