arXiv:2606.14060v1 Announce Type: new 
Abstract: Adversarial conditions such as paraphrasing and targeted style transfer sharply degrade the accuracy of machine text detectors. A document, however, carries multiple complementary signals (e.g., stylistic features, likelihood and rank-order features, and structural features), and an attack that suppresses one may leave others intact. While a parametric classifier can learn to combine these features given sufficient supervision, classifiers are prone to making confidently incorrect predictions when the distribution shifts (e.g., novel attacks or unseen language models). To address this, we propose a multi-view, non-parametric detection framework that extracts complementary feature views from the same document and aggregates per-view evidence through a Gaussian process ensemble. By aggregating evidence across views, an adversary must simultaneously defeat multiple independent axes of detection, substantially raising the cost of evasion. The Gaussian process formulation additionally provides calibrated probabilities and principled abstention on out-of-distribution inputs, supporting reliable deployment in high-stakes settings. We evaluate on three benchmarks spanning diverse generators and attacks: the DetectRL and RAID benchmarks, and the PAN2025 shared task and demonstrate that our multi-view detector maintains strong performance under the considered attacks, outperforming existing approaches against held out attacks.

تم اقتراح إطار جديد للكشف عن النصوص الآلية غير المعتمد على المعلمات، باستخدام عمليات غاوسية متعددة المناظر لتعزيز قوة الكشف ضد الظروف العدائية مثل إعادة الصياغة ونقل الأسلوب المستهدف. تعتمد هذه الطريقة على إشارات تكاملية متعددة من الوثائق لتحسين الدقة.

Se ha propuesto un nuevo marco para la detección de texto máquina no paramétrica, utilizando procesos gaussianos de múltiples vistas para mejorar la robustez de la detección frente a condiciones adversas como el parafraseo y la transferencia de estilo dirigida. Este enfoque aprovecha múltiples señales complementarias de los documentos para mejorar la precisión.

Un nouveau cadre pour la détection de texte machine non paramétrique a été proposé, utilisant des processus gaussiens à vues multiples pour améliorer la robustesse de la détection contre des conditions adversariales telles que le paraphrasage et le transfert de style ciblé. Cette approche s'appuie sur plusieurs signaux complémentaires provenant des documents pour améliorer la précision.

A new framework for non-parametric machine text detection has been proposed, utilizing multi-view Gaussian processes to enhance the robustness of detection against adversarial conditions such as paraphrasing and targeted style transfer. This approach leverages multiple complementary signals from documents to improve accuracy.