NanoSCIDOCS-bm25
收藏魔搭社区2025-11-12 更新2025-03-01 收录
下载链接：
https://modelscope.cn/datasets/sentence-transformers/NanoSCIDOCS-bm25
下载链接
链接失效反馈官方服务：
资源简介：
# NanoBEIR SCIDOCS with BM25 rankings
This dataset is an updated variant of [NanoSCIDOCS](https://huggingface.co/datasets/zeta-alpha-ai/NanoSCIDOCS), which is a subset of the SCIDOCS dataset from the Benchmark for Information Retrieval (BEIR).
SCIDOCS was created as a subset of the rather large BEIR, designed to be more efficient to run. This dataset adds a `bm25-ranked-ids` column to the `relevance` subset, which contains a ranking of every single passage in the corpus to the query.
This dataset is used in Sentence Transformers for evaluating CrossEncoder (i.e. reranker) models on NanoBEIR by reranking the top *k* results from BM25.
# 带BM25排序结果的NanoBEIR SCIDOCS数据集
本数据集是[NanoSCIDOCS](https://huggingface.co/datasets/zeta-alpha-ai/NanoSCIDOCS)的更新变体，而NanoSCIDOCS本身是源自信息检索基准测试集（Benchmark for Information Retrieval, BEIR）的SCIDOCS数据集的子集。
SCIDOCS 是规模庞大的BEIR的子集，专为提升运行效率而设计。本数据集在`relevance`子集中新增了`bm25-ranked-ids`列，该列包含语料库中所有段落针对查询的排序结果。
本数据集被应用于句向量变换器（Sentence Transformers）框架中，用于在NanoBEIR基准上评估交叉编码器（CrossEncoder，即重排序模型），具体方式为对BM25返回的前*k*个结果执行重排序操作。
提供机构：
maas
创建时间：
2025-02-26
