AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Guided Star-Shaped Masked Diffusion

Viacheslav Meshchaninov, Egor Shibaev, Artem Makoian, Ivan Klimov, Nikita Balagansky, Daniil Gavrilov, Aibek Alanov, Dmitry Vetrov

ICML 2026 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration

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
apply pure star-shaped sampler over the whole trajectory (ablation)MDLM sampling · OWT 128 tokensMAUVE: near-zero MAUVE at t_on=1.0Hurt
switch from MDLM to star-shaped sampler late in generation (t_on about 0.3) (ablation)MDLM · OWT 128 tokensMAUVE / perplexity: peak at t_on about 0.3Improved
add learned error predictor to unguided Star+ (G-Star+) (ablation)Star+ · OWT 512 tokens, 32-512 stepsMAUVE: largest gap at 64-256 stepsImproved
use head-only trained 12-block predictor vs fully fine-tuned (ablation)G-Star error predictor · OWT 128 stepsPareto quality-diversityNo clear effect
shrink predictor to a single fine-tuned block (1B,F) (ablation)G-Star error predictor · OWT 128 stepsPareto quality-diversityHurt
lower error-predictor temperature (ablation)G-Star · OWT 512 tokensperplexity vs diversity: low T: best perplexity, lower diversityMixed
use 128-step G-Star-loop instead of 512-step ReMDMG-Star-loop · OWT 512 tokenslatency (s): 9.16 -> 3.43Improved
add G-Star refinement to Dream-Instruct 7BDream-Instruct 7B · IFEvalscore: 56.4 -> 59.3Improved
add G-Star refinement to Dream-Instruct 7BDream-Instruct 7B · MMLUscore: 69.9 -> 71.2Improved
add G-Star refinement to Dream-Instruct 7BDream-Instruct 7B · GSM8Kscore: 81.5 -> 81.6No clear effect
use G-Star instead of ReMDM / MDLMConala conditional MDLM · Conala 128 stepsQwen2.5-Coder ppl: 19.1 -> 16.4Improved
use 12-block full predictor (ablation)G-Star · OWTinference latency: about 50% increaseHurt

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