Class-Incremental Learning on CIFAR-100 B0_Inc5
96.37Average AccuracyTOSCA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TOSCABackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 96.37 | 95.64 | |
| TOSCABackbone=ViT-B/16-IN1K2025.02 | 96.03 | 95.37 | |
| PAM (RN152)Backbone=ResNet-1522026.03 | 94.17 | 93.79 | |
| PAM (RN101)Backbone=ResNet-1012026.03 | 94.16 | 93.05 | |
| NoFA-BCBase classes=0, Incremental step size=52026.05 | 94.04 | 89.21 | |
| NoFA-BC2026.05 | 94.04 | 89.21 | |
| MoTE2026.05 | 93.78 | 88.98 | |
| MOSBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 93.45 | 90.04 | |
| RanPACBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 93.41 | 90.02 | |
| PAM (RN50)Backbone=ResNet-502026.03 | 93.06 | 92.5 | |
| MOSBackbone=ViT-B/16-IN1K2025.02 | 92.71 | 88.82 | |
| RanPACBackbone=ViT-B/16-IN1K2025.02 | 92.43 | 87.92 | |
| ACMapBase classes=0, Incremental step size=52026.05 | 92.04 | 87.81 | |
| EASEBase classes=0, Incremental step size=52026.05 | 91.51 | 85.8 | |
| PAM (RN18)Backbone=ResNet-182026.03 | 91.4 | 88.51 | |
| EASEBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 90.79 | 85.97 | |
| EASEBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 90.79 | 85.97 | |
| APERBase classes=0, Incremental step size=52026.05 | 90.65 | 85.15 | |
| SEMA2026.05 | 90.27 | 85.23 | |
| DS-ALBase classes=0, Incremental step size=52026.05 | 90.24 | 84.12 | |
| EASEBackbone=ViT-B/16-IN1K2025.02 | 89.94 | 84.39 | |
| ACILBase classes=0, Incremental step size=52026.05 | 89.88 | 83.56 | |
| APER-AdapterBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 89.57 | 84.91 | |
| APER-AdapterBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 89.57 | 84.91 | |
| CODABase classes=0, Incremental step size=52026.05 | 89.11 | 81.96 | |
| APER-AdapterBackbone=ViT-B/16-IN1K2025.02 | 88.46 | 83.16 | |
| DualPromptBase classes=0, Incremental step size=52026.05 | 87.87 | 81.15 | |
| CODAPromptBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 87.64 | 81.46 | |
| CODA-PromptBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 87.64 | 81.46 | |
| SimpleCILBase classes=0, Incremental step size=52026.05 | 87.57 | 81.26 | |
| CODAPromptBackbone=ViT-B/16-IN1K2025.02 | 86.67 | 80.68 | |
| SimpleCILBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 86.48 | 81.28 | |
| SimpleCILBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 86.48 | 81.28 | |
| HiDePromptBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 85.99 | 82.95 | |
| L2PBase classes=0, Incremental step size=52026.05 | 85.94 | 79.93 | |
| DualPromptBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 85.61 | 79.92 | |
| DualPromptBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 85.61 | 79.92 | |
| L2PBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 84.9 | 80.06 | |
| L2PBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 84.9 | 80.06 | |
| HiDePromptBackbone=ViT-B/16-IN1K2025.02 | 84.67 | 79.59 | |
| L2PBackbone=ViT-B/16-IN1K2025.02 | 83.37 | 78.64 | |
| DualPromptBackbone=ViT-B/16-IN1K2025.02 | 82.41 | 76.39 | |
| SimpleCILBackbone=ViT-B/16-IN1K2025.02 | 82.21 | 76.24 | |
| DER w/ C-FlatRegularization=false, Memory=false, Expansion=true, C-Flat=true2024.04 | 71.11 | — | |
| DERRegularization=false, Memory=false, Expansion=true, C-Flat=false2024.04 | 69.99 | — | |
| MEMO w/ C-FlatRegularization=false, Memory=false, Expansion=true, C-Flat=true2024.04 | 67.56 | — | |
| MEMORegularization=false, Memory=false, Expansion=true, C-Flat=false2024.04 | 67.42 | — | |
| WA w/ C-FlatRegularization=true, Memory=false, Expansion=false, C-Flat=true2024.04 | 65.7 | — | |
| FOSTER w/ C-FlatRegularization=false, Memory=false, Expansion=true, C-Flat=true2024.04 | 63.58 | — | |
| WARegularization=true, Memory=false, Expansion=false, C-Flat=false2024.04 | 63.36 | — | |
| FOSTERRegularization=false, Memory=false, Expansion=true, C-Flat=false2024.04 | 63.15 | — | |
| FinetuneBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | 60.65 | 48.12 | |
| FinetuneBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 60.65 | 48.12 | |
| Adapter-FTBase classes=0, Incremental step size=52026.05 | 60.51 | 49.32 | |
| Replay w/ C-FlatRegularization=true, Memory=true, Expansion=false, C-Flat=true2024.04 | 59.98 | — | |
| iCaRL w/ C-FlatRegularization=true, Memory=true, Expansion=false, C-Flat=true2024.04 | 59.13 | — | |
| ReplayRegularization=true, Memory=true, Expansion=false, C-Flat=false2024.04 | 58.83 | — | |
| iCaRLRegularization=true, Memory=true, Expansion=false, C-Flat=false2024.04 | 58.66 | — | |
| PODNet w/ C-FlatRegularization=true, Memory=true, Expansion=false, C-Flat=true2024.04 | 49.7 | — | |
| PODNetRegularization=true, Memory=true, Expansion=false, C-Flat=false2024.04 | 48.05 | — | |
| LwFBase classes=0, Incremental step size=52026.05 | 46.29 | 41.07 | |
| FinetuneBackbone=ViT-B/16-IN1K2025.02 | 44.4 | 39.7 | |
| FinetuneBase classes=0, Incremental step size=52026.05 | 38.9 | 20.17 | |
| JointBackbone=ViT-B/16-IN1K2025.02 | — | 95.88 | |
| JointBackbone=ViT-B/16-IN21K, Incremental Setting=B0 Inc52025.02 | — | 96.21 |