Explanation Quality Evaluation on Chest X Pneumothorax (test)
2.2Relevance Rank Accuracy (FeatPerm)ViT-L/16
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ViT-L/16Backbone=ViT-L/16, Resolution=224, Pre-training=IN-21k2026.05 | 2.2 | 1.6 | 2.1 | 2.3 | 1.8 | 50.1 | 0.7 | 50 | 50 | 50 | |
| ResNet-50Backbone=ResNet-50, Pre-training=IN-1k2026.05 | 1.9 | 1.7 | 1.6 | 1.8 | 1.9 | 49.9 | 0.9 | 50 | 50 | 50 | |
| ViT-B/16Backbone=ViT-B/16, Resolution=224, Pre-training=IN-21k2026.05 | 1.9 | 0.9 | 2.4 | 2.6 | 2.3 | 50 | 0.7 | 50.1 | 50.1 | 50 | |
| ResNet-50Backbone=ResNet-50, Pre-training=scratch2026.05 | 1.5 | 1.7 | 1.9 | 2 | 2 | 49.9 | 0.9 | 50.1 | 50.1 | 50 | |
| ResNet-101Backbone=ResNet-1012026.05 | 1.5 | 1.3 | 2.1 | 2.3 | 2 | 49.9 | 0.8 | 50.1 | 50.1 | 50 | |
| DenseNet-121Backbone=DenseNet-121, Pre-training=CheXpert2026.05 | 1.5 | 2 | 1.9 | 1.9 | 1.4 | 50.2 | 0.9 | 50.2 | 50.2 | 49.9 | |
| DenseNet-121Backbone=DenseNet-1212026.05 | 1.4 | 2.4 | 3 | 3.2 | 3.3 | 49.9 | 1 | 50.2 | 50.2 | 50 | |
| DenseNet-169Backbone=DenseNet-1692026.05 | 1.4 | 2.4 | 3.1 | 3.2 | 3.3 | 49.6 | 1 | 50.1 | 50.2 | 50 | |
| ResNet-18Backbone=ResNet-182026.05 | 1.2 | 1.4 | 2.7 | 2.9 | 2.4 | 49.5 | 0.8 | 50.2 | 50.2 | 49.9 | |
| DenseNet-201Backbone=DenseNet-2012026.05 | 1.1 | 1.7 | 2.8 | 3 | 2.8 | 49.6 | 0.9 | 50.2 | 50.2 | 50 | |
| ResNet-34Backbone=ResNet-342026.05 | 0.9 | 0.9 | 2.7 | 3 | 2.2 | 49.4 | 0.7 | 50.2 | 50.2 | 49.9 |