Image Classification on CIFAR-10 (test) (Accuracy, Sparsity, and Effective Rank)
91.8AccuracyPCD
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PCDBackbone=ResNet-34, priority tolerance (τ)=0.022026.06 | 91.8 | 95.2 | 30.9 | |
| PCDBackbone=Inception, priority tolerance (τ)=0.022026.06 | 89.2 | 86.1 | 31.3 | |
| AuxiNashBackbone=ResNet-342026.06 | 75.7 | 99 | 153.1 | |
| CAGradBackbone=ResNet-342026.06 | 64.8 | 98.5 | 159.6 | |
| WSBackbone=ResNet-342026.06 | 63.8 | 90.8 | 155.6 | |
| AuxiNashBackbone=Inception2026.06 | 60.3 | 99 | 18 | |
| WSBackbone=Inception2026.06 | 52.1 | 90.8 | 39.7 | |
| PCGradBackbone=ResNet-342026.06 | 51.4 | 93.1 | 166.8 | |
| FAMOBackbone=ResNet-342026.06 | 43.2 | 98.9 | 185.8 | |
| CAGradBackbone=Inception2026.06 | 38.2 | 99.1 | 51.2 | |
| PCGradBackbone=Inception2026.06 | 33.8 | 91.6 | 55.8 | |
| MGDABackbone=Inception2026.06 | 23.1 | 99.1 | 76.4 | |
| MGDABackbone=ResNet-342026.06 | 13.5 | 99.1 | 226.4 | |
| FAMOBackbone=Inception2026.06 | 10 | 99.2 | 100.3 |