Saved in:
Bibliographic Details
Main Authors: Liu, Zhendong, Nie, Yuanbi, Tan, Yingshui, Yue, Xiangyu, Cui, Qiushi, Wang, Chongjun, Zhu, Xiaoyong, Zheng, Bo
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.13581
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909209035014144
author Liu, Zhendong
Nie, Yuanbi
Tan, Yingshui
Yue, Xiangyu
Cui, Qiushi
Wang, Chongjun
Zhu, Xiaoyong
Zheng, Bo
author_facet Liu, Zhendong
Nie, Yuanbi
Tan, Yingshui
Yue, Xiangyu
Cui, Qiushi
Wang, Chongjun
Zhu, Xiaoyong
Zheng, Bo
contents Benefiting from the powerful capabilities of Large Language Models (LLMs), pre-trained visual encoder models connected to an LLMs can realize Vision Language Models (VLMs). However, existing research shows that the visual modality of VLMs is vulnerable, with attackers easily bypassing LLMs' safety alignment through visual modality features to launch attacks. To address this issue, we enhance the existing VLMs' visual modality safety alignment by adding safety modules, including a safety projector, safety tokens, and a safety head, through a two-stage training process, effectively improving the model's defense against risky images. For example, building upon the LLaVA-v1.5 model, we achieve a safety score of 8.26, surpassing the GPT-4V on the Red Teaming Visual Language Models (RTVLM) benchmark. Our method boasts ease of use, high flexibility, and strong controllability, and it enhances safety while having minimal impact on the model's general performance. Moreover, our alignment strategy also uncovers some possible risky content within commonly used open-source multimodal datasets. Our code will be open sourced after the anonymous review.
format Preprint
id arxiv_https___arxiv_org_abs_2405_13581
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Safety Alignment for Vision Language Models
Liu, Zhendong
Nie, Yuanbi
Tan, Yingshui
Yue, Xiangyu
Cui, Qiushi
Wang, Chongjun
Zhu, Xiaoyong
Zheng, Bo
Computer Vision and Pattern Recognition
Artificial Intelligence
Benefiting from the powerful capabilities of Large Language Models (LLMs), pre-trained visual encoder models connected to an LLMs can realize Vision Language Models (VLMs). However, existing research shows that the visual modality of VLMs is vulnerable, with attackers easily bypassing LLMs' safety alignment through visual modality features to launch attacks. To address this issue, we enhance the existing VLMs' visual modality safety alignment by adding safety modules, including a safety projector, safety tokens, and a safety head, through a two-stage training process, effectively improving the model's defense against risky images. For example, building upon the LLaVA-v1.5 model, we achieve a safety score of 8.26, surpassing the GPT-4V on the Red Teaming Visual Language Models (RTVLM) benchmark. Our method boasts ease of use, high flexibility, and strong controllability, and it enhances safety while having minimal impact on the model's general performance. Moreover, our alignment strategy also uncovers some possible risky content within commonly used open-source multimodal datasets. Our code will be open sourced after the anonymous review.
title Safety Alignment for Vision Language Models
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2405.13581