Revisiting Backdoor Attacks against Large Vision-Language Models from Domain Shift | Read Paper on Bytez