✨ TL;DR
This paper introduces gauge-equivariant graph neural networks that embed local gauge symmetries directly into message passing for learning on lattice gauge theories. The approach enables principled machine learning for systems with site-dependent symmetries and nonlocal observables.
Existing machine-learning approaches lack a general framework for learning under site-dependent local gauge symmetries, which are fundamental to particle physics and strongly correlated quantum systems. Current equivariant neural networks handle global symmetries but struggle with intrinsically nonlocal observables and the complexity of local gauge transformations that vary from site to site.
The authors develop gauge-equivariant graph neural networks that embed non-Abelian symmetry into message passing through matrix-valued, gauge-covariant features and symmetry-compatible update rules. The message passing mechanism implements gauge-covariant transport across the lattice, allowing nonlocal correlations and loop structures to emerge naturally from local operations while respecting the local symmetry constraints.
What the paper shows.
The approach is validated across multiple regimes including pure gauge theories, gauge-matter systems, and dynamical scenarios, demonstrating that gauge-equivariant message passing serves as a general paradigm for learning in systems governed by local symmetry.
The paper does not explicitly discuss computational complexity, scalability to larger lattices, or comparison with traditional numerical methods for lattice gauge theories. The validation scope and specific quantitative performance metrics are not detailed in the abstract, leaving questions about practical applicability and performance advantages over existing approaches.
✨ Generated by Claude · Apr 25, 2026 · Read the PDF for authoritative content.