Image Classification on ImageNet-C 1.0 (test)
68.8Accuracy (Average)STL (FAN-L-Hybrid)
Evaluation results
| Method | Links | ||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| STL (FAN-L-Hybrid)Params=77M, FLOPs=17.0G2024.01 | 68.8 | — | — | — | — | — | — | — | 81.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FAN-L-HybridParams=77M, FLOPs=16.9G2024.01 | 68.3 | — | — | — | — | — | — | — | 81 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STL (FAN-B-Hybrid)Params=51M, FLOPs=11.4G2024.01 | 68.2 | — | — | — | — | — | — | — | 80.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FAN-B-HybridParams=50M, FLOPs=11.3G2024.01 | 66.4 | — | — | — | — | — | — | — | 79.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STL (FAN-S-Hybrid)Params=27M, FLOPs=6.8G2024.01 | 65.5 | — | — | — | — | — | — | — | 78.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FAN-S-HybridParams=26M, FLOPs=6.7G2024.01 | 64.7 | — | — | — | — | — | — | — | 77.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ConvNeXt-BParams=89M, FLOPs=15.4G2024.01 | 63 | — | — | — | — | — | — | — | 75.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeiT-BParams=89M, FLOPs=17.6G2024.01 | 62.7 | — | — | — | — | — | — | — | 76.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| No AdaptationBackbone=ViT-B/16, Quantization=Dynamic2026.07 | 62.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | 30.2 | 88.3 | 65.9 | 61.8 | 57.4 | 49.4 | 70.6 | 70.7 | 69.2 | 48.8 | 64.5 | 60.7 | 70.5 | 62.7 | 69.1 | — | |
| LV-VIT-MParams=56M, FLOPs=16.0G2024.01 | 62 | — | — | — | — | — | — | — | 73.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ConvNeXt-SParams=50M, FLOPs=8.7G2024.01 | 61.7 | — | — | — | — | — | — | — | 74.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Pseudo-labelScenario=CTTA, Reference=Lee (2013)2025.12 | 61.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | 98.5 | 99.9 | 51.3 | 99.9 | 93.8 | 45.7 | 45.2 | 53.9 | 41.6 | 99.6 | 45.6 | 98.9 | 40.4 | 40.4 | 47.7 | -5.4 | |
| Swin-SParams=50M, FLOPs=8.7G2024.01 | 60.4 | — | — | — | — | — | — | — | 72.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-BParams=88M, FLOPs=15.4G2024.01 | 60.4 | — | — | — | — | — | — | — | 72.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LV-VIT-SParams=26M, FLOPs=6.6G2024.01 | 59.7 | — | — | — | — | — | — | — | 71.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CoLABackbone=ViT-B/16, Quantization=Dynamic2026.07 | 59.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | 31.2 | 81 | 63.1 | 58.5 | 50.2 | 51.7 | 69.3 | 62.9 | 67.6 | 45.3 | 59.3 | 56.7 | 68.6 | 61.3 | 66.6 | — | |
| ConvNeXt-TParams=29M, FLOPs=4.5G2024.01 | 59.1 | — | — | — | — | — | — | — | 71.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CDC-TTA (Ours)Backbone=ViT-B/16, Quantization=Dynamic2026.07 | 58.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | 31.3 | 82.1 | 62.6 | 56 | 49.8 | 51.2 | 68.5 | 62.8 | 66.3 | 45.3 | 57.4 | 55.4 | 67.1 | 60 | 65.9 | — | |
| STL (FAN-T-Hybrid)Params=8M, FLOPs=3.6G2024.01 | 58.2 | — | — | — | — | — | — | — | 72.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeiT-SParams=22M, FLOPs=4.6G2024.01 | 58.1 | — | — | — | — | — | — | — | 72.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FAN-T-HybridParams=7M, FLOPs=3.5G2024.01 | 57.4 | — | — | — | — | — | — | — | 71.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SourceScenario=CTTA, Reference=Dosovitskiy et al. (2021)2025.12 | 55.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | 26.4 | 91.4 | 68.5 | 57.5 | 57.7 | 54.2 | 53 | 78.8 | 52.1 | 36.2 | 58.5 | 38 | 51.8 | 49.9 | 63.3 | 0 | |
| Swin-TParams=28M, FLOPs=4.5G2024.01 | 55.4 | — | — | — | — | — | — | — | 68.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| M2AScenario=CTTA, Mode=Frequency2025.12 | 55.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | 26 | 62.4 | 62.3 | 76 | 40.9 | 61.5 | 45.2 | 53.2 | 43.6 | 59.9 | 51.4 | 75 | 42 | 51.8 | 78.7 | 0.5 | |
| CoTTAScenario=CTTA, Reference=Wang et al. (2022)2025.12 | 54.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | 25.9 | 90 | 68.3 | 56.4 | 55.3 | 52.7 | 52.9 | 78.1 | 51.4 | 35.2 | 57.1 | 36.4 | 51.6 | 48.2 | 62 | 1 | |
| PETALScenario=CTTA, Reference=Brahma & Rai (2023)2025.12 | 52.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | 26.4 | 91.3 | 66.8 | 50.7 | 52.7 | 47.2 | 52.1 | 74 | 47.5 | 32 | 56.7 | 32.3 | 48.2 | 46.8 | 59.7 | 3.5 | |
| PVTV2-B1Params=13M, FLOPs=2.1G2024.01 | 51.7 | — | — | — | — | — | — | — | 65.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LV-VIT-TParams=9M, FLOPs=2.1G2024.01 | 51.6 | — | — | — | — | — | — | — | 65.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TentScenario=CTTA, Reference=Wang et al. (2021)2025.12 | 51 | — | — | — | — | — | — | — | — | — | — | — | — | — | 23.9 | 72.8 | 65.8 | 55.7 | 50.3 | 47.7 | 52.2 | 73 | 49.2 | 33.9 | 54.5 | 34.4 | 48.9 | 44 | 58.4 | 4.8 | |
| ResNet50Params=25M, FLOPs=4.1G2024.01 | 50.6 | — | — | — | — | — | — | — | 64.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VDPScenario=CTTA, Reference=Gan et al. (2023)2025.12 | 50 | — | — | — | — | — | — | — | — | — | — | — | — | — | 23.6 | 70.4 | 58.1 | 54.9 | 45.8 | 46.1 | 52.7 | 70.2 | 50.1 | 36.1 | 56.1 | 34.5 | 51.6 | 42.1 | 58.1 | 5.8 | |
| MAE-TTT with entropy minimizationtraining stages=two-stage, batch size=322025.01 | 47.77 | — | — | — | — | — | — | — | — | — | — | — | — | — | 73.8 | 14 | 33.6 | 69 | 47.8 | 64.6 | 38.6 | 42.2 | 36.6 | 68.4 | 32.4 | 67.4 | 41.2 | 51.2 | 35.4 | — | |
| TTT-MAE2025.01 | 45.92 | — | — | — | — | — | — | — | — | — | — | — | — | — | 69.1 | 9.8 | 34.4 | 50.7 | 44.7 | 50.7 | 30.5 | 36.9 | 32.4 | 63 | 41.9 | 63 | 33 | 42.8 | 45.9 | — | |
| SARScenario=CTTA, Reference=Niu et al. (2023)2025.12 | 45.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | 23.8 | 57.2 | 58.2 | 49.9 | 41.8 | 44.1 | 49.3 | 60.9 | 44.9 | 32.7 | 46.1 | 32.9 | 43.8 | 41.3 | 51.8 | 10.6 | |
| ViDAScenario=CTTA, Reference=Liu et al. (2024b)2025.12 | 43.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | 24.3 | 52.8 | 52.2 | 49.1 | 40.7 | 42.7 | 47.7 | 56.9 | 42.9 | 33.1 | 45.5 | 33.5 | 42.5 | 38.9 | 48.9 | 12.4 | |
| Continual-MAEScenario=CTTA, Reference=Liu et al. (2024a)2025.12 | 42.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | 23.4 | 60.3 | 51.4 | 37.5 | 64.3 | 35.8 | 46.3 | 54.9 | 42.5 | 31.4 | 43.3 | 29.2 | 41.9 | 34.2 | 40.7 | 13.3 | |
| EffiNet-B0Params=5M, FLOPs=0.4G2024.01 | 41.1 | — | — | — | — | — | — | — | 53 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| REMScenario=CTTA, Reference=Han et al. (2025)2025.12 | 39.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | 23.2 | 46.8 | 53.2 | 41.6 | 35.4 | 35.2 | 43.5 | 49 | 39.2 | 30.2 | 43.5 | 28.9 | 38.1 | 37.5 | 42.8 | 16.6 | |
| M2AScenario=CTTA, Mode=Spatial2025.12 | 39.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | 22.5 | 47.4 | 54.8 | 40.7 | 35 | 35.9 | 43.4 | 49.4 | 39.4 | 29.9 | 42.9 | 28.5 | 38.3 | 36.1 | 43.1 | 16.6 | |
| SOTTAMode=Online TTA, Batch Size=64, Memory Size=64, BN Momentum=0.2, Learning Rate=0.001, Adaptation Epoch=1, Confidence Threshold=0.332023.10 | 36.8 | — | — | 39.8 | 27.9 | 36.1 | 41.1 | 39 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PVTV2-B0Params=3M, FLOPs=0.6G2024.01 | 36.2 | — | — | — | — | — | — | — | 51.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SupervisedScenario=Fully Supervised Baseline2025.12 | 35.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | 21.5 | 43.3 | 46.4 | 34.4 | 32.5 | 33.1 | 42.5 | 44 | 37.1 | 27.5 | 37.4 | 26.1 | 36.9 | 34.2 | 38.3 | 20.1 | |
| MBV2Params=4M, FLOPs=0.4G2024.01 | 35 | — | — | — | — | — | — | — | 47.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Fine-Tune2025.01 | 34.45 | — | — | — | — | — | — | — | — | — | — | — | — | — | 67.5 | 7.8 | 33.9 | 32.4 | 36.4 | 38.2 | 22 | 15.7 | 23.9 | 51.2 | 37.4 | 51.9 | 23.7 | 37.6 | 37.1 | — | |
| ResNet18Params=11M, FLOPs=1.8G2024.01 | 32.7 | — | — | — | — | — | — | — | 46.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT Probe2025.01 | 32.06 | — | — | — | — | — | — | — | — | — | — | — | — | — | 68.3 | 6.4 | 24.2 | 31.6 | 38.6 | 38.4 | 17.4 | 18.4 | 18.2 | 51.2 | 32.2 | 49.7 | 18.2 | 35.9 | 32.2 | — | |
| ROTTAMode=Online TTA, Batch Size=64, Memory Size=642023.10 | 29.5 | — | — | 29.7 | 25.6 | 29.2 | 32 | 31.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SARMode=Online TTA, Batch Size=64, Sharpness Threshold=0.052023.10 | 27.2 | — | — | 36.1 | 27.6 | 23.5 | 26.8 | 22 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Joint Train2025.01 | 26.88 | — | — | — | — | — | — | — | — | — | — | — | — | — | 62.3 | 4.5 | 26.7 | 39.9 | 25.7 | 30 | 5.8 | 16.3 | 5.8 | 45.3 | 30.9 | 45.9 | 7.1 | 25.1 | 31.8 | — | |
| EATAMode=Online TTA, Batch Size=642023.10 | 26.7 | — | — | 38 | 25.6 | 23.1 | 26.1 | 20.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| COTTAMode=Online TTA, Batch Size=642023.10 | 23.4 | — | — | 32.2 | 23.3 | 17.6 | 28.3 | 16 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BN StatsMode=Online TTA, Batch Size=642023.10 | 18.2 | — | — | 27.1 | 18.9 | 14.8 | 17.4 | 12.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TENTMode=Online TTA, Batch Size=642023.10 | 18.2 | — | — | 27.1 | 18.9 | 14.8 | 17.4 | 12.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SourceTTA=None2023.10 | 14.6 | — | — | 14.6 | 14.6 | 14.6 | 14.6 | 14.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LAMEMode=Online TTA, Batch Size=642023.10 | 14.3 | — | — | 14.4 | 14.4 | 14.4 | 14 | 14.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PLMode=Online TTA, Batch Size=642023.10 | 12.8 | — | — | 30.5 | 6.9 | 5.1 | 18.1 | 3.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AdapterBackbone=ViT-B/162023.12 | — | — | 42.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AdvpropBackbone=ViT-B/16, Resolution=224x2242023.11 | — | — | 51.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AdvpropBackbone=ResNet50, Resolution=224x2242023.11 | — | — | 70.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AdvPropExtra Data=false2021.11 | — | — | 52.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AnchoringModel=SWINv2-T, Number of Parameters=28.4M2024.06 | — | — | — | — | — | — | — | — | — | 72.32 | 65.85 | 58.95 | 49.51 | 37.41 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AnchoringModel=SWINv2-S, Number of Parameters=49.7M2024.06 | — | — | — | — | — | — | — | — | — | 74.75 | 68.87 | 63.12 | 54.7 | 43.14 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AnchoringModel=VITb16, Number of Parameters=86.6M2024.06 | — | — | — | — | — | — | — | — | — | 72.64 | 67.14 | 62.33 | 54.46 | 43.48 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AnchoringModel=SWINv2-B, Number of Parameters=87.8M2024.06 | — | — | — | — | — | — | — | — | — | 74.95 | 69.28 | 63.43 | 55.08 | 43.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AugMax-DuBINBackbone=ResNet182021.10 | — | 35.01 | 82.56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AugMixBackbone=ResNet182021.10 | — | 34.58 | 83.08 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AugReg-ViTBackbone=ViT-B/16, Resolution=224x2242023.11 | — | — | 54.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BATCLIPBackbone=ViT-B-32, Evaluation Protocol=Test-Time Adaptation2026.01 | — | — | 27.37 | — | — | — | — | — | — | — | — | — | — | — | 49.56 | 20.56 | 24.56 | 27.62 | 35.76 | 27.02 | 17.64 | 20.4 | 16.98 | 34.26 | 28.46 | 35.24 | 18.6 | 29.16 | 24.66 | — | |
| Bias TuningBackbone=ViT-B/162023.12 | — | — | 41.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CLIPBackbone=ViT-B/16, Resolution=224x2242023.11 | — | — | 60.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DADBackbone=ViT-B/16, Resolution=224x224, Teacher Model=ViT-L/142023.11 | — | — | 53.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DADBackbone=ResNet50, Resolution=224x224, Teacher Model=ViT-L/142023.11 | — | — | 67.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DATBackbone=ViT-B/16, Resolution=224x224, Teacher Model=ViT-L/142023.11 | — | — | 44.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DATBackbone=ResNet50, Resolution=224x224, Teacher Model=ViT-L/142023.11 | — | — | 74.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DAT + DADBackbone=ViT-B/16, Resolution=224x224, Teacher Model=ViT-L/142023.11 | — | — | 47.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DAT + DADBackbone=ResNet50, Resolution=224x224, Teacher Model=ViT-L/142023.11 | — | — | 65.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DebiasedBackbone=ViT-B/16, Resolution=224x2242023.11 | — | — | 49.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DebiasedBackbone=ResNet50, Resolution=224x2242023.11 | — | — | 67.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepAugment + AugMax-DuBINBackbone=ResNet182021.10 | — | 46.55 | 68.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepAugment + AugMixBackbone=ResNet182021.10 | — | 45.84 | 69.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepAugment+AugMixExtra Data=false2021.11 | — | — | 53.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Discrete ViTExtra Data=false2021.11 | — | — | 46.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Discrete ViTExtra Data=IM-21k2021.11 | — | — | 38.74 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DMNBackbone=ViT-B-32, Evaluation Protocol=Test-Time Adaptation2026.01 | — | — | 22.96 | — | — | — | — | — | — | — | — | — | — | — | 50.5 | 14.78 | 22.92 | 19.92 | 29.62 | 25.72 | 13 | 11.78 | 14.5 | 29.36 | 22.18 | 32.46 | 14.32 | 23.7 | 19.68 | — | |
| ECALPBackbone=ViT-B-32, Evaluation Protocol=Test-Time Adaptation2026.01 | — | — | 23.26 | — | — | — | — | — | — | — | — | — | — | — | 47.92 | 18.26 | 23.28 | 19.12 | 29.5 | 25.38 | 14.4 | 12.18 | 14.68 | 29 | 23.18 | 31.42 | 15.1 | 24.4 | 21.08 | — | |
| Fast AdvpropBackbone=ViT-B/16, Resolution=224x2242023.11 | — | — | 53.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Full TuningBackbone=ViT-B/162023.12 | — | — | 46.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPSBackbone=ViT-B/162023.12 | — | — | 42.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Linear ProbingBackbone=ViT-B/162023.12 | — | — | 46.91 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MEMOBackbone=ViT-B-32, Evaluation Protocol=Test-Time Adaptation2026.01 | — | — | 23.25 | — | — | — | — | — | — | — | — | — | — | — | 50.5 | 17.32 | 24.66 | 19.02 | 30.28 | 26 | 12.84 | 11.88 | 12.78 | 29.14 | 22.82 | 32.48 | 13 | 25.7 | 20.4 | — | |
| MINTBackbone=ViT-B-32, Evaluation Protocol=Test-Time Adaptation2026.01 | — | — | 26.84 | — | — | — | — | — | — | — | — | — | — | — | 50.64 | 20.18 | 24.86 | 24.04 | 32.72 | 27.28 | 18.84 | 19.08 | 19.42 | 33.14 | 27.2 | 34.84 | 19.98 | 27.94 | 22.44 | — | |
| NormalBackbone=ResNet182021.10 | — | 30.91 | 87.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ProposedModel=SWINv2-T, Number of Parameters=28.4M2024.06 | — | — | — | — | — | — | — | — | — | 72.13 | 65.71 | 59.21 | 50.01 | 38.58 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ProposedModel=SWINv2-S, Number of Parameters=49.7M2024.06 | — | — | — | — | — | — | — | — | — | 74.7 | 69.12 | 63.65 | 55.5 | 44.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ProposedModel=VITb16, Number of Parameters=86.6M2024.06 | — | — | — | — | — | — | — | — | — | 72.52 | 67.38 | 62.87 | 55.13 | 44.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ProposedModel=SWINv2-B, Number of Parameters=87.8M2024.06 | — | — | — | — | — | — | — | — | — | 75.24 | 69.63 | 64.05 | 56.08 | 45.19 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Pyramid ATBackbone=ViT-B/16, Resolution=224x224, Teacher Model=ViT-L/142023.11 | — | — | 45 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Pyramid ATBackbone=ResNet50, Resolution=224x2242023.11 | — | — | 76.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PyramidATExtra Data=false, Backbone=ViT-B/162021.11 | — | — | 41.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PyramidATExtra Data=IM-21k, Backbone=ViT-B/162021.11 | — | — | 36.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |