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 2601.20901cs.CR
收藏

面向敏感性感知的语言模型

Towards Sensitivity-Aware Language Models

University of Luxembourg(卢森堡大学);L3S Research Center(L3S研究中心)

Dren Fazlija, Iyiola E. Olatunji, Daniel Kudenko, Sandipan Sikdar

AI总结:

本文提出了一种方法,通过监督微调提升四比特量化LLMs的敏感性感知能力,使其在隐私保护方面表现更优,同时保持其他任务性能。

摘要:

随着大语言模型(LLMs)越来越多地应用于企业数据管理,确保这些模型不泄露敏感信息变得至关重要。在企业数据管理的背景下,引入了敏感性感知的概念,使LLMs能够遵守预定义的访问权限规则。然而,敏感性感知与已建立的隐私概念,如差分隐私(DP)之间的关系仍然不明确,这使得在现实应用中难以有意义地部署。在本工作中,我们正式定义了敏感性感知的概念,并在理论上建立了其与DP的联系。此外,我们开发了一种监督微调配方,使现有的四比特量化LLMs更加敏感性感知。在性能提升高达21.7%的情况下,微调后的LLMs不仅显著优于基线模型,还优于其他全精度开源和商业模型,在实现敏感性感知方面表现更佳,证明了我们所提出方法的有效性。同时,我们的方法在其他任务上也 largely 保持了模型的性能,如通用指令遵循、数学和常识推理。

英文摘要:

With LLMs increasingly deployed in corporate data management, it is crucial to ensure that these models do not leak sensitive information. In the context of corporate data management, the concept of sensitivity awareness has been introduced, enabling LLMs to adhere to predefined access rights rules. However, it remains unclear how sensitivity awareness relates to established notions of privacy, such as differential privacy (DP), thereby making it difficult to deploy meaningfully in real-world applications. In this work, we formalize the notion of sensitivity awareness and theoretically establish its connection to DP. Additionally, we develop a supervised fine-tuning recipe to make existing, four-bit quantized LLMs more sensitivity-aware. With a performance boost of up to 21.7%, the finetuned LLMs not only substantially improve over their baseline but also outperform other full-precision open-source and commercial models of similar size in achieving sensitivity awareness, demonstrating the effectiveness of our proposed approach. At the same time, our method also largely preserves the models' performance on other tasks, such as general instruction-following, mathematical, and common-sense reasoning.

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。

↑