AspGap: Augmented Stellar Parameters and Abundances for 24 million RGB stars from Gaia XP

AspGap is a novel data-driven model to perform regressions with the XP spectra, augmented by transfer learning from high-resolution APOGEE spectra. Using a training set of stellar labels from APOGEE, we demonstrate that AspGap can deliver effective temperature, surface gravity, metallicity, and alpha abundance to accuracies of ~ 1%, 0.15 dex, 0.07 dex, and 0.024 dex, respectively.