arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2408.04237cs.CL
收藏

学习改写:面向泛化的LLM生成文本检测

Learning to Rewrite: Generalized LLM-Generated Text Detection

Columbia University(哥伦比亚大学);Rutgers University(罗格斯大学)

Ran Li, Wei Hao, Weiliang Zhao, Junfeng Yang, Chengzhi Mao

AI总结:

本文提出Learning2Rewrite框架,通过训练LLM放大其改写AI文本与人类文本时的编辑距离差异,实现跨领域、抗对抗攻击的生成文本检测。

摘要:

大型语言模型(LLM)在被用于大规模生成非事实内容和传播虚假信息时会带来重大风险。检测此类LLM生成内容至关重要,但现有检测器往往难以在开放世界场景中实现泛化。我们提出Learning2Rewrite,这是一种用于检测AI生成文本的新颖框架,对未见领域具有出色的泛化能力。我们的方法利用了这样一种洞见:当被要求执行改写任务时,LLM对AI生成内容的固有修改幅度会小于对人类撰写文本的修改幅度。通过训练LLM使其尽量减少对AI生成输入的改动,我们放大了这种差异,从而在多样的文本分布上产生更具区分度且更易泛化的编辑距离。在来自21个独立领域和四大主流LLM(GPT-3.5、GPT-4、Gemini和Llama-3)的数据上进行的大量实验表明,我们的检测器优于最先进检测方法:在分布内测试中AUROC最高提升23.04%,在分布外测试中最高提升37.26%,在对抗攻击下最高提升48.66%。在使用相同参数量的情况下,与直接训练分类模型相比,我们独特的训练目标确保了更好的泛化能力。研究结果表明,强化LLM固有的改写倾向为检测AI生成文本提供了一种稳健且可扩展的解决方案。

英文摘要:

Large language models (LLMs) present significant risks when used to generate non-factual content and spread disinformation at scale. Detecting such LLM-generated content is crucial, yet current detectors often struggle to generalize in open-world contexts. We introduce Learning2Rewrite, a novel framework for detecting AI-generated text with exceptional generalization to unseen domains. Our method leverages the insight that LLMs inherently modify AI-generated content less than human-written text when tasked with rewriting. By training LLMs to minimize alterations on AI-generated inputs, we amplify this disparity, yielding a more distinguishable and generalizable edit distance across diverse text distributions. Extensive experiments on data from 21 independent domains and four major LLMs (GPT-3.5, GPT-4, Gemini, and Llama-3) demonstrate that our detector outperforms state-of-the-art detection methods by up to 23.04% in AUROC for in-distribution tests, 37.26% for out-of-distribution tests, and 48.66% under adversarial attacks. Our unique training objective ensures better generalizability compared to directly training for classification, when leveraging the same amount of parameters. Our findings suggest that reinforcing LLMs' inherent rewriting tendencies offers a robust and scalable solution for detecting AI-generated text.

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。

↑