# Jailbreaks on Vision Language Model via Multimodal Reasoning

Source: [Jailbreaks on Vision Language Model via Multimodal Reasoning](<https://www.sota2.com/research/paper/69802fbba8b01e76cc8e1718>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-03T05:00:04.373Z

Authors: Aarush Noheria, Yuguang Yao

Published: 2026-01-29T23:09:24.000Z

## Abstract

Vision-language models (VLMs) have become central to tasks such as visual question answering, image captioning, and text-to-image generation. However, their outputs are highly sensitive to prompt variations, which can reveal vulnerabilities in safety alignment. In this work, we present a jailbreak framework that exploits post-training Chain-of-Thought (CoT) prompting to construct stealthy prompts capable of bypassing safety filters. To further increase attack success rates (ASR), we propose a ReAct-driven adaptive noising mechanism that iteratively perturbs input images based on model feedback. This approach leverages the ReAct paradigm to refine adversarial noise in regions most likely to activate safety defenses, thereby enhancing stealth and evasion. Experimental results demonstrate that the proposed dual-strategy significantly improves ASR while maintaining naturalness in both text and visual domains.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Jailbreak Attack on VLGuard Text Unsafe](<https://www.sota2.com/research/sota/jailbreak-attack-on-vlguard-text-unsafe.md>): Jailbreak Attack; VLGuard Text Unsafe; ASR: 34.59.

- [Jailbreak Attack on VLGuard Image Unsafe](<https://www.sota2.com/research/sota/jailbreak-attack-on-vlguard-image-unsafe.md>): Jailbreak Attack; VLGuard Image Unsafe; ASR: 52.49.

- [Jailbreak Attack on VLGuard Safe](<https://www.sota2.com/research/sota/jailbreak-attack-on-vlguard-safe.md>): Jailbreak Attack; VLGuard Safe; Attack Success Rate (ASR): 7.53.

- [Jailbreak Attack on SPA-VL Harm](<https://www.sota2.com/research/sota/jailbreak-attack-on-spa-vl-harm.md>): Jailbreak Attack; SPA-VL Harm; ASR: 49.81.

- [Jailbreak Attack on SPA-VL Help](<https://www.sota2.com/research/sota/jailbreak-attack-on-spa-vl-help.md>): Jailbreak Attack; SPA-VL Help; ASR: 32.83.

- [Jailbreak Attack on VLGuard (All)](<https://www.sota2.com/research/sota/jailbreak-attack-on-vlguard-all.md>): Jailbreak Attack; VLGuard (All); ASR: 41.3.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2601.22398>)

[PDF](<https://arxiv.org/pdf/2601.22398.pdf>)
