Training the Untrainable: Introducing Inductive Bias via Representational Alignment | Read Paper on Bytez