lossfunk Research: Diffusion Models Internally Represent Visual Illusions
paraschopra · x · 2026-08-03
New research from @lossfunk reveals that diffusion (denoising) models internally represent visual illusions.
The study tested 9 denoising models across three major architecture families: pixel-space DDPMs, latent diffusion U-Nets, and diffusion transformers.
The findings show that all evaluated models developed internal representations capable of discriminating visual illusions from physically identical control regions.
More from Research
- Transformers Effectively Compute Amortized Gradients of a Non-Conservative Field — teortaxesTex · 2026-08-03
- Surge AI Exec Reveals Why Benchmarks Fail: $15M Cost and Widespread Contamination — alex_verem · 2026-08-03
- Mathematician Uses GPT to Solve Years-Old Gradient Descent Convergence Problem — konstmish · 2026-08-03
- SIGGRAPH Paper Untangles Self-Intersecting Möbius Strips — keenanisalive · 2026-08-03
- Learning a Thousand Robot Tasks in a Day Without Massive Data — chris_j_paxton · 2026-08-03
- Paper Proposes CSED Paradigm: Reshaping Long-term Emotional Support in LLMs — kinamind · 2026-08-03