人类文本是离群点：通过分布外检测识别大语言模型生成的文本
Human Texts Are Outliers: Detecting LLM-generated Texts via Out-of-distribution Detection
AI总结：
本文提出将大语言模型文本检测重新定义为分布外检测问题，将人类文本视为离群点，采用一类学习和基于分数的方法，在DeepFake数据集上达到98.3% AUROC，并验证了跨语言、攻击及未见模型的稳健性。
摘要：
ChatGPT、DeepSeek 和 Claude 等大语言模型（LLM）的快速发展显著增加了数字通信中 AI 生成文本的出现频率。这一趋势增强了人们对可靠检测方法的需求，以区分人类撰写的内容与机器生成的内容。现有方法，无论是零样本方法还是监督分类器，大多将此任务概念化为二元分类问题，这往往导致跨领域和跨模型的泛化性能较差。在本文中，我们认为这种二元表述从根本上误解了检测任务，因为它假设人类撰写的文本具有连贯的表征。实际上，人类文本并不构成统一的分布，其多样性无法通过有限的采样有效捕获。这导致先前的分类器记住观察到的分布外（OOD）特征，而非学习“非分布内”行为的本质，从而限制了对未见人类输入的泛化能力。基于这一观察，我们提出将检测任务重新构建为分布外（OOD）检测问题，将人类撰写的文本视为分布离群点，而机器生成的文本视为分布内（ID）样本。为此，我们开发了一个检测框架，使用一类学习方法（包括 DeepSVDD 和 HRN）以及基于分数的学习技术（如基于能量的方法），以实现稳健且可泛化的性能。跨多个数据集的广泛实验验证了我们基于 OOD 方法的有效性。具体而言，在 DeepFake 数据集上，基于 OOD 的方法实现了 98.3% 的 AUROC 和 AUPR，且 FPR95 仅为 8.9%。此外，我们在多语言、受攻击以及未见模型和未见领域文本设置下测试了我们的检测框架，证明了我们框架的稳健性和泛化能力。代码、预训练权重和演示将发布。
英文摘要：
The rapid advancement of large language models (LLMs) such as ChatGPT, DeepSeek, and Claude has significantly increased the presence of AI-generated text in digital communication. This trend has heightened the need for reliable detection methods to distinguish between human-authored and machine-generated content. Existing approaches both zero-shot methods and supervised classifiers largely conceptualize this task as a binary classification problem, often leading to poor generalization across domains and models. In this paper, we argue that such a binary formulation fundamentally mischaracterizes the detection task by assuming a coherent representation of human-written texts. In reality, human texts do not constitute a unified distribution, and their diversity cannot be effectively captured through limited sampling. This causes previous classifiers to memorize observed OOD characteristics rather than learn the essence of `non-ID' behavior, limiting generalization to unseen human-authored inputs. Based on this observation, we propose reframing the detection task as an out-of-distribution (OOD) detection problem, treating human-written texts as distributional outliers while machine-generated texts are in-distribution (ID) samples. To this end, we develop a detection framework using one-class learning method including DeepSVDD and HRN, and score-based learning techniques such as energy-based method, enabling robust and generalizable performance. Extensive experiments across multiple datasets validate the effectiveness of our OOD-based approach. Specifically, the OOD-based method achieves 98.3% AUROC and AUPR with only 8.9% FPR95 on DeepFake dataset. Moreover, we test our detection framework on multilingual, attacked, and unseen-model and -domain text settings, demonstrating the robustness and generalizability of our framework. Code, pretrained weights, and demo will be released.
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。
