作者位次 #5/11ACL 2026Findings· Findings of ACL
Making MLLM s Blind: Adversarial Smuggling Attacks in MLLM Content Moderation
Zhiheng Li · Zongyang Ma · Yuntong Pan · Ziqi Zhang · Xiaolei Lv · Bo Li · Jun Gao · Jianing Zhang · Chunfeng Yuan · Bing Li · Weiming Hu
摘要
Multimodal Large Language Models (MLLMs) are increasingly being deployed as automated content moderators. Within this landscape, we uncover a critical threat: Adversarial Smuggling Attacks. Unlike adversarial perturbations (for misclassification) and adversarial jailbreaks (for harmful output gen...