List your company
ResearchFormerly WizWand

Image Classification on CIFAR100 (Accuracy)

92AccuracyMLCD
77.781.412585.12588.8375Jun 2, 2021Mar 19, 2022Jan 4, 2023Oct 22, 2023Aug 7, 2024May 25, 2025Mar 12, 2026
Updated 4mo ago

Evaluation results

MethodLinks
MLCDPre-training Data=LAION-400M, Backbone=ViT-L/14, Evaluation Protocol=Linear Probe2024.0792
EfficientNet-B5Params (M)=302022.0291.1
DeiT-BParams (M)=86.62022.0290.8
ViTAE-SParams (M)=23.62022.0290.8
UNICOMPre-training Data=LAION-400M, Backbone=ViT-L/14, Evaluation Protocol=Linear Probe2024.0790.8
LIGOTraining mode=Training from the Pretrained Model: M(12,384) → M(12,768), FLOPs (x 1e18)=5.7, Ratio (Saving)=55.7%2023.1090.52
bert2BERTTraining mode=Training from the Pretrained Model: M(12,384) → M(12,768), FLOPs (x 1e18)=4.6, Ratio (Saving)=64.4%2023.1090.47
MangoTraining mode=Training from the Pretrained Model: M(12,384) → M(12,768), FLOPs (x 1e18)=3.0, Ratio (Saving)=76.4%2023.1090.23
ScratchTraining mode=Training from Scratch, FLOPs (x 1e18)=12.92023.1090.22
StackBERTTraining mode=Training from Scratch, FLOPs (x 1e18)=11.3, Ratio (Saving)=12.6%2023.1090.1
GoGModel=DeiT-B2024.0389.87
Greedy SoupModel=DeiT-B2024.0389.68
GoUModel=DeiT-B2024.0389.68
FGG(best)Model=DeiT-B2024.0389.63
GS(best)Model=DeiT-B2024.0389.19
T2T-ViT-14Params (M)=21.52022.0288.4
OpenCLIPPre-training Data=LAION-400M, Backbone=ViT-L/14, Evaluation Protocol=Linear Probe2024.0787.9
BOFT (b=2, m=2)Backbone=ViT-L, #Params=0.20M2024.0587.84
BOFT (b=4, m=1)Backbone=ViT-L, #Params=0.30M2024.0587.72
CLIPPre-training Data=WIT-400M, Backbone=ViT-L/14, Evaluation Protocol=Linear Probe2024.0787.5
SVFT^B (r=8)Backbone=ViT-L, #Params=2.50M2024.0587.26
ViT-g/14N-shot=10, Probe=linear regression2023.0587.2
CLIP+Pre-training Data=WIT-400M, Backbone=ViT-L/14, Evaluation Protocol=Linear Probe2024.0787.2
ViT-B/16Params (M)=86.52022.0287.1
SVFT^B (r=4)Backbone=ViT-L, #Params=1.32M2024.0587.1
VeRABackbone=ViT-L, #Params=61.4K2024.0586.77
SVFT^PBackbone=ViT-L, #Params=49.2K2024.0586.74
SoViT-400m/14N-shot=10, Probe=linear regression2023.0586.7
SVFT^B (r=2)Backbone=ViT-L, #Params=0.74M2024.0586.59
Full-FTBackbone=ViT-L, #Params=303.3M2024.0586.56
ViT-L/16Params (M)=304.32022.0286.4
X-FM_baseLinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.0186.2
ViTAE-TParams (M)=4.82022.0286
LoRA (r=8)Backbone=ViT-L, #Params=0.35M2024.0586
LoRA (r=1)Backbone=ViT-L, #Params=0.44M2024.0585.97
DOT-CBM2025.0585.83
SVFT^B (r=8)Backbone=ViT-B, #Params=0.94M2024.0585.69
BOFT (b=2, m=2)Backbone=ViT-B, #Params=0.07M2024.0585.55
BOFT (b=4, m=1)Backbone=ViT-B, #Params=0.11M2024.0585.54
Uniform SoupModel=DeiT-B2024.0385.51
Full-FTBackbone=ViT-B, #Params=85.8M2024.0585.35
DoRA (r=8)Backbone=ViT-B, #Params=1.41M2024.0585.03
LoRA (r=1)Backbone=ViT-B, #Params=0.16M2024.0584.86
GoGModel=ResNet502024.0384.77
SparseCBM2025.0584.75
SVFT^B (r=2)Backbone=ViT-B, #Params=0.28M2024.0584.72
CoopCBM2025.0584.66
FGG(best)Model=ResNet502024.0384.64
GoUModel=ResNet502024.0384.57
DoRA (r=1)Backbone=ViT-B, #Params=0.25M2024.0584.46
LoRA (r=8)Backbone=ViT-B, #Params=1.32M2024.0584.41
FLIPPre-training Data=LAION-400M, Backbone=ViT-L/14, Evaluation Protocol=Zero-shot classification2024.0784.1
LaBo2025.0584.1
DoRA (r=1)Backbone=ViT-L, #Params=0.66M2024.0584.06
Greedy SoupModel=ResNet502024.0384.01
SVFT^PBackbone=ViT-B, #Params=18.5K2024.0583.85
GS(best)Model=ResNet502024.0383.8
MLCDPre-training Data=LAION-400M, Backbone=ViT-L/14, Evaluation Protocol=Zero-shot classification2024.0783.7
DoRA (r=8)Backbone=ViT-L, #Params=3.76M2024.0583.55
VeRABackbone=ViT-B, #Params=24.6K2024.0583.38
OmniVLImg-text pairs=14M*, Evaluation protocol=Linear probing2022.0983.2
SVFT^B (r=4)Backbone=ViT-B, #Params=0.50M2024.0583.17
CLIPLinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.0183.1
CLIP-ViT-B/16Img-text pairs=400M, Evaluation protocol=Linear probing2022.0983.1
HeadBackbone=ViT-L2024.0582.95
BASIC-LBackbone=BASIC-L2021.1182.3
Stable SpikeArchitecture=ResNet-19, T=42026.0382.29
ViT-L/16N-shot=10, Probe=linear regression2023.0582.2
STAA-SNNArchitecture=ResNet-19, T=42026.0382.05
Weak2StrongArchitecture=ResNet-19, T=42026.0382.02
DeepTAGEArchitecture=ResNet-19, T=42026.0381.39
CEM2025.0581.23
QKFormerArchitecture=QKFormer, T=42026.0381.15
ALBEFImg-text pairs=14M, Evaluation protocol=Linear probing2022.0980.8
RateBPArchitecture=ResNet-19, T=42026.0380.71
ResNet-50 + DASBackbone=ResNet-50, Attention=DAS2023.1180.54
TS-SNNArchitecture=ResNet-19, T=22026.0380.28
DaVinciLinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.0180.1
SNN-ViTArchitecture=SNN-ViT, T=42026.0380.1
Vanilla-CBM2025.0580.04
Linear Probing2024.0480.03
BLIPImg-text pairs=14M, Evaluation protocol=Linear probing2022.0980
SWIN V2Resolution=64 x 64, Training Resolution=64 x 642022.0579.95
SeTaBackbone=ResNet18, Pruned %=30%2025.0379.4
CLIPEvaluation Protocol=10-shot linear evaluation, Model Scale=L/142023.0679.4
SWEETModel Scale=L6H62026.0179.36
Fully SupervisedData Partitioning=Centralized2021.0679.3
ResNet-50 + Triplet AttentionBackbone=ResNet-50, Attention=Triplet Attention2023.1179.22
InfoBatchBackbone=ResNet18, Pruned %=30%, Training Mode=Same number of epochs2025.0379.2
VIT + TLBResolution=64 x 64, Training Resolution=64 x 642022.0579.17
ResNet-18 + DASBackbone=ResNet-18, Attention=DAS2023.1179.04
SeTaBackbone=ResNet18, Pruned %=50%2025.0379
RényiCLEvaluation Protocol=linear evaluation, Multi-crop Augmentation=true2022.0878.9
Stable SpikeArchitecture=ResNet-18, T=42026.0378.83
RényiCLEvaluation Protocol=linear evaluation, Multi-crop Augmentation=false2022.0878.8
ResNet-18 + Triplet AttentionBackbone=ResNet-18, Attention=Triplet Attention2023.1178.55
ReverB-SNNArchitecture=ResNet-19, T=22026.0378.46
RateBPArchitecture=ResNet-18, T=42026.0378.26
ResNet-18Backbone=ResNet-182023.1178.25
HeadBackbone=ViT-B2024.0578.25