hpprc/reranker-scores

这是一个包含多个日语问答和检索数据集的集合，每个数据集都使用不同的reranker模型对正负例文书的关联度进行评分。每个数据集配置包含锚文本、正例文书的文本和标题、负例文书的文本和标题，以及六种reranker模型（bge、jina、gte、ruri、ruriv3-preview、ja-ce）的平均评分。数据集适用于训练和评估问答和检索系统。

This is a collection of Japanese QA and retrieval datasets, each of which uses different reranker models to score the relevance of positive and negative example documents. Each dataset configuration includes an anchor text, the text and title of the positive example documents, the text and title of the negative example documents, and the average scores from six reranker models (bge, jina, gte, ruri, ruriv3-preview, ja-ce). The datasets are suitable for training and evaluating QA and retrieval systems.