AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Jailbreaking Multimodal Large Language Models via Shuffle Inconsistency

Shiji Zhao, Ranjie Duan, Fengxiang Wang, Chi Chen, Caixin Kang, Shouwei Ruan, Jialing Tao, YueFeng Chen, Hui Xue, Xingxing Wei

ICCV 2025 · Adversarial robustness and security › Attacks on text and graphs · approach: LLM jailbreak and injection defense · 50 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
shuffle harmful text wordsoriginal harmful text prompts · MM-SafetyBench 01-IA, GPT-4otoxic score: 1.51 -> 3.19Improved
shuffle harmful image patchesoriginal harmful images · MM-SafetyBench 01-IA, GPT-4otoxic score: 1.51 -> 2.13Improved
shuffle image and text, vs other mutationstext mutation · MM-SafetyBench 01-IAtoxic score: 2.75 -> 3.19Improved
add image shuffling (ablation)SI-Attack · MM-SafetyBench 01-IA no typography, GPT-4oASR (%): 13.40 -> 35.05Improved
add text shuffling (ablation)SI-Attack · MM-SafetyBench 01-IA no typography, GPT-4oASR (%): 13.40 -> 67.01Improved
add image shuffling to text shuffling (ablation)SI-Attack · MM-SafetyBench 01-IA no typography, GPT-4oASR (%): 67.01 -> 80.41Improved
add query-based optimization to random shuffling (ablation)SI-Attack · MM-SafetyBench 01-IA no typography, GPT-4oASR (%): 28.87 -> 80.41Improved
add query-based optimization to random shuffling (ablation)SI-Attack · MM-SafetyBench 01-IA no typography, GPT-4otoxic score: 2.65 -> 3.96Improved
increase max query iteration from 1 to 10 (ablation)SI-Attack · MM-SafetyBench 01-IA no typography, GPT-4oASR (%): 28.87 -> 80.41Improved
increase max query iteration from 10 to 20 (ablation)SI-Attack · MM-SafetyBench 01-IA no typography, GPT-4oASR (%): 80.41 -> 81.44No clear effect
use SI-Attack instead of QR-Attackjailbreak attack · MM-SafetyBench on GPT-4oASR (%): 20.77 -> 68.57Improved
use SI-Attack instead of QR-Attackjailbreak attack · MM-SafetyBench on VLGuardASR (%): 9.52 -> 40.77Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy