按年被提到
20232023 · 1 次2026
AI 安全、伦理与社会影响›模型窃取、记忆与水印·arXiv:2311.03348 · cs.CL

Scalable and Transferable Black-Box Jailbreaks for Language Models via Persona Modulation

Rusheb Shah, Quentin Feuillade--Montixi, Soroush Pour, Arush Tagade, Stephen Casper, Javier Rando · 2023
开始阅读 ↗PDFalphaXiv 讨论
被提及 · 2 人首次 2023-11-09 · 最近 2024-03-02
2不同的人
2活跃月份
2讨论串
66被引(OpenAlex)
arXiv 摘要(候选层不写导读)

Despite efforts to align large language models to produce harmless responses, they are still vulnerable to jailbreak prompts that elicit unrestricted behaviour. In this work, we investigate persona modulation as a black-box jailbreaking method to steer a target model to take on personalities that are willing to comply with harmful instructions. Rather than manually crafting prompts for each persona, we automate the generation of jailbreaks using a language model assistant. We demonstrate a range of harmful completions made possible by persona modulation, including detailed instructions for synthesising methamphetamine, building a bomb, and laundering money. These automated attacks achieve a harmful completion rate of 42.5% in GPT-4, which is 185 times larger than before modulation (0.23%). These prompts also transfer to Claude 2 and Vicuna with harmful completion rates of 61.0% and 35.9%, respectively. Our work reveals yet another vulnerability in commercial large language models and highlights the need for more comprehensive safeguards.

候选层:至少 2 个不同的人、在 2 个不同月份提到过,还没到主书架的 3 人 3 个月。没有做立场标注,所以只写“被提及”。

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
评论WhatIsAModel · 2024-03

“https://arxiv.org/abs/2311.03348 This seems to make a decent argument that these models are potentially not safe. I prefer criminals don't have access to a PhD bomb making assistants who can explain the process to them like they are 12. While the cat may be out of the bag, you don't just hand out guns to everyone (for free) because a few people misused t…”

Elon Musk sues Sam Altman, Greg Brockman, and OpenAI [pdf]原评论 ↗
被提交为 story · 1 次