Image Classification on CIFAR-100 (Clean, AA, Mean Metrics)
60.74Clean AccuracyMAIL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MAILBackbone=ResNet-18, Threat Model=ℓ∞2026.04 | 60.74 | 22.44 | 41.59 | |
| MMABackbone=ResNet-18, Threat Model=ℓ∞2026.04 | 60.6 | 18.4 | 39.5 | |
| HATBackbone=ResNet-18, Threat Model=ℓ∞2026.04 | 59.19 | 23.75 | 41.47 | |
| RAATBackbone=ResNet-18, Threat Model=ℓ∞2026.04 | 58.53 | 25.65 | 42.09 | |
| GAIRATBackbone=ResNet-18, Threat Model=ℓ∞2026.04 | 58.43 | 17.54 | 37.99 | |
| ADRBackbone=ResNet-18, Threat Model=ℓ∞2026.04 | 56.54 | 26.42 | 41.48 | |
| TEBackbone=ResNet-18, Threat Model=ℓ∞2026.04 | 56.45 | 26.3 | 41.38 | |
| PIATBackbone=ResNet-18, Threat Model=ℓ∞2026.04 | 56.04 | 26.09 | 41.07 | |
| AWPBackbone=ResNet-18, Threat Model=ℓ∞2026.04 | 55.16 | 25.16 | 40.16 | |
| SOVRBackbone=ResNet-18, Threat Model=ℓ∞2026.04 | 52.1 | 24.3 | 38.2 |