Classification and Uncertainty Quantification on MNIST (test) benchmark leaderboard · SOTA2 Research

Classification and Uncertainty Quantification on MNIST (test)

97.9AccuracyLLA* KFAC

Evaluation Results

Method	Links
LLA* KFAC2026.02		97.9	0.07	0.9	0.923	0.928	31.2	17.6
Ensemble2026.02		97.9	0.069	3.8	0.936	0.962	—	11.9
MAP (backbone)2026.02		97.8	0.068	0.5	0.919	0.919	—	1.24
LLA KFAC2026.02		97.8	0.102	4.2	0.971	0.971	130	2.85
LLA*2026.02		97.8	0.07	0.9	0.924	0.924	42	4.6
ELLA2026.02		97.8	0.068	0.5	0.919	0.912	821.8	148.7
VaLLA2026.02		97.8	0.068	0.5	0.919	0.934	2,190	16.4
VaLLA2026.02		97.8	0.068	0.5	0.919	0.934	3,430	18.3
GPP2026.02		97.8	1.648	78.4	0.934	0.904	5,790	23,500
Dropout2026.02		97.8	0.072	0.9	0.923	0.944	—	4.3
DDU (est.)2026.02		97.8	0.068	0.5	0.921	0.919	202.4	8.2
GAPA-Diag2026.02		97.8	0.073	1.6	0.963	0.976	91.9	2.05
GAPA-Full2026.02		97.8	0.072	1.3	0.969	0.983	96.1	8.92
Linear Probing2026.02		97.7	0.117	1.5	0.884	0.883	2,780	3.6
LLA Diag2026.02		97.6	0.177	10.5	0.932	0.941	2,340	2.39