MRAG-Suite: A Diagnostic Evaluation Platform for Visual Retrieval-Augmented Generation

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Ji, Yuelyu, Lan, Wuwei, NG, Patrick
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908760974295040
author Ji, Yuelyu
Lan, Wuwei
NG, Patrick
author_facet Ji, Yuelyu
Lan, Wuwei
NG, Patrick
contents Multimodal Retrieval-Augmented Generation (Visual RAG) significantly advances question answering by integrating visual and textual evidence. Yet, current evaluations fail to systematically account for query difficulty and ambiguity. We propose MRAG-Suite, a diagnostic evaluation platform integrating diverse multimodal benchmarks (WebQA, Chart-RAG, Visual-RAG, MRAG-Bench). We introduce difficulty-based and ambiguity-aware filtering strategies, alongside MM-RAGChecker, a claim-level diagnostic tool. Our results demonstrate substantial accuracy reductions under difficult and ambiguous queries, highlighting prevalent hallucinations. MM-RAGChecker effectively diagnoses these issues, guiding future improvements in Visual RAG systems.
format Preprint
id arxiv_https___arxiv_org_abs_2509_24253
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MRAG-Suite: A Diagnostic Evaluation Platform for Visual Retrieval-Augmented Generation
Ji, Yuelyu
Lan, Wuwei
NG, Patrick
Computation and Language
Multimodal Retrieval-Augmented Generation (Visual RAG) significantly advances question answering by integrating visual and textual evidence. Yet, current evaluations fail to systematically account for query difficulty and ambiguity. We propose MRAG-Suite, a diagnostic evaluation platform integrating diverse multimodal benchmarks (WebQA, Chart-RAG, Visual-RAG, MRAG-Bench). We introduce difficulty-based and ambiguity-aware filtering strategies, alongside MM-RAGChecker, a claim-level diagnostic tool. Our results demonstrate substantial accuracy reductions under difficult and ambiguous queries, highlighting prevalent hallucinations. MM-RAGChecker effectively diagnoses these issues, guiding future improvements in Visual RAG systems.
title MRAG-Suite: A Diagnostic Evaluation Platform for Visual Retrieval-Augmented Generation
topic Computation and Language
url https://arxiv.org/abs/2509.24253