INTERSCRIPT

INTERSCRIPT是由艾伦人工智能研究所创建的数据集，包含8466条用户对模型生成脚本的反馈数据。该数据集旨在通过用户反馈帮助模型理解和修正脚本中的错误，特别是在处理日常任务的复杂结构时。数据集内容涵盖了脚本生成中的多种错误类型，如错误的步骤顺序、缺失的关键步骤等。创建过程中，研究人员通过众包方式收集反馈，确保数据的真实性和多样性。该数据集主要应用于交互式学习领域，旨在通过用户反馈提升模型在结构化预测任务中的表现。

INTERSCRIPT is a dataset developed by the Allen Institute for AI. It contains 8,466 user feedback samples regarding model-generated scripts. This dataset aims to enable models to understand and rectify errors in scripts via user feedback, especially when dealing with complex structures of everyday tasks. The dataset covers a wide range of error categories in script generation, including incorrect step ordering, missing critical steps, and others. During its creation, researchers collected feedback through crowdsourcing to guarantee the authenticity and diversity of the dataset. This dataset is primarily utilized in the field of interactive learning, with the goal of enhancing model performance on structured prediction tasks via user feedback.