Pay Less Attention to Function Words for Free Robustness of Vision-Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tian, Qiwei, Lin, Chenhao, Zhao, Zhengyu, Shen, Chao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915940003741696
author Tian, Qiwei
Lin, Chenhao
Zhao, Zhengyu
Shen, Chao
author_facet Tian, Qiwei
Lin, Chenhao
Zhao, Zhengyu
Shen, Chao
contents To address the trade-off between robustness and performance for robust VLM, we observe that function words could incur vulnerability of VLMs against cross-modal adversarial attacks, and propose Function-word De-Attention (FDA) accordingly to mitigate the impact of function words. Similar to differential amplifiers, our FDA calculates the original and the function-word cross-attention within attention heads, and differentially subtracts the latter from the former for more aligned and robust VLMs. Comprehensive experiments include 2 SOTA baselines under 6 different attacks on 2 downstream tasks, 3 datasets, and 3 models. Overall, our FDA yields an average 18/13/53% ASR drop with only 0.2/0.3/0.6% performance drops on the 3 tested models on retrieval, and a 90% ASR drop with a 0.3% performance gain on visual grounding. We demonstrate the scalability, generalization, and zero-shot performance of FDA experimentally, as well as in-depth ablation studies and analysis. Code is available at https://github.com/michaeltian108/FDA.
format Preprint
id arxiv_https___arxiv_org_abs_2512_07222
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Pay Less Attention to Function Words for Free Robustness of Vision-Language Models
Tian, Qiwei
Lin, Chenhao
Zhao, Zhengyu
Shen, Chao
Machine Learning
Computation and Language
To address the trade-off between robustness and performance for robust VLM, we observe that function words could incur vulnerability of VLMs against cross-modal adversarial attacks, and propose Function-word De-Attention (FDA) accordingly to mitigate the impact of function words. Similar to differential amplifiers, our FDA calculates the original and the function-word cross-attention within attention heads, and differentially subtracts the latter from the former for more aligned and robust VLMs. Comprehensive experiments include 2 SOTA baselines under 6 different attacks on 2 downstream tasks, 3 datasets, and 3 models. Overall, our FDA yields an average 18/13/53% ASR drop with only 0.2/0.3/0.6% performance drops on the 3 tested models on retrieval, and a 90% ASR drop with a 0.3% performance gain on visual grounding. We demonstrate the scalability, generalization, and zero-shot performance of FDA experimentally, as well as in-depth ablation studies and analysis. Code is available at https://github.com/michaeltian108/FDA.
title Pay Less Attention to Function Words for Free Robustness of Vision-Language Models
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2512.07222