Explains how generative models move from noise to an image and how guidance changes that path. His illustrated essays connect a model control to both stronger conditioning and visible distortions.
Why read Sander Dieleman?
Read Dieleman when a diffusion model's guidance setting feels like a mysterious quality slider. His geometry essay uses diagrams to explain how conditioning changes the denoising path and why pushing it too far can damage an image. The visual sections offer a starting point, while the equations need some probability and machine-learning background.
His earlier guidance essay fills in the distinction between classifier guidance and classifier-free guidance. It explains research by other authors rather than claiming he invented either method. Together the articles help you interpret a control exposed by a diffusion workflow; they do not imply every image-generation app uses or exposes that control.
Start with the original
Selected work
-
Illustrated research explainer ·
The geometry of diffusion guidance
What does stronger guidance do to a diffusion model? Dieleman's diagrams connect denoising steps to guidance strength and show why excessive guidance can distort output. Start with the visual sections before the equations.
Cite this selection
XFollowList, “The geometry of diffusion guidance” reading notes in the Sander Dieleman profile. https://xfollowlist.com/people/sander-dieleman#work-e017d7c8572f
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original work published .
Profile-source checking date and references · Dated reading map
-
Research explainer ·
Guidance: a cheat code for diffusion models
How do classifier and classifier-free guidance differ? His own explanation builds the intuition and then the mathematics, with references to the original research. Useful after the geometry essay.
Cite this selection
XFollowList, “Guidance: a cheat code for diffusion models” reading notes in the Sander Dieleman profile. https://xfollowlist.com/people/sander-dieleman#work-77ebbd7249f5
The linked work is the original source; this permalink identifies XFollowList’s reason to read it. Original work published .
Profile-source checking date and references · Dated reading map
Projects & roles
Sander Dieleman’s generative-model essays
His publication on generative models, including illustrated explanations of diffusion sampling and guidance.
Community rating
Sources & further reading
Profile sources checked . This describes checking these references, separately from the publication dates of selected works.
- His own biography identifies his Google DeepMind research and distinguishes contributions to image, video and music work.
- His author site links the @sedielem account.
- Current public profile metadata names Google DeepMind and links sander.ai.
Cite this profile’s source check
XFollowList, Sander Dieleman profile sources. Sources checked 2026-10-11. https://xfollowlist.com/people/sander-dieleman#sources-checked