ZNO-Vision
收藏资源简介:
ZNO-Vision数据集是由乌克兰天主教大学开发的全面多模态乌克兰语基准,源自标准化大学入学考试(ZNO)。该数据集包含4306个专家设计的问题,涵盖数学、物理、化学和人文等12个学术领域。数据集的创建过程包括从Osvita门户收集问题,并过滤掉不适合的问题类型。ZNO-Vision数据集主要用于评估多模态模型在学术和文化领域的乌克兰语理解能力,旨在解决低资源语言中多模态模型评估的不足问题。
The ZNO-Vision dataset is a comprehensive Ukrainian-language multimodal benchmark developed by the Ukrainian Catholic University, derived from the Unified National Entrance Examination (ZNO). It contains 4,306 expert-designed questions spanning 12 academic disciplines including mathematics, physics, chemistry, and humanities. The dataset was constructed by collecting questions from the Osvita Portal and filtering out inappropriate question types. The ZNO-Vision dataset is primarily used to evaluate the Ukrainian language comprehension capabilities of multimodal models in academic and cultural contexts, aiming to address the gap in multimodal model evaluation for low-resource languages.




