Using the Output Embedding to Improve Language Models | Read Paper on Bytez