Aligning the Pretraining and Finetuning Objectives of Language Models | Read Paper on Bytez