VLQA

该数据集名为VLQA，专为视觉-语言问答任务设计，该任务需要联合推理图像和文本对。每个数据点包含一张图片、一段阅读材料、一个问题以及相应的答案。数据集还包括丰富的标注和分类，涵盖图像类型、问题类型、所需的推理能力以及对外部知识的需求等方面。为确保数据质量，该数据集采用了多级别的手动验证流程。规模上，该数据集包含9267个独特的图像-文本-问答条目，其任务定位于视觉-语言问题解答。

This dataset is named VLQA, which is specifically designed for the vision-language question answering (VQA) task that requires joint reasoning over image-text pairs. Each data instance comprises an image, a reading passage, a question, and its corresponding answer. The dataset also includes rich annotations and classifications covering aspects such as image type, question type, required reasoning capabilities, and the demand for external knowledge. To ensure data quality, a multi-level manual verification process is adopted for this dataset. In terms of scale, this dataset contains 9267 unique image-text-question-answer entries, with its task focused on vision-language question answering.