arXiv:2511.16445v1 Announce Type: cross 
Abstract: People living with dementia (PLwD) often show gradual shifts in how they communicate, becoming less expressive, more repetitive, or drifting off-topic in subtle ways. While caregivers may notice these changes informally, most computational tools are not designed to track such behavioral drift over time. This paper introduces PersonaDrift, a synthetic benchmark designed to evaluate machine learning and statistical methods for detecting progressive changes in daily communication, focusing on user responses to a digital reminder system. PersonaDrift simulates 60-day interaction logs for synthetic users modeled after real PLwD, based on interviews with caregivers. These caregiver-informed personas vary in tone, modality, and communication habits, enabling realistic diversity in behavior. The benchmark focuses on two forms of longitudinal change that caregivers highlighted as particularly salient: flattened sentiment (reduced emotional tone and verbosity) and off-topic replies (semantic drift). These changes are injected progressively at different rates to emulate naturalistic cognitive trajectories, and the framework is designed to be extensible to additional behaviors in future use cases. To explore this novel application space, we evaluate several anomaly detection approaches, unsupervised statistical methods (CUSUM, EWMA, One-Class SVM), sequence models using contextual embeddings (GRU + BERT), and supervised classifiers in both generalized and personalized settings. Preliminary results show that flattened sentiment can often be detected with simple statistical models in users with low baseline variability, while detecting semantic drift requires temporal modeling and personalized baselines. Across both tasks, personalized classifiers consistently outperform generalized ones, highlighting the importance of individual behavioral context.

تقدم الورقة PersonaDrift، وهو معيار صناعي مصمم لتقييم أساليب التعلم الآلي لاكتشاف التغيرات السلوكية لدى الأشخاص الذين يعيشون مع الخرف (PLwD). يحاكي سجلات تفاعل لمدة 60 يومًا بناءً على PLwD حقيقية، مع التركيز على استجابات المستخدمين لنظام تذكير رقمي. يبرز المعيار شكلين من التغيرات الطولية التي أشار إليها مقدمو الرعاية على أنها بارزة: انحدار العاطفة وزيادة التكرار في التواصل.

El artículo presenta PersonaDrift, un referente sintético diseñado para evaluar métodos de aprendizaje automático para detectar cambios conductuales en personas que viven con demencia (PLwD). Simula registros de interacción de 60 días basados en PLwD reales, centrándose en las respuestas de los usuarios a un sistema de recordatorio digital. El referente destaca dos cambios significativos: un sentimiento aplanado y un aumento en la repetición de la comunicación, que los cuidadores han señalado como indicadores críticos de deterioro cognitivo.

L'article présente PersonaDrift, une référence synthétique destinée à évaluer les méthodes d'apprentissage automatique pour détecter les changements comportementaux chez les personnes vivant avec la démence (PLwD). Il simule des journaux d'interaction de 60 jours basés sur de véritables PLwD, en se concentrant sur les réponses des utilisateurs à un système de rappel numérique. La référence met en évidence deux changements significatifs : un sentiment aplati et une répétition accrue dans la communication, que les aidants ont notés comme des indicateurs critiques du déclin cognitif.

The paper introduces PersonaDrift, a synthetic benchmark aimed at evaluating machine learning methods for detecting behavioral changes in people living with dementia (PLwD). It simulates 60-day interaction logs based on real PLwD, focusing on user responses to a digital reminder system. The benchmark highlights two significant changes: flattened sentiment and increased repetition in communication, which caregivers have noted as critical indicators of cognitive decline.

PersonaDrift: A Benchmark for Temporal Anomaly Detection in Language-Based Dementia Monitoring

arXiv:2506.08487v2 Announce Type: replace 
Abstract: The rapid adoption of Small Language Models (SLMs) for resource constrained applications has outpaced our understanding of their ethical and fairness implications. To address this gap, we introduce the Vacuous Neutrality Framework (VaNeu), a multi-dimensional evaluation paradigm designed to assess SLM fairness prior to deployment. The framework examines model robustness across four stages - biases, utility, ambiguity handling, and positional bias over diverse social bias categories. To the best of our knowledge, this work presents the first large-scale audit of SLMs in the 0.5-5B parameter range, an overlooked "middle tier" between BERT-class encoders and flagship LLMs. We evaluate nine widely used SLMs spanning four model families under both ambiguous and disambiguated contexts. Our findings show that models demonstrating low bias in early stages often fail subsequent evaluations, revealing hidden vulnerabilities and unreliable reasoning. These results underscore the need for a more comprehensive understanding of fairness and reliability in SLMs, and position the proposed framework as a principled tool for responsible deployment in socially sensitive settings.

يتناول المقال الاعتماد السريع للنماذج اللغوية الصغيرة (SLMs) والتداعيات الأخلاقية المحيطة بها. يقدم إطار الحياد الفارغ (VaNeu)، وهو نموذج تقييم جديد مصمم لتقييم عدالة SLMs قبل نشرها. يقوم الإطار بتقييم قوة النموذج عبر مراحل متعددة، كاشفًا عن نقاط الضعف في النماذج التي تبدو في البداية غير متحيزة. تمثل هذه الدراسة أول تدقيق على نطاق واسع لـ SLMs في نطاق معلمات يتراوح بين 0.5 إلى 5 مليارات.

El artículo aborda la rápida adopción de los Modelos de Lenguaje Pequeños (SLMs) y las implicaciones éticas que los rodean. Se presenta el Marco de Neutralidad Vacía (VaNeu), un nuevo paradigma de evaluación diseñado para evaluar la equidad de los SLMs antes de su implementación. El marco evalúa la robustez del modelo a través de varias etapas, revelando vulnerabilidades en modelos que inicialmente parecen no tener sesgos. Este estudio representa la primera auditoría a gran escala de los SLMs en el rango de parámetros de 0.5 a 5 mil millones.

L'article traite de l'adoption rapide des petits modèles de langage (SLMs) et des implications éthiques qui les entourent. Il introduit le cadre de neutralité vacu (VaNeu), un nouveau paradigme d'évaluation conçu pour évaluer l'équité des SLMs avant leur déploiement. Le cadre évalue la robustesse des modèles à travers diverses étapes, révélant des vulnérabilités dans les modèles qui semblent initialement sans biais. Cette étude représente le premier audit à grande échelle des SLMs dans la plage de paramètres de 0,5 à 5 milliards.

The article discusses the rapid adoption of Small Language Models (SLMs) and the ethical implications surrounding their use. It introduces the Vacuous Neutrality Framework (VaNeu), a new evaluation paradigm designed to assess the fairness of SLMs before deployment. The framework evaluates model robustness across various stages, revealing vulnerabilities in models that initially appear unbiased. This study represents the first large-scale audit of SLMs in the 0.5-5B parameter range.

Beyond Bias Scores: Unmasking Vacuous Neutrality in Small Language Models

arXiv:2511.16018v1 Announce Type: cross 
Abstract: Introduction: The application of Artificial Intelligence in games has evolved significantly, allowing for dynamic content generation. However, its use as a core gameplay co-creation tool remains underexplored. Objective: This paper proposes SpellForger, a game where players create custom spells by writing natural language prompts, aiming to provide a unique experience of personalization and creativity. Methodology: The system uses a supervisedtrained BERT model to interpret player prompts. This model maps textual descriptions to one of many spell prefabs and balances their parameters (damage, cost, effects) to ensure competitive integrity. The game is developed in the Unity Game Engine, and the AI backend is in Python. Expected Results: We expect to deliver a functional prototype that demonstrates the generation of spells in real time, applied to an engaging gameplay loop, where player creativity is central to the experience, validating the use of AI as a direct gameplay mechanic.

تقدم هذه الورقة SpellForger، وهي لعبة تتيح للاعبين إنشاء تعاويذ مخصصة باستخدام مطالبات بلغة طبيعية. باستخدام نموذج BERT المدرب تحت إشراف، يقوم النظام بتفسير هذه المطالبات لتوليد تعاويذ مع معايير متوازنة مثل الضرر والتكلفة. تم تطوير اللعبة في محرك الألعاب Unity مع واجهة خلفية بلغة Python، وتهدف SpellForger إلى تعزيز إبداع اللاعبين وتخصيص تجربتهم في اللعب.

El artículo presenta SpellForger, un juego que permite a los jugadores crear hechizos personalizados mediante indicaciones en lenguaje natural. Utilizando un modelo BERT entrenado de forma supervisada, el juego interpreta estas indicaciones para generar hechizos con parámetros equilibrados como daño y costo. Desarrollado en el motor de juego Unity con un backend en Python, SpellForger busca mejorar la creatividad y personalización de los jugadores en la jugabilidad.

Cet article présente SpellForger, un jeu permettant aux joueurs de créer des sorts personnalisés en utilisant des invites en langage naturel. Grâce à un modèle BERT supervisé, le jeu interprète ces invites pour générer des sorts avec des paramètres équilibrés tels que les dégâts et le coût. Développé dans le moteur de jeu Unity avec un backend Python, SpellForger vise à améliorer la créativité et la personnalisation des joueurs dans le gameplay.

The paper introduces SpellForger, a game that allows players to create custom spells using natural language prompts. Utilizing a supervised-trained BERT model, the game interprets these prompts to generate spells with balanced parameters such as damage and cost. Developed in the Unity Game Engine with a Python backend, SpellForger aims to enhance player creativity and personalization in gameplay.

SpellForger: Prompting Custom Spell Properties In-Game using BERT supervised-trained model

arXiv:2511.04988v2 Announce Type: replace 
Abstract: Accurately forecasting carbon prices is essential for informed energy market decision-making, guiding sustainable energy planning, and supporting effective decarbonization strategies. However, it remains challenging due to structural breaks and high-frequency noise caused by frequent policy interventions and market shocks. Existing studies, including the most recent baseline approaches, have attempted to incorporate breakpoints but often treat denoising and modeling as separate processes and lack systematic evaluation across advanced deep learning architectures, limiting the robustness and the generalization capability. To address these gaps, this paper proposes a comprehensive hybrid framework that integrates structural break detection (Bai-Perron, ICSS, and PELT algorithms), wavelet signal denoising, and three state-of-the-art deep learning models (LSTM, GRU, and TCN). Using European Union Allowance (EUA) spot prices from 2007 to 2024 and exogenous features such as energy prices and policy indicators, the framework constructs univariate and multivariate datasets for comparative evaluation. Experimental results demonstrate that our proposed PELT-WT-TCN achieves the highest prediction accuracy, reducing forecasting errors by 22.35% in RMSE and 18.63% in MAE compared to the state-of-the-art baseline model (Breakpoints with Wavelet and LSTM), and by 70.55% in RMSE and 74.42% in MAE compared to the original LSTM without decomposition from the same baseline study. These findings underscore the value of integrating structural awareness and multiscale decomposition into deep learning architectures to enhance accuracy and interpretability in carbon price forecasting and other nonstationary financial time series.

يقدم هذا البحث إطارًا هجينًا لتوقع أسعار الكربون، حيث يتناول التحديات التي تطرحها الانقطاعات الهيكلية والضوضاء الناتجة عن التدخلات السياسية والصدمات السوقية. يدمج الإطار خوارزميات الكشف عن الانقطاعات الهيكلية مع إزالة الضوضاء باستخدام الموجات وإدماج نماذج التعلم العميق المتقدمة، بما في ذلك LSTM وGRU وTCN. يهدف هذا الإطار إلى تحسين دقة توقعات أسعار الكربون، وهو أمر ضروري لاتخاذ قرارات مستنيرة في سوق الطاقة والتخطيط المستدام.

Este artículo presenta un marco híbrido para la previsión de precios del carbono, abordando los desafíos que plantean los cambios estructurales y el ruido causado por intervenciones políticas y choques de mercado. Integra algoritmos de detección de rupturas estructurales con desruido de señales por wavelet y modelos avanzados de aprendizaje profundo, incluyendo LSTM, GRU y TCN. Este marco busca mejorar la precisión de las predicciones de precios del carbono, que son cruciales para la toma de decisiones en el mercado energético y la planificación sostenible.

Cet article présente un cadre hybride pour la prévision des prix du carbone, abordant les défis posés par les ruptures structurelles et le bruit provenant des changements de politique et des chocs de marché. Il intègre des algorithmes de détection de rupture structurelle avec un débruitage par ondelettes et des modèles avancés d'apprentissage profond, y compris LSTM, GRU et TCN. Ce cadre vise à améliorer la précision des prévisions des prix du carbone, essentielles pour la prise de décision sur le marché de l'énergie et la planification durable.

This paper presents a hybrid framework for forecasting carbon prices, addressing challenges posed by structural breaks and noise from policy changes and market shocks. It integrates structural break detection algorithms with wavelet signal denoising and advanced deep learning models, including LSTM, GRU, and TCN. The framework aims to enhance the accuracy of carbon price predictions, which are crucial for energy market decision-making and sustainable planning.

PersonaDrift: A Benchmark for Temporal Anomaly Detection in Language-Based Dementia Monitoring

Was this article worth reading? Share it