A Few Gaussians Is All You Need: SSOG-Attention That Steers Instead of Scores
Open paper
Benchmark
Model
Rank
Results
image-classification-on-cifar-100
SSOG factorized +μδ
#210
Percentage correct: 70.4
image-classification-on-imagenet
SSOG +μδ+σδ+λ (12.0M)
#1053
Top 1 Accuracy: 72.02%
Rank counts only results with a code link.