FigureQA

该数据集名为FigureQA，专为视觉问答任务设计，要求模型对图表进行推理并据此回答问题。实验表明，无需地面真实程序注释，神经网络模型（NMN）即可进行训练，展现出更好的泛化能力和性能。该数据集所涉及的任务是视觉问题解答。

This dataset, named FigureQA, is specifically developed for the visual question answering (VQA) task, which demands models to conduct reasoning over charts and answer relevant questions accordingly. Experiments have shown that neural module networks (NMN) can be trained without the need for ground-truth program annotations, demonstrating superior generalization ability and performance. The task covered by this dataset is visual question answering.