JailBreakV: A Benchmark for Assessing the Robustness of MultiModal Large Language Models against Jailbreak Attacks

2024-04-03 · Weidi Luo, Siyuan Ma, Xiaogeng Liu, XIAOYU GUO, Chaowei Xiao

With the rapid advancements in Multimodal Large Language Models (MLLMs), securing these models against malicious inputs while aligning them with human values has emerged as a critical challenge. In this paper, we investigate an important and unexplored question of whether techniques that successfully jailbreak Large Language Models (LLMs) can be equally effective in jailbreaking MLLMs. To explore this issue, we introduce JailBreakV-28K, a pioneering benchmark designed to assess the transferability of LLM jailbreak techniques to MLLMs, thereby evaluating the robustness of MLLMs against diverse jailbreak attacks. Utilizing a dataset of 2, 000 malicious queries that is also proposed in this paper, we generate 20, 000 text-based jailbreak prompts using advanced jailbreak attacks on LLMs, alongside 8, 000 image-based jailbreak inputs from recent MLLMs jailbreak attacks, our comprehensive dataset includes 28, 000 test cases across a spectrum of adversarial scenarios. Our evaluation of 10 open-source MLLMs reveals a notably high Attack Success Rate (ASR) for attacks transferred from LLMs, highlighting a critical vulnerability in MLLMs that stems from their text-processing capabilities. Our findings underscore the urgent need for future research to address alignment vulnerabilities in MLLMs from both textual and visual inputs.

📄 PDF Abstract BibTeX arXiv:2404.03027

Code (1)

EddyLuo1232/JailBreakV_28K pytorch

Tasks

LLM Jailbreak

PRISM: Robust VLM Alignment with Principled Reasoning for Integrated Safety in Multimodality

2025-08-26 · Nanxi Li, Zhengyue Zhao, G. Edward Suh, Marco Pavone 외 arxiv

Safeguarding vision-language models (VLMs) is a critical challenge, as existing methods often suffer from over-defense, which harms utility, or rely on shallow alignment, failing to detect complex threats that require de…

Safety Recovery in Reasoning Models Is Only a Few Early Steering Steps Away

2026-02-11 · Soumya Suvra Ghosal, Souradip Chakraborty, Vaibhav Singh, Furong Huang 외 arxiv

Reinforcement learning (RL) based post-training for explicit chain-of-thought (e.g., GRPO) improves the reasoning ability of multimodal large-scale reasoning models (MLRMs). But recent evidence shows that it can simultan…

Reinforcement Learning

Structured Visual Narratives Undermine Safety Alignment in Multimodal Large Language Models

2026-03-23 · Rui Yang Tan, Yujia Hu, Roy Ka-Wei Lee arxiv

Multimodal Large Language Models (MLLMs) extend text-only LLMs with visual reasoning, but also introduce new safety failure modes under visually grounded instructions. We study comic-template jailbreaks that embed harmfu…

Visual Reasoning

Robust Multimodal Safety via Conditional Decoding

2026-03-31 · Anurag Kumar, Raghuveer Peri, Jon Burnsky, Alexandru Nelus 외 arxiv

Multimodal large-language models (MLLMs) often experience degraded safety alignment when harmful queries exploit cross-modal interactions. Models aligned on text alone show a higher rate of successful attacks when extend…

Response Generation

Res-Bench: Benchmarking the Robustness of Multimodal Large Language Models to Dynamic Resolution Input

2025-10-19 · Chenxu Li, Zhicai Wang, Yuan Sheng, Xingyu Zhu 외 arxiv

Multimodal Large Language Models (MLLMs) increasingly support dynamic image resolutions. However, current evaluation paradigms primarily assess semantic performance, overlooking the critical question of resolution robust…