Ks absolute magnitudes from LAMOST for OB stars

The method applies a neural network model trained on stars with precise Gaia parallax to the spectra and predicts K_s_-band absolute magnitudes M_Ks_ with a precision of 0.25mag, which corresponds to a precision of 12% in spectroscopic distance.

For distant stars (e.g., >5kpc), the inclusion of constraints from spectroscopic M_Ks_ significantly improves the distance estimates compared to inferences from Gaia parallax alone. Our method accommodates for emission-line stars by first identifying them via principal component analysis reconstructions and then treating them separately for the M_Ks_ estimation. We also take into account unresolved binary/multiple stars, which we identify through deviations in the spectroscopic M_Ks_ from the geometric M_Ks_ inferred from Gaia parallax.