Adversarial Purification on ImageNet
80.2Standard AccuracyADDT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ADDTDefense model=ADDT (Liu et al., 2021)2026.07 | 80.2 | 35.83 | — | |
| RAMRM-PureMaskDiTDefense model=RAMRM-PureMaskDiT2026.07 | 79.52 | 36.87 | 51.17 | |
| RAMRM-PureMAEDefense model=RAMRM-PureMAE2026.07 | 78.85 | 29.48 | 42.25 | |
| DiffpureDefense model=Diffpure (Nie et al., 2022)2026.07 | 77.51 | 30.15 | 44.15 | |
| AMRM-PureMaskDiTDefense model=AMRM-PureMaskDiT2026.07 | 75.52 | 32.29 | 45.57 | |
| AMRM-PureMAEDefense model=AMRM-PureMAE2026.07 | 67.53 | 24.75 | 35.95 |