Speaker
Description
Identified light-hadron production in Z→qq̄ events — K_S^0, Λ, φ, K⁰ and ρ⁰ — is sensitive to quark fragmentation, strangeness suppression and vector-meson production, and its precision at CEPC is limited by the reconstruction of the corresponding two-body decays from charged tracks. This contribution presents a deep-learning implementation of the "leaves-to-tree" step at CEPC: neutral hadrons are reconstructed directly from charged tracks using ParT. Focusing on the benchmark channels K_S^0→π⁺π⁻ and Λ→pπ⁻, the reconstruction efficiency, the purity and their product are compared with the conventional cut-based reconstruction, both integrated and as a function of the reconstructed energy. Extensions to the prompt resonances φ, K⁰ and ρ⁰ and to multi-body decays, and the prospects for precision measurements of identified-hadron production at the Z pole, are also discussed.
