Continual Learning on CIFAR-100 (Training Time and Accuracy)
8.31Training Time per TaskFly-CL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Fly-CLBackbone=ViT-B/16, Task Sequence=Longer2025.10 | 8.31 | 1 | 94.22 | |
| Fly-CLBackbone=ResNet-502025.10 | 14.28 | 5.25 | 84.61 | |
| F-OALBackbone=ViT-B/16, Task Sequence=Longer2025.10 | 16.32 | 8.91 | 92.63 | |
| Fly-CLBackbone=ViT-B/16, Pre-trained=true2025.10 | 19.07 | 5.38 | 93.89 | |
| RanPACBackbone=ViT-B/16, Task Sequence=Longer2025.10 | 44.53 | 37.03 | 93.68 | |
| RanPACBackbone=ResNet-502025.10 | 55.68 | 46.65 | 82.72 | |
| F-OALBackbone=ViT-B/16, Pre-trained=true2025.10 | 71.05 | 57.13 | 91.96 | |
| F-OALBackbone=ResNet-502025.10 | 80.74 | 71.78 | 66.63 | |
| RanPACBackbone=ViT-B/16, Pre-trained=true2025.10 | 98.62 | 84.42 | 94.21 | |
| InfLoRABackbone=ViT-B/16, Task Sequence=Longer2025.10 | 124.8 | 84.36 | 88.18 | |
| SEMABackbone=ViT-B/16, Task Sequence=Longer2025.10 | 128.95 | 88.57 | 88.93 | |
| DualpromptBackbone=ViT-B/16, Task Sequence=Longer2025.10 | 130.12 | 91.05 | 83.42 | |
| MoE-AdapterBackbone=ViT-B/16, Task Sequence=Longer2025.10 | 132.67 | 92.31 | 88.42 | |
| L2PBackbone=ViT-B/16, Task Sequence=Longer2025.10 | 147.47 | 107.59 | 82.69 | |
| MoE-AdapterBackbone=ViT-B/16, Pre-trained=true2025.10 | 187.91 | 106.37 | 90.43 | |
| InfLoRABackbone=ViT-B/16, Pre-trained=true2025.10 | 220.82 | 140.31 | 91.1 | |
| DualPromptBackbone=ViT-B/16, Pre-trained=true2025.10 | 231.89 | 153.66 | 87.47 | |
| SEMABackbone=ViT-B/16, Pre-trained=true2025.10 | 241.6 | 160.87 | 92.04 | |
| L2PBackbone=ViT-B/16, Pre-trained=true2025.10 | 263.54 | 183.56 | 87.74 | |
| EASEBackbone=ViT-B/16, Task Sequence=Longer2025.10 | 350.92 | 331.76 | 90.14 | |
| EASEBackbone=ViT-B/16, Pre-trained=true2025.10 | 621.26 | 583.4 | 92.96 |