PreprintCC BY 4.0arXiv
MadNIS, Ready for Production at LO
Abstract
We extend neural multi-channel importance sampling in MadNIS to full proton-proton processes in a production-ready event-generation setup. We benchmark a wide range of LHC processes, including multi-jet, vector-boson, top-quark, and Higgs production, and obtain event-generation speed-ups of up to a factor of 250 without process-specific hyperparameter tuning. We quantify how the speed-up scales with training time and final-state multiplicity, and when the training cost amortizes. MadNIS will become the default for phase-space integration and event generation at leading order in MadGraph.
Related
- Toward FAIR and AI-Ready Data: An Assessment of the GlueX Experiment's Data Ecosystem at Jefferson Lab
- On Revolutionizing Power Systems: Suitability of Quantum Computing-Ready Pipelines for Stochastic Unit Commitment Optimization
- Are Molecular Foundation Models Ready for Free Energy Simulations?
- Validation of a FLASH-ready 25 MeV proton beamline for preclinical radiobiology experiments
- Exascale-ready HemeLB for large-eddy simulation of full-human arterial blood flow
BibTeX
@misc{heimel2026madnis,
title = {{MadNIS, Ready for Production at LO}},
author = {Theo Heimel and Olivier Mattelaer and Daniele Massaro and Ramon Winterhalder},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.17177},
eprint = {2610.17177},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.17177},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - MadNIS, Ready for Production at LO
AU - Theo Heimel
AU - Olivier Mattelaer
AU - Daniele Massaro
AU - Ramon Winterhalder
PY - 2026
DA - 2026/10/12
PB - arXiv
DO - 10.48550/arxiv.2610.17177
UR - https://arxiv.org/abs/2610.17177
AB - We extend neural multi-channel importance sampling in MadNIS to full proton-proton processes in a production-ready event-generation setup. We benchmark a wide range of LHC processes, including multi-jet, vector-boson, top-quark, and Higgs production, and obtain event-generation speed-ups of up to a factor of 250 without process-specific hyperparameter tuning. We quantify how the speed-up scales with training time and final-state multiplicity, and when the training cost amortizes. MadNIS will become the default for phase-space integration and event generation at leading order in MadGraph.
ER -