Graph Neural Network Flavor Tagger and measurement of $\mathrm{s}\mathrm{i}\mathrm{n}2\beta$ at Belle II

We present GFlaT, a new algorithm that uses a graph-neural-network to determine the flavor of neutral B mesons produced in $\mathrm{{\rm Y}}(4\mathrm{S})$ decays. We evaluate its performance using $B$ decays to flavor-specific hadronic final states reconstructed in a $362$ ${\mathrm{f}\mathrm{b}}^{-1}$ sample of electron-positron collisions recorded at the $\mathrm{{\rm Y}}(4\mathrm{S})$ resonance with the Belle II detector at the SuperKEKB collider. We achieve an effective tagging efficiency of $(37.40\pm 0.43\pm 0.36)\mathrm{\%}$, where the first uncertainty is statistical and the second systematic, which is $18\mathrm{\%}$ better than the previous Belle II algorithm. Demonstrating the algorithm, we use $B^0\to J\mathrm{/}\psi K_{\mathrm{S}}^0$ decays to measure the direct and mixing-induced CP violation parameters, $C=(-0.035\pm 0.026\pm 0.013)$ and $S=(0.724\pm 0.035\pm 0.014)$, from which we obtain $\beta =(23.2\pm 1.5\pm 0.6{)}^{\circ }$.