diversity guidance

A strategy used in machine learning, particularly in generative models, where the training process includes mechanisms to encourage variety in outputs, preventing mode collapse, and ensuring a richer set of generated samples.