PAVE: Patching and Adapting Video Large Language Models | Read Paper on Bytez