Image Classification on ImageNet-1K (test val) (Backdoor Attack Metrics)
79.55Clean Accuracy (C-ACC)Fine-Pruning
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Fine-PruningBackbone=DeiT-Small, Attack Configuration=2-of-32026.02 | 79.55 | — | — | 0.67 | 92.51 | 100 | |
| Baseline (no defense)Backbone=DeiT-Small, Attack Configuration=2-of-32026.02 | 78.88 | — | — | 0 | 98.3 | 100 | |
| BDVTBackbone=DeiT-Small, Attack Configuration=2-of-32026.02 | 78.58 | — | — | -0.3 | 14.38 | 98.82 | |
| Patch ProcessingBackbone=DeiT-Small, Attack Configuration=2-of-32026.02 | 76.07 | — | — | -2.81 | 93.14 | 100 | |
| Neural CleanseBackbone=DeiT-Small, Attack Configuration=2-of-32026.02 | — | — | 3 | — | — | — |