Image Classification on Cars (test)
96.868AccuracyEfficient Adaptive Ensembling
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Efficient Adaptive Ensemblingensemble size=2 weak models2022.06 | 96.868 | — | 0.548 | — | |
| SOTA2022.06 | 96.32 | — | — | — | |
| TesNetBackbone Architecture=DenseNet-161, Number of Parameters=~28.68M2024.10 | 92.6 | — | — | — | |
| ProtoViTBackbone Architecture=CaiT-XXS 24, Number of Parameters=~11.9M, K=4, r=12024.10 | 92.4 | — | — | — | |
| ProtoViTBackbone Architecture=DeiT-Small, Number of Parameters=~22M, K=4, r=12024.10 | 91.84 | — | — | — | |
| ViT-NetBackbone Architecture=CaiT-XXS 24, Number of Parameters=~11.9M2024.10 | 91.54 | — | — | — | |
| ViT-NetBackbone Architecture=DeiT-Small, Number of Parameters=~22M2024.10 | 91.34 | — | — | — | |
| ProtoPFormerBackbone Architecture=CaiT-XXS 24, Number of Parameters=~11.9M2024.10 | 91.04 | — | — | — | |
| ProtoPFormerBackbone Architecture=DeiT-Small, Number of Parameters=~22M2024.10 | 90.86 | — | — | — | |
| BaselineBackbone Architecture=CaiT-XXS 24, Number of Parameters=~11.9M2024.10 | 90.19 | — | — | — | |
| BaselineBackbone Architecture=DeiT-Small, Number of Parameters=~22M2024.10 | 90.06 | — | — | — | |
| ProtoPoolBackbone Architecture=DenseNet-161, Number of Parameters=~28.68M2024.10 | 90 | — | — | — | |
| ProtoPNetBackbone Architecture=DenseNet-161, Number of Parameters=~28.68M2024.10 | 89.5 | — | — | — | |
| ProtoViTBackbone Architecture=DeiT-Tiny, Number of Parameters=~5M, K=4, r=12024.10 | 89.02 | — | — | — | |
| Def. ProtoPNet(2x2)Backbone Architecture=DenseNet-161, Number of Parameters=~28.68M2024.10 | 88.7 | — | — | — | |
| ProtoPFormerBackbone Architecture=DeiT-Tiny, Number of Parameters=~5M2024.10 | 88.48 | — | — | — | |
| ViT-NetBackbone Architecture=DeiT-Tiny, Number of Parameters=~5M2024.10 | 88.41 | — | — | — | |
| FullBackbone=ViT-Large, # Params=303M2024.07 | 88.15 | — | — | — | |
| BaseBackbone Architecture=DeiT-Tiny, Number of Parameters=~5M2024.10 | 86.21 | — | — | — | |
| LoRABackbone=ViT-Large, Rank=r=16, # Params=1.57M2024.07 | 86.11 | — | — | — | |
| FullBackbone=ViT-Base, # Params=85.8M2024.07 | 85.1 | — | — | — | |
| C3ABackbone=ViT-Large, Config=b=1024/16, # Params=0.79M2024.07 | 84.94 | — | — | — | |
| CLIP + EB βTraining Protocol=Zero-shot SSL compressor2021.06 | 79.6 | — | — | 131 | |
| C3ABackbone=ViT-Base, Config=b=768/12, # Params=0.22M2024.07 | 79.05 | — | — | — | |
| LoRABackbone=ViT-Base, Rank=r=16, # Params=0.59M2024.07 | 78.04 | — | — | — | |
| MultitaskBackbone=ViT-B/32, #Params=x1, Merging Strategy=N/A2026.05 | 77.9 | — | — | — | |
| DiDi-Merging-MBackbone=ViT-B/32, #Params=x1.4, Merging Strategy=Dynamic Merging2026.05 | 77.9 | — | — | — | |
| SMILEBackbone=ViT-B/32, #Params=x3.07, Merging Strategy=Dynamic Merging2026.05 | 77.8 | — | — | — | |
| IndividualsBackbone=ViT-B/32, #Params=x8, Merging Strategy=N/A2026.05 | 77.7 | — | — | — | |
| FREE-MergingBackbone=ViT-B/32, #Params=x2.16, Merging Strategy=Dynamic Merging2026.05 | 77.6 | — | — | — | |
| TSV-CBackbone=ViT-B/32, #Params=x2.08, Merging Strategy=Dynamic Merging2026.05 | 77.3 | — | — | — | |
| DiDi-Merging-SBackbone=ViT-B/32, #Params=x1.24, Merging Strategy=Dynamic Merging2026.05 | 77.2 | — | — | — | |
| WEMoEBackbone=ViT-B/32, #Params=x6.27, Merging Strategy=Dynamic Merging2026.05 | 76.8 | — | — | — | |
| ISO-CTSBackbone=ViT-B/32, #Params=x1, Merging Strategy=Static Merging2026.05 | 74.4 | — | — | — | |
| EMR-MergingBackbone=ViT-B/32, #Params=x4, Merging Strategy=Dynamic Merging2026.05 | 72.7 | — | — | — | |
| TALL-Mask+TIESBackbone=ViT-B/32, #Params=x2.5, Merging Strategy=Dynamic Merging2026.05 | 72.5 | — | — | — | |
| CLEANTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 72.33 | — | — | — | |
| Twin-MergingBackbone=ViT-B/32, #Params=x2.25, Merging Strategy=Dynamic Merging2026.05 | 71.7 | — | — | — | |
| WUDI-MergingBackbone=ViT-B/32, #Params=x1, Merging Strategy=Static Merging2026.05 | 71 | — | — | — | |
| TSV-MergingBackbone=ViT-B/32, #Params=x1, Merging Strategy=Static Merging2026.05 | 70.7 | — | — | — | |
| CLEANTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 67.18 | — | — | — | |
| MANO-tinyParams.=28M, Complexity=O(N), Evaluation Protocol=Linear Probing2025.07 | 65.68 | — | — | — | |
| CAT-MergingBackbone=ViT-B/32, #Params=x1, Merging Strategy=Static Merging2026.05 | 65.4 | — | — | — | |
| CLEANTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 63.84 | — | — | — | |
| Weight AveragingBackbone=ViT-B/32, #Params=x1, Merging Strategy=Static Merging2026.05 | 63.4 | — | — | — | |
| TinyViTParams.=21M, Complexity=O(N^2), Evaluation Protocol=Linear Probing2025.07 | 61.7 | — | — | — | |
| SYNPERTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 58.34 | — | — | — | |
| EMAXNTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 55.64 | — | — | — | |
| EMINNTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 54.43 | — | — | — | |
| EMINNTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 54.23 | — | — | — | |
| SYNPERTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 53.5 | — | — | — | |
| EMAXNTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 52.95 | — | — | — | |
| ADVPOISONTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 51.91 | — | — | — | |
| DEEPCONFUSETarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 51.11 | — | — | — | |
| ADVPOISONTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 50.08 | — | — | — | |
| Supervised CLIPTraining Protocol=Supervised2021.06 | 49.1 | — | — | — | |
| DEEPCONFUSETarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 47.15 | — | — | — | |
| ADVPOISONTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 46.06 | — | — | — | |
| SYNPERTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 45.54 | — | — | — | |
| EMAXNTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 43.4 | — | — | — | |
| ViT-baseParams.=86M, Complexity=O(N^2), Evaluation Protocol=Linear Probing2025.07 | 41.95 | — | — | — | |
| DEEPCONFUSETarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 41.15 | — | — | — | |
| EMINNTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 39.67 | — | — | — | |
| SwinV2-TParams.=28M, Complexity=O(N), Evaluation Protocol=Linear Probing2025.07 | 38.36 | — | — | — | |
| HeadBackbone=ViT-Large2024.07 | 37.91 | — | — | — | |
| DeiT-smallParams.=22M, Complexity=O(N^2), Evaluation Protocol=Linear Probing2025.07 | 36.38 | — | — | — | |
| UCTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 33.57 | — | — | — | |
| UCTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 29.46 | — | — | — | |
| HeadBackbone=ViT-Base2024.07 | 25.76 | — | — | — | |
| UC-CLIPTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 15.33 | — | — | — | |
| UCTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 13.92 | — | — | — | |
| UC-CLIPTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 4.74 | — | — | — | |
| UC-CLIPTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 4.18 | — | — | — | |
| BYOLEvaluation Protocol=Linear2021.06 | — | 66.7 | — | — | |
| BYOLEvaluation Protocol=Fine-tuned2021.06 | — | 91.6 | — | — | |
| CLIP-B32 (Last)Model=CLIP-B32, Selection Strategy=Last Layer2026.05 | — | 50.53 | — | — | |
| CLIP-B32 (LOES)Model=CLIP-B32, Selection Strategy=LOES2026.05 | — | 59.95 | — | — | |
| DeiT-B/16 (Last)Model=DeiT-B/16, Selection Strategy=Last Layer2026.05 | — | 38.88 | — | — | |
| DeiT-B/16 (LOES)Model=DeiT-B/16, Selection Strategy=LOES2026.05 | — | 57.98 | — | — | |
| DINOv2-S (Last)Model=DINOv2-S, Selection Strategy=Last Layer2026.05 | — | 48.61 | — | — | |
| DINOv2-S (LOES)Model=DINOv2-S, Selection Strategy=LOES2026.05 | — | 60.86 | — | — | |
| DINOv3-S/16 (Last)Model=DINOv3-S/16, Selection Strategy=Last Layer2026.05 | — | 77.54 | — | — | |
| DINOv3-S/16 (LOES)Model=DINOv3-S/16, Selection Strategy=LOES2026.05 | — | 84.53 | — | — | |
| Fine-tuningBackbone=ResNet-101, Pre-trained=ImageNet (ILSVRC-2012), Evaluation Protocol=Fine-tuning2021.11 | — | 89.28 | — | — | |
| l2-NormBackbone=ResNet-101, Pre-trained=ImageNet (ILSVRC-2012), Evaluation Protocol=Fine-tuning2021.11 | — | 88.96 | — | — | |
| l2-PGMBackbone=ResNet-101, Pre-trained=ImageNet (ILSVRC-2012), Evaluation Protocol=Fine-tuning2021.11 | — | 88.92 | — | — | |
| l2-SPBackbone=ResNet-101, Pre-trained=ImageNet (ILSVRC-2012), Evaluation Protocol=Fine-tuning2021.11 | — | 88.96 | — | — | |
| LSBackbone=ResNet-101, Pre-trained=ImageNet (ILSVRC-2012), Evaluation Protocol=Fine-tuning2021.11 | — | 89.66 | — | — | |
| MAE-B/16 (Last)Model=MAE-B/16, Selection Strategy=Last Layer2026.05 | — | 7.42 | — | — | |
| MAE-B/16 (LOES)Model=MAE-B/16, Selection Strategy=LOES2026.05 | — | 15.45 | — | — | |
| Random initEvaluation Protocol=Fine-tuned2021.06 | — | 91.4 | — | — | |
| REGSLBackbone=ResNet-101, Pre-trained=ImageNet (ILSVRC-2012), Evaluation Protocol=Fine-tuning2021.11 | — | 89.14 | — | — | |
| SimCLREvaluation Protocol=Linear2021.06 | — | 50.3 | — | — | |
| SimCLREvaluation Protocol=Fine-tuned2021.06 | — | 91.3 | — | — | |
| SSL-HSIC (w/ target)Evaluation Protocol=Linear2021.06 | — | 62.6 | — | — | |
| SSL-HSIC (w/ target)Evaluation Protocol=Fine-tuned2021.06 | — | 91.8 | — | — | |
| SSL-HSIC (w/o target)Evaluation Protocol=Linear2021.06 | — | 59.3 | — | — | |
| SSL-HSIC (w/o target)Evaluation Protocol=Fine-tuned2021.06 | — | 91.6 | — | — | |
| Supervised-INEvaluation Protocol=Linear2021.06 | — | 67.8 | — | — | |
| Supervised-INEvaluation Protocol=Fine-tuned2021.06 | — | 92.1 | — | — |