To address this issue, researchers at ETH Zurich have unveiled a revised version of the transformer, the deep learning architecture underlying language models. The new design reduces the size of ...