arXiv:2606.20437v1 Announce Type: cross Abstract: Charged-particle tracking -- reconstructing trajectories from sparse detector measurements -- is a fundamental high-energy-physics inference problem and a canonical example of learning under extreme combinatorial ambiguity. At the High-Luminosity Large Hadron Collider (HL-LHC), tracking must remain accurate and efficient despite unprecedented collision densities. Graph neural networks perform strongly, but incur substantial costs from graph construction and processing, while transformer-based approaches rely on auxiliary stages that prevent end
The continuous growth in data density from experiments like HL-LHC necessitates more efficient and robust computational methods for particle reconstruction, driving innovation in AI approaches.
This development proposes a new transformer-based AI method that could significantly improve the efficiency and accuracy of charged particle reconstruction, a foundational task in high-energy physics with implications for data processing in other complex scientific fields.
The proposed HEPTv2 moves towards end-to-end transformer-based solutions, potentially reducing the reliance on laborious graph construction and auxiliary stages, streamlining complex data analysis.
- · High-Energy Physics Research
- · AI/ML Research in Scientific Computing
- · Accelerator Laboratories
- · Traditional Graph Neural Network Approaches
- · Auxiliary Stage Development Teams
More accurate and faster analysis of high-luminosity collider data will accelerate discovery in fundamental physics.
The efficiency gains from end-to-end transformer models could be adapted to other scientific domains facing combinatorial complexity.
Reduced computational overhead could allow for more ambitious experimental designs and larger datasets, pushing the boundaries of scientific inquiry.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG