Class-Incremental Learning on TinyImageNet 20 Steps (test)
86.39Average AccuracySimE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SimEBackbone=CLIP ViT-L/14, Pre-trained Dataset=Laion-2B2026.03 | 86.39 | 83.33 | |
| SimE2026.03 | 79.29 | 75.37 | |
| ZSCL2026.03 | 77.18 | 68.3 | |
| Continual-CLIP2026.03 | 70.51 | 66.43 | |
| LwF-VR2026.03 | 69.94 | 63.89 | |
| iCaRL2026.03 | 69.65 | 64.68 | |
| CLIP2026.03 | 69.49 | 65.3 | |
| LwF2026.03 | 54.79 | 42.26 | |
| Fine-tune2026.03 | 54.62 | 44.55 | |
| DyTox2026.03 | 46.18 | 36.21 | |
| UCIR2026.03 | 42.84 | 30.85 | |
| PASS2026.03 | 42.01 | 32.93 |