Class-Incremental Learning on VTAB5T small Uniformly Abrupt
1.85Average ForgettingRanPAC
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RanPACScenario=Uniformly Abrupt, Backbone=ViT-B/16, Pre-trained=ImageNet-21k fine-tuned on ImageNet-1K2026.01 | 1.85 | 1 | 2.5 | |
| PROTEUSScenario=Uniformly Abrupt, Backbone=ViT-B/16, Pre-trained=ImageNet-21k fine-tuned on ImageNet-1K, LoRA rank r=162026.01 | 4.12 | 2 | 2.17 | |
| SPromptScenario=Uniformly Abrupt, Backbone=ViT-B/16, Pre-trained=ImageNet-21k fine-tuned on ImageNet-1K2026.01 | 6.24 | 3 | 2.67 | |
| CODAPromptScenario=Uniformly Abrupt, Backbone=ViT-B/16, Pre-trained=ImageNet-21k fine-tuned on ImageNet-1K2026.01 | 6.55 | 4 | 3.17 | |
| DualPromptScenario=Uniformly Abrupt, Backbone=ViT-B/16, Pre-trained=ImageNet-21k fine-tuned on ImageNet-1K2026.01 | 9.17 | 5 | 4.83 | |
| L2PScenario=Uniformly Abrupt, Backbone=ViT-B/16, Pre-trained=ImageNet-21k fine-tuned on ImageNet-1K2026.01 | 12.3 | 6 | 5.67 |