List your company
ResearchFormerly WizWand

Image Classification on ImageNet 1k (10% labels)

84.9Top-1 AccSemi-SST
47.14856.94966.7576.551Jun 17, 2020May 26, 2021May 4, 2022Apr 12, 2023Mar 20, 2024Feb 26, 2025Feb 5, 2026
Updated 8mo ago

Evaluation results

MethodLinks
Semi-SSTArchitecture Type=Transformer, Architecture=ViT-Huge, Param=632M2025.0584.9——
Semi-SSTArchitecture=ViT-Huge, Param=632M2025.0584.9——
Super-SSTArchitecture Type=Transformer, Architecture=ViT-Huge, Param=632M2025.0584.8——
Super-SSTArchitecture=ViT-Huge, Param=632M2025.0584.8——
Semi-ViTArchitecture Type=Transformer, Architecture=ViT-Huge, Param=632M2025.0584.3——
Semi-ViTArchitecture Type=Transformer, Architecture=ViT-Large, Param=307M2025.0583.3——
Super-SSTArchitecture Type=Transformer, Architecture=ViT-Small, Param=22M, distilled=true2025.0580.3——
Super-SST (distilled)Architecture=ViT-Small, Param=22M2025.0580.3——
SimCLR v2Arch.=RN152w3+SK, Param.=795M2022.0280.1——
Semi-ViTArchitecture Type=Transformer, Architecture=ViT-Base, Param=86M2025.0579.7——
SEERArch.=RG10B, Param.=10B2022.0278.8——
Semi-SSTArchitecture Type=CNN, Architecture=ConvNeXt-T, Param=28M2025.0578.6——
Semi-SSTArchitecture Type=Transformer, Architecture=ViT-Small, Param=22M2025.0578.6——
Semi-SSTArchitecture=ViT-Small, Param=22M2025.0578.6——
Super-SSTArchitecture Type=Transformer, Architecture=ViT-Small, Param=22M2025.0578.3——
Super-SSTArchitecture=ViT-Small, Param=22M2025.0578.3——
SEERArch.=RG256, Param.=1.5B2022.0277.9——
Super-SSTArchitecture Type=CNN, Architecture=ConvNeXt-T, Param=28M2025.0577.9——
SimMatchV2Architecture Type=CNN, Architecture=ConvNeXt-T, Param=28M2025.0577.8——
BYOLArch.=RN200, Param.=250M2022.0277.7——
Semi-ViTArchitecture Type=Transformer, Architecture=ViT-Small, Param=22M2025.0577.1——
SEERArch.=RG128, Param.=693M2022.0276.7——
SimMatchV2Architecture Type=CNN, Architecture=ResNet-50, Param=26M2025.0576.2——
SimMatchV2Architecture Type=Transformer, Architecture=ViT-Small, Param=22M2025.0575.9——
SemiformerArchitecture=ViT-S + Conv, Params=40M, Learning Paradigm=Semi-supervised2021.1175.5——
PAWSArchitecture Type=CNN, Architecture=ResNet-50, Param=26M2025.0575.5——
SemiFormerArchitecture Type=Transformer, Architecture=ViT-S+Conv, Param=42M2025.0575.5——
iBOTArch.=ViT-S/162021.1175.1——
ShrinkMatchPre-training=None, Method=ShrinkMatch, Epochs=400, Params (train / test)=31.8M / 25.6M2023.0874.591.9—
SimMatchArchitecture Type=CNN, Architecture=ResNet-50, Param=26M2025.0574.4——
DINOArch.=ViT-S/162021.1174.3——
SimMatch+Pre-training=None, Method=SimMatch+ [45], Epochs=400, Params (train / test)=30.0M / 25.6M2023.0874.191.5—
Semi-ViTArchitecture Type=CNN, Architecture=ConvNeXt-T, Param=28M2025.0574.1——
FixMatch-EMANPre-training=MoCo-EMAN [5], Method=FixMatch-EMAN [5], Epochs=1100, Params (train / test)=30.0M / 25.6M2023.087490.9—
EMANArchitecture Type=CNN, Architecture=ResNet-50, Param=26M2025.0574——
CowMixArch.=RN152, Param.=265M2022.0273.9——
MPLArchitecture=ResNet-50, Params=24M, Learning Paradigm=Semi-supervised2021.1173.9——
CowMixArchitecture=ResNet-152, Params=60M, Learning Paradigm=Semi-supervised2021.1173.9——
MPLArchitecture Type=CNN, Architecture=ResNet-50, Param=26M2025.0573.9——
CoMatchPre-training=MoCo v2 [9], Method=CoMatch [20], Epochs=1200, Params (train / test)=30.0M / 25.6M2023.0873.791.4—
CoMatchPre-training=None, Method=CoMatch [20], Epochs=400, Params (train / test)=30.0M / 25.6M2023.0873.691.6—
CoMatchArchitecture Type=CNN, Architecture=ResNet-50, Param=26M2025.0573.6——
BYOLArchitecture=ResNet-50 (2x), Params=94M, Learning Paradigm=Self-supervised pretraining2021.1173.5——
S4LArchitecture=ResNet-50 (4x), Params=375M, Learning Paradigm=Semi-supervised2021.1173.2——
BYOL multi-taskBackbone=ResNet-50, Evaluation Protocol=Semi-supervised2026.0272.591.3—
ReLIC v2Backbone=ResNet-50, Evaluation Protocol=Semi-supervised2026.0272.491.2—
DINOArchitecture=ViT-S, Params=21M, Learning Paradigm=Self-supervised pretraining2021.1172.2——
WCLPre-training=WCL [44], Method=Fine-tune, Epochs=800, Params (train / test)=34.2M / 25.6M2023.087291.2—
SimCLRArchitecture=ResNet-50 (2×), Params=94M, Learning Paradigm=Self-supervised pretraining2021.1171.7——
FixMatchBackbone=ResNet-50, Evaluation protocol=Fine-tuning, Method category=Methods using label-propagation, Data augmentation=RandAugment2020.0671.589.1—
FixMatchArch.=RN50, Param.=24M2022.0271.5——
CPCArchitecture=ResNet-161, Params=305M, Learning Paradigm=Self-supervised pretraining2021.1171.5——
FixMatchArchitecture=ResNet-50, Params=24M, Learning Paradigm=Semi-supervised2021.1171.5——
FixMatchArchitecture Type=CNN, Architecture=ResNet-50, Param=26M2025.0571.5——
SimCLRv2Arch.=RN50, self-distillation=true2021.1170.5——
SimCLRv2Architecture Type=CNN, Architecture=ResNet-50, Param=26M2025.0570.5——
C-BYOLBackbone=ResNet-50, Evaluation Protocol=Semi-supervised2026.0270.590—
SwAVBackbone=ResNet-50, Evaluation protocol=Fine-tuning, Method category=Methods using self-supervision only2020.0670.289.9—
SwAVArch.=RN502021.1170.2——
SwAVArch.=RN50, Param.=24M2022.0270.2——
SwAVBackbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Protocol=Fine-tuned2023.0470.289.9—
SwAVPre-training=SwAV [6], Method=Fine-tune, Epochs=800, Params (train / test)=30.4M / 25.6M2023.0870.289.9—
SwAVBackbone=ResNet-50, Evaluation Protocol=Semi-supervised2026.0270.289.9—
NNCLRBackbone=ResNet-50, Evaluation Protocol=Semi-supervised2026.0269.889.3—
Barlow TwinsBackbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Protocol=Fine-tuned2023.0469.789.3—
Barlow TwinsBackbone=ResNet-50, Evaluation Protocol=Semi-supervised2026.0269.789.3—
VICRegBackbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Protocol=Fine-tuned2023.0469.589.5—
I-VNE+Backbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Protocol=Fine-tuned2023.0469.189.9—
Fine-tuningBackbone=MAE ViT-B, Evaluation Protocol=Fine-tuning (FT)2025.0668.9——
UDABackbone=ResNet-50, Evaluation protocol=Fine-tuning, Method category=Methods using label-propagation, Data augmentation=RandAugment2020.0668.888.5—
BYOLArch.=RN502021.1168.8——
BYOLArchitecture=ResNet-50, Params=24M, Learning Paradigm=Self-supervised pretraining2021.1168.8——
UDAArchitecture=ResNet-50, Params=24M, Learning Paradigm=Semi-supervised2021.1168.8——
BYOLBackbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Protocol=Fine-tuned2023.0468.889—
UDAArchitecture Type=CNN, Architecture=ResNet-50, Param=26M2025.0568.8——
BYOLBackbone=ResNet-50, Evaluation Protocol=Semi-supervised2026.0268.889—
SimCLR v2Pre-training=SimCLR v2 [8], Method=Fine-tune, Epochs=800, Params (train / test)=34.2M / 25.6M2023.0868.489.2—
SimCLRv2Arch.=RN50, self-distillation=false2021.1168.1——
MoCo v2 (Fine-tune)Pre-training=MoCo v2 [9], Method=Fine-tune, Epochs=800, Params (train / test)=30.0M / 25.6M2023.0866.187.9—
SimCLRBackbone=ResNet-50, Evaluation protocol=Fine-tuning, Method category=Methods using self-supervision only2020.0665.687.8—
SimCLRArch.=RN50, Param.=24M2022.0265.6——
SimCLRArchitecture=ResNet-50, Params=24M, Learning Paradigm=Self-supervised pretraining2021.1165.6——
SimCLRBackbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Protocol=Fine-tuned2023.0465.687.8—
SimCLRBackbone=ResNet-50, Evaluation Protocol=Semi-supervised2026.0265.687.8—
Efficient ProbingBackbone=MAE ViT-B, Evaluation Protocol=Efficient Probing (EP)2025.0665.2—71.5
PIRLBackbone=ResNet-50, Evaluation protocol=Fine-tuning, Method category=Methods using self-supervision only2020.0660.483.8—
Supervised Baseline (Conv)Architecture=Conv, Params=13M, Learning Paradigm=Supervised2021.1160.2——
SupervisedBackbone=ResNet-50, Evaluation protocol=Fine-tuning2020.0656.480.4—
SupervisedBackbone=ResNet-50, Pre-training=ImageNet-1K, Evaluation Protocol=Fine-tuned2023.0456.480.4—
supervisedBackbone=ResNet-50, Evaluation Protocol=Semi-supervised2026.0256.480.4—
Linear ProbingBackbone=MAE ViT-B, Evaluation Protocol=Linear Probing (LP)2025.0655.9——
Supervised Baseline (ViT-S)Architecture=ViT-S, Params=23M, Learning Paradigm=Supervised2021.1148.6——
PCLBackbone=ResNet-50, Evaluation protocol=Fine-tuning, Method category=Methods using self-supervision only2020.06—86.2—