cross-encoder/ettin-reranker-v1-data

这是用于训练cross-encoder/ettin-reranker系列CrossEncoder重排器的训练数据集。它混合了广泛领域的文本对数据，以及通过强教师重排器重新评分的检索对，所有标签均由自动评分系统生成，而非人工标注。数据集包含多个子集，如agnews、altlex、amazon_qa等，每个子集都有查询（query）、文档（document）和标签（label）三列，其中标签是标量相关性或相似性分数。数据来源于多个公开数据集，如sentence-transformers、nomic、beir等，并通过lightonai/embeddings-pre-training等渠道处理。

This is the training dataset for the cross-encoder/ettin-reranker series of CrossEncoder rerankers. It comprises text pair data across a wide range of domains, alongside retrieval pairs rescored by a robust teacher reranker, with all labels generated by an automated scoring system rather than manual annotations. The dataset includes multiple subsets such as agnews, altlex, amazon_qa, and others. Each subset contains three columns: query, document, and label, where the label represents a scalar correlation or similarity score. The data is sourced from multiple public datasets including sentence-transformers, nomic, beir, and more, and processed via channels such as lightonai/embeddings-pre-training.