I only know the basics about flow matching/diffusion, but this is really nice! Moore, Escher, Penrose: A Conformal Golden Braid arxiv.org/abs/2610.02210
AI Weekly's analysis
→
- Sophia Feldman and Assaf Shocher use a frozen text-to-image diffusion model to generate self-referential scenes modelled on Escher's Print Gallery.
- They construct a generalized inverse T† of the non-invertible transformation and apply the Penrose identity TT†T = T to project onto admissible images.
- Denoising steps are braided between the source geometry and the transformed geometry so the scene and its distortion develop together.
Read full analysis →
