Test-Time Backdoor Attacks on Multimodal Large Language Models

Open in new window