@article {2006.16236, archivePrefix = {arXiv}, eprint = {2006.16236}, primaryClass = {cs.LG}, author = {Angelos Katharopoulos and Apoorv Vyas and Nikolaos Pappas and François Fleuret}, title = {Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention}, year = {2020}, url = {http://arxiv.org/abs/2006.16236} }