VAST27M
收藏arXiv2025-09-30 收录
下载链接：
https://ispamm.github.io/GRAM/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了150,000个样本，是VAST模型的一个子集，专门用于对GRAM模型进行预训练。此外，该数据集还用于重塑GRAM模型的潜在空间，其规模为150,000个样本，任务是对多模态表征学习进行预训练。
This dataset contains 150,000 samples, which acts as a subset of the VAST model and is specifically tailored for pretraining the GRAM model. Furthermore, this dataset is employed to reshape the latent space of the GRAM model, with its primary task being pretraining for multimodal representation learning.
提供机构：
VAST
