Domain Generalization on Office-Home (test)
92.9Average AccuracyText-Anchored Information Bottleneck
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Text-Anchored Information BottleneckBackbone=Swin Transformer, Pre-training source=IN-21K2026.07 | 92.9 | — | — | — | — | — | — | — | — | |
| Text-Anchored Information BottleneckBackbone=DINOv2, Pre-training source=LVD-142M2026.07 | 90.2 | — | — | — | — | — | — | — | — | |
| Text-Anchored Information BottleneckBackbone=DeiT, Pre-training source=IN-1K2026.07 | 88.7 | — | — | — | — | — | — | — | — | |
| Text-Anchored Information BottleneckBackbone=RegNetY-16GF, Pre-training source=IG-3B2026.07 | 88.4 | — | — | — | — | — | — | — | — | |
| Text-Anchored Information BottleneckBackbone=RegNetY-16GF, Pre-training source=IN-12K2026.07 | 87.2 | — | — | — | — | — | — | — | — | |
| LPBackbone=Swin Transformer, Pre-training source=IN-21K2026.07 | 86.6 | — | — | — | — | — | — | — | — | |
| MetaPromptFew-shot=5-shot, CLIP=true2022.12 | 85.1 | — | — | — | — | — | — | — | — | |
| MIROBackbone=DINOv2, Pre-training source=LVD-142M2026.07 | 85.1 | — | — | — | — | — | — | — | — | |
| VL2VBackbone=DINOv2, Pre-training source=LVD-142M2026.07 | 85.1 | — | — | — | — | — | — | — | — | |
| DAMPEvaluation Protocol=Prompt Learning, Backbone=ViT-B/162024.03 | 85 | — | — | — | — | — | — | — | — | |
| VL2VBackbone=Swin Transformer, Pre-training source=IN-21K2026.07 | 84.8 | — | — | — | — | — | — | — | — | |
| LPBackbone=DINOv2, Pre-training source=LVD-142M2026.07 | 84.5 | — | — | — | — | — | — | — | — | |
| DPLEvaluation Protocol=Prompt Learning, Backbone=ViT-B/162024.03 | 84.2 | — | — | — | — | — | — | — | — | |
| CoCoOpFew-shot=5-shot, CLIP=true2022.12 | 84.1 | — | — | — | — | — | — | — | — | |
| MetaPromptFew-shot=1-shot, CLIP=true2022.12 | 84 | — | — | — | — | — | — | — | — | |
| VL2VBackbone=RegNetY-16GF, Pre-training source=IG-3B2026.07 | 84 | — | — | — | — | — | — | — | — | |
| EoAEnsemble=true, CLIP=true2022.12 | 83.9 | — | — | — | — | — | — | — | — | |
| VL2VBackbone=RegNetY-16GF, Pre-training source=IN-12K2026.07 | 83.6 | — | — | — | — | — | — | — | — | |
| MIROBackbone=Swin Transformer, Pre-training source=IN-21K2026.07 | 83.5 | — | — | — | — | — | — | — | — | |
| CLIPoodBackbone=RegNetY-16GF, Pre-training source=IG-3B2026.07 | 83.3 | — | — | — | — | — | — | — | — | |
| RISEBackbone=Swin Transformer, Pre-training source=IN-21K2026.07 | 82.8 | — | — | — | — | — | — | — | — | |
| CLIP + DPLBackbone=CLIP ViT-B16, Test-time batch size=642021.11 | 82.3 | — | — | — | — | — | — | — | — | |
| CLIP (Teacher)Backbone=ViT B/16, Ens/MA=No, Inference Mode=Zero-shot2023.09 | 82.3 | — | — | — | — | — | — | — | — | |
| CLIPEvaluation Protocol=Zero-shot, Backbone=ViT-B/162024.03 | 82.3 | — | — | — | — | — | — | — | — | |
| CLIP (template)Zero-shot=true, CLIP=true2022.12 | 82.1 | — | — | — | — | — | — | — | — | |
| CLIPoodBackbone=DINOv2, Pre-training source=LVD-142M2026.07 | 81.6 | — | — | — | — | — | — | — | — | |
| RISEBackbone=RegNetY-16GF, Pre-training source=IG-3B2026.07 | 81.5 | — | — | — | — | — | — | — | — | |
| LPBackbone=RegNetY-16GF, Pre-training source=IG-3B2026.07 | 81.3 | — | — | — | — | — | — | — | — | |
| CLIPoodBackbone=RegNetY-16GF, Pre-training source=IN-12K2026.07 | 81.1 | — | — | — | — | — | — | — | — | |
| CLIPoodBackbone=Swin Transformer, Pre-training source=IN-21K2026.07 | 80.9 | — | — | — | — | — | — | — | — | |
| ERM + SHOTBackbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=SHOT2021.11 | 80.7 | — | — | — | — | — | — | — | — | |
| CORAL + SHOTBackbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=SHOT2021.11 | 80.7 | — | — | — | — | — | — | — | — | |
| CORAL + SHOT**Backbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=SHOT, Update target=feature extractor2021.11 | 80.7 | — | — | — | — | — | — | — | — | |
| SEDGEEnsemble=true, CLIP=true2022.12 | 80.7 | — | — | — | — | — | — | — | — | |
| VL2VBackbone=DeiT, Pre-training source=IN-1K2026.07 | 80.5 | — | — | — | — | — | — | — | — | |
| MIROBackbone=RegNetY-16GF, Pre-training source=IG-3B2026.07 | 80.4 | — | — | — | — | — | — | — | — | |
| CORAL + Pseudo LabelBackbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=Pseudo Label2021.11 | 79.8 | — | — | — | — | — | — | — | — | |
| ERM + SHOT**Backbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=SHOT, Update target=feature extractor2021.11 | 79.4 | — | — | — | — | — | — | — | — | |
| CLIPZero-shot=true, CLIP=true2022.12 | 79.4 | — | — | — | — | — | — | — | — | |
| CORAL + T3ABackbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=T3A2021.11 | 79.1 | — | — | — | — | — | — | — | — | |
| CORALEvaluation Protocol=Fine-tuning (CLIP), Backbone=ViT-B/162024.03 | 78.9 | — | — | — | — | — | — | — | — | |
| CORAL + Pseudo Label*Backbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=Pseudo Label, Update target=linear classifier2021.11 | 78.8 | — | — | — | — | — | — | — | — | |
| DANNEvaluation Protocol=Fine-tuning (CLIP), Backbone=ViT-B/162024.03 | 78.8 | — | — | — | — | — | — | — | — | |
| CORALBackbone=CLIP ViT-B16, Test-time batch size=642021.11 | 78.7 | — | — | — | — | — | — | — | — | |
| CORAL + Tent*Backbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=Tent, Update target=linear classifier2021.11 | 78.6 | — | — | — | — | — | — | — | — | |
| ERMBackbone=CLIP ViT-B16, Test-time batch size=642021.11 | 78.4 | — | — | — | — | — | — | — | — | |
| ERMEvaluation Protocol=Fine-tuning (CLIP), Backbone=ViT-B/162024.03 | 78.1 | — | — | — | — | — | — | — | — | |
| MIROBackbone=DeiT, Pre-training source=IN-1K2026.07 | 78.1 | — | — | — | — | — | — | — | — | |
| ERM + Pseudo LabelBackbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=Pseudo Label2021.11 | 78 | — | — | — | — | — | — | — | — | |
| LPBackbone=DeiT, Pre-training source=IN-1K2026.07 | 77.6 | — | — | — | — | — | — | — | — | |
| CLIPoodBackbone=DeiT, Pre-training source=IN-1K2026.07 | 77.3 | — | — | — | — | — | — | — | — | |
| ERM + T3ABackbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=T3A2021.11 | 76.9 | — | — | — | — | — | — | — | — | |
| ERM + Pseudo Label*Backbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=Pseudo Label, Update target=linear classifier2021.11 | 76.6 | — | — | — | — | — | — | — | — | |
| ERM + Tent*Backbone=CLIP ViT-B16, Test-time batch size=64, Test-time adaptation method=Tent, Update target=linear classifier2021.11 | 76.5 | — | — | — | — | — | — | — | — | |
| Text-Anchored Information BottleneckBackbone=EfficientNet, Pre-training source=IN-1K2026.07 | 75.7 | — | — | — | — | — | — | — | — | |
| VL2VBackbone=EfficientNet, Pre-training source=IN-1K2026.07 | 75.3 | — | — | — | — | — | — | — | — | |
| PADGBackbone=ResNet-502025.07 | 75.09 | 75.37 | 81.97 | 71.04 | 71.99 | — | — | — | — | |
| RISEBackbone=RegNetY-16GF, Pre-training source=IN-12K2026.07 | 74.9 | — | — | — | — | — | — | — | — | |
| RISEBackbone=DeiT, Pre-training source=IN-1K2026.07 | 74.8 | — | — | — | — | — | — | — | — | |
| RISE (ERM + Hint + AD + RD + MT)Backbone=ResNet50, Ens/MA=Yes2023.09 | 72.6 | — | — | — | — | — | — | — | — | |
| EoABackbone=ResNet50, Ens/MA=Yes2023.09 | 72.5 | — | — | — | — | — | — | — | — | |
| MIROBackbone=EfficientNet, Pre-training source=IN-1K2026.07 | 72.4 | — | — | — | — | — | — | — | — | |
| CLIPBackbone=ResNet-502025.07 | 71.86 | 71.89 | 79.71 | 68.03 | 67.81 | — | — | — | — | |
| RISE (ERM + Hint + AD + RD)Backbone=ResNet50, Ens/MA=No2023.09 | 71.6 | — | — | — | — | — | — | — | — | |
| ERM + Hint + ADBackbone=ResNet50, Ens/MA=No2023.09 | 71.3 | — | — | — | — | — | — | — | — | |
| ERM + Hint + RDBackbone=ResNet50, Ens/MA=No2023.09 | 71.1 | — | — | — | — | — | — | — | — | |
| CLIPoodBackbone=EfficientNet, Pre-training source=IN-1K2026.07 | 71.1 | — | — | — | — | — | — | — | — | |
| Single sample generalizationBackbone=ResNet-502022.02 | 71.07 | 67.21 | 57.97 | 78.61 | 80.47 | — | — | — | — | |
| Single sample generalizationBackbone=ResNet-502022.02 | 71.07 | — | — | — | — | — | — | — | — | |
| EnsembleBackbone=ResNet50, Ens/MA=Yes2023.09 | 70.8 | — | — | — | — | — | — | — | — | |
| SWADEnsemble=true2022.12 | 70.6 | — | — | — | — | — | — | — | — | |
| SWADBackbone=ResNet50, Ens/MA=Yes2023.09 | 70.6 | — | — | — | — | — | — | — | — | |
| MIROBackbone=RegNetY-16GF, Pre-training source=IN-12K2026.07 | 70.5 | — | — | — | — | — | — | — | — | |
| ERM + HintBackbone=ResNet50, Ens/MA=No2023.09 | 70.2 | — | — | — | — | — | — | — | — | |
| GENIEBackbone=ResNet-50, Pre-training=ImageNet, Training Iterations=5,000, Evaluation Protocol=DomainBed (Training-domain Validation Set)2026.06 | 69.7 | — | — | — | — | — | — | — | — | |
| CROValidation Protocol=DomainBed-style training-domain validation2026.05 | 69.3 | 64.8 | 54.9 | 77.9 | 79.4 | — | — | — | — | |
| FADBackbone=ResNet-50, Pre-training=ImageNet, Training Iterations=5,000, Evaluation Protocol=DomainBed (Training-domain Validation Set)2026.06 | 69.2 | — | — | — | — | — | — | — | — | |
| LPBackbone=EfficientNet, Pre-training source=IN-1K2026.07 | 69 | — | — | — | — | — | — | — | — | |
| RISEBackbone=DINOv2, Pre-training source=LVD-142M2026.07 | 69 | — | — | — | — | — | — | — | — | |
| Dubey et al.Backbone=ResNet-502022.02 | 68.9 | — | — | — | — | — | — | — | — | |
| Dubey et al. (2021)Backbone=ResNet-502022.02 | 68.9 | — | — | — | — | — | — | — | — | |
| ADAMWBackbone=ResNet-50, Pre-training=ImageNet, Training Iterations=5,000, Evaluation Protocol=DomainBed (Training-domain Validation Set)2026.06 | 68.8 | — | — | — | — | — | — | — | — | |
| Sun & SaenkoBackbone=ResNet-502022.02 | 68.7 | 65.3 | 54.4 | 76.5 | 78.4 | — | — | — | — | |
| CORAL2022.12 | 68.7 | — | — | — | — | — | — | — | — | |
| Best SOTA competitorBackbone=ResNet50, Ens/MA=No2023.09 | 68.7 | — | — | — | — | — | — | — | — | |
| CORALValidation Protocol=DomainBed-style training-domain validation2026.05 | 68.7 | 65.3 | 54.4 | 76.5 | 78.4 | — | — | — | — | |
| Fish2022.12 | 68.6 | — | — | — | — | — | — | — | — | |
| SGDBackbone=ResNet-50, Pre-training=ImageNet, Training Iterations=5,000, Evaluation Protocol=DomainBed (Training-domain Validation Set)2026.06 | 68.5 | — | — | — | — | — | — | — | — | |
| ADAHESSIANBackbone=ResNet-50, Pre-training=ImageNet, Training Iterations=5,000, Evaluation Protocol=DomainBed (Training-domain Validation Set)2026.06 | 68.4 | — | — | — | — | — | — | — | — | |
| RISEBackbone=EfficientNet, Pre-training source=IN-1K2026.07 | 68.4 | — | — | — | — | — | — | — | — | |
| LPBackbone=RegNetY-16GF, Pre-training source=IN-12K2026.07 | 68.4 | — | — | — | — | — | — | — | — | |
| YOGIBackbone=ResNet-50, Pre-training=ImageNet, Training Iterations=5,000, Evaluation Protocol=DomainBed (Training-domain Validation Set)2026.06 | 68.3 | — | — | — | — | — | — | — | — | |
| MLDG2022.12 | 68.2 | — | — | — | — | — | — | — | — | |
| GAMBackbone=ResNet-50, Pre-training=ImageNet, Training Iterations=5,000, Evaluation Protocol=DomainBed (Training-domain Validation Set)2026.06 | 68.2 | — | — | — | — | — | — | — | — | |
| CoOpVenue=IJCV’22, Backbone=ResNet-502025.07 | 68.17 | 69.32 | 72.25 | 62.62 | 68.5 | — | — | — | — | |
| MixupValidation Protocol=DomainBed-style training-domain validation2026.05 | 68.1 | 62.4 | 54.8 | 76.9 | 78.3 | — | — | — | — | |
| SagNetValidation Protocol=DomainBed-style training-domain validation2026.05 | 68.1 | 63.4 | 54.8 | 75.8 | 78.3 | — | — | — | — | |
| Algorithm 1 SelectionSelection Criterion=Algorithm 12021.06 | 68.09 | 65.76 | 55.07 | 75.2 | 76.31 | 1 | — | — | — | |
| ADABELIEFBackbone=ResNet-50, Pre-training=ImageNet, Training Iterations=5,000, Evaluation Protocol=DomainBed (Training-domain Validation Set)2026.06 | 68 | — | — | — | — | — | — | — | — | |
| SAMBackbone=ResNet-50, Pre-training=ImageNet, Training Iterations=5,000, Evaluation Protocol=DomainBed (Training-domain Validation Set)2026.06 | 68 | — | — | — | — | — | — | — | — |