Information-Theoretic Scaling Laws Predict Memorization and FID in Diffusion Models

A new paper by Luca Ambrogioni, Druv Pai, and Yi Ma introduces information-theoretic scaling laws for diffusion models. By quantifying how much information a model learns from training data, the work rigorously predicts when memorization begins and how FID evolves.

2026-10-29 ~ 2026-10-29 · 3 related posts

• Information-theoretic scaling laws predict memorization and FID in diffusion models — LucaAmb · 2026-10-29
• Information-theoretic scaling laws for diffusion models predict when memorization begins — davidad · 2026-10-29

1 near-duplicate retellings: LucaAmb