Continual Learning on CIFAR-100
92.4AccuracyDVC
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DVCReplay=true, Backbone=ViT, Pre-trained=true2024.03 | 92.4 | — | 8.2 | — | — | |
| SCRReplay=true, Backbone=ViT, Pre-trained=true2024.03 | 91.9 | — | 6.2 | — | — | |
| iCaRLReplay=true, Backbone=ViT, Pre-trained=true2024.03 | 91.6 | — | 3.2 | — | — | |
| EASEReplay=false, Backbone=ViT, Pre-trained=true2024.03 | 91.1 | — | 6.1 | — | — | |
| F-OALReplay=false, Backbone=ViT, Pre-trained=true2024.03 | 91.1 | — | 5.5 | — | — | |
| SLCAReplay=false, Backbone=ViT, Pre-trained=true2024.03 | 90.4 | — | 7.1 | — | — | |
| ERReplay=true, Backbone=ViT, Pre-trained=true2024.03 | 90.1 | — | 20.7 | — | — | |
| PCRReplay=true, Backbone=ViT, Pre-trained=true2024.03 | 89.1 | — | 9.2 | — | — | |
| ASERReplay=true, Backbone=ViT, Pre-trained=true2024.03 | 87.2 | — | 16.5 | — | — | |
| SEMAType=Expandable Param Size, Params (M)=0.6452024.03 | 86.98 | — | — | — | — | |
| CODA-PType=Expandable Param Size, Params (M)=3.9172024.03 | 86.11 | — | — | — | — | |
| ANCLScenario=Clean2023.11 | 81.5 | -0.1 | — | — | — | |
| DualPromptType=Expandable Param Size, Params (M)=1.0222024.03 | 80.43 | — | — | — | — | |
| ANCLepsilon=0.1, Scenario=Uniform2023.11 | 80.4 | -0.2 | — | — | — | |
| LAEReplay=false, Backbone=ViT, Pre-trained=true2024.03 | 79.1 | — | 11.8 | — | — | |
| L2PType=Fixed Param Size, Params (M)=0.1232024.03 | 77.87 | — | — | — | — | |
| RWALKScenario=Clean2023.11 | 70.1 | -6 | — | — | — | |
| LwFReplay=false, Backbone=ViT, Pre-trained=true2024.03 | 69.3 | — | 1.3 | — | — | |
| ANCLepsilon=0.3, Scenario=Uniform2023.11 | 68.5 | -3.6 | — | — | — | |
| RWALKepsilon=0.1, Scenario=Uniform2023.11 | 68.5 | -5.1 | — | — | — | |
| EWCScenario=Clean2023.11 | 68.3 | -5.2 | — | — | — | |
| MASepsilon=0.1, Scenario=Uniform2023.11 | 67.7 | -1.8 | — | — | — | |
| MASepsilon=0.3, Scenario=Uniform2023.11 | 67.5 | -9.6 | — | — | — | |
| EWCepsilon=0.1, Scenario=Uniform2023.11 | 67 | -5.1 | — | — | — | |
| AFECScenario=Clean2023.11 | 65.6 | -2.9 | — | — | — | |
| ANCLepsilon=0.1, Scenario=Reckless (BrainWash)2023.11 | 64.9 | -4.4 | — | — | — | |
| AFECepsilon=0.1, Scenario=Uniform2023.11 | 64.1 | -3.6 | — | — | — | |
| RWALKepsilon=0.1, Scenario=Reckless (BrainWash)2023.11 | 62.8 | -14.5 | — | — | — | |
| MASScenario=Clean2023.11 | 62.6 | -1.8 | — | — | — | |
| ANCLepsilon=0.1, Scenario=Cautious (BrainWash)2023.11 | 61.7 | -0.9 | — | — | — | |
| MASepsilon=0.1, Scenario=Cautious (BrainWash)2023.11 | 61.4 | -5.8 | — | — | — | |
| EWCepsilon=0.1, Scenario=Cautious (BrainWash)2023.11 | 58.8 | -9.5 | — | — | — | |
| EWCepsilon=0.3, Scenario=Uniform2023.11 | 57.5 | -12.2 | — | — | — | |
| AFECepsilon=0.1, Scenario=Cautious (BrainWash)2023.11 | 57.2 | -7.9 | — | — | — | |
| RWALKepsilon=0.1, Scenario=Cautious (BrainWash)2023.11 | 55.8 | -16.3 | — | — | — | |
| AFECepsilon=0.1, Scenario=Reckless (BrainWash)2023.11 | 55.4 | -8.8 | — | — | — | |
| ANCLepsilon=0.3, Scenario=Cautious (BrainWash)2023.11 | 53.8 | -6.2 | — | — | — | |
| MASepsilon=0.1, Scenario=Reckless (BrainWash)2023.11 | 53.7 | -6.4 | — | — | — | |
| RWALKepsilon=0.3, Scenario=Reckless (BrainWash)2023.11 | 53.5 | -21.6 | — | — | — | |
| AFECepsilon=0.3, Scenario=Uniform2023.11 | 52.4 | -14.4 | — | — | — | |
| EWCepsilon=0.1, Scenario=Reckless (BrainWash)2023.11 | 51 | -12.6 | — | — | — | |
| MASepsilon=0.3, Scenario=Cautious (BrainWash)2023.11 | 50.7 | -22.8 | — | — | — | |
| EWCReplay=false, Backbone=ViT, Pre-trained=true2024.03 | 49.9 | — | 67.4 | — | — | |
| AFECepsilon=0.3, Scenario=Cautious (BrainWash)2023.11 | 49 | -24.2 | — | — | — | |
| RWALKepsilon=0.3, Scenario=Uniform2023.11 | 48.8 | -25.5 | — | — | — | |
| RWALKepsilon=0.3, Scenario=Cautious (BrainWash)2023.11 | 47 | -32.5 | — | — | — | |
| MASepsilon=0.3, Scenario=Reckless (BrainWash)2023.11 | 45 | -17.9 | — | — | — | |
| ANCLepsilon=0.3, Scenario=Reckless (BrainWash)2023.11 | 44.3 | -30.4 | — | — | — | |
| EWCepsilon=0.3, Scenario=Cautious (BrainWash)2023.11 | 42.2 | -24.7 | — | — | — | |
| AFECepsilon=0.3, Scenario=Reckless (BrainWash)2023.11 | 36.9 | -24.6 | — | — | — | |
| RanDumbrehearsal-free=true2024.02 | 28.6 | — | — | — | — | |
| EWCepsilon=0.3, Scenario=Reckless (BrainWash)2023.11 | 25.5 | -29.1 | — | — | — | |
| PredKDrehearsal-free=true2024.02 | 24.6 | — | — | — | — | |
| PredKD + EWCrehearsal-free=true2024.02 | 23.3 | — | — | — | — | |
| PredKD + L2rehearsal-free=true2024.02 | 21.5 | — | — | — | — | |
| PredKD + FeatKDrehearsal-free=true2024.02 | 12.4 | — | — | — | — | |
| ACMapNumber of tasks (T)=10, Setting=Long-tailed step-imbalance2026.03 | — | — | — | 87.11 | 80.5 | |
| Aper-adapterNumber of tasks (T)=10, Setting=Long-tailed step-imbalance2026.03 | — | — | — | 85.24 | 78.91 | |
| AttriCLIP2024.03 | — | -0.128 | — | — | — | |
| CL-LoRANumber of tasks (T)=10, Setting=Long-tailed step-imbalance2026.03 | — | — | — | 86.02 | 76.17 | |
| CLAP4CLIPbase_method=Continual-CLIP2024.03 | — | -0.106 | — | — | — | |
| CLAP4CLIPbase_method=CoOp2024.03 | — | -0.129 | — | — | — | |
| CLAP4CLIPbase_method=MaPLe2024.03 | — | -0.105 | — | — | — | |
| CLAP4CLIPbase_method=AttriCLIP2024.03 | — | -0.143 | — | — | — | |
| Continual-CLIP2024.03 | — | -0.086 | — | — | — | |
| CoOp2024.03 | — | -0.257 | — | — | — | |
| DualPromptBackbone=ViT-B/162025.10 | — | -7.06 | — | 81.45 | — | |
| EASEBackbone=ViT-B/162025.10 | — | -4.66 | — | 87.63 | — | |
| EASENumber of tasks (T)=10, Setting=Long-tailed step-imbalance2026.03 | — | — | — | 88.5 | 81.2 | |
| F-OALBackbone=ViT-B/162025.10 | — | -4.9 | — | 87.04 | — | |
| Fly-CLBackbone=ViT-B/162025.10 | — | -3.82 | — | 89.85 | — | |
| InfLoRABackbone=ViT-B/162025.10 | — | -5.07 | — | 86.56 | — | |
| InfLoRANumber of tasks (T)=10, Setting=Long-tailed step-imbalance2026.03 | — | — | — | 79.81 | 71.69 | |
| L2PBackbone=ViT-B/162025.10 | — | -6.23 | — | 83.55 | — | |
| MaPLe2024.03 | — | -0.209 | — | — | — | |
| MoE-AdapterBackbone=ViT-B/162025.10 | — | -4.97 | — | 86.88 | — | |
| One-ANumber of tasks (T)=10, Setting=Long-tailed step-imbalance2026.03 | — | — | — | 86.65 | 81.05 | |
| RanPACBackbone=ViT-B/162025.10 | — | -3.47 | — | 90.83 | — | |
| SEMABackbone=ViT-B/162025.10 | — | -4.7 | — | 87.47 | — | |
| SEMANumber of tasks (T)=10, Setting=Long-tailed step-imbalance2026.03 | — | — | — | 74.98 | 64.41 |