@inproceedings{chen-etal-2026-side,
    title = "The Side Effects of Being Smart: Safety Risks in {MLLM}s' Multi-Image Reasoning",
    author = "Chen, Renmiao  and
      Lu, Yida  and
      Cui, Shiyao  and
      Ouyang, Xuan  and
      Huang, Victor Shea-Jay  and
      Zhang, Shumin  and
      Pan, Chengwei  and
      Qiu, Han  and
      Huang, Minlie",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-long.1710/",
    pages = "36866--36882",
    ISBN = "979-8-89176-390-6",
    abstract = "As Multimodal Large Language Models (MLLMs) acquire stronger reasoning capabilities to handle complex, multi-image instructions, this advancement may pose new safety risks. We study this problem by introducing MIR-SafetyBench, the first benchmark focused on multi-image reasoning safety, which consists of 2,676 instances across a taxonomy of 9 multi-image relations. Our extensive evaluations on 19 MLLMs reveal a troubling trend: models with more advanced multi-image reasoning can be more vulnerable on MIR-SafetyBench. Beyond attack success rates, we find that many responses labeled as safe are superficial, often driven by misunderstanding or evasive, non-committal replies. We further observe that unsafe generations exhibit lower attention entropy than safe ones on average. This internal signature suggests a possible risk that models may over-focus on task solving while neglecting safety constraints."
}