1 paper
Ron Keuth, Paul Kaftan, Mattias P. Heinrich
The generalization of the Transformer architecture via MetaFormer has reshaped our understanding of its success in computer vision. By replacing self-attention with simpler token m…