Class Incremental Learning on OmniBenchmark (OB) 10 Tasks
87.33AN (Avg Novelty/Accuracy)GDDSG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GDDSGBackbone=ViT-B/16, Pre-training=ImageNet-21K2025.02 | 87.33 | 0.96 | |
| RanPACBackbone=ViT-B/16, Pre-training=ImageNet-21K2025.02 | 78.86 | 7.62 | |
| CODA-PromptBackbone=ViT-B/16, Pre-training=ImageNet-21K2025.02 | 77.59 | 9.05 | |
| EASEBackbone=ViT-B/16, Pre-training=ImageNet-21K2025.02 | 74.56 | 8.6 | |
| ADAMBackbone=ViT-B/16, Pre-training=ImageNet-21K2025.02 | 72.84 | 8.23 | |
| SimpleCILBackbone=ViT-B/16, Pre-training=ImageNet-21K2025.02 | 72.15 | 8.31 | |
| L2PBackbone=ViT-B/16, Pre-training=ImageNet-21K2025.02 | 70.1 | 13.7 | |
| DualPromptBackbone=ViT-B/16, Pre-training=ImageNet-21K2025.02 | 69.73 | 11.91 | |
| FinetuneBackbone=ViT-B/16, Pre-training=ImageNet-21K2025.02 | 61.73 | 31.84 |