What is classifier free guidance?

Classifier guiding is a newly developed strategy for balancing mode coverage and sample fidelity in post-training conditional diffusion models, similar to low temperature sampling or truncation in other kinds of generative models.