List your company
ResearchFormerly WizWand

Image Classification on CIFAR-10 (test) (NLL vs Sample Size)

0.36NLL (N=100)DE ELBO
0.2640.9121.562.208Feb 3, 2025
Updated 8mo ago

Evaluation results

MethodLinks
DE ELBOBackbone=ViT-B/16, Model=L2-SP2025.020.360.240.120.07
MAP + GSBackbone=ViT-B/16, Model=L2-SP2025.020.460.250.10.06
MAP + GSBackbone=ConNeXt-Tiny, Model=L2-SP2025.020.460.220.110.07
DE ELBOBackbone=ConNeXt-Tiny, Model=L2-SP2025.020.540.230.120.07
MAP + GSBackbone=ResNet-50, Model=PTYL2025.020.90.360.150.1
MAP + LaplaceBackbone=ResNet-50, Model=L2-SP2025.020.940.40.150.09
MAP + LaplaceBackbone=ResNet-50, Model=L2-zero2025.020.970.410.190.1
DE ELBOBackbone=ResNet-50, Model=PTYL, Order=22025.0210.460.230.12
MAP + GSBackbone=ResNet-50, Model=PTYL (SSL)2025.021.010.440.240.12
DE ELBOBackbone=ResNet-50, Model=L2-zero2025.021.050.460.20.12
MAP + GSBackbone=ResNet-50, Model=Linear probing2025.021.220.750.550.48
MAP + LaplaceBackbone=ResNet-50, Model=PTYL2025.021.340.50.230.11
DE ELBOBackbone=ResNet-50, Model=PTYL, Order=12025.021.360.760.290.12
MAP + GSBackbone=ResNet-50, Model=L2-zero2025.021.891.060.550.48
DE ELBOBackbone=ResNet-50, Model=L2-SP2025.022.31.180.780.18
MAP + GSBackbone=ResNet-50, Model=L2-SP2025.022.760.540.390.28