Domain Generalization on OfficeHome
96.15AccuracyTask Experts
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Task ExpertsBackbone=ViT-L-142026.03 | 96.15 | — | |
| Text-Anchored Information BottleneckBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=T2026.07 | 93.2 | — | |
| Task ExpertsBackbone=ViT-B-322026.03 | 91.65 | — | |
| MagMaxBackbone=ViT-L-142026.03 | 91.41 | — | |
| SCOREBackbone=ViT-L-142026.03 | 91.28 | — | |
| TIESBackbone=ViT-L-142026.03 | 91.26 | — | |
| TSVBackbone=ViT-L-142026.03 | 91.22 | — | |
| PCBBackbone=ViT-L-142026.03 | 91.13 | — | |
| Iso-CBackbone=ViT-L-142026.03 | 90.32 | — | |
| Task. Arithm.Backbone=ViT-L-142026.03 | 89.97 | — | |
| Iso-CTSBackbone=ViT-L-142026.03 | 89.94 | — | |
| ZeroshotBackbone=ViT-L-142026.03 | 88.82 | — | |
| SIMPLE+Backbone=ViT-B/16, Venue=ICLR’232025.07 | 87.7 | — | |
| CLIPCEILBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=V,T2026.07 | 87.7 | — | |
| CLIP-DTPBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=V,T2026.07 | 87.7 | — | |
| VL2VBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=V,T2026.07 | 87.4 | — | |
| Ours (PADG)Backbone=ViT-B/162025.07 | 87.06 | — | |
| CLIPoodBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=V,T2026.07 | 87 | — | |
| Text-Anchored Information BottleneckBackbone=ViT-B/16, Pre-training=ImageNet-1k, Guidance (G)=T2026.07 | 86.4 | — | |
| Ours (DPR)Backbone=ViT-B/16, Venue=CVPR’242025.07 | 86.13 | — | |
| CLIP-DPRBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=V,T2026.07 | 86.1 | — | |
| VL2VBackbone=ViT-B/16, Pre-training=ImageNet-1k, Guidance (G)=V,T2026.07 | 85.7 | — | |
| CAR-FTBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=V,T2026.07 | 85.7 | — | |
| VL2V-SDBackbone=ViT-B/16, Venue=CVPR’242025.07 | 85.44 | — | |
| MetaPromptFew-shot=true, Shots=5, CLIP-based=true2022.12 | 85.1 | — | |
| SCOREBackbone=ViT-B-322026.03 | 84.23 | — | |
| DPLBackbone=ViT-B/162025.07 | 84.2 | — | |
| GESTURBackbone=ViT-B/16, Venue=ICCV’232025.07 | 84.2 | — | |
| GESTURBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=-2026.07 | 84.2 | — | |
| CoCoOpFew-shot=true, Shots=5, CLIP-based=true2022.12 | 84.1 | — | |
| MetaPromptFew-shot=true, Shots=1, CLIP-based=true2022.12 | 84 | — | |
| EoAEnsemble=true, CLIP-based=true2022.12 | 83.9 | — | |
| CoOpBackbone=ViT-B/16, Venue=IJCV’222025.07 | 83.9 | — | |
| CLIP-LoRABackbone=ViT-B/16, Venue=CVPR’242025.07 | 83.9 | — | |
| TSVBackbone=ViT-B-322026.03 | 83.74 | — | |
| SPGBackbone=CLIP ViT-B/162024.04 | 83.6 | — | |
| DUPRGBackbone=ViT-B/16, Venue=ICLR’232025.07 | 83.6 | — | |
| SPGBackbone=ViT-B/16, Venue=ECCV’242025.07 | 83.6 | — | |
| PCBBackbone=ViT-B-322026.03 | 83.59 | — | |
| CoCoOpBackbone=CLIP ViT-B/162024.04 | 83.4 | — | |
| MaPLeBackbone=CLIP ViT-B/162024.04 | 83.4 | — | |
| MaPLeBackbone=ViT-B/16, Venue=CVPR’232025.07 | 83.4 | — | |
| TIESBackbone=ViT-B-322026.03 | 83.27 | — | |
| VPTBackbone=CLIP ViT-B/162024.04 | 83.2 | — | |
| CoOpBackbone=CLIP ViT-B/162024.04 | 83 | — | |
| DPLBackbone=CLIP ViT-B/162024.04 | 83 | — | |
| LPBackbone=ViT-B/16, Pre-training=ImageNet-1k, Guidance (G)=-2026.07 | 82.8 | — | |
| MIRO (ViT-B/16)Backbone=ViT-B/16, Venue=ECCV’222025.07 | 82.5 | — | |
| MIROBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=-2026.07 | 82.5 | — | |
| Iso-CBackbone=ViT-B-322026.03 | 82.34 | — | |
| LPBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=-2026.07 | 82.3 | — | |
| CLIP-ZSBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=-2026.07 | 82.3 | — | |
| CLIP (template)Zero-shot=true, CLIP-based=true2022.12 | 82.1 | — | |
| Sparse and Shared Feature ActivationsN-shot=102023.04 | 82 | — | |
| Iso-CTSBackbone=ViT-B-322026.03 | 82 | — | |
| CLIPBackbone=ViT-B/162025.07 | 82 | — | |
| ZS-CLIPBackbone=CLIP ViT-B/16, Evaluation Protocol=Zero-shot2024.04 | 81.8 | — | |
| OracleBackbone=ViT-b162023.06 | 81.23 | — | |
| VPBackbone=CLIP ViT-B/162024.04 | 81.2 | — | |
| ERMN-shot=102023.04 | 81 | — | |
| CLIPood*Backbone=ViT-B/16, Pre-training=ImageNet-1k, Guidance (G)=T2026.07 | 80.9 | — | |
| SEDGEEnsemble=true, CLIP-based=true2022.12 | 80.7 | — | |
| SEDGEBackbone=ViT-B/162025.07 | 80.7 | — | |
| NAC-MEBackbone=ViT-b162023.06 | 80.52 | 63.17 | |
| MagMaxBackbone=ViT-B-322026.03 | 80.37 | — | |
| RISE*Backbone=ViT-B/16, Pre-training=ImageNet-1k, Guidance (G)=V,T2026.07 | 80.3 | — | |
| ValidationBackbone=ViT-b162023.06 | 80.11 | 58.29 | |
| CADGEnsemble status=Non-ensemble2022.03 | 79.9 | — | |
| ZeroshotBackbone=ViT-B-322026.03 | 79.82 | — | |
| CLIPZero-shot=true, CLIP-based=true2022.12 | 79.4 | — | |
| Lin. Prob.Backbone=CLIP ViT-B/16, Evaluation Protocol=Linear Probing2024.04 | 79.3 | — | |
| Text-Anchored Information BottleneckBackbone=ResNet-50, Pre-training=ImageNet-1k, Guidance (G)=T2026.07 | 79 | — | |
| ERM (ViT-B/16)Backbone=ViT-B/16, Venue=ICLR’212025.07 | 78.5 | — | |
| RISEBackbone=CLIP-ViT-B/16, Pre-training=private dataset (400M), Guidance (G)=V,T2026.07 | 78.4 | — | |
| SMABackbone=ViT-B/16, Venue=NIPS’222025.07 | 78.1 | — | |
| Sparse and Shared Feature ActivationsN-shot=52023.04 | 77 | — | |
| SWADBackbone=ViT-B/16, Venue=NIPS’212025.07 | 76.9 | — | |
| ERMN-shot=52023.04 | 75.7 | — | |
| MIRO*Backbone=ViT-B/16, Pre-training=ImageNet-1k, Guidance (G)=-2026.07 | 74.9 | — | |
| VL2VBackbone=ResNet-50, Pre-training=ImageNet-1k, Guidance (G)=V,T2026.07 | 74.4 | — | |
| GMoE-S/16Backbone=GMoE-S/16, Selection Criterion=leave-one-domain-out2022.06 | 73.9 | — | |
| SPGBackbone=CLIP RN502024.04 | 73.8 | — | |
| DPLBackbone=CLIP RN502024.04 | 73.6 | — | |
| CoOpBackbone=CLIP RN502024.04 | 73.5 | — | |
| CoCoOpBackbone=CLIP RN502024.04 | 73.4 | — | |
| INSUREBackbone=ResNet-50, Pre-training=ImageNet-1k, Guidance (G)=-2026.07 | 72 | — | |
| ViT-S/16Backbone=ViT-S/16, Selection Criterion=leave-one-domain-out2022.06 | 71.9 | — | |
| RESBackbone=ResNet-50, Venue=ECCV’242025.07 | 71.8 | — | |
| MIRO + SWAD + POEMSetting=DomainBed settings (reproduced)2023.05 | 71.7 | — | |
| SMOSBackbone=ResNet-50, Venue=CVPR’242025.07 | 71.6 | — | |
| RISEBackbone=ResNet-50, Pre-training=ImageNet-1k, Guidance (G)=V,T2026.07 | 71.6 | — | |
| MIRO + POEMSetting=DomainBed settings (reproduced)2023.05 | 71.4 | — | |
| Previous SOTA [31] + SWAD2021.02 | 71.3 | — | |
| MIRO + SWADSetting=DomainBed settings (reproduced)2023.05 | 71.3 | — | |
| FGMix + SWAModel Selection Strategy=training domain validation, Optimizer=SWA, Training Time (ms/iter)=319.82022.09 | 71.3 | — | |
| GESTURBackbone=ResNet-50, Pre-training=ImageNet-1k, Guidance (G)=-2026.07 | 71.1 | — | |
| ZS-CLIPBackbone=CLIP RN50, Evaluation Protocol=Zero-shot2024.04 | 70.8 | — | |
| SWADSetting=DomainBed settings (reproduced)2023.05 | 70.7 | — | |
| MIROSetting=DomainBed settings (reproduced)2023.05 | 70.7 | — | |
| GMDG2026.01 | 70.7 | — |