czh-up/CoMT
收藏官方服务:
资源简介:
CoMT是一个为大型视觉语言模型设计的新型多模态思维链基准数据集,包含视觉创作、视觉删除、视觉更新和视觉选择四个任务类别,旨在通过多模态输入和多模态输出模仿人类的视觉操作整合思维过程。
CoMT is a novel benchmark designed for large vision-language models, featuring four task categories: Visual Creation, Visual Deletion, Visual Update, and Visual Selection. It aims to mimic human-like visual operation integration through multi-modal input and multi-modal output.
提供机构:
czh-up


