arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2501.02997cs.AIcs.CL
收藏

CALM:面向大语言模型的好奇心驱动审计

CALM: Curiosity-Driven Auditing for Large Language Models

City University of Hong Kong(香港城市大学);Fudan University(复旦大学)

Xiang Zheng, Longxiang Wang, Yi Liu, Xingjun Ma, Chao Shen, Cong Wang

AI总结:

该研究将黑盒 LLM 审计建模为离散黑盒优化问题,提出 CALM,通过内在动机强化学习微调审计智能体,以自动发现目标模型的有害、偏见及幻觉输入-输出对。

摘要:

审计大语言模型(LLM)是一项关键且具有挑战性的任务。在本研究中,我们聚焦于在无法获取其参数、只能访问所提供服务的情况下审计黑盒 LLM。我们将此类审计视为一个黑盒优化问题,目标是自动发现目标 LLM 表现出非法、不道德或不安全行为的输入-输出对。例如,我们可能寻找一个本身无毒、但目标 LLM 会对其作出有毒回复的输入,或者寻找一个会诱发目标 LLM 产生包含政治敏感人物的幻觉回复的输入。由于可行点稀少、提示词空间具有离散性以及搜索空间巨大,这种黑盒优化颇具挑战。为应对这些挑战,我们提出了面向大语言模型的好奇心驱动审计方法 CALM,它利用内在动机强化学习来微调一个作为审计智能体的 LLM,以发现目标 LLM 潜在的有害且有偏见的输入-输出对。CALM 在黑盒设置下成功识别出涉及名人的贬损性补全,并发现了可诱发特定名字的输入。本工作为审计黑盒 LLM 提供了一个有前景的方向。我们的代码可在 https://github.com/x-zheng16/CALM.git 获取。

英文摘要:

Auditing Large Language Models (LLMs) is a crucial and challenging task. In this study, we focus on auditing black-box LLMs without access to their parameters, only to the provided service. We treat this type of auditing as a black-box optimization problem where the goal is to automatically uncover input-output pairs of the target LLMs that exhibit illegal, immoral, or unsafe behaviors. For instance, we may seek a non-toxic input that the target LLM responds to with a toxic output or an input that induces the hallucinative response from the target LLM containing politically sensitive individuals. This black-box optimization is challenging due to the scarcity of feasible points, the discrete nature of the prompt space, and the large search space. To address these challenges, we propose Curiosity-Driven Auditing for Large Language Models (CALM), which uses intrinsically motivated reinforcement learning to finetune an LLM as the auditor agent to uncover potential harmful and biased input-output pairs of the target LLM. CALM successfully identifies derogatory completions involving celebrities and uncovers inputs that elicit specific names under the black-box setting. This work offers a promising direction for auditing black-box LLMs. Our code is available at https://github.com/x-zheng16/CALM.git.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑