arXiv:2511.12520v1 Announce Type: new 
Abstract: Retrieval-Augmented Generation (RAG) improves large language models by retrieving external knowledge, often truncated into smaller chunks due to the input context window, which leads to information loss, resulting in response hallucinations and broken reasoning chains. Moreover, traditional RAG retrieves unstructured knowledge, introducing irrelevant details that hinder accurate reasoning. To address these issues, we propose TAdaRAG, a novel RAG framework for on-the-fly task-adaptive knowledge graph construction from external sources. Specifically, we design an intent-driven routing mechanism to a domain-specific extraction template, followed by supervised fine-tuning and a reinforcement learning-based implicit extraction mechanism, ensuring concise, coherent, and non-redundant knowledge integration. Evaluations on six public benchmarks and a real-world business benchmark (NowNewsQA) across three backbone models demonstrate that TAdaRAG outperforms existing methods across diverse domains and long-text tasks, highlighting its strong generalization and practical effectiveness.

تقدم الورقة المعنونة 'TAdaRAG: Task Adaptive Retrieval-Augmented Generation via On-the-Fly Knowledge Graph Construction' إطارًا جديدًا يهدف إلى تحسين أساليب الجيل المعزز بالاسترجاع (RAG). غالبًا ما يعاني RAG التقليدي من فقدان المعلومات بسبب قيود سياق الإدخال ويسترجع معرفة غير منظمة قد تُدخل تفاصيل غير ذات صلة. يعالج TAdaRAG هذه المشكلات من خلال بناء رسوم بيانية للمعرفة تتكيف مع المهام في الوقت الفعلي، باستخدام آلية توجيه مدفوعة بالنية وتعلم تعزيز لدمج المعرفة بشكل فعال. تشير التقييمات إلى تحسينات كبيرة عبر معايير متعددة.

El artículo titulado 'TAdaRAG: Task Adaptive Retrieval-Augmented Generation via On-the-Fly Knowledge Graph Construction' presenta un nuevo marco destinado a mejorar los métodos de Generación Aumentada por Recuperación (RAG). El RAG tradicional a menudo sufre de pérdida de información debido a las limitaciones de contexto de entrada y recupera conocimientos no estructurados que pueden introducir detalles irrelevantes. TAdaRAG aborda estos problemas construyendo gráficos de conocimiento adaptativos a la tarea en tiempo real, utilizando un mecanismo de enrutamiento basado en la intención y un apr…

L'article intitulé 'TAdaRAG: Task Adaptive Retrieval-Augmented Generation via On-the-Fly Knowledge Graph Construction' présente un nouveau cadre visant à améliorer les méthodes de génération augmentée par récupération (RAG). Le RAG traditionnel souffre souvent de pertes d'information dues aux limitations du contexte d'entrée et récupère des connaissances non structurées qui peuvent introduire des détails non pertinents. TAdaRAG aborde ces problèmes en construisant des graphes de connaissances adaptatifs aux tâches en temps réel, en utilisant un mécanisme de routage basé sur l'intention et un a…

The paper titled 'TAdaRAG: Task Adaptive Retrieval-Augmented Generation via On-the-Fly Knowledge Graph Construction' introduces a new framework aimed at enhancing Retrieval-Augmented Generation (RAG) methods. Traditional RAG often suffers from information loss due to input context limitations and retrieves unstructured knowledge that can introduce irrelevant details. TAdaRAG addresses these issues by constructing task-adaptive knowledge graphs in real-time, utilizing an intent-driven routing mechanism and reinforcement learning for effective knowledge integration. Evaluations indicate signific…

TAdaRAG: Task Adaptive Retrieval-Augmented Generation via On-the-Fly Knowledge Graph Construction

arXiv:2601.08311v1 Announce Type: new 
Abstract: Large Multimodal Models (LMMs) have recently shown remarkable promise in low-level visual perception tasks, particularly in Image Quality Assessment (IQA), demonstrating strong zero-shot capability. However, achieving state-of-the-art performance often requires computationally expensive fine-tuning methods, which aim to align the distribution of quality-related token in output with image quality levels. Inspired by recent training-free works for LMM, we introduce IQARAG, a novel, training-free framework that enhances LMMs' IQA ability. IQARAG leverages Retrieval-Augmented Generation (RAG) to retrieve some semantically similar but quality-variant reference images with corresponding Mean Opinion Scores (MOSs) for input image. These retrieved images and input image are integrated into a specific prompt. Retrieved images provide the LMM with a visual perception anchor for IQA task. IQARAG contains three key phases: Retrieval Feature Extraction, Image Retrieval, and Integration & Quality Score Generation. Extensive experiments across multiple diverse IQA datasets, including KADID, KonIQ, LIVE Challenge, and SPAQ, demonstrate that the proposed IQARAG effectively boosts the IQA performance of LMMs, offering a resource-efficient alternative to fine-tuning for quality assessment.

تم تقديم إطار عمل جديد يسمى IQARAG لتعزيز قدرات تقييم جودة الصورة (IQA) للنماذج متعددة الوسائط الكبيرة (LMM) من خلال نهج بدون تدريب يستخدم الجيل المعزز بالاسترجاع (RAG). تستخدم هذه الطريقة صور مرجعية مشابهة دلاليًا مع درجات رأي متوسطة (MOS) لتحسين أداء LMMs في مهام IQA.

Se ha introducido un nuevo marco llamado IQARAG para mejorar las capacidades de Evaluación de Calidad de Imagen (IQA) de los Modelos Multimodales Grandes (LMM) a través de un enfoque sin entrenamiento que utiliza la Generación Aumentada por Recuperación (RAG). Este método recupera imágenes de referencia semánticamente similares con puntajes de opinión media (MOS) correspondientes para mejorar el rendimiento de los LMM en tareas de IQA.

Un nouveau cadre nommé IQARAG a été introduit pour améliorer les capacités d'évaluation de la qualité d'image (IQA) des grands modèles multimodaux (LMM) grâce à une approche sans formation qui utilise la génération augmentée par récupération (RAG). Cette méthode récupère des images de référence sémantiquement similaires avec des scores d'opinion moyens (MOS) correspondants pour améliorer les performances des LMM dans les tâches d'IQA.

A novel framework named IQARAG has been introduced to enhance the Image Quality Assessment (IQA) capabilities of Large Multimodal Models (LMMs) through a training-free approach that utilizes Retrieval-Augmented Generation (RAG). This method retrieves semantically similar reference images with corresponding Mean Opinion Scores (MOSs) to improve the LMM's performance in IQA tasks.

Enhancing Image Quality Assessment Ability of LMMs via Retrieval-Augmented Generation

One More Thing in AI – Your Shortcut to AI Mastery

TAdaRAG: Task Adaptive Retrieval-Augmented Generation via On-the-Fly Knowledge Graph Construction

Was this article worth reading? Share it

One More Thing in AI

LucidQuery AI

Airparser

Magicley AI

Supametas.AI

Scop.ai

Ready to build your own newsroom?