{"slug": "hitting-a-moving-target-test-time-adaptation-for-ai-text-detection-under-shift", "title": "Hitting a Moving Target: Test-Time Adaptation for AI Text Detection under Continual Distribution Shift", "summary": "Researchers propose a test-time adaptation method using semi-supervised learning to improve AI text detection under continual distribution shifts, such as adversarial humanization and new LLMs. The approach leverages inference-time homogeneity among unlabeled samples, outperforming state-of-the-art detectors like Pangram (24.1% vs. 90.5% detection of adversarial AI text).", "body_md": "arXiv:2606.25152v1 Announce Type: new\nAbstract: Deployed approaches for AI text detection often rely 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 temporal drift in human writing. Simultaneously, existing approaches do not leverage a key signal of LLM usage: inference-time homogeneity. We propose a test-time adaptation (TTA) approach, using semi-supervised learning, 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 test-time adaptation with semi-supervised learning is largely robust; e.g., the commercial model Pangram detects just 24.1% of our adversarial AI-generated text, compared to 90.5% for our test-time approach. We establish that test-time adaptation is a promising framework for AI text detection in the wild. We publicly release our code (which includes code for model training, evaluation, and plots) at https://github.com/kkr36/llm_detection.", "url": "https://wpnews.pro/news/hitting-a-moving-target-test-time-adaptation-for-ai-text-detection-under-shift", "canonical_source": "https://arxiv.org/abs/2606.25152", "published_at": "2026-06-25 04:00:00+00:00", "updated_at": "2026-06-25 04:15:22.569367+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-safety", "machine-learning"], "entities": ["Pangram", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/hitting-a-moving-target-test-time-adaptation-for-ai-text-detection-under-shift", "markdown": "https://wpnews.pro/news/hitting-a-moving-target-test-time-adaptation-for-ai-text-detection-under-shift.md", "text": "https://wpnews.pro/news/hitting-a-moving-target-test-time-adaptation-for-ai-text-detection-under-shift.txt", "jsonld": "https://wpnews.pro/news/hitting-a-moving-target-test-time-adaptation-for-ai-text-detection-under-shift.jsonld"}}