AutoDCWorkflow Benchmark
收藏资源简介:
AutoDCWorkflow Benchmark是一个用于评估大型语言模型(LLM)自动生成数据清洗工作流能力的基准数据集。该数据集包含四个真实世界的数据集,分别是Menu、Dish、Paycheck Protection Program (PPP) loan data和Chicago Food Inspection data,每个数据集被注入了不同类型的数据错误,并准备了20到30个数据清洗目的。数据集的创建过程基于不同的表格内容和列模式,生成了67个实例级别的目的,这些目的根据相关列的数量和数据的“脏度”状态分为不同的难度级别。该数据集主要用于测试LLM在自动生成数据清洗工作流方面的表现,旨在解决数据清洗任务中的自动化问题。
AutoDCWorkflow Benchmark is a benchmark dataset for evaluating the capability of Large Language Models (LLMs) to automatically generate data cleaning workflows. This dataset includes four real-world datasets, namely Menu, Dish, Paycheck Protection Program (PPP) loan data, and Chicago Food Inspection data. Each dataset is injected with distinct types of data errors, and 20 to 30 data cleaning objectives are prepared for each. The dataset construction process generates 67 instance-level objectives based on diverse table contents and column schemas, which are categorized into different difficulty levels according to the number of relevant columns and the dirtiness status of the data. This dataset is primarily used to test the performance of LLMs when automatically generating data cleaning workflows, with the goal of addressing automation issues in data cleaning tasks.




