RealCQA

该数据集名为RealCQA，专为图表问答任务设计，旨在通过结构化的前提条件来评估模型的分析推理能力。它强调将推理能力与视觉理解相结合，用于评估不同模型在处理与图表相关各类问题时的表现。该数据集包含150万个问答对，30,000个挑战性JSON文件以及52,000个结构化前提条件，任务集中在图表问答上。

The dataset, named RealCQA, is exclusively designed for chart question answering tasks, aiming to evaluate the analytical reasoning abilities of models via structured premises. It highlights the integration of reasoning capabilities and visual understanding, and is utilized to assess the performance of diverse models when dealing with various chart-related questions. This dataset encompasses 1.5 million question-answer pairs, 30,000 challenging JSON files, and 52,000 structured premises, with its core task focused on chart question answering.