Class-incremental learning on ImageNet-R T=10
75.78AccuracyJanus-LoRA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Janus-LoRABackbone=ViT-B/162026.05 | 75.78 | 81.64 | 6.05 | |
| BiLoRABackbone=ViT-B/162026.05 | 73.93 | 77.92 | 2.07 | |
| InfLoRABackbone=ViT-B/162026.05 | 73.9 | 80.01 | 6.33 | |
| LoRA-DRSBackbone=ViT-B/162026.05 | 72.58 | 78.34 | 3.72 | |
| Coda-PromptBackbone=ViT-B/162026.05 | 71.59 | 76.72 | 4.48 | |
| LoRA-GPMBackbone=ViT-B/162026.05 | 71.56 | 79.34 | 10.95 | |
| L2PBackbone=ViT-B/162026.05 | 65.99 | 72.7 | 6.95 | |
| Dual-PromptBackbone=ViT-B/162026.05 | 64.45 | 69.23 | 4.45 | |
| Fine-TuningBackbone=ViT-B/162026.05 | 64.24 | 74.68 | 21.85 |