Magneto/qa-dataset-llm-judge-flattened

本数据集包含了5008个高质量的问答对，这些问答对是从官方监管文件中提取的，并使用LLM-as-Judge方法进行了分析和质量评估。每个问答对都根据立足性、相关性、完整性和准确性四个关键标准进行了评估，并提供了改进后的答案和详细的评估解释。数据集采用扁平化的表格结构，易于分析和机器学习应用。

This dataset contains 5,008 high-quality question-answer pairs extracted from regulatory and policy documents, analyzed and quality-assessed using LLM-as-Judge methodology. Each Q&A pair is evaluated based on four key criteria: groundedness, relevance, completeness, and accuracy, with improved answers and detailed evaluation explanations provided. The dataset is structured in a flattened tabular format, suitable for analysis and machine learning applications.