Combinatorial Residual Network: a decay-chain-aware classifier without explicit jet assignment
Abstract
Final states with several jets from the decays of heavy particles pose a combinatorial problem, since the jets must be assigned, at least implicitly, to their parent particles. We describe the Combinatorial Residual Network (CoReNet), a classifier that encodes the assumed decay chain of the final state in its architecture. CoReNet builds its event representation level by level along the decay hierarchy, evaluates every jet assignment hypothesis with weights shared across objects, and combines the hypotheses with learned weights using residual networks. The exchange symmetries of the decay are built into the construction, the size of the network does not depend on the number of hypotheses, and the network is trained with the classification loss alone, without generator-level jet assignments. The architecture was developed for the bb̅bb̅ final state and has been used in several CMS searches. On fast-simulation samples of Higgs boson pair production in the bb̅bb̅ final state against QCD multijet production, CoReNet with 1508 trainable parameters performs well above kinematic baselines and trains in minutes on a single GPU. Its hypothesis weights select the truth-matched jet pairing in most of the signal events in which it is defined, although no assignment target is used in training. Since only the encoded decay chain is specific to the final state, the construction carries over to other topologies, which we test with HH → bb̅W+W- against semileptonic tt̅. There, encoding the decay chain of the background as well as that of the signal further improves the classification, at a cost of about 103 parameters. In both final states its performance is comparable to that of a state-of-the-art transformer-based network more than two orders of magnitude larger. The implementation is publicly available at https://github.com/cms-cmu/CoReNet.
Related
- Linear Analytical Model for Static Traffic Assignment: Explicit Formulation of OD Travel Times and Link Flows
- Automated AFGL quantum number assignment for CO2 isotopologues using a graph neural network
- Magnonic Combinatorial Memory based on a network of coupled active ring circuits
- A Magnetism-Aware Crystal Graph Neural Network with Explicit Magnetic-Sublattice and Exchange-Geometry Representations
- Coordinate-Residual Physics-Driven Neural Network for Inverse Scattering Imaging
BibTeX
@misc{alison2026combinatorial,
title = {{Combinatorial Residual Network: a decay-chain-aware classifier without explicit jet assignment}},
author = {John Alison and Patrick Bryant and Alejandro Gomez Espinosa and Christian Guinto-Brody and Aniket Khanal and Chuyuan Liu and Sindhu Murthy and Guanglei Zhao},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.40749},
eprint = {2610.40749},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.40749},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Combinatorial Residual Network: a decay-chain-aware classifier without explicit jet assignment
AU - John Alison
AU - Patrick Bryant
AU - Alejandro Gomez Espinosa
AU - Christian Guinto-Brody
AU - Aniket Khanal
AU - Chuyuan Liu
AU - Sindhu Murthy
AU - Guanglei Zhao
PY - 2026
DA - 2026/10/29
PB - arXiv
DO - 10.48550/arxiv.2610.40749
UR - https://arxiv.org/abs/2610.40749
AB - Final states with several jets from the decays of heavy particles pose a combinatorial problem, since the jets must be assigned, at least implicitly, to their parent particles. We describe the Combinatorial Residual Network (CoReNet), a classifier that encodes the assumed decay chain of the final state in its architecture. CoReNet builds its event representation level by level along the decay hierarchy, evaluates every jet assignment hypothesis with weights shared across objects, and combines the hypotheses with learned weights using residual networks. The exchange symmetries of the decay are built into the construction, the size of the network does not depend on the number of hypotheses, and the network is trained with the classification loss alone, without generator-level jet assignments. The architecture was developed for the bb̅bb̅ final state and has been used in several CMS searches. On fast-simulation samples of Higgs boson pair production in the bb̅bb̅ final state against QCD multijet production, CoReNet with 1508 trainable parameters performs well above kinematic baselines and trains in minutes on a single GPU. Its hypothesis weights select the truth-matched jet pairing in most of the signal events in which it is defined, although no assignment target is used in training. Since only the encoded decay chain is specific to the final state, the construction carries over to other topologies, which we test with HH → bb̅W^+W^- against semileptonic tt̅. There, encoding the decay chain of the background as well as that of the signal further improves the classification, at a cost of about 10^3 parameters. In both final states its performance is comparable to that of a state-of-the-art transformer-based network more than two orders of magnitude larger. The implementation is publicly available at https://github.com/cms-cmu/CoReNet.
ER -