Speaker
Description
Future Higgs factories provide a clean experimental environment and large Higgs samples, offering a unique opportunity for both precision Higgs measurements and direct searches for additional scalar states. In this talk, I will present a particle-level analysis strategy for Higgs physics based on the More-Interaction Particle Transformer (MIParT), which aims to maximize the use of reconstructed final-state information. I will first discuss its application to hadronic Higgs measurements in $e^+e^- \to ZH$ at $\sqrt{s}=240~\mathrm{GeV}$, where improved jet-flavor discrimination leads to projected precisions of $0.18\%$, $1.07\%$, and $0.52\%$ for $H\to b\bar b$, $H\to c\bar c$, and $H\to gg$, respectively, in the $Z\to \nu\bar{\nu}$ channel, together with sensitivity to $H\to s\bar s$. I will then present its application to the search for a light scalar around $95~\mathrm{GeV}$ in $e^+e^- \to ZS$, focusing on the $S\to \tau^+\tau^-$ and $S\to b\bar b$ final states. Relative to conventional cut-based analyses, the particle-level approach improves the expected statistical precision by factors of about $2.4$ and $1.4$ in the two channels, respectively. These results show that particle-level deep learning can strengthen both the precision and discovery frontiers of Higgs physics.
