遇见数据集

<b>CMiLBench</b>

收藏
DataCite Commons2025-06-20 更新2025-09-08 收录
官方服务:

资源简介:

CMiLBench is a hierarchical multitask evaluation benchmark specifically designed for Chinese minority languages (Tibetan `bo`, Mongolian `mn`, Uyghur `ug`). This benchmark aims to systematically evaluate large language models' understanding, generation, and safety alignment capabilities in low-resource language environments.CMiLBench contains the following three major task categories with a total of 17 subtasks, covering linguistic foundational capabilities, cultural knowledge abilities, and multilingual safety.

提供机构:
figshare
创建时间:
2025-06-20
二维码
社区交流群
二维码
科研交流群
商业服务