AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Insertion Based Sequence Generation with Learnable Order Dynamics

Dhruvesh Patel, Benjamin Rozonoyer, Gaurav Pandey, Tahira Naseem, Ramón Fernandez Astudillo, Andrew McCallum

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
add learnable order dynamics (LoFlexMDM instead of FlexMDM, best decoding)FlexMDM · De novo moleculesquality (%): 62.0 -> 79.5Improved
add learnable order dynamicsFlexMDM · De novo moleculesdiversity: 0.890 -> 0.850Hurt
add learnable order dynamicsFlexMDM · De novo moleculesvalidity (%): 98.9 -> 99.7Improved
add learnable order dynamicsFlexMDM · Fragment-constrained: motif extensionquality (%): 46.9 -> 53.6Improved
add learnable order dynamicsFlexMDM · Fragment-constrained: linker designquality (%): 50.8 -> 51.7Improved
add learnable order dynamicsFlexMDM · Fragment-constrained: scaffold decorationquality (%): 39.0 -> 40.5Improved
add learnable order dynamicsFlexMDM · Fragment-constrained: superstructure generationquality (%): 35.8 -> 37.0Improved
use confidence-based position selection for unmasking (ablation)FlexMDM · De novo moleculesquality (%): 62.0 -> 5.5Hurt
use confidence-based position selection for unmasking (ablation)LoFlexMDM medium aux, p=0.5 · De novo moleculesquality (%): 79.5 -> 69.3Hurt
lower nucleus cutoff p from 1.0 to 0.5 (ablation)LoFlexMDM medium aux, no conf. · De novo moleculesquality (%): 55.1 -> 79.5Improved
increase auxiliary rate network size from xtiny to medium (ablation)LoFlexMDM, no conf., p=0.5 · De novo moleculesquality (%): 79.4 -> 79.5No clear effect
fix unmasking rate b_unmask instead of trainable (ablation)LoFlexMDM · Star graph hardexact match accuracyImproved

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