rlhn/hn-remove-680K
收藏Hugging Face2025-05-27 更新2025-11-01 收录
下载链接：
https://hf-mirror.com/datasets/rlhn/hn-remove-680K
下载链接
链接失效反馈官方服务：
资源简介：
RLHN（难负样本清理）数据集是一个包含680K训练对的数据集，它从现有的BGE训练集中选取了7个数据集的查询、正例和经过清理的难负样本。该数据集用于微调嵌入、ColBERT或多元向量以及重排模型，以提高信息检索的鲁棒性。
The RLHN (Hard Negative Cleaning) dataset is a dataset containing 680K training pairs, which selects queries, positives, and cleaned hard negatives from 7 datasets in the existing BGE training collection. This dataset is used for fine-tuning embeddings, ColBERT or multi-vector, and reranker models to improve the robustness of information retrieval.
提供机构：
rlhn
