What Nabat built
Seth Jacob Nabat, an 18-year-old senior at William Howard Taft Charter High School in Woodland Hills, California, developed a machine-learning program to interpret high-energy particle collisions. Physicists normally rely on computer models that assume symmetry in results, which reduces computing time and energy while potentially doubling measurement errors. Nabat's project replaces that single assumption with a three-part system designed to avoid the trade-off, according to the Society for Science.
How the three-part model works
The first network is aware of symmetry and uses it to find the quickest approximation of collision results. A second, unconstrained network catches camera and measurement errors, and a third network finds patterns in those errors. In tests, Nabat found that the combined model could navigate imperfect data without losing the efficiency advantage, allowing researchers to examine directly what breaks symmetry, a problem described as fundamental to physics and especially quantum field theory.
Competition and prize
Nabat placed tenth among the top ten winners of the 2026 Regeneron Science Talent Search, earning a $40,000 award. The competition, run by the Society for Science, is the oldest science and mathematics competition in the United States and traces its origins to the 1942 Westinghouse Science Talent Search. Roughly 1,900 high school seniors submit original research each year, and the program culminates in a public exhibition of projects followed by a formal awards ceremony where the top ten are announced.
Background beyond the lab
Nabat is co-captain of his school's varsity debate team and volunteers as an instructor with the UCLA Math Circle, where he teaches elementary and middle school students college-level mathematics. He is the son of Robyn Brook-Nabat and Scott Nabat.
Open questions
The Society for Science account does not specify whether Nabat's model has been tested against established particle-collision datasets outside the competition, and it does not describe the computing resources or training data used. It also does not state whether the project will be released as open-source software or submitted for publication in a peer-reviewed venue.
Follow-up
The 2026 Regeneron Science Talent Search awards ceremony has taken place, and Nabat's next verifiable milestone will be his stated next step of taking his education and particle-collision research further, though the source set does not name a specific institution, fellowship, or publication.
Share this article
