A Lorentz-equivariant transformer for all of the LHC
We show that the Lorentz-Equivariant Geometric Algebra Transformer (L-GATr) yields state-of-the-art performance for a wide range of machine learning tasks at the Large Hadron Collider. L-GATr represents data in a geometric algebra over space-time and is equivariant under Lorentz transformations. The underlying architecture is a versatile and scalable transformer, which is able to break symmetries if needed. We demonstrate the power of L-GATr for amplitude regression and jet classification, and then benchmark it as the first Lorentz-equivariant generative network. For all three LHC tasks, we find significant improvements over previous architectures.
Authors
- Pim de Haan (ORCID: https://orcid.org/0000-0002-3481-8594)
- Johann Brehmer (ORCID: https://orcid.org/0000-0003-3344-4209)
- Tilman Plehn (ORCID: https://orcid.org/0000-0001-5660-7790)
- Jonas Spinner (ORCID: https://orcid.org/0000-0003-4046-948X)
- Jesse Thaler (ORCID: https://orcid.org/0000-0002-2406-8160)
- Víctor Bresó
- H. Qu (ORCID: https://orcid.org/0000-0002-0250-8655)
Institutions
- York College, City University of New York (US)
- Heidelberg University (DE)
- The NSF AI Institute for Artificial Intelligence and Fundamental Interactions (US)
- Massachusetts Institute of Technology (US)
- European Organization for Nuclear Research (CH)
Publication Details
- Journal
- SciPost Physics
- Published
- 2025-10-23
- DOI
- https://doi.org/10.21468/scipostphys.19.4.108
- Citations
- 12
- Primary Topic
- Particle physics theoretical and experimental studies
- Type
- article
- Field-Weighted Citation Impact
- 22.44
Funders
- National Science Foundation
- U.S. Department of Energy
- Carl-Zeiss-Stiftung
- Deutsche Forschungsgemeinschaft
- Bundesministerium für Bildung und Forschung
- Generalitat Valenciana
- High Energy Physics