duongtr/ReasonVQA

ReasonVQA是一个具有结构知识的视觉问答多跳推理基准数据集。数据集包含了训练和验证的JSONL格式的问题数据，每个问题包括问题ID、问题内容、图片ID、图片名称、图片路径、来源、跳数、答案列表、正确性标记、类别列表、属性ID、属性标签、实体ID、实体标签、路由信息和是否包含场景图信息。

ReasonVQA is a Multi-hop Reasoning Benchmark with Structural Knowledge for Visual Question Answering. The dataset includes questions in JSONL format for training and validation, each containing question ID, question content, image ID, image name, image path, source, number of hops, list of answers, correctness indicators, list of categories, property ID, property label, entity ID, entity label, route information, and whether it includes scene graph information.