WikiTabGen

该数据集是一个全新的基准，包含了100个精心挑选的维基百科表格，旨在评估大型语言模型生成表格的能力。这些表格在事实准确性方面经过了处理，并且附有简短的 自然语言描述，以便更好地进行表格生成评估任务。This dataset is a novel benchmark containing 100 carefully curated Wikipedia tables, which is designed to evaluate the table generation capabilities of large language models (LLMs). These tables have been vetted for factual accuracy, and each is paired with a brief natural language description to better support the table generation evaluation task.