Adversarial Robustness on CIFAR-100 sr=90% (test)
64.37Clean AccuracyAT
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ATBackbone=ResNet-50, Sparsity rate=90%, Params=25.6M2025.04 | 64.37 | 36.29 | 28.07 | 30.18 | 39.8 | 1 | |
| EEDBackbone=ResNet-50, Sparsity rate=90%, Params=2.6M2025.04 | 63.6 | 36.29 | 26.79 | 28.01 | 37.14 | 2.9 | |
| HARPBackbone=ResNet-50, Sparsity rate=90%, Params=2.6M2025.04 | 62.51 | 33.4 | 25.36 | 27.25 | 34.2 | 2.87 | |
| TwinRepBackbone=ResNet-50, Sparsity rate=90%, Params=2.6M2025.04 | 62.31 | 34.08 | 24.44 | 28.92 | 33.73 | 2.86 | |
| HYDRABackbone=ResNet-50, Sparsity rate=90%, Params=2.6M2025.04 | 62.1 | 33.52 | 24.12 | 26.2 | 33.94 | 2.8 | |
| R-ADMMBackbone=ResNet-50, Sparsity rate=90%, Params=2.6M2025.04 | 61.38 | 31.23 | 21.85 | 24.04 | 30.42 | 2.43 | |
| RSTBackbone=ResNet-50, Sparsity rate=90%, Params=2.6M2025.04 | 61.14 | 29.81 | 20.15 | 21.38 | 28.45 | 2.98 | |
| Flying BirdBackbone=ResNet-50, Sparsity rate=90%, Params=2.6M2025.04 | 60.03 | 33.18 | 24.91 | 24.19 | 32.26 | 2.75 | |
| MADBackbone=ResNet-50, Sparsity rate=90%, Params=2.6M2025.04 | 56.88 | 30.59 | 21.53 | 23.55 | 30.97 | 2.55 |