遇见数据集

opensearch-project/msmarco-hard-negatives-llm-scores

收藏
Hugging Face2025-08-14 更新2025-09-13 收录
官方服务:

资源简介:

MS MARCO硬负样本LLM分数数据集,用于训练检索系统。该数据集从MS MARCO训练集派生而来,包含每个查询的顶部100个候选文档,并附有来自双编码器和交叉编码器教师的重排分数。

MS MARCO Hard Negatives LLM Scores dataset for training retrieval systems. Derived from the MS MARCO train split, it includes the top-100 candidate documents for each query with re-ranking scores from various bi-encoder and cross-encoder teachers.

提供机构:
opensearch-project
二维码
社区交流群
二维码
科研交流群
商业服务