CrystalAR: Autoregressive Modeling for Symmetric Crystal Generation
What it is
CrystalAR is a unified autoregressive framework for generating crystalline structures that explicitly incorporates space-group symmetry. It reformulates continuous coordinate generation as a hierarchy of discrete spatial localization decisions, allowing it to jointly model Wyckoff symmetry templates and complete 3D structures within a single sequence.
The problem it solves
Existing methods face a fundamental tension between flexible exploration of discrete crystallographic templates and faithful realization of these templates as continuous 3D structures, while also suffering from increased combinatorial complexity in stoichiometric relationships.
What you could use it for
- Generating novel crystalline materials with specific symmetry constraints
- Template-free crystal structure prediction (CSP)
- Exploring novel symmetry templates in materials discovery
Why it matters
It provides a promising alternative paradigm for symmetry-preserving crystal generation that retains strong template generation capabilities while achieving competitive de novo generation and CSP accuracy without relying on retrieved prior data.
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.