HEP-PHLGHEP-EXNov 1, 2024

A Lorentz-Equivariant Transformer for All of the LHC

arXiv:2411.00446v364 citationsh-index: 18SciPost Physics
Originality Highly original
AI Analysis

This addresses the need for efficient and accurate models in high-energy physics, offering a novel architecture that is not incremental but provides broad improvements for LHC applications.

The paper tackled machine learning tasks at the Large Hadron Collider by introducing L-GATr, a Lorentz-equivariant transformer, and achieved state-of-the-art performance with significant improvements in amplitude regression, jet classification, and generative modeling.

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.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes