New paper links Föllmer process to DDPM denoising diffusion samplers

michaelchchoi · x · 2026-09-29

In an ongoing paper-recommendation thread, michaelchchoi shares Yuta Koike's open-access paper (Japanese Journal of Statistics and Data Science):

• The Föllmer process — Brownian motion conditioned to hit a target distribution at time 1 — is an "augmented" time-compressed version of the reverse SDE behind DDPM
• The paper clarifies how direct discretization of the Föllmer process relates to the DDPM sampler
• It yields natural hyperparameter settings for DDPM and accommodates a broader class of variance schedules than discretized reverse SDEs
• This systematically recovers state-of-the-art DDPM sampling error bounds with slight improvements

Original post →

More from Research

• Amazon AGI's AutoGym auto-generates tasks, environments, and verifiers for agent RL training — omarsar0 · 2026-09-29
• RECLAIM preprint: best AI agent reproduces only 41% of papers with code, 15% without — VraserX · 2026-09-29
• Diffusion Schrödinger Bridge paper solves entropy-regularized optimal transport to speed up generative modeling — skoularidou · 2026-09-29
• NeurIPS paper: AI protein folding models converge on a shared two-stage mechanism — rishabh16_ · 2026-09-29
• Stanford's CME 295 Transformers & LLMs course returns, Lecture 1 on YouTube — afshinea · 2026-09-29
• An LLM beat Zork by cheating: analysis shows Astra played like it had a walkthrough — rajammanabrolu · 2026-09-29

Research channel →