GlobalBench

GlobalBench是一个旨在推动全球自然语言处理（NLP）技术均衡发展的基准数据集。该数据集由卡内基梅隆大学等机构创建，包含966个数据集，覆盖190种语言，旨在动态追踪所有NLP数据集的进展。GlobalBench不仅关注准确性，还追踪技术在所有语言中的估计用户效用和公平性，提供多维度的语言技术服务全球人民的视角。此外，GlobalBench设计用于识别最未被充分服务的语言，并奖励针对这些语言的研究努力。

GlobalBench is a benchmark dataset developed to promote the balanced advancement of global natural language processing (NLP) technologies. Created by institutions including Carnegie Mellon University, this dataset comprises 966 subsets covering 190 languages, and is designed to dynamically track the progress of all NLP-related datasets. Beyond focusing on model accuracy, GlobalBench also tracks the estimated user utility and fairness of technologies across all covered languages, providing a multi-dimensional perspective on how language technologies serve people worldwide. Additionally, GlobalBench is intended to identify the most under-served languages and incentivize research efforts targeting these languages.