Salvato in:
Dettagli Bibliografici
Autori principali: Luo, Yifan, Zhou, Zhennan, Wang, Meitan, Dong, Bin
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2410.10150
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929540916314112
author Luo, Yifan
Zhou, Zhennan
Wang, Meitan
Dong, Bin
author_facet Luo, Yifan
Zhou, Zhennan
Wang, Meitan
Dong, Bin
contents In this paper, we investigate the safety mechanisms of instruction fine-tuned large language models (LLMs). We discover that re-weighting MLP neurons can significantly compromise a model's safety, especially for MLPs in end-of-sentence inferences. We hypothesize that LLMs evaluate the harmfulness of prompts during end-of-sentence inferences, and MLP layers plays a critical role in this process. Based on this hypothesis, we develop 2 novel white-box jailbreak methods: a prompt-specific method and a prompt-general method. The prompt-specific method targets individual prompts and optimizes the attack on the fly, while the prompt-general method is pre-trained offline and can generalize to unseen harmful prompts. Our methods demonstrate robust performance across 7 popular open-source LLMs, size ranging from 2B to 72B. Furthermore, our study provides insights into vulnerabilities of instruction-tuned LLM's safety and deepens the understanding of the internal mechanisms of LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2410_10150
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Jailbreak Instruction-Tuned LLMs via end-of-sentence MLP Re-weighting
Luo, Yifan
Zhou, Zhennan
Wang, Meitan
Dong, Bin
Computation and Language
Artificial Intelligence
In this paper, we investigate the safety mechanisms of instruction fine-tuned large language models (LLMs). We discover that re-weighting MLP neurons can significantly compromise a model's safety, especially for MLPs in end-of-sentence inferences. We hypothesize that LLMs evaluate the harmfulness of prompts during end-of-sentence inferences, and MLP layers plays a critical role in this process. Based on this hypothesis, we develop 2 novel white-box jailbreak methods: a prompt-specific method and a prompt-general method. The prompt-specific method targets individual prompts and optimizes the attack on the fly, while the prompt-general method is pre-trained offline and can generalize to unseen harmful prompts. Our methods demonstrate robust performance across 7 popular open-source LLMs, size ranging from 2B to 72B. Furthermore, our study provides insights into vulnerabilities of instruction-tuned LLM's safety and deepens the understanding of the internal mechanisms of LLMs.
title Jailbreak Instruction-Tuned LLMs via end-of-sentence MLP Re-weighting
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2410.10150