Skip to content
Search
paperAugust 2026Unreviewed

A Multimodal Automatic Redteaming Evaluation based on Atomic Jailbreak Strategy Decoupling and Combination

Shiji Zhao, Yuxuan Zhou, Chen Xiong, Dongxian Wu, Yang Bai, Xun Chen

Abstract

Multimodal Large Language Models (MLLMs) have achieved impressive progress in image-text comprehension and generation, yet they remain susceptible to jailbreak attacks that can trigger harmful outputs and pose serious safety concerns. Existing multimodal jailbreak attacks have shown the feasibility of such attacks, but they still face two fundamental challenges: the lack of a atomic multi-modal strategy space, the absence of a concise and efficient executable framework beyond human-craft experie

Categories

Framework mappings

MITRE ATLAS
  • AML.T0054LLM Jailbreak

Suggested from the entry's categories.

Cite

@misc{zhao2026multimodal,
  title = {{A Multimodal Automatic Redteaming Evaluation based on Atomic Jailbreak Strategy Decoupling and Combination}},
  author = {Shiji Zhao and Yuxuan Zhou and Chen Xiong and Dongxian Wu and Yang Bai and Xun Chen},
  year = {2026},
  month = aug,
  eprint = {2608.04034},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2608.04034}
}