LabelBench

Welcome to LabelBench, where we evaluate label-efficient learning performance with a concerted combination of large pretrained models, semi-supervised learning and active learning algorithms. We encourage researchers to contribute datasets, pretrained models, semi-supervised training algorithms and active learning algorithms to this repo. Additionally, results and findings can be reported under the results directory.

Audience: Machine learning researchers working on active learning, adaptive labeling, and data-efficient training.