Can neural networks spot four top quarks and reveal new physics?

Researchers used a specialized neural network called a hyper-graph neural network to identify four-top-quark events among the noise of similar background processes at the LHC. The network learned patterns in how particles cluster and correlate, achieving a detection significance of 9.11 sigma with 140 inverse femtobarns—beating existing methods by ~10%. This improved sensitivity translates to tighter limits on five hypothetical new physics operators that could affect top quark interactions.