SPARTQA

SPARTQA是由密歇根州立大学创建的一个文本问题回答基准，专注于空间推理。该数据集包含1100个独特的QA对，均匀分布在四种类型的问题中：寻找关系（FR）、寻找块（FB）、选择对象（CO）和是/否（YN）。SPARTQA旨在解决现有数据集在空间现象描述上的不足，通过更自然和复杂的故事以及需要深度推理的问题，推动对文本空间推理更复杂模型的研究。SPARTQA is a text-based question answering benchmark focused on spatial reasoning, developed by Michigan State University. This dataset contains 1,100 unique QA pairs, evenly distributed across four question types: Finding Relationships (FR), Finding Blocks (FB), Choosing Objects (CO), and Yes/No (YN). SPARTQA is designed to address the shortcomings of existing datasets in describing spatial phenomena, and it promotes research on more sophisticated models for textual spatial reasoning through more natural and complex narratives as well as questions requiring in-depth reasoning.