Dataset for WDMS binary candidates from Gaia DR3 BP/RP (XP) spectra

Using Gaia DR3's low-resolution BP/RP (XP) spectra, we have constructed an extensive catalog of WDMS binaries through a method that is entirely independent of models. Our approach integrates a model-independent neural network for spectral modelling with Gaussian Process Classification to accurately identify WDMS binaries among over 10 million stars within 1 kpc. We have identified ~ 30,000 WDMS binary candidates, including 1,700 high-confidence systems confirmed through spectral fitting. Our technique is highly effective at detecting systems where the main-sequence star dominates the spectrum—cases that have historically challenged conventional, model-dependent detection methods. Validation using GALEX photometry reinforces the reliability of our classifications: 70% of candidates with an absolute magnitude M_G>7 exhibit UV excess, a characteristic signature of white dwarf companions.