Image Classification on CIFAR-10 (Adversarial Attack Robustness)
60.68Robustness (FGSM)Poincare ResNet-32
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Poincare ResNet-32Model=PRN 32, ε=2.4/255 (ε1)2025.11 | 60.68 | 51.09 | 28.96 | 18.69 | — | |
| Poincare ResNet-32Model=PRN 32, ε=3.2/255 (ε2)2025.11 | 59.1 | 48.05 | 18.43 | 11.44 | 86.21 | |
| Poincare ResNet-20Model=PRN 20, ε=2.4/255 (ε1)2025.11 | 56.59 | 47.63 | 31.83 | 22.42 | — | |
| Poincare ResNet-20Model=PRN 20, ε=3.2/255 (ε2)2025.11 | 54.43 | 44.53 | 21.4 | 14.25 | 84.76 | |
| Poincare ResNet-32Model=PRN 32, ε=8.0/255 (ε3)2025.11 | 54.19 | 41.56 | 8.05 | 7.77 | — | |
| Poincare ResNet-20Model=PRN 20, ε=8.0/255 (ε3)2025.11 | 49.63 | 36.69 | 8.86 | 8.35 | — |