Learning-to-Defer Classification on CIFAR-10
82.67Confident Accuracy (C.Acc)Mao et al. (2024c)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Mao et al. (2024c)Classifier=ResNet-4, Experts=3 ResNet-16 models, Learning Rate=0.005, Attack=PGD (epsilon = 0.03137)2025.10 | 82.67 | 27.47 | 19.8 | 0.75 | |
| RERM-CClassifier=ResNet-4, Experts=3 ResNet-16 models, Learning Rate=0.01, Attack=PGD (epsilon = 0.03137), rho=1.0, nu=0.002, loss=logistic (u = 1)2025.10 | 75.6 | 52 | 68.67 | 0.52 |