Multi-Task Incremental Learning on MTIL
85.9Average AccuracyCHEEM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CHEEMBackbone=ViT-Base2023.03 | 85.9 | 1.7 | 62.3 | |
| LoRA-CLBackbone=ViT-Base2023.03 | 84.7 | 1.6 | 68.2 | |
| S-PromptsBackbone=ViT-Base2023.03 | 81.6 | 1.6 | 67.6 | |
| MoALBackbone=ViT-Base2023.03 | 78.3 | 3.7 | 33.7 | |
| DIKIBackbone=ViT-Base2023.03 | 76.4 | 2 | 42.5 | |
| CHEEMBackbone=DEiT-Tiny2023.03 | 74.5 | 1.9 | 4.5 | |
| LoRA-CLBackbone=DEiT-Tiny2023.03 | 71.1 | 1.9 | 4.5 | |
| TunaBackbone=ViT-Base2023.03 | 70.7 | 17.4 | 405 | |
| DIKIBackbone=DEiT-Tiny2023.03 | 67.6 | 1.8 | 2.8 | |
| S-PromptsBackbone=DEiT-Tiny2023.03 | 67.3 | 1.8 | 4.4 | |
| MoALBackbone=DEiT-Tiny2023.03 | 63.8 | 2.8 | 2.2 | |
| EWCBackbone=ViT-Base2023.03 | 44.6 | 23.8 | 33.7 | |
| TunaBackbone=DEiT-Tiny2023.03 | 44.5 | 30.2 | 26 | |
| CODA-PBackbone=ViT-Base2023.03 | 40.2 | 25.3 | 70.3 | |
| EWCBackbone=DEiT-Tiny2023.03 | 35.3 | 7.3 | 2.2 | |
| DualPBackbone=ViT-Base2023.03 | 33.8 | 22.1 | 70.3 | |
| DualPBackbone=DEiT-Tiny2023.03 | 30.9 | 17.5 | 5 | |
| L2PBackbone=ViT-Base2023.03 | 26.6 | 31 | 70.3 | |
| L2PBackbone=DEiT-Tiny2023.03 | 23.2 | 25.8 | 5.1 | |
| CODA-PBackbone=DEiT-Tiny2023.03 | 5.6 | 42.6 | 5 |