ProvenanceAI/financial-data-engineering-sample

这是Provenance AI金融数据工程数据集的免费样本，包含用于RLHF和微调的专家级偏好数据。数据集涵盖了将Excel Power Query管道迁移到Python/Polars、USPS DARS财务系统约定和调试实际管道错误等内容。数据独特之处在于它来自实际的生产管道迁移，包含真实的错误和领域约定。数据集提供SFT和DPO格式，完整数据集还包括更多示例和案例研究。README还提到了其他可用的数据集，并提供了引用信息。

Free sample from the Provenance AI Financial Data Engineering dataset, containing expert-level preference data for RLHF and fine-tuning in financial data engineering. The dataset covers migrating Excel Power Query pipelines to Python/Polars, USPS DARS financial system conventions, and debugging real pipeline bugs. The data is unique as it comes from an actual production pipeline migration, with real bugs and domain conventions. The dataset is available in SFT and DPO formats, and the full dataset includes additional examples and a case study. The README also mentions other available datasets and provides citation information.