aiqbench analysis
AtomSteer adds inference-time constraint correction to pretrained diffusion and flow-matching models for crystals, molecules, and protein-ligand complexes, without retraining. Its reported gains in space-group coverage and E/Z stereochemistry are benchmark results on three specific backbones, so the practical value depends on how well they transfer to other models and real materials or drug-discovery workflows. Readers should watch whether the constraint projections hold up beyond these test cases.
AtomSteer: Test-Time Constraint Steering for Atomistic Generative Models
arXiv:2610.38517v1 Announce Type: cross Abstract: Diffusion and flow matching models offer scalable approaches to generating crystalline materials, molecular conformers, and protein-ligand complexes. However, satisfying crystallographic symmetry, structural geometry, and accurate stereochemistry remains challenging. Satisfying these requirements calls for corrections that preserve the geometry already captured by the pretrained model. In this work, we propose AtomSteer, a test-time inference framework for steering pretrained atomistic generative models with structural constraints. During sampling, AtomSteer corrects intermediate structure predictions with bounded local projections, graph-rigid stereochemical transformations, and projections over periodic symmetries, accepting updates that improve constraint satisfaction while limiting changes to the rest of the structure. We demonstrate AtomSteer across three pretrained backbones, FlowMM, ET-Flow, and Boltz-2, spanning de novo crystal generation and crystal structure prediction, molecular conformer generation, and protein-ligand complex structure prediction. AtomSteer improves space-group distribution coverage in de novo crystal generation and improves target space-group recovery in crystal structure prediction. AtomSteer improves molecular validity and E/Z stereochemical accuracy while retaining conformer ensemble fidelity. For protein-ligand complexes, AtomSteer reduces ligand E/Z errors relative to experimental reference structures while achieving competitive docking success and PoseBusters validity. Our method provides a unified framework for adding chemical and crystallographic control to existing atomistic generative models without retraining or architectural changes.
Abstract:Diffusion and flow matching models offer scalable approaches to generating crystalline materials, molecular conformers, and protein-ligand complexes. However, satisfying crystallographic symmetry, structural geometry, and accurate stereochemistry remains challenging. Satisfying these requirements calls for corrections that preserve the geometry already captured by the pretrained model. In this work, we propose AtomSteer, a test-time inference framework for steering pretrained atomistic generative models with structural constraints. During sampling, AtomSteer corrects intermediate structure predictions with bounded local projections, graph-rigid stereochemical transformations, and projections over periodic symmetries, accepting updates that improve constraint satisfaction while limiting changes to the rest of the structure. We demonstrate AtomSteer across three pretrained backbones, FlowMM, ET-Flow, and Boltz-2, spanning de novo crystal generation and crystal structure prediction, molecular conformer generation, and protein-ligand complex structure prediction. AtomSteer improves space-group distribution coverage in de novo crystal generation and improves target space-group recovery in crystal structure prediction. AtomSteer improves molecular validity and E/Z stereochemical accuracy while retaining conformer ensemble fidelity. For protein-ligand complexes, AtomSteer reduces ligand E/Z errors relative to experimental reference structures while achieving competitive docking success and PoseBusters validity. Our method provides a unified framework for adding chemical and crystallographic control to existing atomistic generative models without retraining or architectural changes.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2610.38517 [cs.LG]
(or arXiv:2610.38517v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.38517 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 295 words
Related coverage
Metamodel-Guided Model Generation with Layered Constraints
arXiv · Sep 24, 2026
Learning Local Constraints for Reinforcement-Learned Content Generators
arXiv · Oct 2, 2026
CIT-CAD: Constraint Intent Tree-based CAD Code Generation and Verification
arXiv · Sep 9, 2026
A Constraint-Aware Generative Framework for Synthetic Origin-Destination Demand in Logistics Networks
arXiv · Sep 7, 2026