Zero-shot Image Classification on Caltech 101 (test)
84.73Accuracy (Zero-shot)ET3
Evaluation results
| Method | Links | |
|---|---|---|
| ET3Backbone=ViT-L/14, Pre-training=FARE, epsilon_t=4/255, Evaluation Scenario=Clean2026.03 | 84.73 | |
| BaseBackbone=ViT-L/14, Pre-training=FARE, epsilon_t=4/255, Evaluation Scenario=Clean2026.03 | 84.7 | |
| BaseBackbone=ViT-L/14, Pre-training=TeCoA, epsilon_t=4/255, Evaluation Scenario=Clean2026.03 | 78.36 | |
| ET3Backbone=ViT-L/14, Pre-training=TeCoA, epsilon_t=4/255, Evaluation Scenario=Clean2026.03 | 78.16 | |
| ET3Backbone=ViT-L/14, Pre-training=FARE, epsilon_t=4/255, Evaluation Scenario=Robust (AA)2026.03 | 69.4 | |
| ET3Backbone=ViT-L/14, Pre-training=TeCoA, epsilon_t=4/255, Evaluation Scenario=Robust (AA)2026.03 | 65.1 | |
| BaseBackbone=ViT-L/14, Pre-training=FARE, epsilon_t=4/255, Evaluation Scenario=Robust (AA)2026.03 | 64.2 | |
| BaseBackbone=ViT-L/14, Pre-training=TeCoA, epsilon_t=4/255, Evaluation Scenario=Robust (AA)2026.03 | 60.9 |