Interpretable Lightweight Transformer via Unrolling of Learned Graph Smoothness Priors | Read Paper on Bytez