Simple, but not simplified: A new approach for optimising beyond-Standard Model physics searches at the Large Hadron Collider

Abstract: Searches for beyond-Standard Model physics scenarios, such as supersymmetry (SUSY), at the Large Hadron Collider (LHC) are frequently optimised on simplified models. After assuming particular particle production and decay processes, analyses are optimised by tuning event selections on benchmark models generated in 2D planes of parameters, with all other parameters held fixed. Motivated by recent evidence that this removes sensitivity to a large volume of viable SUSY models, we propose an alternative approach based on dimensional reduction of global fit results. Starting from the results of a global fit of the 4D electroweak minimal supersymmetric standard model performed by the GAMBIT collaboration, we show how to define a 2D plane by using a variational autoencoder to map points in the original 4D parameter space to a 2D latent space. This allows for easy visualisation of the 4D global fit results, which generates insights into what we may be missing at the LHC. Furthermore, the invertible nature of the map to the 2D plane allows experimentalists to choose and simulate benchmark models in a 2D plane (as they could for a simplified model) whilst still accessing the full range of phenomenological scenarios identified by the global fit in the 4D model. We provide a demonstration of how our visualisation and benchmark model simulation process works, and develop an analysis that is able to exclude four benchmark models not excluded by the ATLAS and CMS collaborations in the set of results used for the global fit.