Automated Model Card Generation Dataset

本数据集名为Automated Model Card Generation Dataset，由印度理工学院甘地讷格尔分校创建，旨在自动化生成机器学习模型的卡片。数据集包含500个问题-答案对，涉及25个机器学习模型，覆盖模型训练配置、数据集、偏差、架构细节和训练资源等关键方面。数据集通过标注者从原始论文中提取答案，并经过初步和专家两阶段标注流程确保质量。该数据集应用于训练模型，以自动化从论文文本生成模型卡片，减少人工在模型卡片编制过程中的努力，主要解决模型文档自动化生成的问题。This dataset, named Automated Model Card Generation Dataset, was developed by the Indian Institute of Technology Gandhinagar for the purpose of automating the generation of machine learning model cards. It consists of 500 question-answer pairs across 25 machine learning models, covering critical aspects including model training configurations, datasets, biases, architectural details, and training resources. The answers within the dataset are extracted by annotators from original research papers, with its quality guaranteed via a two-stage annotation workflow comprising preliminary and expert review phases. This dataset is utilized to train models that automatically generate model cards from scholarly paper texts, reducing manual labor in model card compilation and primarily addressing the challenge of automated model documentation generation.