InfoChartQA

该数据集是一个用于评估多模态大型语言模型在信息图表理解方面的基准，包含了5,642对信息图表和平面图表。每一对图表虽然视觉呈现方式不同，但共享相同的基础数据。此外，该数据集还包括基于视觉元素的问题，能够为多模态大型语言模型进行细粒度的错误分析和消融研究。规模上，该数据集共有5,642对图表，任务类型为多模态问答。

This dataset is a benchmark for evaluating multimodal large language models in infographic comprehension, comprising 5,642 pairs of infographics and planar charts. Each pair of charts differs in visual presentation but shares identical underlying data. Furthermore, the dataset includes questions grounded in visual elements, enabling fine-grained error analysis and ablation studies for multimodal large language models. In terms of scale, this dataset contains a total of 5,642 chart pairs, and the task type is multimodal question answering.