Representatively Distilled Dataset
收藏arXiv2025-09-30 收录
数据链接：
官方服务：
资源简介：
该数据集是通过一种新颖的基于扩散的框架（D^3HR）生成的精炼数据集，它确保了原始数据集的高代表性。该数据集的生成采用了三个阶段的流程，包括领域映射、分布匹配和群体采样，旨在保留信息并确保结构的一致性。该数据集的任务是数据集精炼。
This dataset is a refined dataset generated via a novel diffusion-based framework (D^3HR), which ensures high representativeness of the original dataset. It is constructed through a three-stage workflow including domain mapping, distribution matching and population sampling, aiming to retain information and guarantee structural consistency. The task of this dataset is dataset refinement.
