Classification on CIFAR-10 (test) (Robust Accuracy)
86.84Robust AccuracyZePAD (W-MSE)
Evaluation results
| Method | Links | |
|---|---|---|
| ZePAD (W-MSE)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=W-MSE, Backbone=ResNet-18, Setting=Semi-black-box2026.02 | 86.84 | |
| ZePAD (RESSL)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=RESSL, Backbone=ResNet-182026.02 | 84.58 | |
| ZePAD (DINO)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=DINO, Backbone=ResNet-182026.02 | 80.44 | |
| ZePAD (BYOL)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=BYOL, Backbone=ResNet-182026.02 | 80.18 | |
| ZePAD (NNCLR)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=NNCLR, Backbone=ResNet-182026.02 | 79.84 | |
| ZePAD (VibCreg)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=VibCreg, Backbone=ResNet-182026.02 | 79.62 | |
| ZePAD (SupCon)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=SupCon, Backbone=ResNet-182026.02 | 77.99 | |
| ZePAD (SimCLR)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=SimCLR, Backbone=ResNet-182026.02 | 74.7 | |
| ZePAD (SwAV)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=SwAV, Backbone=ResNet-182026.02 | 74.61 | |
| ZePAD (W-MSE)Pre-training Dataset (Dp)=ImageNet, SSL Encoder=W-MSE, Backbone=ResNet-182026.02 | 74 | |
| ZePAD (MoCo v3)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=MoCo v3, Backbone=ResNet-182026.02 | 72.49 | |
| ZePAD (MoCo v2+)Pre-training Dataset (Dp)=CIFAR10, SSL Encoder=MoCo v2+, Backbone=ResNet-182026.02 | 69.7 | |
| SimCLR (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=SimCLR, Backbone=ResNet-182026.02 | 47.13 | |
| MoCo v2+ (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=MoCo v2+, Backbone=ResNet-182026.02 | 45.01 | |
| MoCo v3 (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=MoCo v3, Backbone=ResNet-182026.02 | 36.86 | |
| SwAV (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=SwAV, Backbone=ResNet-182026.02 | 31.67 | |
| SupCon (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=SupCon, Backbone=ResNet-182026.02 | 16 | |
| VibCreg (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=VibCreg, Backbone=ResNet-182026.02 | 14.07 | |
| BYOL (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=BYOL, Backbone=ResNet-182026.02 | 12.47 | |
| RESSL (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=RESSL, Backbone=ResNet-182026.02 | 12.05 | |
| W-MSE (Baseline)Pre-training Dataset (Dp)=ImageNet, SSL Method=W-MSE, Backbone=ResNet-182026.02 | 11.78 | |
| NNCLR (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=NNCLR, Backbone=ResNet-182026.02 | 10.71 | |
| DINO (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=DINO, Backbone=ResNet-182026.02 | 10.46 | |
| W-MSE (Baseline)Pre-training Dataset (Dp)=CIFAR10, SSL Method=W-MSE, Backbone=ResNet-18, Setting=Semi-black-box2026.02 | 10.09 |