Image Classification on CIFAR-100 (test)
94.55AccuracyVision Transformer (ViT-H/14)
Evaluation results
| Method | Links | |||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Vision Transformer (ViT-H/14)Pre-training Dataset=JFT-300M, Architecture=ViT-H/142020.10 | 94.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Vision Transformer (ViT-L/16)Pre-training Dataset=JFT-300M, Architecture=ViT-L/162020.10 | 93.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BiT-LArchitecture=ResNet152x42020.10 | 93.51 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Vision Transformer (ViT-L/16)Pre-training Dataset=ImageNet-21k, Architecture=ViT-L/162020.10 | 93.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Res-TuningParam. (M)=0.48 (0.55%), Mem.=6.85G2023.10 | 93.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SAMPa-0.2Backbone=ViT-B/16, Pre-trained=Yes, Evaluation Protocol=Fine-tuning, Epochs=10, Optimizer=AdamW2024.10 | 93.06 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientNetV2-LParams=121M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=finetuned2021.04 | 92.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientNetV2-MParams=55M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=finetuned2021.04 | 92.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AdaptFormerParam. (M)=1.26 (1.46%), Mem.=6.32G2023.10 | 91.86 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SAMBackbone=ViT-B/16, Pre-trained=Yes, Evaluation Protocol=Fine-tuning, Epochs=10, Optimizer=AdamW2024.10 | 91.79 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| R50+ViT-B_16Backbone=R50+ViT-B_16, Pre-trained=ImageNet-21k, Evaluation Protocol=Fine-tuned2021.06 | 91.71 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientNet-B7Params=66M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=finetuned2021.04 | 91.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CMT-Spre-training=true, batch size=10242022.09 | 91.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAM-AdapterParam. (M)=10.08 (11.72%), Mem.=9.57G2023.10 | 91.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BiT-M R101x3Backbone=BiT-M R101x3, Pre-trained=ImageNet-21k, Evaluation Protocol=Fine-tuned2021.06 | 91.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientNetV2-SParams=24M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=finetuned2021.04 | 91.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPipeParams=556M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=finetuned2021.04 | 91.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoCo v3Backbone=ViT-H, Pre-training=ImageNet-1k Self-Supervised, Evaluation Protocol=End-to-end fine-tuning2021.04 | 91.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoCo v3Backbone=ViT-L, Pre-training=ImageNet-1k Self-Supervised, Evaluation Protocol=End-to-end fine-tuning2021.04 | 91.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ASF-former-BParams (M)=56.72022.04 | 91 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-B_16Backbone=ViT-B_16, Pre-trained=ImageNet-21k, Evaluation Protocol=Fine-tuned2021.06 | 90.95 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeiT-BParams=86M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=finetuned2021.04 | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeiT-B-384Params=86M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=finetuned2021.04 | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViTAE-S+384Params (M)=23.62022.04 | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CeiT-S+384Params (M)=24.22022.04 | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeiT-B+384Params (M)=86.62022.04 | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoCo v3Backbone=ViT-B, Pre-training=ImageNet-1k Self-Supervised, Evaluation Protocol=End-to-end fine-tuning2021.04 | 90.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MLP-Mixer-B_16Backbone=MLP-Mixer-B_16, Pre-trained=ImageNet-21k, Evaluation Protocol=Fine-tuned2021.06 | 90.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ASF-former-SParams (M)=19.32022.04 | 90.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Ens.Model=ViT-B/162025.12 | 90.27 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TNT-S+384Params (M)=23.82022.04 | 90.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ASAMBackbone=PyramidNet-272, rho=12021.02 | 89.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + matched inference + fuzzy relaxationpre-training=true, batch size=1282022.09 | 89.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DPSModel=ViT-B/162025.12 | 89.54 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + smoothed focuspre-training=true, batch size=1282022.09 | 89.53 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + matched inferencepre-training=true, batch size=1282022.09 | 89.39 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SAMBackbone=PyramidNet-272, rho=0.12021.02 | 89.36 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PS-KDModel=ViT-B/162025.12 | 89.36 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| KDModel=ViT-B/162025.12 | 89.35 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Res-Tuning-BypassParam. (M)=0.46 (0.53%), Mem.=4.72G2023.10 | 89.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + matched inference + Ldrlocpre-training=true, batch size=1282022.09 | 89.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + Ldrlocpre-training=true, batch size=1282022.09 | 89.27 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + relaxed self-attentionpre-training=true, batch size=1282022.09 | 89.21 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T (resimulated baseline)pre-training=true, batch size=1282022.09 | 89.16 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BaselineModel=ViT-B/162025.12 | 89.16 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FullParam. (M)=85.9 (100%), Mem.=9.02G2023.10 | 89.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TEModel=ViT-B/162025.12 | 89.11 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-B/16-SAMBackbone=ViT-B/16, SAM Pre-training=true, Resolution=224x2242021.06 | 89.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SGDBackbone=PyramidNet-2722021.02 | 88.91 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + matched inferencepre-training=true, batch size=10242022.09 | 88.73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LSTParam. (M)=0.93 (1.08%), Mem.=5.26G2023.10 | 88.72 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DLBModel=ViT-B/162025.12 | 88.72 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + relaxed self-attentionpre-training=true, batch size=10242022.09 | 88.64 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + Ldrlocpre-training=true, batch size=10242022.09 | 88.63 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + matched inference + fuzzy relaxationpre-training=true, batch size=10242022.09 | 88.63 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + matched inference + Ldrlocpre-training=true, batch size=10242022.09 | 88.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T (resimulated baseline)pre-training=true, batch size=10242022.09 | 88.53 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + smoothed focuspre-training=true, batch size=10242022.09 | 88.44 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T - stochastic depthpre-training=true, batch size=1282022.09 | 88.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + matched inference - stochastic depthpre-training=true, batch size=1282022.09 | 88.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| T2T-ViT-14pre-training=true, batch size=10242022.09 | 88.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T (+ Ldrloc)pre-training=true, batch size=10242022.09 | 88.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| T2T-ViT-14Params (M)=21.52022.04 | 88.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T (vanilla)pre-training=true, batch size=10242022.09 | 88.22 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SHiRABackbone=ViT-Base, Evaluation Protocol=Fine-tuned2024.06 | 88.15 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Global DEPre-train=Self-supervised (SSL), Backbone=ResNet-50, Ensemble Size=52023.03 | 88.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientNet-B0MAdds (M)=400, Params (M)=4, Search Method=manual, Year=20192026.02 | 88.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MOEA-BUS-XLMAdds (M)=657, Params (M)=6.3, Search Cost (GPU Days)=1.2, Search Method=EA2026.02 | 88.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LoRABackbone=ViT-Base, Evaluation Protocol=Fine-tuned2024.06 | 87.97 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ResNet-152-SAMBackbone=ResNet-152, SAM Pre-training=true, Resolution=224x2242021.06 | 87.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| StarSSE (opt.)Pre-train=Self-supervised (SSL), Backbone=ResNet-50, Ensemble Size=52023.03 | 87.63 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T - stochastic depthpre-training=true, batch size=10242022.09 | 87.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-S/16-SAMBackbone=ViT-S/16, SAM Pre-training=true, Resolution=224x2242021.06 | 87.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-B/16Backbone=ViT-B/16, SAM Pre-training=false, Resolution=224x2242021.06 | 87.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MOEA-BUS-LMAdds (M)=458, Params (M)=6.24, Search Cost (GPU Days)=1.2, Search Method=EA2026.02 | 87.53 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-T + matched inference - stochastic depthpre-training=true, batch size=10242022.09 | 87.49 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Local DEPre-train=Self-supervised (SSL), Backbone=ResNet-50, Ensemble Size=52023.03 | 87.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FairNAS-AInput Size=224x224, x+ (M)=391, SE and Swish=true2019.07 | 87.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FairNAS-AMAdds (M)=391, Search Cost (GPU Days)=12, Search Method=EA, Year=20212026.02 | 87.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SPNASParams (M)=6.7, Search Cost (GPU Days)=1.6, Search Method=EA, Year=20252026.02 | 87.26 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MOEA-BUS-MMAdds (M)=273, Params (M)=4.77, Search Cost (GPU Days)=1.2, Search Method=EA2026.02 | 87.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Side-TuningParam. (M)=9.62 (11.18%), Mem.=3.48G2023.10 | 87.16 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SSE (opt.)Pre-train=Self-supervised (SSL), Backbone=ResNet-50, Ensemble Size=52023.03 | 87.11 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-B/16Params=87M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=finetuned2021.04 | 87.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-L/32Params=306M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=finetuned2021.04 | 87.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MixNet-MInput Size=224x224, x+ (M)=3592019.07 | 87.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-S-16pre-training=true, batch size=10242022.09 | 87.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-B/16Params (M)=86.52022.04 | 87.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImNet supervisedBackbone=ViT-B, Pre-training=ImageNet-1k Supervised, Evaluation Protocol=End-to-end fine-tuning2021.04 | 87.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BiT-M R50x1Backbone=BiT-M R50x1, Pre-trained=ImageNet-21k, Evaluation Protocol=Fine-tuned2021.06 | 87.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FairNAS-BInput Size=224x224, x+ (M)=348, SE and Swish=true2019.07 | 87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FairNAS-BMAdds (M)=348, Search Cost (GPU Days)=12, Search Method=EA, Year=20212026.02 | 87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Global DEPre-train=Supervised (SV), Backbone=ResNet-50, Ensemble Size=52023.03 | 86.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientNet-B0Input Size=224x224, x+ (M)=3872019.07 | 86.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MOEA-BUS-SMAdds (M)=198, Params (M)=4.19, Search Cost (GPU Days)=1.2, Search Method=EA2026.02 | 86.71 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NASNet-A LargeInput Size=331x331, x+ (M)=120312019.07 | 86.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FairNAS-CInput Size=224x224, x+ (M)=324, SE and Swish=true2019.07 | 86.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NASNet-A LargeMAdds (M)=12031, Search Cost (GPU Days)=1800, Search Method=RL, Year=20182026.02 | 86.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FairNAS-CMAdds (M)=324, Search Cost (GPU Days)=12, Search Method=EA, Year=20212026.02 | 86.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SGD + GAMBackbone=PyramidNet110, Data Augmentation=RA2023.03 | 86.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |