List your company
ResearchFormerly WizWand

Image Classification on Tiny ImageNet 200 (test)

66.16Test AccuracyBMLR-P
11.071225.373139.67553.9769Mar 31, 2022Dec 7, 2022Aug 15, 2023Apr 22, 2024Dec 29, 2024Sep 6, 2025May 15, 2026
Updated 5mo ago

Evaluation results

MethodLinks
BMLR-PSpace=P^n_K, Backbone=ResNet-18, Fit time (s/epoch)=71.98, #Params=102.80K2026.0266.16——
BMLR-LSpace=L^n_K, Backbone=ResNet-18, Fit time (s/epoch)=69.19, #Params=102.80K2026.0265.99——
DIFFUSEMIXBackbone=PreactResNet-18, Training Epochs=300 epochs2024.0465.7783.66—
LMLRSpace=L^n_K, Backbone=ResNet-18, Fit time (s/epoch)=69.27, #Params=102.60K2026.0265.63——
PBMLR-PSpace=P^n_K, Backbone=ResNet-18, Fit time (s/epoch)=336.58, #Params=204.80K2026.0265.43——
MLRSpace=R^n, Backbone=ResNet-18, Fit time (s/epoch)=69.17, #Params=102.60K2026.0265.19——
PMLRSpace=P^n_K, Backbone=ResNet-18, Fit time (s/epoch)=71.90, #Params=102.80K2026.0264.93——
LLQR+SAMArchitecture=WRN-28-10, Induced geometry=NGD-induced2026.0564.65——
Guided-APBackbone=PreactResNet-18, Training Epochs=300 epochs2024.0464.6382.49—
LLQR+SAMArchitecture=WRN-28-10, Induced geometry=Newton-induced2026.0564.59——
Co-MixupBackbone=PreactResNet-18, Training Epochs=300 epochs2024.0464.15——
SAMArchitecture=WRN-28-102026.0563.91——
PuzzleMixBackbone=PreactResNet-18, Training Epochs=300 epochs2024.0463.4875.52—
LLQR+SAMArchitecture=ResNet-18, Induced geometry=NGD-induced2026.0562.24——
LLQR+SAMArchitecture=ResNet-18, Induced geometry=Newton-induced2026.0561.98——
SAMArchitecture=ResNet-182026.0561.17——
SGDMArchitecture=WRN-28-10, Geometry=Base2026.0561.06——
LLQRArchitecture=WRN-28-10, Induced geometry=Newton-induced2026.0560.37——
LLQRArchitecture=WRN-28-10, Induced geometry=NGD-induced2026.0560.08——
SGDMArchitecture=ResNet-18, Geometry=Base2026.0557.97——
LLQRArchitecture=ResNet-18, Induced geometry=Newton-induced2026.0557.62——
VanillaBackbone=PreactResNet-18, Training Epochs=300 epochs2024.0457.2373.65—
LLQRArchitecture=ResNet-18, Induced geometry=NGD-induced2026.0557.02——
SaliencyMixBackbone=PreactResNet-18, Training Epochs=300 epochs2024.0456.5476.14—
Guided-SRBackbone=PreactResNet-18, Training Epochs=300 epochs2024.0455.9774.68—
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.5, Label Budget (b)=50%2026.0355.89——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.7, Label Budget (b)=50%2026.0355.63——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.6, Label Budget (b)=50%2026.0355.29——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.8, Label Budget (b)=50%2026.0355.18——
Baseline (AL)Architecture=ResNet-50, Label Budget (b)=50%2026.0354.65——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.6, Label Budget (b)=40%2026.0352.09——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.7, Label Budget (b)=40%2026.0352.02——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.5, Label Budget (b)=40%2026.0351.84——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.8, Label Budget (b)=40%2026.0351.77——
Baseline (AL)Architecture=ResNet-50, Label Budget (b)=40%2026.0349.86——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.5, Label Budget (b)=30%2026.0347.41——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.8, Label Budget (b)=30%2026.0347.15——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.6, Label Budget (b)=30%2026.0347.02——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.7, Label Budget (b)=30%2026.0346.45——
Baseline (AL)Architecture=ResNet-50, Label Budget (b)=30%2026.0343.96——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.5, Label Budget (b)=20%2026.0339.7——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.7, Label Budget (b)=20%2026.0339.24——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.6, Label Budget (b)=20%2026.0338.84——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.8, Label Budget (b)=20%2026.0337.86——
Baseline (AL)Architecture=ResNet-50, Label Budget (b)=20%2026.0333.62——
V-MCR2Objective=V-MCR2, Training Epochs=2000, d=5002022.0326.65——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.6, Label Budget (b)=10%2026.0319.25——
CEObjective=CE, Training Epochs=2000, d=5002022.0319.07——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.5, Label Budget (b)=10%2026.0318.71——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.8, Label Budget (b)=10%2026.0318.34——
PruneFuseArchitecture=ResNet-50, Pruning ratio (p)=0.7, Label Budget (b)=10%2026.0318.32——
Baseline (AL)Architecture=ResNet-50, Label Budget (b)=10%2026.0314.86——
MCR2Objective=MCR2, Training Epochs=200, d=5002022.0313.19——