Class-incremental learning on TinyImageNet 100 base classes (20-step)
80.01Average AccuracyiGSP
Evaluation results
| Method | Links | ||
|---|---|---|---|
| iGSP2026.05 | 80.01 | 76 | |
| MoE-Adapter(++)2026.05 | 79.96 | 75.77 | |
| ZSCL2026.05 | 77.18 | 68.3 | |
| LwF-VR2026.05 | 69.94 | 63.89 | |
| iCaRL2026.05 | 69.65 | 64.68 | |
| CLIPEvaluation protocol=zero-shot2026.05 | 69.49 | 65.3 | |
| LwF2026.05 | 54.79 | 42.26 | |
| Fine-tune2026.05 | 54.62 | 44.55 | |
| DyTox+2026.05 | 46.18 | 36.21 | |
| UCIR2026.05 | 42.84 | 30.85 | |
| PASS2026.05 | 42.01 | 32.93 | |
| EEIL2026.05 | 40.41 | 29.72 | |
| MUC2026.05 | 21.89 | 10.32 | |
| EWC2026.05 | 12.35 | 4.73 |