Class-Incremental Learning on CIFAR100 50 steps (test)
85.35Last AccuracySimE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SimEBackbone=CLIP ViT-L/14, Pre-trained Dataset=Laion-2B2026.03 | 85.35 | 91.64 | |
| ResNet18 Joint#P=11.22M2021.11 | 81.74 | — | |
| Transf. Joint#P=10.72M2021.11 | 76.12 | — | |
| MoE-Adapter(++)2026.05 | 75.24 | 83.6 | |
| iGSP2026.05 | 75.13 | 83.82 | |
| SimE2026.03 | 73.88 | 84.16 | |
| ZSCL2026.03 | 67.36 | 79.92 | |
| ZSCL2026.05 | 67.36 | 79.92 | |
| Continual-CLIP2026.03 | 66.72 | 76.49 | |
| CLIP2026.03 | 65.94 | 75.67 | |
| CLIP zero-shotmode=zero-shot2026.05 | 65.94 | 75.67 | |
| DER2026.03 | 59.76 | 72.05 | |
| DER2026.05 | 59.76 | 72.05 | |
| DyTox++#P=10.77M, variant=row 92021.11 | 59.51 | 75.53 | |
| LwF-VR2026.03 | 59.45 | 71.02 | |
| LwF-VR2026.05 | 59.45 | 71.02 | |
| DyTox+#P=10.77M, variant=row 82021.11 | 59.1 | 74.16 | |
| DER w/o P#P=561.39M2021.11 | 59.08 | 72.41 | |
| iCaRL2026.03 | 59.07 | 71.28 | |
| iCaRL2026.05 | 59.07 | 71.28 | |
| DyTox++#P=10.77M, variant=row 72021.11 | 58.76 | 75.45 | |
| DyTox+#P=10.77M, variant=row 62021.11 | 57.09 | 74.35 | |
| DyTox#P=10.77M2021.11 | 52.34 | 70.2 | |
| DyTox2026.05 | 51.09 | 68.9 | |
| DyTox2026.03 | 43.47 | 64.39 | |
| WA#P=11.22M2021.11 | 42.14 | 64.32 | |
| BiC2026.05 | 41.04 | 62.09 | |
| UCIR2026.03 | 37.09 | 56.86 | |
| UCIR2026.05 | 37.09 | 56.86 | |
| PODNet2026.05 | 32.99 | 51.19 | |
| LwF2026.03 | 32.9 | 47.69 | |
| LwF2026.05 | 32.9 | 47.69 | |
| Fine-tune2026.03 | 18.89 | 39.23 | |
| Fine-tune2026.05 | 18.89 | 39.23 |