发表机构
University of Bucharest（布加勒斯特大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究揭示AI文本检测器依赖统计特征而非语义，导致“人性化”改写反而易被识别，且对正式人类文本误报率高，结构抽象未能提升鲁棒性。
AI 中文摘要
受监督的AI文本检测器报告了较高的基准准确率，但其决策依据尚不明确。我们基于语义、结构和分词器层面的扰动，使用M4数据集（N=10,000）和受控生成（N=300）分析了一个基于RoBERTa的检测器。当Mistral-7B-Instruct被要求让机器文本听起来更人性化时，动词多样性从0.77升至0.92，而输出反而更容易被检测。检测分数似乎追踪统计复杂度，这也导致在正式人类写作上出现76.3%的假阳性率。作为对照，我们评估了基于事件的潜在空间检测。改写改变了其87%的事件序列（Jaccard=0.067），同形字改变了70%的提取动词，尽管提取仍能运行（Jaccard=0.30）。其最佳领域AUC为0.577。RoBERTa的鲁棒性似乎特定于其所用的特征，而结构抽象并未使检测更加鲁棒。
英文摘要
Supervised AI-text detectors report high benchmark accuracy, but it is not clear what their decisions are based on. We analyze a RoBERTa-based detector under semantic, structural, and tokenizer-level perturbations, using the M4 dataset (N = 10,000) and controlled generations (N = 300). When Mistral-7B-Instruct was asked to make machine text sound more human, Verb Diversity rose from 0.77 to 0.92 and the outputs became easier to detect. Detection scores appear to track statistical complexity, which also leads to a 76.3% false-positive rate on formal human writing. As a control, we evaluate event-based Latent Space detection. Paraphrasing changed 87% of its event sequences (Jaccard = 0.067), and homoglyphs altered 70% of the extracted verbs even though extraction still ran (Jaccard = 0.30). Its best domain AUC was 0.577. RoBERTa's robustness seems specific to the features it uses, and structural abstraction did not make detection more robust.
