MNIST
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
MNIST MT (test)
99.2Accuracy
8
MNIST-M
95.4Accuracy
8
MNIST MT
99.2Accuracy
8
RotMNIST
99.25Accuracy
8
MNIST-10 multi-class (test)
5Weighted Accuracy Drop
8
MNIST-2 binarized (test)
0.006Weighted Accuracy Drop
8
MNIST 2
0.837AUC
8
ColorMNIST
0.96MSSSIM
8
MNIST binarized (val)
-90.1Mean Log-Likelihood
8
MNIST (test)
17.49AU-MSE
8
MNIST
0.948AUAC
8
MNIST
10Time (s)
8
MNIST (test)
0.094Test Loss
8
MNIST ε = 1.58 (test)
99Clean Accuracy
8
MNIST synthetic label noise (test)
98.74Accuracy (Symmetric, 20%)
8
MNIST to MNIST-M
99.5Accuracy
8
MNIST non-permutation (test)
99.2Accuracy
8
MNIST 1K
917.9Index Score
8
MNIST Sequence (test)
104.3CLL Score
8
MNIST Gaussian
100AUROC
8
MNIST (test)
-1,030.3Avg Log Likelihood (nats)
8
MNIST (test)
-1,038.5Avg Test Log Likelihood (nats)
8
MNIST
1.39Bits per Pixel (BPP)
8
MNIST
1.41Bits per Pixel
8
permuted sequential MNIST (test)
98.4Test Accuracy
8