ConceptClassifier-Free Guidance / CFG

Classifier Guidance

A technique for strengthening conditional control in generative models. During training, conditions are randomly dropped so that the network learns both conditional and unconditional predictions; during inference, the conditional signal is amplified by extrapolating from the difference between the conditional and unconditional predictions.

Timeline (last 90 days)

Oct 12

Guided diffusion applies gradients at each step of the reverse denoising process; common guidance approaches include classifier guidance and classifier-free guidance.

Unverified50%

All Facts (1)

Unverified

Guided diffusion applies gradients at each step of the reverse denoising process; common guidance approaches include classifier guidance and classifier-free guidance.

50%

Source Articles

RFD4-Proteina: Designing Proteins with Reward-Guided Inference-Time SearchOct 21 SteerTTS: A 23M-Parameter TTS Model with Hybrid Attention and Bidirectional Mamba-3 for Fine-Grained Prosody ControlOct 19