Speaker
Description
The Higgs self-coupling constant is a key probe of the shape of the Higgs potential and a sensitive window into physics beyond the Standard Model, and can ultimately be constrained by measuring di-Higgs (HH) production. This work targets the HH→bbWW decay channel in CMS Run 3 data, where one Higgs decays to two b quarks and the other decays to two W bosons, each of which decays hadronically to a pair of quarks — a fully hadronic, six-jet final state. This channel benefits from the largest branching ratio among the bbWW final states and full kinematic reconstructibility since no neutrinos escape detection. However, it also introduces a severe combinatorial jet-assignment problem and a large QCD multijet background. We address this assignment problem using SPANet (Symmetry Preserving Attention Network), a neural network designed to assign final-state jets to their parent particles. Preliminary results show SPANet reconstructing the on-shell W boson mass more accurately than a baseline method that selects the two highest ParticleNet quark-vs-gluon discriminator-scored non-b jets as the W candidate, though further validation is required. This jet-assignment method is a necessary step toward a future measurement of the Higgs self-coupling in this channel.
