This repository contains the implementation of Con Instruction, introduced in the ACL 2025 paper: Con Instruction: Universal Jailbreaking of Multimodal Large Language Models via Non-Textual Modalities ...