arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2604.06154cs.CL
收藏

排除性反学习

Exclusive Unlearning

Tohoku University(东北大学);NII LLMC(国立信息学研究所 LLMC);The University of Tokyo(东京大学)

Mutsumi Sasaki, Kouta Nakayama, Yusuke Miyao, Yohei Oseki, Masaru Isonuma

AI总结:

本文提出排除性反学习方法,通过广泛遗忘非目标知识以实现广泛危害消除,确保模型在多种输入下安全,同时保持对特定领域指令的响应能力。

摘要:

在将大型语言模型(LLMs)引入工业应用,如医疗和教育领域时,生成有害内容的风险成为重大挑战。尽管现有机器反学习方法可以删除特定有害知识和表达,但多样化有害内容使全面删除变得困难。在本研究中,我们提出排除性反学习(EU),旨在通过广泛遗忘一切非保留的知识和表达来实现广泛危害消除。我们证明,通过排除性反学习,可以获取一个在广泛输入下确保安全的模型,包括对抗性攻击,同时保持对如医学和数学等特定领域指令的响应能力。

英文摘要:

When introducing Large Language Models (LLMs) into industrial applications, such as healthcare and education, the risk of generating harmful content becomes a significant challenge. While existing machine unlearning methods can erase specific harmful knowledge and expressions, diverse harmful content makes comprehensive removal difficult. In this study, instead of individually listing targets for forgetting, we propose Exclusive Unlearning (EU), which aims for broad harm removal by extensively forgetting everything except for the knowledge and expressions we wish to retain. We demonstrate that through Exclusive Unlearning, it is possible to obtain a model that ensures safety against a wide range of inputs, including jailbreaks, while maintaining the ability to respond to diverse instructions related to specific domains such as medicine and mathematics.

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。

↑