The Spacetime of Diffusion Models: An Information Geometry Perspective

ICLROral2026

Authors
Rafal Karczewski, Markus Heinonen, Alison Pouplin, Søren Hauberg, Vikas K Garg
Affiliation
Aalto University
Venue
ICLR 2026
Track
Oral

TL;DR

We present a novel geometric perspective on the latent space of diffusion models. We first show that the standard pullback approach, utilizing the deterministic probability flow ODE decoder, is fundamentally flawed.

Opening excerpt from the authors’ abstract. source

Read the paper

Topics

diffusion geometric

← All ICLR 2026 Oral papers · Browse the whole archive