arXiv:2511.16690v1 Announce Type: new 
Abstract: Many AI detection models have been developed to counter the presence of articles created by artificial intelligence (AI). However, if a human-authored article is slightly polished by AI, a shift will occur in the borderline decision of these AI detection models, leading them to consider it AI-generated article. This misclassification may result in falsely accusing authors of AI plagiarism and harm the credibility of AI detector models. In English, some efforts were made to meet this challenge, but not in Arabic. In this paper, we generated two datasets. The first dataset contains 800 Arabic articles, half AI-generated and half human-authored. We used it to evaluate 14 Large Language models (LLMs) and commercial AI detectors to assess their ability in distinguishing between human-authored and AI-generated articles. The best 8 models were chosen to act as detectors for our primary concern, which is whether they would consider slightly polished human text as AI-generated. The second dataset, Ar-APT, contains 400 Arabic human-authored articles polished by 10 LLMs using 4 polishing settings, totaling 16400 samples. We use it to evaluate the 8 nominated models and determine whether slight polishing will affect their performance. The results reveal that all AI detectors incorrectly attribute a significant number of articles to AI. The best performing LLM, Claude-4 Sonnet, achieved 83.51%, their performance decreased to 57.63% for articles slightly polished by LLaMA-3. Whereas for the best performing commercial model, originality.AI, that achieves 92% accuracy, dropped to 12% for articles slightly polished by Mistral or Gemma-3.

أظهرت دراسة حديثة أوجه القصور في نماذج الكشف عن الذكاء الاصطناعي في تصنيف المقالات العربية المصقولة قليلاً بدقة، مما يكشف أن هذه النماذج قد تحكم خطأً على المحتوى الذي كتبه البشر على أنه مولد بواسطة الذكاء الاصطناعي. تضمنت الدراسة إنشاء مجموعتين من البيانات من المقالات العربية، وتقييم 14 نموذجًا للغة الكبيرة وكاشفات الذكاء الاصطناعي التجارية لتقييم دقة تصنيفها.

Un estudio reciente destacó las deficiencias de los modelos de detección de IA en la clasificación precisa de artículos árabes ligeramente pulidos, revelando que tales modelos pueden juzgar erróneamente el contenido escrito por humanos como generado por IA. La investigación implicó la creación de dos conjuntos de datos de artículos árabes, evaluando 14 Modelos de Lenguaje Grande y detectores de IA comerciales para evaluar su precisión de clasificación.

Une étude récente a mis en lumière les lacunes des modèles de détection d'IA dans la classification précise des articles arabes légèrement polis, révélant que ces modèles peuvent juger à tort le contenu rédigé par des humains comme étant généré par l'IA. La recherche a impliqué la création de deux ensembles de données d'articles arabes, évaluant 14 modèles de langage de grande taille et des détecteurs d'IA commerciaux pour évaluer leur précision de classification.

A recent study highlighted the shortcomings of AI detection models in accurately classifying slightly polished Arabic articles, revealing that such models may misjudge human-authored content as AI-generated. The research involved creating two datasets of Arabic articles, evaluating 14 Large Language Models and commercial AI detectors to assess their classification accuracy.

Falsely Accused: How AI Detectors Misjudge Slightly Polished Arabic Articles

arXiv:2511.17256v1 Announce Type: cross 
Abstract: As Large Language Models (LLMs) increasingly influence high-stakes decision-making across global contexts, ensuring their alignment with diverse cultural values has become a critical governance challenge. This study presents a Multi-Layered Auditing Platform for Responsible AI that systematically evaluates cross-cultural value alignment in China-origin and Western-origin LLMs through four integrated methodologies: Ethical Dilemma Corpus for assessing temporal stability, Diversity-Enhanced Framework (DEF) for quantifying cultural fidelity, First-Token Probability Alignment for distributional accuracy, and Multi-stAge Reasoning frameworK (MARK) for interpretable decision-making. Our comparative analysis of 20+ leading models, such as Qwen, GPT-4o, Claude, LLaMA, and DeepSeek, reveals universal challenges-fundamental instability in value systems, systematic under-representation of younger demographics, and non-linear relationships between model scale and alignment quality-alongside divergent regional development trajectories. While China-origin models increasingly emphasize multilingual data integration for context-specific optimization, Western models demonstrate greater architectural experimentation but persistent U.S.-centric biases. Neither paradigm achieves robust cross-cultural generalization. We establish that Mistral-series architectures significantly outperform LLaMA3-series in cross-cultural alignment, and that Full-Parameter Fine-Tuning on diverse datasets surpasses Reinforcement Learning from Human Feedback in preserving cultural variation...

قدمت دراسة حديثة منصة تدقيق متعددة الطبقات للذكاء الاصطناعي المسؤول، تهدف إلى تقييم توافق القيم الثقافية عبر نماذج اللغة الكبيرة (LLMs) من الصين والغرب. تسلط هذه الأبحاث الضوء على التحديات المتعلقة بالحكم التي تطرحها نماذج اللغة في اتخاذ القرارات عالية المخاطر، كاشفة عن عدم استقرار أساسي في أنظمة القيم وتمثيل ديموغرافي ناقص بين النماذج الرائدة مثل Qwen وGPT-4o.

Un estudio reciente ha presentado una Plataforma de Auditoría Multicapa para la IA Responsable, destinada a evaluar la alineación de valores interculturales en Modelos de Lenguaje de Gran Tamaño (LLMs) de origen chino y occidental. Esta investigación destaca los desafíos de gobernanza que plantean los LLMs en la toma de decisiones de alto riesgo, revelando inestabilidades fundamentales en los sistemas de valores y una subrepresentación demográfica entre modelos líderes como Qwen y GPT-4o.

Une étude récente a introduit une plateforme d'audit multicouche pour une IA responsable, visant à évaluer l'alignement des valeurs interculturelles dans les modèles de langage de grande taille (LLMs) originaires de Chine et de l'Occident. Cette recherche met en lumière les défis de gouvernance posés par les LLMs dans la prise de décision à enjeux élevés, révélant des instabilités fondamentales dans les systèmes de valeurs et une sous-représentation démographique parmi les modèles leaders comme Qwen et GPT-4o.

A recent study has introduced a Multi-Layered Auditing Platform for Responsible AI, aimed at evaluating cross-cultural value alignment in Large Language Models (LLMs) from China and the West. This research highlights the governance challenges posed by LLMs in high-stakes decision-making, revealing fundamental instabilities in value systems and demographic under-representation among leading models like Qwen and GPT-4o.

Falsely Accused: How AI Detectors Misjudge Slightly Polished Arabic Articles

Was this article worth reading? Share it

polygrai

Grubby.AI

Translate a Book with AI