Benchmarking Multimodal RAG through a Chart-based Document Question-Answering Generation Framework
Overview
Multimodal Retrieval-Augmented Generation (MRAG) enhances reasoning capabilities by integrating external knowledge. However, existing benchmarks primarily focus on simple image-text interactions, overlooking complex visual formats like charts that are prevalent in real-world applications. In this work, we introduce a novel task, Chart-based MRAG, to address this… See the full description on the dataset page: https://huggingface.co/datasets/ymyang/Chart-MRAG.
