List your company
ResearchFormerly WizWand

Classification on CIFAR10 (test)

97.42AccuracyreSGHMC
82.75686.56390.3794.177Nov 17, 2018Jan 20, 2020Mar 25, 2021May 29, 2022Aug 2, 2023Oct 5, 2024Dec 9, 2025
Updated 5mo ago

Evaluation results

MethodLinks
reSGHMCModel Architecture=WRN-28-102020.0897.42
M-SGDModel Architecture=WRN-28-102020.0897.33
SGHMCModel Architecture=WRN-28-102020.0897.32
reSGHMCModel Architecture=WRN-16-82020.0896.87
SGHMCModel Architecture=WRN-16-82020.0896.73
M-SGDModel Architecture=WRN-16-82020.0896.71
reSGHMCModel Architecture=ResNet-562020.0896.12
M-SGDModel Architecture=ResNet-562020.0896.01
MS-ResNet-18Architecture=MS-ResNet-18, Coding Schemes=CSGC Coding, Time Steps=62025.0695.98
SGHMCModel Architecture=ResNet-562020.0895.95
reSGHMCModel Architecture=ResNet-322020.0895.35
SGHMCModel Architecture=ResNet-322020.0895.18
M-SGDModel Architecture=ResNet-322020.0895.15
StandardModel=Resnet182022.0395.1
MS-ResNet-18Architecture=MS-ResNet-18, Coding Schemes=Direct Coding, Time Steps=62025.0694.92
reSGHMCModel Architecture=ResNet-202020.0894.62
SOCModel=Resnet182022.0394.24
SGHMCModel Architecture=ResNet-202020.0894.22
M-SGDModel Architecture=ResNet-202020.0894.21
VGG192018.1193.6
PointConvArchitecture=VGG192018.1193.19
ResNet-19Architecture=ResNet-19, Coding Schemes=Rate Coding, Time Steps=62025.0693.16
VGG-16Architecture=VGG-16, Coding Schemes=Temporal Coding, Time Steps=1002025.0692.68
AugerinoBase Group=SE(2), Number of Elements=2, Partial Equivariance=false, Augerino=true2021.1092.48
BCOPModel=Resnet182022.0392.38
AugerinoBase Group=SE(2), Number of Elements=4, Partial Equivariance=false, Augerino=true2021.1092.28
StandardModel=Resnet92022.0392.26
AugerinoBase Group=SE(2), Number of Elements=8, Partial Equivariance=false, Augerino=true2021.1091.99
I-NeurALlabel budget=20%, T (rounds)=100002022.1091.8
I-NeurALlabel budget=10%2022.1091.75
DnCBackbone=ResNet-50, Pre-training dataset=JFT, Pre-training epochs=4500 ImageNet-equivalent, Evaluation protocol=Linear transfer2022.0191.7
Partial G-CNNBase Group=SE(2), Number of Elements=4, Partial Equivariance=true, Augerino=false2021.1091.66
ResNet-19Architecture=ResNet-19, Coding Schemes=Phase Coding, Time Steps=82025.0691.4
CNNBase Group=T(2), Number of Elements=1, Architecture=13-layer CNN2021.1091.21
LAMMSetting=Fine-tuned2023.0691.2
F-EDL2025.1091.19
DAEDL2025.1091.11
NeuAL-NTK-Dlabel budget=20%, T (rounds)=100002022.1090.9
G-CNNBase Group=SE(2), Number of Elements=8, Partial Equivariance=false, Augerino=false2021.1090.55
ReLICv2Backbone=ResNet-50, Pre-training dataset=JFT, Pre-training epochs=1000 ImageNet-equivalent, Evaluation protocol=Linear transfer2022.0190.2
R-EDL2025.1090.09
Partial G-CNNBase Group=SE(2), Number of Elements=8, Partial Equivariance=true, Augerino=false2021.1089.96
ReLICv2Backbone=ResNet-50, Pre-training dataset=JFT, Pre-training epochs=5000 ImageNet-equivalent, Evaluation protocol=Linear transfer2022.0189.9
BYOLBackbone=ResNet-50, Pre-training dataset=JFT, Pre-training epochs=5000 ImageNet-equivalent, Evaluation protocol=Linear transfer2022.0189.8
Marginlabel budget=20%, T (rounds)=100002022.1089.75
G-CNNBase Group=SE(2), Number of Elements=4, Partial Equivariance=false, Augerino=false2021.1089.73
Proj-TModel=Resnet182022.0389.59
Randomlabel budget=20%, T (rounds)=100002022.1089.5
SupervisedClass Prior=0.22025.0889.5
Marginlabel budget=10%2022.1089.45
Proj-DModel=Resnet182022.0389.43
ALPSlabel budget=10%2022.1089.35
I-EDL2025.1089.2
PointConvArchitecture=5-layer2018.1189.13
NeuAL-NTK-Dlabel budget=10%2022.1089.05
AlexNet2018.1189
I-NeurALLabel Budget=3%, T (rounds)=100002022.1089
Randomlabel budget=10%2022.1089
ALPSlabel budget=20%, T (rounds)=100002022.1088.95
Image Convolution2018.1188.52
SconfConfDiff-ABSClass Prior=0.22025.0888.4
MarginLabel Budget=3%, T (rounds)=100002022.1088.25
GMDBackbone=ResNet-18, Setting=Federated, Corruption Type=Bit Flip, Corruption (%)=20%2021.0688.07
GALAXYinitial_labels=100, query_schedule=10 x 102025.1087.9
Partial G-CNNBase Group=SE(2), Number of Elements=2, Partial Equivariance=true, Augerino=false2021.1087.78
SupervisedClass Prior=0.52025.0887.6
Confidenceinitial_labels=100, query_schedule=10 x 102025.1087.31
SconfConfDiff-ReLUClass Prior=0.22025.0887.1
Convex(γ = 0.2)-ABSClass Prior=0.22025.0887
Convex(γ = 0.8)-ABSClass Prior=0.22025.0887
Convex(γ = 0.5)-ABSClass Prior=0.22025.0886.9
ConfDiff-ABSClass Prior=0.22025.0886.8
SconfConfDiff-UnbiasedClass Prior=0.22025.0886.5
RandomLabel Budget=3%, T (rounds)=100002022.1086.4
GMDBackbone=ResNet-18, Setting=Federated, Corruption Type=Clean, Corruption (%)=0%2021.0685.65
ParBaLS EPIGinitial_labels=100, query_schedule=10 x 102025.1085.52
G-CNNBase Group=SE(2), Number of Elements=2, Partial Equivariance=false, Augerino=false2021.1085.51
CMDBackbone=ResNet-18, Setting=Federated, Corruption Type=Clean, Corruption (%)=0%2021.0685.5
ALPSLabel Budget=3%, T (rounds)=100002022.1085.4
ParBaLS-MAP EPIGinitial_labels=100, query_schedule=10 x 102025.1085.39
BGMDBackbone=ResNet-18, Setting=Federated, Corruption Type=Clean, Corruption (%)=0%2021.0684.82
WholeIPC=Full, Ratio=100%2025.1284.8
BGMDBackbone=ResNet-18, Setting=Federated, Corruption Type=Bit Flip, Corruption (%)=20%2021.0684.56
ConfDiff-ABSClass Prior=0.52025.0884.2
NeuAL-NTK-DLabel Budget=3%, T (rounds)=100002022.1084.05
BGMDBackbone=ResNet-18, Setting=Federated, Corruption Type=Additive, Corruption (%)=20%2021.0683.84
SconfConfDiff-UnbiasedClass Prior=0.52025.0883.7
SconfConfDiff-ReLUClass Prior=0.52025.0883.7
SconfConfDiff-ABSClass Prior=0.52025.0883.7
Poisoned triggerLoss scaling=1, Clip type=l2-norm, Clip bound=0.12021.0683.68
Poisoned triggerLoss scaling=20, Clip type=l2-norm, Clip bound=0.12021.0683.68
Poisoned triggerLoss scaling=10, Clip type=l2-norm, Clip bound=0.12021.0683.66
Poisoned triggerLoss scaling=50, Clip type=l2-norm, Clip bound=0.12021.0683.66
EDL2025.1083.55
Poisoned triggerLoss scaling=1, Clip type=l2-norm, Clip bound=12021.0683.39
Poisoned triggerLoss scaling=20, Clip type=l2-norm, Clip bound=12021.0683.39
Poisoned triggerLoss scaling=1, Clip type=None2021.0683.32
Poisoned triggerLoss scaling=1, Clip type=l2-norm, Clip bound=102021.0683.32
Poisoned triggerLoss scaling=1, Clip type=l-infinity-norm, Clip bound=12021.0683.32
Poisoned triggerLoss scaling=1, Clip type=l-infinity-norm, Clip bound=0.12021.0683.32