Model Efficiency on CIFAR-100 Image Classification
8.37GFLOPsVanilla
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VanillaBackbone=ResNet50, Architecture=Vanilla, Params (M)=1.242026.01 | 8.37 | — | 33.077 | |
| RecurrentBackbone=ResNet50, Architecture=Recurrent, Params (M)=0.112026.01 | 8.37 | — | 32.898 | |
| FROSTBackbone=ViT-B/16, Architecture=FROST, Params (M)=20.81, halting quantile (q)=0.52026.01 | 24.71 | — | 42.292 | |
| VanillaBackbone=ViT-B/16, Architecture=Vanilla, Params (M)=85.802026.01 | 35.13 | — | 21.484 | |
| RecurrentBackbone=ViT-B/16, Architecture=Recurrent, Params (M)=7.832026.01 | 35.13 | — | 21.976 | |
| FROSTBackbone=ResNet50, Architecture=FROST, Params (M)=1.49, halting quantile (q)=0.52026.01 | 39.11 | — | 47.183 | |
| BasicSSMBackbone=ViT-B/16, Architecture=BasicSSM, Params (M)=20.812026.01 | 65.45 | — | 15.986 | |
| BasicSSMBackbone=ResNet50, Architecture=BasicSSM, Params (M)=1.492026.01 | 82.77 | — | 23.334 |