Humberto Reyes-Gonzalez

Humberto Reyes-Gonzalez

PhD in Physics. AI for science.

Sitemap

A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.

Pages

Posts

Future Blog Post

less than 1 minute read

Published:

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Blog Post number 4

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 3

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 2

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 1

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

portfolio

publications

Unitarity and Stability Conditions in a 4-Higgs Doublet Model with an S3 Family Symmetry

Published in J. Phys.: Conf. Ser., 2017

In order to obtain a dark matter candidate we propose an extension of the S (3) symmetric 3-Higgs doublet model, adding a new scalar doublet, occupying all the irreducible representations of the discrete symmetry. To ensure stability of the dark matter particle we impose an extra Z (2) symmetry. We find the analytical masses of the scalar particles and constrain their values using stability and unitarity conditions.

Recommended citation: Espinoza, C., Garcés, E. A., Mondragón, M., Reyes-González, H. (2017). 'Unitarity and Stability Conditions in a 4-Higgs Doublet Model with an S3 Family Symmetry.' J. Phys.: Conf. Ser. 912.
Download Paper

LHC Limits on Gluinos and Squarks in the Minimal Dirac Gaugino Model

Published in Journal of High Energy Physics, 2019

At the LHC, gluino-pair production is enhanced while squark production is suppressed as compared to the MSSM, and the decay signatures are altered by a more complex chargino and neutralino spectrum. We investigate how this impacts current gluino and squark mass limits from Run 2 of the LHC.

Recommended citation: Chalons, G., Goodsell, M., Kraml, S., Reyes-González, H., Williamson, S. (2019). 'LHC Limits on Gluinos and Squarks in the Minimal Dirac Gaugino Model.' JHEP 04 (2019) 113.
Download Paper

The S3 Symmetric Model with a Dark Scalar

Published in Physics Letters B, 2019

We study the S 3 symmetric extension of the Standard Model in which all the irreducible representations of the permutation group are occupied by SU(2) scalar doublets, one of which is taken as inert and can lead to dark matter candidates. We perform a scan over parameter space probing points against physical constraints ranging from unitarity tests to experimental Higgs searches limits. We find that the latter constraints severely restrict the parameter space of the model. For acceptable points we compute the value of the relic density of the dark scalar candidates and find that it has a region for low dark matter masses which complies with the Higgs searches bounds and lies within the experimental Planck limit. For masses ≳80 GeV the value of the relic density is below the Planck bound, and it reaches values close to it for very heavy masses ∼5 TeV. In this heavy mass region, this opens the interesting possibility of extending the dark sector of the model with additional particles.

Recommended citation: Espinoza, C., Garcés, E. A., Mondragón, M., Reyes-González, H. (2019). 'The S3 Symmetric Model with a Dark Scalar.' Phys. Lett. B 788, 185–191.
Download Paper

Constraining Electroweakinos in the Minimal Dirac Gaugino Model

Published in SciPost Physics, 2020

Supersymmetric models with Dirac instead of Majorana gaugino masses have distinct phenomenological consequences. In this paper, we investigate the electroweakino sector of the Minimal Dirac Gaugino Supersymmetric Standard Model (MDGSSM) with regards to dark matter (DM) and collider constraints. We delineate the parameter space where the lightest neutralino of the MDGSSM is a viable DM candidate, that makes for at least part of the observed relic abundance while evading constraints from DM direct detection, LEP and lowenergy data, and LHC Higgs measurements. The collider phenomenology of the thus emerging scenarios is characterised by the richer electroweakino spectrum as compared to the Minimal Supersymmetric Standard Model (MSSM) – 6 neutralinos and 3 charginos instead of 4 and 2 in the MSSM, naturally small mass splittings, and the frequent presence of long-lived particles, both charginos and/or neutralinos. Reinterpreting ATLAS and CMS analyses with the help of SModelS and MadAnalysis 5, we discuss the sensitivity of existing LHC searches for new physics to these scenarios and show which cases can be constrained and which escape detection. Finally, we propose a set of benchmark points which can be useful for further studies, designing dedicated experimental analyses and/or investigating the potential of future experiments.

Recommended citation: Goodsell, M., Kraml, S., Reyes-González, H., Williamson, S. (2020). 'Constraining Electroweakinos in the Minimal Dirac Gaugino Model.' SciPost Phys. 9.
Download Paper

SModelS v1.2: Long-Lived Particles, Combination of Signal Regions, and Other Novelties

Published in Computer Physics Communications, 2020

SModelS is an automatized tool enabling the fast interpretation of simplified model results from the LHC within any model of new physics respecting a Z2 symmetry. With the version 1.2 we announce several new features. First, previous versions were restricted to missing energy signatures and assumed prompt decays within each decay chain. SModelS v1.2 considers the lifetime of each Z2 -odd particle and appropriately takes into account missing energy, heavy stable charged particle and R-hadron signatures. Second, the current version allows for a combination of signal regions in efficiency map results whenever a covariance matrix is available from the experiment. This is an important step towards fully exploiting the constraining power of efficiency map results. Several other improvements increase the user-friendliness, such as the use of wildcards in the selection of experimental results, and a faster database which can be given as a URL. Finally, smodelsTools provides an interactive plots maker to conveniently visualize the results of a model scan.

Recommended citation: Ambrogi, F., et al. (2020). 'SModelS v1.2: Long-Lived Particles, Combination of Signal Regions, and Other Novelties.' Comput. Phys. Commun. 251, 106848.
Download Paper

New Signatures of Dirac Neutralino Dark Matter

Published in Rencontres de Moriond, 2021

Supersymmetric dark matter has been studied extensively in the context of the MSSM, where gauginos have Majorana masses. Introducing Dirac gaugino masses, we obtain an enriched phenomenology from which considerable differences in, e.g., LHC signatures can be expected. Concretely, in the Minimal Dirac Gaugino Model (MDGSSM) we have an electroweakino sector extended by two extra neutralinos and one extra chargino. The bino- and wino-like states bring about small mass splittings leading to the frequent presence of scenarios with Long Lived Particles (LLPs). In this contribution, we delineate the parameter space of the electroweakino sector of the MDGSSM, where the lightest neutralino is a viable dark matter candidate that escapes current dark matter direct detection. We then focus on the allowed regions that contain LLPs and confront them against the corresponding LHC searches. Finally, we discuss the predominant case of long-lived neutralinos, to which no search is currently sensitive.

Recommended citation: Goodsell, M., Kraml, S., Reyes-González, H., Williamson, S. (2021). 'New Signatures of Dirac Neutralino Dark Matter.' Moriond EW 2021.
Download Paper

New Developments in SModelS

Published in PoS TOOLS 2020, 2021

SModelS is an automatized tool enabling the fast interpretation of simplified model results from the LHC within any model of new physics respecting a Z_2 symmetry. In this contribution, we report on two important updates of SModelS during 2020: the extension of the SModelS’ database with 13 ATLAS and 10 CMS analyses, including 5 ATLAS and 1 CMS analyses at full Run~2 luminosity, and the ability to use full likelihoods now provided by ATLAS in the form of pyhf JSON files. Moreover, we briefly explain how to use SModelS and give an overview of ongoing developments.

Recommended citation: Alguero, G., et al. (2021). 'New Developments in SModelS.' PoS TOOLS 2020, 022.
Download Paper

CaloMan: Fast Generation of Calorimeter Showers with Density Estimation on Learned Manifolds

Published in NeurIPS ML for Physical Sciences Workshop, 2022

Precision measurements and new physics searches at the Large Hadron Collider require efficient simulations of particle propagation and interactions within the detectors. The most computationally expensive simulations involve calorimeter showers. Advances in deep generative modelling - particularly in the realm of high-dimensional data - have opened the possibility of generating realistic calorimeter showers orders of magnitude more quickly than physics-based simulation. However, the high-dimensional representation of showers belies the relative simplicity and structure of the underlying physical laws. This phenomenon is yet another example of the manifold hypothesis from machine learning, which states that high-dimensional data is supported on low-dimensional manifolds. We thus propose modelling calorimeter showers first by learning their manifold structure, and then estimating the density of data across this manifold. Learning manifold structure reduces the dimensionality of the data, which enables fast training and generation when compared with competing methods.

Recommended citation: Caterini, A. L., et al., Reyes-González, H. (2022). 'CaloMan: Fast Generation of Calorimeter Showers with Density Estimation on Learned Manifolds.' NeurIPS ML4PS.
Download Paper

Constraining New Physics with SModelS Version 2

Published in Journal of High Energy Physics, 2022

We present version 2 of SModelS, a program package for the fast reinterpretation of LHC searches for new physics on the basis of simplified model results. The major novelty of the SModelS v2 series is an extended topology description with a flexible number of particle attributes, such as spin, charge, decay width, etc. This enables, in particular, the treatment of a wide range of signatures with long-lived particles. Moreover, constraints from prompt and long-lived searches can be evaluated simultaneously in the same run. The current database includes results from searches for heavy stable charged particles, disappearing tracks, displaced jets and displaced leptons, in addition to a large number of prompt searches. The capabilities of the program are demonstrated by two physics applications: constraints on long-lived charged scalars in the scotogenic model, and constraints on the electroweak-ino sector in the Minimal Supersymmetric Standard Model.

Recommended citation: Alguero, G., et al., Reyes-González, H. (2022). 'Constraining New Physics with SModelS Version 2.' JHEP 08 (2022) 068.
Download Paper

Testing the Boundaries: Normalizing Flows for Higher Dimensional Data Sets

Published in J. Phys.: Conf. Ser., 2022

Normalizing Flows (NFs) are emerging as a powerful class of generative models, as they not only allow for efficient sampling, but also deliver, by construction, density estimation. They are of great potential usage in High Energy Physics (HEP), where complex high dimensional data and probability distributions are everyday’s meal. However, in order to fully leverage the potential of NFs it is crucial to explore their robustness as data dimensionality increases. Thus, in this contribution, we discuss the performances of some of the most popular types of NFs on the market, on some toy data sets with increasing number of dimensions.

Recommended citation: Reyes-González, H., Torre, R. (2022). 'Testing the Boundaries: Normalizing Flows for Higher Dimensional Data Sets.' J. Phys.: Conf. Ser. 2438.
Download Paper

Strength in Numbers: Optimal and Scalable Combination of LHC New-Physics Searches

Published in SciPost Physics, 2023

Normalizing flows have emerged as a powerful brand of generative models, as they not only allow for efficient sampling of complicated target distributions but also deliver density estimation by construction. We propose here an in-depth comparison of coupling and autoregressive flows, both based on symmetric (affine) and non-symmetric (rational quadratic spline) bijectors, considering four different architectures: real-valued non-Volume preserving (RealNVP), masked autoregressive flow (MAF), coupling rational quadratic spline (C-RQS), and autoregressive rational quadratic spline (A-RQS). We focus on a set of multimodal target distributions of increasing dimensionality ranging from 4 to 400. The performances were compared by means of different test statistics for two-sample tests, built from known distance measures: the sliced Wasserstein distance, the dimension-averaged one-dimensional Kolmogorov–Smirnov test, and the Frobenius norm of the difference between correlation matrices. Furthermore, we included estimations of the variance of both the metrics and the trained models. Our results indicate that the A-RQS algorithm stands out both in terms of accuracy and training speed. Nonetheless, all the algorithms are generally able, without too much fine-tuning, to learn complicated distributions with limited training data and in a reasonable time of the order of hours on a Tesla A40 GPU. The only exception is the C-RQS, which takes significantly longer to train, does not always provide good accuracy, and becomes unstable for large dimensionalities. All algorithms were implemented using TensorFlow2 and TensorFlow Probability and have been made available on GitHub.

Recommended citation: Araz, J., et al., Reyes-González, H. (2023). 'Strength in Numbers: Optimal and Scalable Combination of LHC New-Physics Searches.' SciPost Phys. 14, 077.
Download Paper

CaloChallenge 2022: A Community Challenge for Fast Calorimeter Simulation

Published in Reports on Progress in Physics, 2024

We present the results of the ‘Fast Calorimeter Simulation Challenge 2022’—the CaloChallenge. We study state-of-the-art generative models on four calorimeter shower datasets of increasing dimensionality, ranging from a few hundred voxels to a few tens of thousand voxels. The 31 individual submissions span a wide range of current popular generative architectures, including variational autoencoders (VAEs), generative adversarial networks (GANs), normalizing flows, diffusion models, and models based on conditional flow matching. We compare all submissions in terms of quality of generated calorimeter showers, as well as shower generation time and model size. To assess the quality we use a broad range of different metrics including differences in one-dimensional histograms of observables, KPD/FPD scores, AUCs of binary classifiers, and the log-posterior of a multiclass classifier. The results of the CaloChallenge provide the most complete and comprehensive survey of cutting-edge approaches to calorimeter fast simulation to date. In addition, our work provides a uniquely detailed perspective on the important problem of how to evaluate generative models. As such, the results presented here should be applicable for other domains that use generative AI and require fast and faithful generation of samples in a large phase space.

Recommended citation: Krause, C., Reyes-González, H., et al. (2024). 'CaloChallenge 2022: A Community Challenge for Fast Calorimeter Simulation.' Rep. Prog. Phys.
Download Paper

Comparative Study of Coupling and Autoregressive Flows through Robust Statistical Tests

Published in Symmetry, 2024

Normalizing flows have emerged as a powerful brand of generative models, as they not only allow for efficient sampling of complicated target distributions but also deliver density estimation by construction. We propose here an in-depth comparison of coupling and autoregressive flows, both based on symmetric (affine) and non-symmetric (rational quadratic spline) bijectors, considering four different architectures: real-valued non-Volume preserving (RealNVP), masked autoregressive flow (MAF), coupling rational quadratic spline (C-RQS), and autoregressive rational quadratic spline (A-RQS). We focus on a set of multimodal target distributions of increasing dimensionality ranging from 4 to 400. The performances were compared by means of different test statistics for two-sample tests, built from known distance measures: the sliced Wasserstein distance, the dimension-averaged one-dimensional Kolmogorov–Smirnov test, and the Frobenius norm of the difference between correlation matrices. Furthermore, we included estimations of the variance of both the metrics and the trained models. Our results indicate that the A-RQS algorithm stands out both in terms of accuracy and training speed. Nonetheless, all the algorithms are generally able, without too much fine-tuning, to learn complicated distributions with limited training data and in a reasonable time of the order of hours on a Tesla A40 GPU. The only exception is the C-RQS, which takes significantly longer to train, does not always provide good accuracy, and becomes unstable for large dimensionalities. All algorithms were implemented using TensorFlow2 and TensorFlow Probability and have been made available on GitHub.

Recommended citation: Coccaro, A., Letizia, M., Reyes-González, H., Torre, R. (2024). 'Comparative Study of Coupling and Autoregressive Flows through Robust Statistical Tests.' Symmetry 16, 942.
Download Paper

The Fundamental Limit of Jet Tagging

Published in Physical Review D, 2024

Identifying the origin of high-energy hadronic jets (jet tagging) has been a critical benchmark problem for machine learning in particle physics. Jets are ubiquitous at colliders and are complex objects that serve as prototypical examples of collections of particles to be categorized. Over the last decade, machine learning-based classifiers have replaced classical observables as the state of the art in jet tagging. Increasingly complex machine learning models are leading to increasingly more effective tagger performance. Our goal is to address the question of convergence—are we getting close to the fundamental limit on jet tagging or is there still potential for computational, statistical, and physical insights for further improvements? We address this question using state-of-the-art generative models to create a realistic, synthetic dataset with a known jet tagging optimum. Various state-of-the-art taggers are deployed on this dataset, showing that there is a significant gap between their performance and the optimum. Our dataset and software are made public to provide a benchmark task for future developments in jet tagging and other areas of particle physics.

Recommended citation: Geuskens, J., et al., Reyes-González, H. (2024). 'The Fundamental Limit of Jet Tagging.' Phys. Rev. D 112, L091901.
Download Paper

Les Houches Guide to Reusable ML Models in LHC Analyses

Published in SciPost Physics Communication Reports, 2024

With the increasing usage of machine-learning in high-energy physics analyses, the publication of the trained models in a reusable form has become a crucial question for analysis preservation and reuse. The complexity of these models creates practical issues for both reporting them accurately and for ensuring the stability of their behaviours in different environments and over extended timescales. In this note we discuss the current state of affairs, highlighting specific practical issues and focusing on the most promising technical and strategic approaches to ensure trustworthy analysis-preservation. This material originated from discussions in the LHC Reinterpretation Forum and the 2023 PhysTeV workshop at Les Houches.

Recommended citation: Araz, J., et al., Reyes-González, H. (2024). 'Les Houches Guide to Reusable ML Models in LHC Analyses.' SciPost Phys. Comm. Rep. 3.
Download Paper

The NFLikelihood: an unsupervised DNNLikelihood from Normalizing Flows

Published in SciPost Physics Core, 2024

We propose the NFLikelihood, an unsupervised version, based on Normalizing Flows, of the DNNLikelihood proposed in [Eur. Phys. J. C 80, 664 (2020)]. We show, through realistic examples, how Autoregressive Flows, based on affine and rational quadratic spline bijectors, are able to learn complicated high-dimensional Likelihoods arising in High Energy Physics (HEP) analyses. We focus on a toy LHC analysis example already considered in the literature and on two Effective Field Theory fits of flavor and electroweak observables, whose samples have been obtained through the HEPFit code. We discuss advantages and disadvantages of the unsupervised approach with respect to the supervised one and discuss a possible interplay between the two.

Recommended citation: Reyes-González, H., Torre, R. (2024). 'The NFLikelihood: an unsupervised DNNLikelihood from Normalizing Flows.' SciPost Phys. Core 7, 048.
Download Paper

Aspen Open Jets: Unlocking LHC Data for Foundation Models in Particle Physics

Published in Machine Learning: Science and Technology, 2025

Foundation models are deep learning models pre-trained on large amounts of data which are capable of generalizing to multiple datasets and/or downstream tasks. This work demonstrates how data collected by the CMS experiment at the Large Hadron Collider can be useful in pre-training foundation models for HEP. Specifically, we introduce the AspenOpenJets (AOJs) dataset, consisting of approximately 178 M high pt jets derived from CMS 2016 Open Data. We show how pre-training the OmniJet-α foundation model on AOJs improves performance on generative tasks with significant domain shift: generating boosted top and QCD jets from the simulated JetClass dataset. In addition to demonstrating the power of pre-training of a jet-based foundation model on actual proton–proton collision data, we provide the ML-ready derived AOJs dataset for further public use.

Recommended citation: Amram, O., et al., Reyes-González, H. (2025). 'Aspen Open Jets: Unlocking LHC Data for Foundation Models in Particle Physics.' Mach. Learn.: Sci. Technol. 6, 030601.
Download Paper

Learning to Validate Generative Models: a Goodness-of-Fit Approach

Published in arXiv preprint, 2025

We introduce a goodness-of-fit framework to rigorously evaluate generative models in high-dimensional physics applications.

Recommended citation: Cappelli, P., Grosso, G., Letizia, M., Reyes-González, H., Zanetti, M. (2025). "Learning to Validate Generative Models: a Goodness-of-Fit Approach." arXiv:2511.09118.
Download Paper

On the Codesign of Scientific Experiments and Industrial Systems.

Published in arXiv, 2026

The optimization of large experiments in fundamental science, such as detectors for subnuclear physics at particle colliders, shares with the optimization of complex systems for industrial or societal applications the common issue of addressing the inter-relation between parameters describing the hardware used in data production and parameters used to analyse those data. While in many cases this coupling can be ignored – when the problem can be successfully factored into simpler sub-tasks and the latter addressed serially – there are situations in which that approach fails to converge to the absolute maximum of expected performance, as it results in a mis-alignment of the optimized hardware and software solutions. In this work we consider a few use cases of interest in fundamental science collected primarily from particle physics and related areas, and a pot-pourri of industrial and societal applications where the matter is similarly of relevance. We discuss the emergence of strong hardware-software coupling in some of those systems, as well as co-design procedures that may be deployed to identify the global maximum of their relevant utility functions. We observe how numerous opportunities exist to advance methods and tools for hardware-software co-design optimization, bridging fundamental science and industry through application- and challenge-driven projects, and shaping the future of scientific experiments and industrial systems.

Recommended citation: Dorigo, T., Reyes-Gonzalez, H., et al. (2026). 'On the Codesign of Scientific Experiments and Industrial Systems.' arXiv:2603.26613.
Download Paper

supervision

talks

teaching