Implicit Multimodal Alignment: On the Generalization of Frozen LLMs to Multimodal Inputs | Read Paper on Bytez