Balancing Fidelity and Diversity in Diffusion Models via Symmetric Attention Decomposition: Hopfield Perspective

Hyunmin Cho, Woo Kyoung Han, Kyong Hwan Jin

We characterize the pre-softmax attention matrix $\mathbf{Q}{\mathbf{K}}^{\top }$ in transformers as an associative memory matrix encoding pairwise associations between input features. By decomp…