arXiv subjects

Max Hart

Publications and source records attributed to Max Hart.

4 recordsLinked to original sources

arXiv2026

Better Queries, Cheaper Attention: Adapting Transformers for Efficient Sparse Reconstruction

Query-based transformer decoders are effective for object reconstruction from sparse scientific sensor measurements, but their scalability to high-multiplicity data is limited by fixed, input-independent query sets and costly decoder cross-attention. We introduce a geometry-aware dynamic-query decoder that couples input-conditioned query construction with structured sparse cross-attention. Decoder queries are initialised from selected encoder-level measurement representations that serve as candidate trajectory seeds, making both query content and query multiplicity input-dependent. Local Strided Cross-Attention (LSCA) exploits the induced geometric ordering by replacing learned mask-gated cross-attention with a geometry-defined local support that restricts attention to physically plausible query-hit interactions and exposes sparsity for block-sparse execution. We study this architecture for charged-particle trajectory reconstruction in a simplified High-Luminosity Large Hadron Collider detector, where thousands of trajectories must be reconstructed from tens of thousands of sparse measurements. In the nominal configuration, the dynamic-query (DQ) architecture increases trajectory reconstruction efficiency from 94.1% to 98.1% and reduces the fake rate by more than a factor of two relative to the fixed-query baseline. The DQ+LSCA model reduces end-to-end inference latency by nearly 50% and peak allocated inference memory by more than a factor of 10 relative to the fixed-query baseline.

hep-ex↗

arXiv2025

GLOW: A Unified Particle Flow Transformer

We present GLOW, a transformer-based particle flow model that combines incidence matrix supervision from HGPflow with a MaskFormer architecture. Evaluated on CLIC detector simulations, GLOW achieves state-of-the-art performance and, together with prior work, demonstrates that a single unified transformer architecture can effectively address diverse reconstruction tasks in particle physics.

hep-ex↗

arXiv2024

Transformers for Charged Particle Track Reconstruction in High Energy Physics

Reconstructing charged particle tracks is a fundamental task in modern collider experiments. The unprecedented particle multiplicities expected at the High-Luminosity Large Hadron Collider (HL-LHC) pose significant challenges for track reconstruction, where traditional algorithms become computationally infeasible. To address this challenge, we present a novel learned approach to track reconstruction that adapts recent advances in computer vision and object detection. Our architecture combines a Transformer hit filtering network with a MaskFormer reconstruction model that jointly optimises hit assignments and the estimation of the charged particles' properties. Evaluated on the TrackML dataset, our best performing model achieves state-of-the-art tracking performance with 97% efficiency for a fake rate of 0.6%, and inference times of 100ms. Our tunable approach enables specialisation for specific applications like triggering systems, while its underlying principles can be extended to other reconstruction challenges in high energy physics. This work demonstrates the potential of modern deep learning architectures to address emerging computational challenges in particle physics while maintaining the precision required for groundbreaking physics analysis.

hep-ex↗

arXiv2023

Secondary Vertex Reconstruction with MaskFormers

In high-energy particle collisions, the reconstruction of secondary vertices from heavy-flavour hadron decays is crucial for identifying and studying jets initiated by $b$- or $c$-quarks. Traditional methods, while effective, require extensive manual optimisation and struggle to perform consistently across wide regions of phase space. Meanwhile, recent advancements in machine learning have improved performance but are unable to fully reconstruct multiple vertices. In this work we propose a novel approach to secondary vertex reconstruction based on recent advancements in object detection and computer vision. Our method directly predicts the presence and properties of an arbitrary number of vertices in a single model. This approach overcomes the limitations of existing techniques. Applied to simulated proton-proton collision events, our approach demonstrates significant improvements in vertex finding efficiency, achieving a 10% improvement over an existing state-of-the-art method. Moreover, it enables vertex fitting, providing accurate estimates of key vertex properties such as transverse momentum, radial flight distance, and angular displacement from the jet axis. When integrated into a flavour tagging pipeline, our method yields a 50% improvement in light-jet rejection and a 15% improvement in $c$-jet rejection at a $b$-jet selection efficiency of 70%. These results demonstrate the potential of adapting advanced object detection techniques for particle physics, and pave the way for more powerful and flexible reconstruction tools in high-energy physics experiments.

hep-ex↗