AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Learn to Guide Your Diffusion Model

Alexandre Galashov, Ashwini Pokle, Arnaud Doucet, Arthur Gretton, Mauricio Delbracio, Valentin De Bortoli

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Guidance for conditional diffusion · 11 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
learn conditioning-dependent guidance weights via self-consistency instead of constant guidancediffusion model · ImageNet 64x64FID: 2.40 -> 1.99Improved
learn guidance weights via self-consistency instead of unguided samplingdiffusion model · ImageNet 64x64FID: 4.46 -> 1.99Improved
learned self-consistency weights instead of limited interval guidancediffusion model · ImageNet 64x64FID: 2.11 -> 1.99Improved
condition guidance weights on class c in self-consistency (ablation)self-consistency guidance · ImageNet 64x64FID: 2.07 -> 1.99Improved
condition guidance weights on class c in self-consistency (ablation)self-consistency guidance · ImageNet 64x64Inception Score: 76.7 -> 73.62Hurt
use l2 objective instead of self-consistency MMD objective (ablation)learned guidance · ImageNet 64x64FID: 1.99 -> 2.09Hurt
learned self-consistency weights instead of unguided samplingdiffusion model · CelebA 64x64FID: 2.44 -> 2.10Improved
add conditioning to self-consistency guidance weights (ablation)self-consistency guidance · CelebA 64x64FID: 2.28 -> 2.10Improved
constant guidance w=0.01 instead of nonediffusion model · CelebA 64x64FID: 2.44 -> 2.45No clear effect
learned self-consistency guidance with negative prompt instead of unguidedflow matching T2I model · MS COCO 512x512FID: 24.74 -> 18.01Improved
learned self-consistency guidance instead of constant w=7.5flow matching T2I model · MS COCO 512x512FID: 31.2 -> 18.01Improved
learned self-consistency guidance instead of constant w=7.5flow matching T2I model · MS COCO 512x512CLIP Score: 0.306 -> 0.295Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy