Class-Incremental Learning on ImageNet-R 50 tasks
82.76Last AccuracyJoint Training
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Joint TrainingBackbone=ViT-B/16-IN21K2026.05 | 82.76 | — | |
| E2-LoRABackbone=ViT-B/16-IN21K2026.05 | 78.58 | 83.96 | |
| MOSBackbone=ViT-B/16-IN21K2026.05 | 77.29 | 83.06 | |
| TUNABackbone=ViT-B/16-IN21K2026.05 | 75.35 | 80.95 | |
| LORA-DRSBackbone=ViT-B/16-IN21K2026.05 | 72.12 | 77.94 | |
| EASEBackbone=ViT-B/16-IN21K2026.05 | 68.54 | 75.77 | |
| DualPromptBackbone=ViT-B/16-IN21K2026.05 | 61.5 | 68.63 | |
| InfLoRABackbone=ViT-B/16-IN21K2026.05 | 60.49 | 69.95 | |
| L2PBackbone=ViT-B/16-IN21K2026.05 | 55.89 | 62.98 | |
| CODA-PromptBackbone=ViT-B/16-IN21K2026.05 | 48.89 | 55.59 |