INSTRUCTEXCEL

INSTRUCTEXCEL是一个大规模的基准数据集，旨在评估大型语言模型在自然语言指令下生成Excel OfficeScripts代码的能力。该数据集由微软主导创建，利用Excel的‘自动化’功能从用户的操作中自动生成OfficeScripts。数据集包含超过10,000个样本，涵盖170多种Excel操作，覆盖2,000个公开可用的Excel工作表。这些样本通过众包方式收集，每个样本包括自然语言描述和相应的OfficeScript代码。INSTRUCTEXCEL的应用领域主要集中在帮助非专家用户通过自然语言指令执行复杂的Excel任务，从而提高工作效率和学习Excel功能。

INSTRUCTEXCEL is a large-scale benchmark dataset designed to evaluate the ability of large language models (LLMs) to generate Excel OfficeScripts code under natural language instructions. This dataset was primarily created by Microsoft, which leverages Excel's "Automate" feature to automatically generate OfficeScripts from user operations. It contains over 10,000 samples covering more than 170 types of Excel operations, and spans 2,000 publicly available Excel worksheets. These samples were collected via crowdsourcing, with each sample consisting of a natural language description and its corresponding OfficeScript code. The main application scenarios of INSTRUCTEXCEL focus on helping non-expert users perform complex Excel tasks through natural language instructions, thereby improving work efficiency and facilitating the learning of Excel functions.