Layer-Wise Perturbations via Sparse Autoencoders for Adversarial Text Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Shu, Huizhen, Li, Xuying, Wang, Qirui, Kosuga, Yuji, Tian, Mengqiu, Li, Zhuo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915446422241280
author Shu, Huizhen
Li, Xuying
Wang, Qirui
Kosuga, Yuji
Tian, Mengqiu
Li, Zhuo
author_facet Shu, Huizhen
Li, Xuying
Wang, Qirui
Kosuga, Yuji
Tian, Mengqiu
Li, Zhuo
contents With the rapid proliferation of Natural Language Processing (NLP), especially Large Language Models (LLMs), generating adversarial examples to jailbreak LLMs remains a key challenge for understanding model vulnerabilities and improving robustness. In this context, we propose a new black-box attack method that leverages the interpretability of large models. We introduce the Sparse Feature Perturbation Framework (SFPF), a novel approach for adversarial text generation that utilizes sparse autoencoders to identify and manipulate critical features in text. After using the SAE model to reconstruct hidden layer representations, we perform feature clustering on the successfully attacked texts to identify features with higher activations. These highly activated features are then perturbed to generate new adversarial texts. This selective perturbation preserves the malicious intent while amplifying safety signals, thereby increasing their potential to evade existing defenses. Our method enables a new red-teaming strategy that balances adversarial effectiveness with safety alignment. Experimental results demonstrate that adversarial texts generated by SFPF can bypass state-of-the-art defense mechanisms, revealing persistent vulnerabilities in current NLP systems.However, the method's effectiveness varies across prompts and layers, and its generalizability to other architectures and larger models remains to be validated.
format Preprint
id arxiv_https___arxiv_org_abs_2508_10404
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Layer-Wise Perturbations via Sparse Autoencoders for Adversarial Text Generation
Shu, Huizhen
Li, Xuying
Wang, Qirui
Kosuga, Yuji
Tian, Mengqiu
Li, Zhuo
Computation and Language
Artificial Intelligence
With the rapid proliferation of Natural Language Processing (NLP), especially Large Language Models (LLMs), generating adversarial examples to jailbreak LLMs remains a key challenge for understanding model vulnerabilities and improving robustness. In this context, we propose a new black-box attack method that leverages the interpretability of large models. We introduce the Sparse Feature Perturbation Framework (SFPF), a novel approach for adversarial text generation that utilizes sparse autoencoders to identify and manipulate critical features in text. After using the SAE model to reconstruct hidden layer representations, we perform feature clustering on the successfully attacked texts to identify features with higher activations. These highly activated features are then perturbed to generate new adversarial texts. This selective perturbation preserves the malicious intent while amplifying safety signals, thereby increasing their potential to evade existing defenses. Our method enables a new red-teaming strategy that balances adversarial effectiveness with safety alignment. Experimental results demonstrate that adversarial texts generated by SFPF can bypass state-of-the-art defense mechanisms, revealing persistent vulnerabilities in current NLP systems.However, the method's effectiveness varies across prompts and layers, and its generalizability to other architectures and larger models remains to be validated.
title Layer-Wise Perturbations via Sparse Autoencoders for Adversarial Text Generation
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2508.10404