arXiv cs.CL· Kevin Ren, Manish Raghavan, Nikhil Garg·· 2 天前AI 评分31
面向持续分布偏移的 AI 文本检测测试时自适应方法
Hitting a Moving Target: Test-Time Adaptation for AI Text Detection under Continual Distribution Shift
AI 导读
研究提出一种测试时自适应（TTA）方法，利用推理时无标注样本的同质性来应对 AI 文本检测中的持续分布偏移，包括对抗性人类化、新 LLM 发布和人类写作的时间漂移。实验显示，SOTA 监督检测器在分布偏移下系统性失效，而结合半监督学习的 TTA 基本保持稳健；例如商业模型 Pangram 仅检出 27.0% 的对抗性 AI 生成文本，该 TTA 方法则达 93.3%。
正文 · 原文
Abstract:Deployed AI text detection approaches are supervised, relying on training-time access to labeled datasets of both human-written and AI-generated text. This approach is vulnerable to three types of distribution shifts that occur continually post-deployment, and for which labeled data is often unavailable: adversarial humanization, new LLMs being released, and unidentifiable temporal drift in human writing after widespread LLM usage. Simultaneously, existing approaches do not leverage a key signal of LLM usage: inference-time homogeneity. We propose a test-time adaptation (TTA) approach that adapts to distribution shifts by leveraging homogeneity among unlabeled samples observed at inference time. Empirically, we find that state-of-the-art supervised detectors systematically fail when they encounter distribution shifts in AI-generated and human writing, both adversarial and natural, while TTA with semi-supervised learning is largely robust; e.g., the commercial model Pangram detects just 27.0% of our adversarial AI-generated text, compared to 93.3% for our test-time approach. We establish that TTA is a promising framework for AI text detection in the wild.
| Comments: | NeurIPS 2026. We release our code at: this https URL |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.25152 [cs.CL] |
| (or arXiv:2606.25152v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2606.25152 arXiv-issued DOI via DataCite |
Submission history
From: Kevin Ren [view email]
[v1]
Tue, 23 Jun 2026 20:37:18 UTC (900 KB)
[v2]
Wed, 14 Oct 2026 18:51:21 UTC (2,087 KB)
来源：arXiv cs.CL · arxiv.org