Image Classification on Pets (test)
98.22AccuracyEfficient Adaptive Ensembling
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Efficient Adaptive Ensemblingensemble size=2 weak models2022.06 | 98.22 | — | 1.12 | — | — | — | |
| SOTA2022.06 | 97.1 | — | — | — | — | — | |
| DINO.txt#param=304M, Evaluation Protocol=Zero-shot2026.04 | 95.3 | — | — | — | — | — | |
| LoRABackbone=ViT-Large, Rank=r=16, # Params=1.57M2024.07 | 94.62 | — | — | — | — | — | |
| C3ABackbone=ViT-Large, Config=b=1024/16, # Params=0.79M2024.07 | 94.48 | — | — | — | — | — | |
| FullBackbone=ViT-Large, # Params=303M2024.07 | 94.3 | — | — | — | — | — | |
| C3ABackbone=ViT-Base, Config=b=768/12, # Params=0.22M2024.07 | 93.88 | — | — | — | — | — | |
| LoRABackbone=ViT-Base, Rank=r=16, # Params=0.59M2024.07 | 93.76 | — | — | — | — | — | |
| AdaVFM (BASE)#param=36M, Evaluation Protocol=Zero-shot2026.04 | 93.3 | — | — | — | — | — | |
| AdaVFM (LARGE)#param=50M, Evaluation Protocol=Zero-shot2026.04 | 93.3 | — | — | — | — | — | |
| FullBackbone=ViT-Base, # Params=85.8M2024.07 | 92.82 | — | — | — | — | — | |
| AdaVFM (SMALL)#param=29M, Evaluation Protocol=Zero-shot2026.04 | 92.8 | — | — | — | — | — | |
| UnrestrictedMoESeed Model=CLIP ViT-B/322026.05 | 92.45 | — | — | — | — | — | |
| Expert I (Pets)Seed Model=CLIP ViT-B/322026.05 | 92.4 | — | — | — | — | — | |
| AdaVFM (TINY)#param=16M, Evaluation Protocol=Zero-shot2026.04 | 92.1 | — | — | — | — | — | |
| MetaMoESeed Model=CLIP ViT-B/322026.05 | 91.91 | — | — | — | — | — | |
| AdaVFM (MIN)#param=6M, Evaluation Protocol=Zero-shot2026.04 | 91.4 | — | — | — | — | — | |
| FlexOlmoSeed Model=CLIP ViT-B/322026.05 | 91.36 | — | — | — | — | — | |
| HeadBackbone=ViT-Large2024.07 | 91.11 | — | — | — | — | — | |
| BTMSeed Model=CLIP ViT-B/322026.05 | 90.81 | — | — | — | — | — | |
| Supervised CLIPTraining Protocol=Supervised2021.06 | 90.4 | — | — | — | — | — | |
| HeadBackbone=ViT-Base2024.07 | 90.28 | — | — | — | — | — | |
| CLIP + EB βTraining Protocol=Zero-shot SSL compressor2021.06 | 89.5 | — | — | 150 | — | — | |
| NoLA#param=86M, Evaluation Protocol=Zero-shot2026.04 | 89.3 | — | — | — | — | — | |
| ProText#param=86M, Evaluation Protocol=Zero-shot2026.04 | 89 | — | — | — | — | — | |
| BTXSeed Model=CLIP ViT-B/322026.05 | 88.44 | — | — | — | — | — | |
| Waffle#param=86M, Evaluation Protocol=Zero-shot2026.04 | 88.1 | — | — | — | — | — | |
| ModelSoupSeed Model=CLIP ViT-B/322026.05 | 87.9 | — | — | — | — | — | |
| ZeroShotSeed Model=CLIP ViT-B/322026.05 | 85.77 | — | — | — | — | — | |
| CLIP#param=86M, Evaluation Protocol=Zero-shot2026.04 | 85 | — | — | — | — | — | |
| Expert II (Flowers)Seed Model=CLIP ViT-B/322026.05 | 84.25 | — | — | — | — | — | |
| OpenCLIP#param=304M, Evaluation Protocol=Zero-shot2026.04 | 83.9 | — | — | — | — | — | |
| LaFTer#param=86M, Evaluation Protocol=Zero-shot2026.04 | 82.7 | — | — | — | — | — | |
| Expert III (EuroSAT)Seed Model=CLIP ViT-B/322026.05 | 82.64 | — | — | — | — | — | |
| CLEANTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 62.31 | — | — | — | — | — | |
| EMAXNTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 54.7 | — | — | — | — | — | |
| DEEPCONFUSETarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 53.72 | — | — | — | — | — | |
| EMINNTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 52.96 | — | — | — | — | — | |
| SYNPERTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 52.6 | — | — | — | — | — | |
| ADVPOISONTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 50.86 | — | — | — | — | — | |
| CLEANTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 48.68 | — | — | — | — | — | |
| CLEANTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 44.86 | — | — | — | — | — | |
| ADVPOISONTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 37.99 | — | — | — | — | — | |
| EMINNTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 37.04 | — | — | — | — | — | |
| EMINNTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 36.88 | — | — | — | — | — | |
| DEEPCONFUSETarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 35.54 | — | — | — | — | — | |
| SYNPERTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 34.51 | — | — | — | — | — | |
| ADVPOISONTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 34.29 | — | — | — | — | — | |
| EMAXNTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 34.26 | — | — | — | — | — | |
| EMAXNTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 33.71 | — | — | — | — | — | |
| DEEPCONFUSETarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 33.71 | — | — | — | — | — | |
| SYNPERTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 28.02 | — | — | — | — | — | |
| UCTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 17.06 | — | — | — | — | — | |
| UCTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 12.21 | — | — | — | — | — | |
| UC-CLIPTarget Model=EfficientNet-B1, Evaluation Setting=Label-agnostic2022.12 | 6.49 | — | — | — | — | — | |
| UC-CLIPTarget Model=ResNet-18, Evaluation Setting=Label-agnostic2022.12 | 4.69 | — | — | — | — | — | |
| UCTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 4.28 | — | — | — | — | — | |
| UC-CLIPTarget Model=RegNetX-1.6GF, Evaluation Setting=Label-agnostic2022.12 | 3.87 | — | — | — | — | — | |
| BaselineModel=SwinT, Memory Budget (B_mem)=31889952, Evaluation Protocol=fine-tuned2025.10 | — | 90.18 | — | — | — | — | |
| BiT-LPre-training strategy=Generalist, Hyperparameter setting=BiT-HyperRule, Evaluation protocol=Fine-tuning2019.12 | — | 96.62 | — | — | — | — | |
| BYOLEvaluation Protocol=Linear2021.06 | — | — | — | — | 90.4 | — | |
| BYOLEvaluation Protocol=Fine-tuned2021.06 | — | — | — | — | 91.7 | — | |
| CorruptEncoderBackbone=ResNet-18, CL algorithm=MoCo-v2, Evaluation Protocol=linear evaluation2022.11 | — | — | — | — | — | 72.1 | |
| CTRLBackbone=ResNet-18, CL algorithm=MoCo-v2, Evaluation Protocol=linear evaluation2022.11 | — | — | — | — | — | 35.4 | |
| D-Det RGNModel=SwinT, Memory Budget (B_mem)=27946, Evaluation Protocol=fine-tuned2025.10 | — | 90.91 | — | — | — | — | |
| D-Det RGNModel=SwinT, Memory Budget (B_mem)=112640, Evaluation Protocol=fine-tuned2025.10 | — | 90.78 | — | — | — | — | |
| D-Det RGNModel=SwinT, Memory Budget (B_mem)=633859, Evaluation Protocol=fine-tuned2025.10 | — | 91.39 | — | — | — | — | |
| D-Det RGNModel=SwinT, Memory Budget (B_mem)=2767686, Evaluation Protocol=fine-tuned2025.10 | — | 91.2 | — | — | — | — | |
| D-Full RandomModel=SwinT, Memory Budget (B_mem)=27946, Evaluation Protocol=fine-tuned2025.10 | — | 90.76 | — | — | — | — | |
| D-Full RandomModel=SwinT, Memory Budget (B_mem)=112640, Evaluation Protocol=fine-tuned2025.10 | — | 90.61 | — | — | — | — | |
| D-Full RandomModel=SwinT, Memory Budget (B_mem)=633859, Evaluation Protocol=fine-tuned2025.10 | — | 90.95 | — | — | — | — | |
| D-Full RandomModel=SwinT, Memory Budget (B_mem)=2767686, Evaluation Protocol=fine-tuned2025.10 | — | 91.55 | — | — | — | — | |
| DeiT-smallParams.=22M, Complexity=O(N^2), Evaluation Protocol=Linear Probing2025.07 | — | 87.56 | — | — | — | — | |
| Generalist SOTAPre-training strategy=Generalist2019.12 | — | 95.9 | — | — | — | — | |
| MANO-tinyParams.=28M, Complexity=O(N), Evaluation Protocol=Linear Probing2025.07 | — | 88.31 | — | — | — | — | |
| MobileNetV2Bit-width (W/A)=32/322024.01 | — | 86 | — | — | — | — | |
| MobileNetV2Bit-width (W/A)=4MP/4MP2024.01 | — | 86.1 | — | — | — | — | |
| MobileNetV2Bit-width (W/A)=3MP/3MP2024.01 | — | 84.1 | — | — | — | — | |
| No AttackBackbone=ResNet-18, CL algorithm=MoCo-v2, Evaluation Protocol=linear evaluation2022.11 | — | — | — | — | — | 1.5 | |
| Poisoned-EncoderBackbone=ResNet-18, CL algorithm=MoCo-v2, Evaluation Protocol=linear evaluation2022.11 | — | — | — | — | — | 45.8 | |
| Random initEvaluation Protocol=Fine-tuned2021.06 | — | — | — | — | 81.5 | — | |
| ResNet18Bit-width (W/A)=32/322024.01 | — | 88.9 | — | — | — | — | |
| ResNet18Bit-width (W/A)=4MP/4MP2024.01 | — | 88.7 | — | — | — | — | |
| ResNet18Bit-width (W/A)=3MP/3MP2024.01 | — | 87.9 | — | — | — | — | |
| SimCLREvaluation Protocol=Linear2021.06 | — | — | — | — | 83.6 | — | |
| SimCLREvaluation Protocol=Fine-tuned2021.06 | — | — | — | — | 89.2 | — | |
| Specialist SOTAPre-training strategy=Specialist2019.12 | — | 97.1 | — | — | — | — | |
| SSL-BackdoorBackbone=ResNet-18, CL algorithm=MoCo-v2, Evaluation Protocol=linear evaluation2022.11 | — | — | — | — | — | 4.6 | |
| SSL-HSIC (w/ target)Evaluation Protocol=Linear2021.06 | — | — | — | — | 91.2 | — | |
| SSL-HSIC (w/ target)Evaluation Protocol=Fine-tuned2021.06 | — | — | — | — | 91.7 | — | |
| SSL-HSIC (w/o target)Evaluation Protocol=Linear2021.06 | — | — | — | — | 88.2 | — | |
| SSL-HSIC (w/o target)Evaluation Protocol=Fine-tuned2021.06 | — | — | — | — | 90.6 | — | |
| Supervised-INEvaluation Protocol=Linear2021.06 | — | — | — | — | 91.5 | — | |
| Supervised-INEvaluation Protocol=Fine-tuned2021.06 | — | — | — | — | 92.1 | — | |
| SwinV2-TParams.=28M, Complexity=O(N), Evaluation Protocol=Linear Probing2025.07 | — | 87.14 | — | — | — | — | |
| TinyViTParams.=21M, Complexity=O(N^2), Evaluation Protocol=Linear Probing2025.07 | — | 86.5 | — | — | — | — | |
| TRaDyModel=SwinT, Memory Budget (B_mem)=27946, Evaluation Protocol=fine-tuned2025.10 | — | 91.15 | — | — | — | — | |
| TRaDyModel=SwinT, Memory Budget (B_mem)=112640, Evaluation Protocol=fine-tuned2025.10 | — | 91.11 | — | — | — | — | |
| TRaDyModel=SwinT, Memory Budget (B_mem)=633859, Evaluation Protocol=fine-tuned2025.10 | — | 90.95 | — | — | — | — | |
| TRaDyModel=SwinT, Memory Budget (B_mem)=2767686, Evaluation Protocol=fine-tuned2025.10 | — | 91.55 | — | — | — | — |