(Translated by https://www.hiragana.jp/)
[2405.20015] Efficient LLM-Jailbreaking by Introducing Visual Modality