Classifier Free Guidance

Classifier-free guidance (CFG) is a technique used to improve the quality and controllability of conditional generative models, particularly diffusion models, by strategically combining conditional and unconditional model predictions during sampling. Current research focuses on refining CFG methods to mitigate issues like oversaturation and mode collapse, exploring adaptive weighting schemes and novel guidance architectures, and extending CFG's application to diverse domains including recommender systems, molecular generation, and speech synthesis. This approach offers a powerful, training-free method for enhancing the performance of pre-trained generative models, impacting various fields by enabling more precise control over generated outputs and improving the efficiency of sampling processes.