Mixture Density Networks for Neutrino Reconstruction at Hadron Colliders
Abstract
Neutrino momentum reconstruction at hadron colliders is intrinsically ambiguous because the longitudinal momentum is not directly observed. We study this problem in semileptonic tt̅ events using \monster{} (Mixture of Neutrino Solutions with Transformer Event Representation), a mixture density network that predicts a multivariate normal mixture for the conditional distribution of neutrino momentum from reconstructed event objects. The model uses a Transformer-based event encoder and yields a sampling-free point estimate of the neutrino momentum from the closed-form density. On the public benchmark introduced with \nuflows, \monster{} reduces the 68th percentile of the three-momentum residual by 5\% and the 95th percentile by 6\% relative to the empirical-mode \nuflows{} baseline, with per-variable root-mean-square errors smaller by 2 to 9\% at comparable bias, while running 2 to 8.5 times faster at inference, depending on the device and batch size. These results indicate that mixture density networks are a competitive alternative to normalizing flows for neutrino reconstruction.
Versions
Related
- Probing the Tau Neutrino Magnetic Moment via the μ−γ→ μ−νν̅ Process at Future Muon Colliders
- Axion-Like Particle Searches at Muon Colliders: Photon-Neutrino Signatures and Heavy-Neutral-Lepton Feasibility
- Thermodynamic Signatures of Phase Separation in Mass Imbalanced Fermi Mixtures: Superfluid Density of States and Quasiparticle Specific Heat in the 163Dy 40K Atomic Mixture
- Probing Neutrino-Energy Reconstruction with New Superscaling-Based Observables
- Earth-density effects in long-baseline neutrino experiments in a four-flavor (3+1) sterile-neutrino framework
BibTeX
@misc{yang2026mixture,
title = {{Mixture Density Networks for Neutrino Reconstruction at Hadron Colliders}},
author = {Seungjin Yang and Jason S. H. Lee and Junghwan Goh},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.22979},
eprint = {2609.22979},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.22979},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Mixture Density Networks for Neutrino Reconstruction at Hadron Colliders
AU - Seungjin Yang
AU - Jason S. H. Lee
AU - Junghwan Goh
PY - 2026
DA - 2026/09/19
PB - arXiv
DO - 10.48550/arxiv.2609.22979
UR - https://arxiv.org/abs/2609.22979
AB - Neutrino momentum reconstruction at hadron colliders is intrinsically ambiguous because the longitudinal momentum is not directly observed. We study this problem in semileptonic tt̅ events using \monster{} (Mixture of Neutrino Solutions with Transformer Event Representation), a mixture density network that predicts a multivariate normal mixture for the conditional distribution of neutrino momentum from reconstructed event objects. The model uses a Transformer-based event encoder and yields a sampling-free point estimate of the neutrino momentum from the closed-form density. On the public benchmark introduced with \nuflows, \monster{} reduces the 68th percentile of the three-momentum residual by 5\% and the 95th percentile by 6\% relative to the empirical-mode \nuflows{} baseline, with per-variable root-mean-square errors smaller by 2 to 9\% at comparable bias, while running 2 to 8.5 times faster at inference, depending on the device and batch size. These results indicate that mixture density networks are a competitive alternative to normalizing flows for neutrino reconstruction.
ER -