Scaling Collider Event Generation with Residual-Quantized Tokens

Abstract

Full detector simulation and reconstruction of collider events are projected to become major bottlenecks at the High-Luminosity Large Hadron Collider, motivating the development of fast, ML-based surrogates. At the same time, LLMs have driven fast progress in generative discrete modeling: autoregressive transformers trained on tokenized data now represent the state of the art across a range of generative tasks. We extend the discrete modeling paradigm by introducing a particle-level generative model trained on residual-quantized full-event data. We demonstrate the ability of this model family to perform conditional generation from detector-stable particles; we study its scaling behavior across a range of dataset and model sizes, characterize the effects of repeated data exposure and demonstrate that token-level loss systematically predicts downstream physical fidelity. These results provide an empirical framework for scalable collider full-event generation based on residual-quantized representations.

Scaling RIFT 1: Extending RIFT to third-generation gravitational-wave analyses
R. O'Shaughnessy, A. Jan, J. Lange +1
arXiv29 Sep 2026PreprintarXiv licenceFull textInstrumentation & methods Gravitation & general relativityarXiv
Neural Scaling Laws for Jet Generation
Oz Amram, Darius A. Faroughy, Tjarko Gerdes +5
arXiv6 Oct 2026PreprintCC BY 4.0Full textComputational physics High-energy & particle physicsarXiv
Dynamical superfluid and Bose-insulator phases in quantized microcavity lattices
Sanjib Ghosh
Communications Physics24 Oct 2026Peer-reviewedCC BY-NC-ND 4.0Full textAtomic, molecular & optical physicsNatureCrossref
Quantized heat flow in the Hofstadter butterfly
Aifei Zhang, Gibril Aissani, Quan Dong +11
Physical Review Letters29 Sep 2026Peer-reviewedLicence not statedFull textAtomic, molecular & optical physicsHAL
A collider as a quantum computer
Wei Xie, Ji-Chong Yang
Physics Letters B30 Oct 2026Peer-reviewedCC BY 4.0High-energy & particle physicsCrossref

Cite

BibTeX RIS

BibTeX

@misc{godi2026scaling,
  title = {{Scaling Collider Event Generation with Residual-Quantized Tokens}},
  author = {Dan Godi and Dmitrii Kobylianskii and Eilam Gross},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.00569},
  eprint = {2610.00569},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.00569},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Scaling Collider Event Generation with Residual-Quantized Tokens
AU  - Dan Godi
AU  - Dmitrii Kobylianskii
AU  - Eilam Gross
PY  - 2026
DA  - 2026/10/01
PB  - arXiv
DO  - 10.48550/arxiv.2610.00569
UR  - https://arxiv.org/abs/2610.00569
AB  - Full detector simulation and reconstruction of collider events are projected to become major bottlenecks at the High-Luminosity Large Hadron Collider, motivating the development of fast, ML-based surrogates. At the same time, LLMs have driven fast progress in generative discrete modeling: autoregressive transformers trained on tokenized data now represent the state of the art across a range of generative tasks. We extend the discrete modeling paradigm by introducing a particle-level generative model trained on residual-quantized full-event data. We demonstrate the ability of this model family to perform conditional generation from detector-stable particles; we study its scaling behavior across a range of dataset and model sizes, characterize the effects of repeated data exposure and demonstrate that token-level loss systematically predicts downstream physical fidelity. These results provide an empirical framework for scalable collider full-event generation based on residual-quantized representations.
ER  -

Abstract

Related

Cite