RAT: Bridging RNN Efficiency and Attention Accuracy in Language Modeling | Xiaol.x | Podwise