Reconstructing Sparticle masses at the LHC using Generative Machine Learning

Abstract

We explore a generative model framework to infer the masses of heavy particles from detector-level data over a broad parameter space. Our model combines a transformer-based detector encoder and a diffusion neural network. We first apply our model to a new physics scenario involving the pair production of wino-like chargino-neutralino, $pp\to \tilde{\ddot{I}}{‡}_1^{\pm }\tilde{\ddot{I}}{‡}_2^0$, in the $1\ell +2{\hat{I}}^3+jets$ channel at the high luminosity LHC~(HL-LHC). We find that our framework can achieve mass reconstruction efficiency of $\gtrsim 70\%$ for the lightest neutralino $\tilde{\ddot{I}}{‡}_1^0$ and $\gtrsim 40\%$ for the second lightest neutralino $\tilde{\ddot{I}}{‡}_2^0$, for a mass tolerance of $\hat{I}”m=30$GeV, across the entire parameter space accessible at the HL-LHC. We further extend our analysis to a different scenario with $pp\to \tilde{\ddot{I}}{‡}_1^{\pm }\tilde{\ddot{I}}{‡}_1^{\mp }+\tilde{\ddot{I}}{‡}_1^{\pm }\tilde{\ddot{I}}{‡}_2^0$ pair production at the HL-LHC in the $4\ell +\mathrm{E}{/}_{\mathrm{T}}$ channel, and for a fixed value of $m_{\tilde{\ddot{I}}{‡}_2^0}$, we obtain reconstruction efficiencies $\gtrsim 80\%$ over a wide range of $m_{\tilde{\ddot{I}}{‡}_1^0}$ for $\hat{I}”m=30$GeV.

17 pages, 7 figures and 1 table, New results and discussions added