mFollowIR-cross-lingual-parquet-mteb
收藏魔搭社区2025-12-05 更新2025-09-13 收录
下载链接:
https://modelscope.cn/datasets/jhu-clsp/mFollowIR-cross-lingual-parquet-mteb
下载链接
链接失效反馈官方服务:
资源简介:
# mFollowIR-cross-lingual-mteb
This is a new version of the mFollowIR-cross-lingual dataset modified to fit the new MTEB format.
1. Restructured queries to include both original and changed versions
2. Separated instructions into a dedicated configuration
3. Reorganized qrels into default (original) and qrel_diff configurations
## Dataset Structure
The dataset contains the following configurations:
### Language: fas
- corpus-fas: Original corpus documents
- queries-fas: Queries with both original and changed versions
- instruction-fas: Instructions for both original and changed queries
- default-fas: Original relevance judgments
- qrel_diff-fas: Changes in relevance judgments
- top_ranked-fas: Top ranked documents for each query
### Language: rus
- corpus-rus: Original corpus documents
- queries-rus: Queries with both original and changed versions
- instruction-rus: Instructions for both original and changed queries
- default-rus: Original relevance judgments
- qrel_diff-rus: Changes in relevance judgments
- top_ranked-rus: Top ranked documents for each query
### Language: zho
- corpus-zho: Original corpus documents
- queries-zho: Queries with both original and changed versions
- instruction-zho: Instructions for both original and changed queries
- default-zho: Original relevance judgments
- qrel_diff-zho: Changes in relevance judgments
- top_ranked-zho: Top ranked documents for each query
# mFollowIR-cross-lingual-mteb
本数据集为适配全新MTEB格式而修改的mFollowIR跨语言数据集新版本。
1. 重构查询结构,同时保留原始版本与修改版本的查询内容;
2. 将指令拆分至独立配置项中;
3. 重新组织相关度标注集(qrels),将其分为默认(原始)配置与qrel_diff配置。
## 数据集结构
本数据集包含如下配置项:
### 语言:fas
- corpus-fas:原始语料库文档
- queries-fas:同时包含原始版本与修改版本的查询
- instruction-fas:适配原始与修改后查询的指令
- default-fas:原始相关度标注结果
- qrel_diff-fas:相关度标注的变更内容
- top_ranked-fas:各查询对应的Top排序文档
### 语言:rus
- corpus-rus:原始语料库文档
- queries-rus:同时包含原始版本与修改版本的查询
- instruction-rus:适配原始与修改后查询的指令
- default-rus:原始相关度标注结果
- qrel_diff-rus:相关度标注的变更内容
- top_ranked-rus:各查询对应的Top排序文档
### 语言:zho
- corpus-zho:原始语料库文档
- queries-zho:同时包含原始版本与修改版本的查询
- instruction-zho:适配原始与修改后查询的指令
- default-zho:原始相关度标注结果
- qrel_diff-zho:相关度标注的变更内容
- top_ranked-zho:各查询对应的Top排序文档
提供机构:
maas
创建时间:
2025-09-10



