Building Models and Model Building: Machine Learning Uncertainties for Collider Physics and Dynamical Solutions to the Hierarchy Problem
Name
benevedes-seanmb-phd-physics-2026-thesis.pdf
Size
5 MB
Format
Adobe PDF
Checksum (MD5)
3ac20077c717db83b00d2b316d68f43a
Author(s)
Benevedes, Sean
Advisor(s)
Thaler, Jesse
Date Issued
May 2026
Publisher
Massachusetts Institute of Technology
Abstract
This thesis pursues the search for fundamental physics along two complementary directions: by proposing new physics beyond the Standard Model, and by developing statistical tools to test such theories more precisely at colliders.
On the model building side, we study new approaches to the hierarchy problem based on metastability of the Higgs vacuum. We first develop a mechanism in which the observed electroweak scale is linked to the requirement that the Higgs remains in a metastable vacuum. We then extend this idea by recasting the assumption of metastability as a statement about the allowed energies of Higgs vacua in the far ultraviolet, motivated by conjectures in quantum gravity. Each of these mechanisms can be readily tested at future colliders, giving us the opportunity to discover the presence of new principles underlying the Standard Model.
On the methodological side, we develop machine-learning-based statistical tools for simulation-based inference in high energy physics. We introduce wifi ensembles, a framework for neural density ratio estimation with frequentist uncertainty quantification, and show how these uncertainties can be propagated to downstream inference. We also develop a method for inference in the presence of domain shift, in which parametrized combinations of individually misspecified simulations are used to enable more trustworthy extraction of signal fractions. We show that each of these frameworks yields well-calibrated confidence intervals for parameters of interest in their domains of applicability.
Taken together, these works suggest several promising directions for future progress, both in the search for new principles beyond the Standard Model and in the development of the tools needed to test them.
MIT Department
Massachusetts Institute of Technology. Department of Physics
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
