AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Incomplete Data, Complete Dynamics: A Diffusion Approach

Zihan Zhou, Chenguang Wang, Hongyi Ye, Yongtao Guan, Tianshu Yu

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Diffusion theory and design space

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
Use block-wise context-query partitioning matching block masks instead of pixel-level (ablation)context-query diffusion · Navier-Stokes 8/9 blockserror: 1.4925 -> 0.7592Improved
Use block-wise instead of pixel-level partitioning (ablation)context-query diffusion · Advection 5/9 blockserror: 0.1894 -> 0.1407Improved
Use block-wise instead of pixel-level partitioning (ablation)context-query diffusion · Shallow Water 5/9 blockserror: 0.0989 -> 0.0865Improved
Train on context-query split of incomplete data vs AmbientDiffdiffusion imputation · Navier-Stokes 20% observedMSE x1e-3: 2.043 -> 1.931Improved
Use ours vs AmbientDiff at extreme sparsitydiffusion imputation · ERA5 1% observedMSE x1e-2: 1.234 -> 1.229No clear effect
Switch MissDiff to data matchingMissDiff baseline · physical dynamics imputationMSEImproved
Train on fewer observed points (20% instead of 80%)Ours · Navier-Stokes test at 80%MSE x1e-3: 0.2229 -> 0.3363Hurt
Use moderate context ratio (ablation)Ours · physical dynamics imputationreconstruction errorImproved

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