ReadingTimeMachine/visual_qa_multipanel

这是一个用于测试大型语言/多模态模型（LMMs）的多面板视觉问答（VQA）数据集。数据集包含了合成的多面板图形以及与之相关的问答对。图形从易于人类/LMMs解析到因数据标记重叠而非常棘手的都有。

This is a multi-panel visual question and answer (VQA) dataset for testing large language/multimodal models (LMMs). The dataset contains synthetically generated figures along with associated question-answer pairs. The figures range from easy to decipher by humans/LMMs to very tricky due to overlapping data marks.