Qiuchen-Wang/NavQA_Revised

NavQA Revised是NaVQA数据集的重新注释版本，主要用于导航视频问答任务。数据集包含两个JSONL文件，每个文件有210个问答示例，总计420行数据。数据集的语言为英语，任务类别包括问答、视频问答和具身导航推理。数据集的主要变化包括时区校正和2 fps匹配。每个示例包含多个字段，如id、seq_id、question、type、answers等。数据集旨在评估导航导向的视频问答系统和具身代理的长时空间和时间记忆推理能力。

NavQA Revised is a re-annotated version of the NaVQA dataset, primarily used for navigation video question answering tasks. The dataset includes two JSONL files, each containing 210 question-answer examples, totaling 420 rows of data. The datasets language is English, and its task categories include question answering, video question answering, and embodied navigation reasoning. The main changes in the dataset include timezone correction and 2 fps matching. Each example contains multiple fields such as id, seq_id, question, type, answers, etc. The dataset is intended to evaluate navigation-oriented video question answering systems and embodied agents long-horizon spatial and temporal memory reasoning capabilities.