MiLiC-Eval
收藏资源简介:
MiLiC-Eval是由北京大学团队创建的第一个针对中国少数民族语言的标准化基准,包含9个任务和24000个实例,专注于4种少数民族语言。该数据集涵盖了词汇理解、主题分类、阅读理解、响应选择、标题生成、机器翻译和数学推理等多种语言和问题解决技能的任务,旨在评估大型语言模型在少数民族语言上的表现,特别是那些使用不常见书写系统的语言。
MiLiC-Eval is the first standardized benchmark for Chinese minority languages, developed by a research team from Peking University. It includes 9 tasks and 24,000 instances, focusing on 4 minority languages. This dataset covers tasks involving various linguistic and problem-solving skills, such as vocabulary comprehension, topic classification, reading comprehension, response selection, title generation, machine translation, and mathematical reasoning. It aims to evaluate the performance of large language models on Chinese minority languages, especially those that use uncommon writing systems.

- 1MiLiC-Eval: Benchmarking Multilingual LLMs for China's Minority Languages北京大学 · 2025年



