Efficient LLM-Jailbreaking via Multimodal-LLM Jailbreak
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (42) · 35527-35535
DOI: 10.1609/aaai.v40i42.40863 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Haoxuan Ji und 4 weiteren Autor:innen.
Abstract (Originalsprache)
This paper focuses on jailbreaking attacks against large language models (LLMs), eliciting them to generate objectionable content in response to harmful user queries. Unlike previous LLM-jailbreak methods that directly orient to LLMs, our approach begins by constructing a multimodal large language model (MLLM) built upon the target LLM. Subsequently, we perform an efficient MLLM jailbreak and obtain a jailbreaking embedding. Finally, we convert the embedding into a textual jailbreaking suffix to carry out the jailbreak of target LLM. Compared to the direct LLM-jailbreak methods, our indirect jailbreaking approach is more efficient, as MLLMs are more vulnerable to jailbreak than pure LLMs. Additionally, to improve the attack success rate of jailbreak, we propose an image-text semantic matching scheme to identify a suitable initial input. Extensive experiments demonstrate that our approach surpasses current state-of-the-art jailbreak methods in terms of both efficiency and effectiveness. Moreover, our approach exhibits superior cross-class generalization abilities.
Zitieren
APA
Ji, H., Lin, Z., Niu, Z., Gao, X., & Hua, G. (2026). Efficient LLM-Jailbreaking via Multimodal-LLM Jailbreak. Proceedings of the AAAI Conference on Artificial Intelligence, 40(42), 35527-35535. https://doi.org/10.1609/aaai.v40i42.40863
Vancouver
Ji H, Lin Z, Niu Z, Gao X, Hua G. Efficient LLM-Jailbreaking via Multimodal-LLM Jailbreak. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(42):35527-35535. doi:10.1609/aaai.v40i42.40863
BibTeX
@article{Ji2026,
author = {Ji, Haoxuan and Lin, Zheng and Niu, Zhenxing and Gao, Xinbo and Hua, Gang},
title = {Efficient LLM-Jailbreaking via Multimodal-LLM Jailbreak},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {42},
pages = {35527--35535},
doi = {10.1609/aaai.v40i42.40863}
}
EUVIMED