List your company
ResearchFormerly WizWand

Image Classification on ImageNet 1k (ACC)

86.3AccuracyDINOv2
56.86864.50972.1579.791Apr 19, 2025Jun 21, 2025Aug 24, 2025Oct 26, 2025Dec 29, 2025Mar 2, 2026May 5, 2026
Updated 6mo ago

Evaluation results

MethodLinks
DINOv2ViT=L/14, Pre-training Data=LVD-142M, Evaluation Protocol=Linear, Notes=distilled from DINOv2-G2026.0586.3
LOOPE (XG + XC)Model=Cross-ViT [2]2025.0484.8
LOOPE (XG + XC)Model=ViT-Base [10]2025.0484.7
Static (XG)Model=ViT-Base [10]2025.0484.5
Static (XG)Model=Cross-ViT [2]2025.0484.5
FrancaViT=L/14, Pre-training Data=IN-21k, Evaluation Protocol=Linear2026.0584.5
SinusoidModel=ViT-Base [10]2025.0484.4
DINOv2ViT=L/14, Pre-training Data=IN-21k, Evaluation Protocol=Linear2026.0584
LearnableModel=ViT-Base [10]2025.0483.6
LOOPE (XG + Xc)Backbone=DeiT-Base, Positional Encoding=LOOPE (XG + Xc)2025.0483.5
LOOPE (XG + XC)Model=DeiT-Base [28]2025.0483.5
LearnableModel=Cross-ViT [2]2025.0483.5
LFF(Fourier)Backbone=DeiT-Base, Positional Encoding=LFF(Fourier)2025.0483.4
2D SinusoidBackbone=DeiT-Base, Positional Encoding=2D Sinusoid2025.0483.2
Static (XG)Backbone=DeiT-Base, Positional Encoding=Static (XG)2025.0483.1
Static (XG)Model=DeiT-Base [28]2025.0483.1
LOOPE (XG + XC)Model=CaiT [29]2025.0483.1
SinusoidModel=Cross-ViT [2]2025.0483.1
LOOPE (XG + XC)Model=DeiT-Small [28]2025.0483
AS2DRoPEBackbone=DeiT-Base, Positional Encoding=AS2DRoPE2025.0482.9
SinusoidModel=DeiT-Small [28]2025.0482.8
Static (XG)Model=DeiT-Small [28]2025.0482.8
Static (XG)Model=CaiT [29]2025.0482.8
CPEBackbone=DeiT-Base, Positional Encoding=CPE2025.0482.7
No PEModel=ViT-Base [10]2025.0482.7
CMAEvaluation Protocol=Methods requiring training (or fine-tuning)2026.0382.64
SigLIP2ViT=L/16, Pre-training Data=WebLI, Evaluation Protocol=Linear2026.0582.5
No PEModel=Cross-ViT [2]2025.0482.4
Web-DINOViT=L/16, Pre-training Data=MC-2B, Evaluation Protocol=Linear2026.0582.4
iBOTViT=L/16, Pre-training Data=IN-21k, Evaluation Protocol=Linear2026.0582.3
LearnableModel=DeiT-Small [28]2025.0482.2
LearnableModel=DeiT-Base [28]2025.0482.1
TC-JEPAViT=L/16, Pre-training Data=IN-21k, Evaluation Protocol=Linear2026.0582.1
SinusoidModel=DeiT-Base [28]2025.0482
TC-JEPABackbone=ViT-L/14, Pre-training Data=CC27M, Protocol=linear classification2026.0581.9
No PEModel=DeiT-Base [28]2025.0481.7
LearnableModel=CaiT [29]2025.0481.6
SinusoidModel=CaiT [29]2025.0481.6
TC-JEPAViT=L/16, Pre-training Data=CC27M, Evaluation Protocol=Linear2026.0581.4
No PEModel=DeiT-Small [28]2025.0480.5
SigLIPViT=L/16, Pre-training Data=WebLI, Evaluation Protocol=Linear2026.0580.5
RPEBackbone=DeiT-Base, Positional Encoding=RPE2025.0480.3
No PEModel=CaiT [29]2025.0480.3
CAPIViT=L/14, Pre-training Data=IN-21k, Evaluation Protocol=Linear2026.0579.3
BSViTArchitecture=BSViT-10-512, Param (M)=38.65, Time Step=12026.0478.98
A²OS²AArchitecture=Spiking Transformer-10-512, Param (M)=36.01, Time Step=42026.0478.66
DreamLIPBackbone=ViT-B/16, Pre-training Data=CC30M, Protocol=linear classification2026.0578.6
STAttenArchitecture=SDT-8-768, Param (M)=66.34, Time Step=42026.0478.11
AugMix + ReActEvaluation Protocol=Methods requiring training (or fine-tuning)2026.0377.63
AugMix + ASHEvaluation Protocol=Methods requiring training (or fine-tuning)2026.0377.63
TC-JEPABackbone=ViT-B/16, Pre-training Data=CC27M, Protocol=linear classification2026.0577.3
I-JEPAViT=L/16, Pre-training Data=IN-21k, Evaluation Protocol=Linear2026.0577.2
TC-JEPABackbone=ViT-B/16, Pre-training Data=YFCC15M, Protocol=linear classification2026.0577.1
SCALEEvaluation Protocol=Methods requiring training (or fine-tuning)2026.0376.18
STAttenArchitecture=SDT-8-512, Param (M)=29.68, Time Step=42026.0476.18
SpikingformerArchitecture=Spikingformer-8-768, Param (M)=66.34, Time Step=42026.0475.85
DreamLIPBackbone=ViT-B/16, Pre-training Data=YFCC15M, Protocol=linear classification2026.0575.2
SDTArchitecture=SDT-10-512, Param (M)=36.01, Time Step=42026.0474.66
BSViTArchitecture=BSViT-8-384, Param (M)=18.01, Time Step=12026.0474.37
SPARCBackbone=ViT-B/16, Pre-training Data=CC27M, Protocol=linear classification2026.0574.1
A²OS²AArchitecture=Spiking Transformer-8-384, Param (M)=16.81, Time Step=42026.0474.04
MaskCLIPBackbone=ViT-B/16, Pre-training Data=YFCC15M, Protocol=linear classification2026.0573.7
SpikformerArchitecture=Spikformer-10-512, Param (M)=36.01, Time Step=42026.0473.68
SPARCBackbone=ViT-B/16, Pre-training Data=YFCC15M, Protocol=linear classification2026.0573.4
Cluster-aware UpcyclingArch.=MoE, MoE Init=Cluster-aware Upcycling, Evaluation Protocol=FT2026.0473.3
CLIP-MoEArch.=MoE, MoE Init=CLIP-MoE, Evaluation Protocol=FT2026.0473.2
Drop-UpcyclingArch.=MoE, MoE Init=Drop-Upcycling, Evaluation Protocol=FT2026.0473.1
Sparse UpcyclingArch.=MoE, MoE Init=Sparse Upcycling, Evaluation Protocol=FT2026.0473
DenseArch.=Dense, MoE Init=-, Evaluation Protocol=FT2026.0472.8
Spiking ResNetArchitecture=ResNet-50, Param (M)=25.56, Time Step=3502026.0472.75
SpikingformerArchitecture=Spikingformer-8-384, Param (M)=16.81, Time Step=42026.0472.45
SDTArchitecture=SDT-8-512, Param (M)=29.68, Time Step=12026.0471.68
BLIPBackbone=ViT-B/16, Pre-training Data=Merged14M, Protocol=linear classification2026.0571.2
SpikformerArchitecture=Spikformer-8-384, Param (M)=16.81, Time Step=42026.0470.24
GroupViTBackbone=ViT-S/16, Pre-training Data=CC27M, Protocol=linear classification2026.0569.8
SEW ResNetArchitecture=SEW-ResNet-152, Param (M)=60.19, Time Step=42026.0469.26
TETArchitecture=SEW-ResNet-34, Param (M)=21.79, Time Step=42026.0468
LAPTEvaluation Protocol=Methods requiring training (or fine-tuning)2026.0367.86
AdaNegEvaluation Protocol=Training-free & test-time adaptation methods2026.0367.13
SEW ResNetArchitecture=SEW-ResNet-34, Param (M)=21.79, Time Step=42026.0467.04
NegLabelEvaluation Protocol=Training-free & non-adaptive methods2026.0366.82
TANLEvaluation Protocol=Training-free & test-time adaptation methods2026.0366.82
CLIPBackbone=ViT-B/16, Pre-training Data=YFCC15M, Protocol=linear classification2026.0566.5
MCMEvaluation Protocol=Training-free & non-adaptive methods2026.0366.28
TETArchitecture=Spiking-ResNet-34, Param (M)=21.79, Time Step=62026.0464.79
STBP-tdBNArchitecture=Spiking-ResNet-34, Param (M)=21.79, Time Step=62026.0463.72
DsCoBackbone=EfficientNet-B0, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Accessible2026.0363
Minimax-IGDBackbone=EfficientNet-B0, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Accessible2026.0362.7
DiT-IGDBackbone=EfficientNet-B0, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Accessible2026.0362
DFDsCoBackbone=EfficientNet-B0, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Free2026.0361.8
Hybrid trainingArchitecture=ResNet-34, Param (M)=21.79, Time Step=2502026.0461.48
DsCoBackbone=ResNet-18, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Accessible2026.0360.4
Minimax-IGDBackbone=ResNet-18, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Accessible2026.0360.3
DiT-IGDBackbone=ResNet-18, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Accessible2026.0359.8
DsCoBackbone=MobileNet-V2, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Accessible2026.0359
RDEDBackbone=EfficientNet-B0, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Accessible2026.0358.5
Minimax-IGDBackbone=MobileNet-V2, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Accessible2026.0358.5
DFDsCoBackbone=ResNet-18, IPC (Images Per Class)=50, Data Access Category (Data-Accessible/Data-Free)=Data-Free2026.0358.2
Cluster-aware UpcyclingArch.=MoE, MoE Init=Cluster-aware Upcycling, Evaluation Protocol=10-shot2026.0458.2
CLIP-MoEArch.=MoE, MoE Init=CLIP-MoE, Evaluation Protocol=10-shot2026.0458