BertilBraun/competency-extraction-dpo-v2

Competency Extraction DPO数据集是一个专门设计用于从科学出版物中提取能力概况的数据集。该数据集专注于从学术论文的摘要中识别和结构化能力。它包含6,179个样本，提供了关于自动化能力提取的宝贵见解。数据集采用标准DPO格式，包含prompt、chosen和rejected条目。prompt提示模型从摘要列表中提取能力，chosen和rejected条目仅包含提取的能力概况。数据集内容基于科学出版物摘要的能力概况提取，使用了Mixtral-8x7B-Instruct-v0.1模型，并采用了锦标赛风格的排名方法来确定最相关的能力概况。

The Competency Extraction DPO Dataset is a specialized dataset designed to extract competency profiles from scientific publications. The dataset focuses on identifying and structuring competencies found within the abstracts of academic papers, containing 6,179 samples and providing valuable insights into automated competency extraction. The dataset uses the Mixtral-8x7B-Instruct-v0.1 model for extracting competency profiles and determines the most relevant profiles through a tournament-style ranking method. The dataset follows the standard DPO dataset format, including prompt, chosen, and rejected entries.