New transformer architecture can make language models faster and resource-efficient

ETH Zurich's new transformer architecture enhances language model efficiency, preserving accuracy while reducing size and computational demands.