UMVM
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/zjukg/UMAEA
下载链接
链接失效反馈官方服务：
资源简介：
该数据集（MMEA-UMVM）旨在分析在多模态实体对齐场景中不确定缺失视觉模态的影响，它是通过从现有的MMEA数据集中随机删除实体图像而生成的。数据集包含了从0.05到原始数据集最大值不等的视觉模态缺失程度。其规模涵盖了97个基准分割，任务专注于多模态实体对齐。
This dataset (MMEA-UMVM) is designed to analyze the impact of uncertain missing visual modalities in multimodal entity alignment scenarios. It is generated by randomly deleting entity images from an existing MMEA dataset. The dataset covers visual modality missing rates ranging from 0.05 to the maximum value of the original dataset, includes 97 benchmark splits, and focuses on the task of multimodal entity alignment.
提供机构：
Authors of the paper
