arXiv:2512.03582v1 Announce Type: new 
Abstract: Narratives are the cognitive and emotional scaffolds of propaganda. They organize isolated persuasive techniques into coherent stories that justify actions, attribute blame, and evoke identification with ideological camps. In this paper, we propose a novel fine-grained narrative classification in biased news articles. We also explore article-bias classification as the precursor task to narrative classification and fine-grained persuasive technique identification. We develop INDI-PROP, the first ideologically grounded fine-grained narrative dataset with multi-level annotation for analyzing propaganda in Indian news media. Our dataset INDI-PROP comprises 1,266 articles focusing on two polarizing socio-political events in recent times: CAA and the Farmers' protest. Each article is annotated at three hierarchical levels: (i) ideological article-bias (pro-government, pro-opposition, neutral), (ii) event-specific fine-grained narrative frames anchored in ideological polarity and communicative intent, and (iii) persuasive techniques. We propose FANTA and TPTC, two GPT-4o-mini guided multi-hop prompt-based reasoning frameworks for the bias, narrative, and persuasive technique classification. FANTA leverages multi-layered communicative phenomena by integrating information extraction and contextual framing for hierarchical reasoning. On the other hand, TPTC adopts systematic decomposition of persuasive cues via a two-stage approach. Our evaluation suggests substantial improvement over underlying baselines in each case.

دراسة جديدة تقترح نظام تصنيف سردي مفصل للمقالات الإخبارية المتحيزة، مع التركيز على الجوانب المعرفية والعاطفية للدعاية. تقدم البحث مجموعة بيانات INDI-PROP، التي تتضمن 1,266 مقالة تتعلق بقانون الجنسية الهندي واحتجاجات المزارعين، مع توضيح التحيز الإيديولوجي وإطارات السرد.

Un nuevo estudio propone un sistema de clasificación narrativa detallada para artículos de noticias sesgados, centrándose en los aspectos cognitivos y emocionales de la propaganda. La investigación presenta INDI-PROP, un conjunto de datos que comprende 1,266 artículos relacionados con la CAA y la protesta de los agricultores, anotados por sesgo ideológico y marcos narrativos.

Une nouvelle étude propose un système de classification narrative fine pour les articles de presse biaisés, en se concentrant sur les aspects cognitifs et émotionnels de la propagande. La recherche introduit INDI-PROP, un ensemble de données comprenant 1 266 articles liés à la CAA et à la protestation des agriculteurs, annotés pour le biais idéologique et les cadres narratifs.

A new study proposes a fine-grained narrative classification system for biased news articles, focusing on propaganda's cognitive and emotional aspects. The research introduces INDI-PROP, a dataset comprising 1,266 articles related to the CAA and Farmers' protest, annotated for ideological bias and narrative frames.

Fine-grained Narrative Classification in Biased News Articles

arXiv:2509.17701v2 Announce Type: replace 
Abstract: Large Language Models (LLMs) are increasingly used for educational support, yet their response quality varies depending on the language of interaction. This paper presents an automated multilingual pipeline for generating, solving, and evaluating math problems aligned with the German K-10 curriculum. We generated 628 math exercises and translated them into English, German, and Arabic. Three commercial LLMs (GPT-4o-mini, Gemini 2.5 Flash, and Qwen-plus) were prompted to produce step-by-step solutions in each language. A held-out panel of LLM judges, including Claude 3.5 Haiku, evaluated solution quality using a comparative framework. Results show a consistent gap, with English solutions consistently rated highest, and Arabic often ranked lower. These findings highlight persistent linguistic bias and the need for more equitable multilingual AI systems in education.

قدمت دراسة حديثة خط أنابيب متعدد اللغات لتوليد وحل وتقييم مسائل الرياضيات باستخدام نماذج اللغة الكبيرة (LLMs)، والتي تتماشى بشكل خاص مع المنهج الدراسي الألماني K-10. قامت الدراسة بتوليد 628 تمرين رياضي وترجمته إلى الإنجليزية والألمانية والعربية، مما يكشف عن تفاوتات كبيرة في جودة الحلول عبر اللغات، حيث تم تقييم الحلول الإنجليزية بشكل أعلى باستمرار وغالبًا ما تم تصنيف العربية بشكل أقل.

Un estudio reciente presentó un pipeline multilingüe para generar, resolver y evaluar problemas matemáticos utilizando Modelos de Lenguaje de Gran Tamaño (LLMs), específicamente alineado con el currículo alemán K-10. La investigación generó 628 ejercicios matemáticos y los tradujo al inglés, alemán y árabe, revelando disparidades significativas en la calidad de las soluciones según el idioma, siendo el inglés el mejor valorado y el árabe a menudo el peor.

Une étude récente a présenté un pipeline multilingue pour générer, résoudre et évaluer des problèmes mathématiques à l'aide de modèles de langage de grande taille (LLMs), spécifiquement aligné sur le programme scolaire allemand K-10. La recherche a généré 628 exercices mathématiques et les a traduits en anglais, allemand et arabe, révélant des disparités significatives dans la qualité des solutions selon les langues, l'anglais étant systématiquement mieux noté et l'arabe souvent moins bien noté.

A recent study introduced a multilingual pipeline for generating, solving, and evaluating math problems using Large Language Models (LLMs), specifically aligned with the German K-10 curriculum. The research generated 628 math exercises and translated them into English, German, and Arabic, revealing significant disparities in solution quality across languages, with English consistently rated highest and Arabic often rated lower.

Investigating Bias: A Multilingual Pipeline for Generating, Solving, and Evaluating Math Problems with LLMs

arXiv:2502.13127v2 Announce Type: replace 
Abstract: Recent advances in Large Language Models (LLMs) have enabled them to process increasingly longer sequences, ranging from 2K to 2M tokens and even beyond. However, simply extending the input sequence length does not necessarily lead to effective long-context understanding. In this study, we integrate Chain-of-Thought (CoT) reasoning into LLMs in a supervised manner to facilitate effective long-context understanding. To achieve this, we introduce LongFinanceQA, a synthetic dataset in the financial domain designed to improve long-context reasoning. Unlike existing long-context synthetic data, LongFinanceQA includes intermediate CoT reasoning before the final conclusion, which encourages LLMs to perform explicit reasoning, improving accuracy and interpretability in long-context understanding. To generate synthetic CoT reasoning, we propose Property-based Agentic Inference (PAI), an agentic framework that simulates human-like reasoning steps, including property extraction, retrieval, and summarization. We evaluate PAI's reasoning capabilities by assessing GPT-4o-mini w/ PAI on the Loong benchmark, outperforming standard GPT-4o-mini by 20.0%. Furthermore, we fine-tune LLaMA-3.1-8B-Instruct on LongFinanceQA, achieving a 28.0% gain on Loong's financial subset.

أدت التطورات الأخيرة في نماذج اللغة الكبيرة (LLMs) إلى تطوير نهج للتفكير المتسلسل (CoT) تحت إشراف، مما يعزز فهم السياقات الطويلة. يتجلى ذلك في تقديم LongFinanceQA، وهو مجموعة بيانات اصطناعية مصممة للقطاع المالي، والتي تتضمن تفكير CoT الوسيط لتحسين الدقة وقابلية التفسير في مخرجات LLM.

Los avances recientes en Modelos de Lenguaje de Gran Tamaño (LLMs) han llevado al desarrollo de un enfoque de razonamiento en Cadena de Pensamiento (CoT) supervisado, mejorando la comprensión de contextos largos. Esto se ejemplifica con la introducción de LongFinanceQA, un conjunto de datos sintético diseñado para el sector financiero, que incorpora razonamiento CoT intermedio para mejorar la precisión y la interpretabilidad en las salidas de los LLM.

Les avancées récentes dans les modèles de langage de grande taille (LLMs) ont conduit au développement d'une approche de raisonnement supervisé en chaîne de pensée (CoT), améliorant la compréhension des contextes longs. Cela se manifeste par l'introduction de LongFinanceQA, un ensemble de données synthétiques conçu pour le secteur financier, qui intègre un raisonnement CoT intermédiaire pour améliorer la précision et l'interprétabilité des résultats des LLM.

Recent advancements in Large Language Models (LLMs) have led to the development of a supervised Chain-of-Thought (CoT) reasoning approach, enhancing long-context understanding. This is exemplified by the introduction of LongFinanceQA, a synthetic dataset tailored for the financial sector, which incorporates intermediate CoT reasoning to improve accuracy and interpretability in LLM outputs.

Fine-grained Narrative Classification in Biased News Articles

Was this article worth reading? Share it

Airparser

LucidQuery AI

NewsperAI