AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Improving Sampling for Masked Diffusion Models via Information Gain

Kaisen Yang, Jayden Teoh, Kaicheng Yang, Yitong Zhang, Alex Lamb

ICML 2026 · Image generation › Fast diffusion sampling · approach: Masked diffusion language model decoding · 0 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
replace greedy entropy selection with Info-Gain SamplerDream-7B-Instruct · Reasoning avg, K=1average accuracy: 54.4 -> 62.0Improved
replace greedy entropy selection with Info-Gain SamplerDream-7B-Instruct · Reasoning avg, K=2average accuracy: 38.6 -> 53.4Improved
add information-gain look-ahead vs PC-SamplerDream-7B-Instruct · Reasoning avg, K=2average accuracy: 49.8 -> 53.4Improved
add information-gain look-aheadSDAR-8B-Chat · Math+code avg, K=1average accuracy: 54.6 -> 66.2Improved
add information-gain look-aheadTraDo-8B-Instruct · Math+code avg, K=1average accuracy: 56.9 -> 68.2Improved
use Info-Gain instead of LookUMDream-7B-Instruct · MATH500, K=2accuracy: 35.2 -> 34.2Hurt
use Info-Gain instead of LookUMDream-7B-Instruct · MATH500, K=1accuracy: 51.4 -> 51.3No clear effect
use Info-Gain instead of Margin samplingMMaDa · GenEval avgscore: 56.3 -> 58.2Improved
use Info-Gain instead of Confidence samplingMMaDa · ImageNet-512FID: 43.3 -> 38.1Improved
use Info-Gain at high token temperature 1.5, K=2SDAR-8B-Chat · AlpacaEvalwin rate vs Entropy: 80.3Improved
let the sampler resolve reasoning before answer (AR order vs greedy entropy)MDM greedy samplers · Binary judgment case studyaccuracy: 67 -> 90Improved
use Info-Gain beam search with larger beam size (ablation)Info-Gain Sampler · MATH-500cumulative uncertaintyNo clear effect

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