rlhn/remove-100K

RLHN数据集是一个包含100K训练对的Tevatron数据集，它包含了原始的查询、正样本和经过清理后的难负样本。这个数据集可以用来微调嵌入、ColBERT或多元向量以及重排模型。原始数据集（包含错误负样本，质量不佳）可以在rlhn/default-100K找到。注意：RLHN数据集不是新的训练数据集，而是对现有BGE集合训练数据集中的难负样本进行清理后的数据集。

The RLHN dataset is a Tevatron dataset (100K training pairs) containing the original queries, positives, and hard negatives after cleaning each training pair with a single false negative. This dataset can be used to fine-tune embedding, ColBERT, or multi-vector reranker models. The original dataset (with false negatives and bad quality) can be found at rlhn/default-100K. Note: RLHN datasets are not new training datasets but existing BGE collection training datasets with hard negatives cleaned!