Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Hua, Jiaqi, Wei, Wanxu
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2501.07959
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913673689169920
author Hua, Jiaqi
Wei, Wanxu
author_facet Hua, Jiaqi
Wei, Wanxu
contents Recently, several works have been conducted on jailbreaking Large Language Models (LLMs) with few-shot malicious demos. In particular, Zheng et al. focus on improving the efficiency of Few-Shot Jailbreaking (FSJ) by injecting special tokens into the demos and employing demo-level random search, known as Improved Few-Shot Jailbreaking (I-FSJ). Nevertheless, we notice that this method may still require a long context to jailbreak advanced models e.g. 32 shots of demos for Meta-Llama-3-8B-Instruct (Llama-3) \cite{llama3modelcard}. In this paper, we discuss the limitations of I-FSJ and propose Self-Instruct Few-Shot Jailbreaking (Self-Instruct-FSJ) facilitated with the demo-level greedy search. This framework decomposes the FSJ attack into pattern and behavior learning to exploit the model's vulnerabilities in a more generalized and efficient way. We conduct elaborate experiments to evaluate our method on common open-source models and compare it with baseline algorithms. Our code is available at https://github.com/iphosi/Self-Instruct-FSJ.
format Preprint
id arxiv_https___arxiv_org_abs_2501_07959
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Self-Instruct Few-Shot Jailbreaking: Decompose the Attack into Pattern and Behavior Learning
Hua, Jiaqi
Wei, Wanxu
Artificial Intelligence
Recently, several works have been conducted on jailbreaking Large Language Models (LLMs) with few-shot malicious demos. In particular, Zheng et al. focus on improving the efficiency of Few-Shot Jailbreaking (FSJ) by injecting special tokens into the demos and employing demo-level random search, known as Improved Few-Shot Jailbreaking (I-FSJ). Nevertheless, we notice that this method may still require a long context to jailbreak advanced models e.g. 32 shots of demos for Meta-Llama-3-8B-Instruct (Llama-3) \cite{llama3modelcard}. In this paper, we discuss the limitations of I-FSJ and propose Self-Instruct Few-Shot Jailbreaking (Self-Instruct-FSJ) facilitated with the demo-level greedy search. This framework decomposes the FSJ attack into pattern and behavior learning to exploit the model's vulnerabilities in a more generalized and efficient way. We conduct elaborate experiments to evaluate our method on common open-source models and compare it with baseline algorithms. Our code is available at https://github.com/iphosi/Self-Instruct-FSJ.
title Self-Instruct Few-Shot Jailbreaking: Decompose the Attack into Pattern and Behavior Learning
topic Artificial Intelligence
url https://arxiv.org/abs/2501.07959