Dataset distillation is the task of synthesizing a small dataset such that models trained on it achieve high performance on the original large dataset. A dataset distillation algorithm takes as input a large real dataset to be distilled (training set), and outputs a small synthetic distilled dataset, which is evaluated via testing models trained on this distilled dataset on a separate real dataset (validation/test set). A good small distilled dataset is not only useful in dataset understanding, but has various applications (e.g., continual learning, privacy, neural architecture search, etc.). This task was first introduced in the paper Dataset Distillation [Tongzhou Wang et al., '18], along with a proposed algorithm using backpropagation through optimization steps. Then the task was first extended to the real-world datasets in the paper Medical Dataset Distillation [Guang Li et al., '19], which also explored the privacy preservation possibilities of dataset distillation. In the paper Dataset Condensation [Bo Zhao et al., '20], gradient matching was first introduced and greatly promoted the development of the dataset distillation field.

In recent years (2022-now), dataset distillation has gained increasing attention in the research community, across many institutes and labs. More papers are now being published each year. These wonderful researches have been constantly improving dataset distillation and exploring its various variants and applications.