AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Improving Classifier-Free Guidance in Masked Diffusion: Low-Dim Theoretical Insights with High-Dim Impact

Kevin Rojas, Ye He, Chieh-Hsin Lai, Yuhta Takida, Yuki Mitsufuji, Molei Tao

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Guidance for conditional diffusion

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
column-normalize the guided rate matrix (normalized guidance) (ablation)MaskGIT with constant guidance, tau-leaping · ImageNet-256 (50K samples, 50 steps)FIDImproved
use normalized guidance instead of unlocking or simple guidance at moderate wMaskGIT · ImageNet-256 precision/recallprecisionImproved
add normalization to guidance (ablation)Show-o and Meissonic · GenEvalprompt alignment scoreImproved
add normalization to guidance (ablation)LLaDA-8B-Instruct, Euler sampler without remasking · MATH-500accuracyImproved
use increasing guidance schedules (Right Interval, Ramp-Up) (ablation)MaskGIT · ImageNet-256 (10K samples)FIDImproved
use Ramp-Up instead of Right Interval schedule (ablation)MaskGIT · ImageNet-256 (10K samples)FIDImproved
use decreasing guidance schedules (Left Interval, Ramp-Down) (ablation)MaskGIT · ImageNet-256 (10K samples)FIDHurt
add normalized guidance in uniform diffusionconditional QM9 uniform diffusion model · QM9 (ring-count conditioning)validity, uniqueness, noveltyImproved

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