Caption :

MIT researchers developed a technique that reduces the tendency for contrastive learning models to use shortcuts, by forcing the model to focus on features in the data that it hadn’t considered before.