StructFact

该数据集名为StructFact，旨在评估大型语言模型（LLM）从结构化数据中推断事实知识时的结构化推理能力。它包含了8,340个事实性问题，覆盖了不同的任务、领域、时间线和地区。此外，StructFact还有助于探究当前LLM在从结构化数据中推断事实知识方面的局限性，并推动相关现实应用领域的进步。该数据集的任务是评估LLM在五个基于结构化数据的事实性任务中的推理能力。

This dataset, termed StructFact, is developed to evaluate the structured reasoning capabilities of large language models (LLMs) during their inference of factual knowledge from structured data. It encompasses 8,340 factual questions spanning diverse tasks, domains, timelines and regions. Furthermore, StructFact enables the investigation of the limitations of contemporary LLMs in inferring factual knowledge from structured data, while advancing progress in relevant real-world application domains. The core objective of this dataset is to assess the reasoning abilities of LLMs across five structured data-grounded factual tasks.