Class-Incremental Learning on ImageNet-A CIL T=10 (test)
62.4Avg AccuracyRanPAC
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RanPACAlgorithm=12023.07 | 62.4 | 21 | |
| RanPAC (No Phase 1)Phase 1=false2023.07 | 58.2 | — | |
| RanPAC (No RPs)RP layer=false2023.07 | 57.7 | — | |
| Joint linear probeJoint training=true2023.07 | 56.6 | — | |
| RanPAC (No RPs or Phase 1)RP layer=false, Phase 1=false2023.07 | 54.4 | — | |
| ADaM2023.07 | 52.6 | — | |
| RanPAC (NCM with Phase 1)Evaluation=NCM, Phase 1=true2023.07 | 49.7 | 11 | |
| RanPAC (NCM only)Evaluation=NCM, Phase 1=false2023.07 | 49.3 | 8 | |
| DualPrompt2023.07 | 45.4 | — | |
| CODA-Prompt2023.07 | 44.5 | — | |
| L2P2023.07 | 42.5 | — |