Few-Shot Adversarial Prompt Learning on Vision-Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhou, Yiwei, Xia, Xiaobo, Lin, Zhiwei, Han, Bo, Liu, Tongliang
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910661959745536
author Zhou, Yiwei
Xia, Xiaobo
Lin, Zhiwei
Han, Bo
Liu, Tongliang
author_facet Zhou, Yiwei
Xia, Xiaobo
Lin, Zhiwei
Han, Bo
Liu, Tongliang
contents The vulnerability of deep neural networks to imperceptible adversarial perturbations has attracted widespread attention. Inspired by the success of vision-language foundation models, previous efforts achieved zero-shot adversarial robustness by aligning adversarial visual features with text supervision. However, in practice, they are still unsatisfactory due to several issues, including heavy adaptation cost, suboptimal text supervision, and uncontrolled natural generalization capacity. In this paper, to address these issues, we propose a few-shot adversarial prompt framework where adapting input sequences with limited data makes significant adversarial robustness improvement. Specifically, we achieve this by providing adversarially correlated text supervision that is end-to-end learned from adversarial examples. We also propose a novel training objective that enhances the consistency of multi-modal features while encourages differentiated uni-modal features between natural and adversarial examples. The proposed framework gives access to learn adversarial text supervision, which provides superior cross-modal adversarial alignment and matches state-of-the-art zero-shot adversarial robustness with only 1% training data. Code is available at: https://github.com/lionel-w2/FAP.
format Preprint
id arxiv_https___arxiv_org_abs_2403_14774
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Few-Shot Adversarial Prompt Learning on Vision-Language Models
Zhou, Yiwei
Xia, Xiaobo
Lin, Zhiwei
Han, Bo
Liu, Tongliang
Computer Vision and Pattern Recognition
Computation and Language
Cryptography and Security
Machine Learning
The vulnerability of deep neural networks to imperceptible adversarial perturbations has attracted widespread attention. Inspired by the success of vision-language foundation models, previous efforts achieved zero-shot adversarial robustness by aligning adversarial visual features with text supervision. However, in practice, they are still unsatisfactory due to several issues, including heavy adaptation cost, suboptimal text supervision, and uncontrolled natural generalization capacity. In this paper, to address these issues, we propose a few-shot adversarial prompt framework where adapting input sequences with limited data makes significant adversarial robustness improvement. Specifically, we achieve this by providing adversarially correlated text supervision that is end-to-end learned from adversarial examples. We also propose a novel training objective that enhances the consistency of multi-modal features while encourages differentiated uni-modal features between natural and adversarial examples. The proposed framework gives access to learn adversarial text supervision, which provides superior cross-modal adversarial alignment and matches state-of-the-art zero-shot adversarial robustness with only 1% training data. Code is available at: https://github.com/lionel-w2/FAP.
title Few-Shot Adversarial Prompt Learning on Vision-Language Models
topic Computer Vision and Pattern Recognition
Computation and Language
Cryptography and Security
Machine Learning
url https://arxiv.org/abs/2403.14774