Starkly Speaking on Monday: Classifier-Free Guidance: From High-Dimensional Analysis to Generalized Guidance Forms

[Hannes Stärk]

Hi together,

Let us understand a core ingredient of diffusion models better - classifier free guidance!

Paper:
Classifier-Free Guidance: From High-Dimensional Analysis to Generalized Guidance Forms https://arxiv.org/abs/2502.07849 (Krunoslav Lehman Pavasovic, Jakob Verbeek, Giulio Biroli, Marc Mezard)
Classifier-Free Guidance (CFG) is a widely adopted technique in diffusion and flow-based generative models, enabling high-quality conditional generation. A key theoretical challenge is characterizing the distribution induced by CFG, particularly in high-dimensional settings relevant to real-world data. Previous works have shown that CFG modifies the target distribution, steering it towards a distribution sharper than the target one, more shifted towards the boundary of the class. In this work, we provide a high-dimensional analysis of CFG, showing that these distortions vanish as the data dimension grows. We present a blessing-of-dimensionality result demonstrating that in sufficiently high and infinite dimensions, CFG accurately reproduces the target distribution. Using our high-dimensional theory, we show that there is a large family of guidances enjoying this property, in particular non-linear CFG generalizations. We study a simple non-linear power-law version, for which we demonstrate improved robustness, sample fidelity and diversity. Our findings are validated with experiments on class-conditional and text-to-image generation using state-of-the-art diffusion and flow-matching models.

Speaker:
Krunoslav Lehman Pavasovic who is a PhD Student at Meta & ENS Paris.

Meeting Details:
Every Monday at 12:00 ET / 9:00 PT / 18:00 CE(S)T.  
https://zoom.us/j/5775722530?pwd=ZzlGTXlDNThhUDZOdU4vN2JRMm5pQT09

Add it to your calendar:
Subscribe via Google Calendar, or subscribe via iCal.
Alternatively, add the events, or add this single event.

Slack Workspace for discussion and paper voting:
https://join.slack.com/t/logag/shared_invite/zt-2zuxi7gd1-rLUgxg6gnCkhO7WlRsyElg

All information: Schedule of upcoming papers, recordings, etc.:
https://portal.valencelabs.com/starklyspeaking