We present the Checkfor.ai text classifier, a transformer-based neural
network trained to distinguish text written by large language models from text
written by humans. Checkfor.ai outperforms zero-shot methods such as DetectGPT
as well as leading commercial AI detection tools with over 9 times lower error
rates on a comprehensive benchmark comprised of ten text domains (student
writing, creative writing, scientific writing, books, encyclopedias, news,
email, scientific papers, short-form Q\&A) and 8 open- and closed-source large
language models. We propose a training algorithm, hard negative mining with
synthetic mirrors, that enables our classifier to achieve orders of magnitude
lower false positive rates on high-data domains such as reviews. Finally, we
show that Checkfor.ai is not biased against nonnative English speakers and
generalizes to domains and models unseen during training.

通过使用基于转换器的神经网络的 Checkfor.ai 文本分类器，我们能够识别大型语言模型写的文本和人类写的文本。Checkfor.ai 在十个文本领域（学生写作、创意写作、科学写作、书籍、百科全书、新闻、电子邮件、科学论文、问答简述）和 8 个开源和闭源大型语言模型的综合基准测试中，相较于 DetectGPT 和其他商业 AI 检测工具，错误率降低了九倍以上。我们提出了一种训练算法，即使用合成镜像进行负样本训练，使我们的分类器在高数据领域（例如评论）能够实现数量级更低的误报率。最后，我们证明 Checkfor.ai 不对非英语母语人士存在偏见，并且在训练期间未见过的领域和模型中具有普适性。