Survey of Surveys Data Release 3. Homogenised stellar parameters across nine spectroscopic surveys

Large spectroscopic surveys have measured atmospheric parameters for millions of Milky Way stars, but differences in instruments, wavelength ranges, analysis methods, and assumptions introduce survey-dependent systematics that complicate their joint use. We aim to place effective temperature, surface gravity, and metallicity estimates from major spectroscopic surveys onto a common, externally validated scale, and to provide a homogenised catalogue for large-scale Galactic studies. We construct the calibration hierarchically. We first define an internal reference scale from the best-populated and best-overlapped surveys (APOGEE, GALAH, and LAMOST LRS), and model survey-to-reference corrections and their uncertainties with machine-learning regression. The scale is then extended to LAMOST MRS, Gaia RVS, RAVE, DESI, Gaia-ESO, and SEGUE, and anchored to high-quality measurements from the PASTEL database. The calibration reduces survey-to-survey offsets and improves consistency in independent overlap samples: independent calibration of GALAH and LAMOST LRS temperatures onto the APOGEE scale reduces their median offset from 44.5 K to 6.2 K. The final catalogue contains about 10.13 million stars, including 6.10 million in the recommended high-quality subset and 3.69 million in the metallicity-optimised subset. For the high-quality subset, Gaia benchmark stars give median residuals of 30 K in Teff, 0.03 dex in log g, and $-0.02$ dex in [Fe/H]. The [Fe/H]-optimised subset gives median residuals of 0.003 dex against Gaia benchmark stars, 0.05 dex for globular clusters, and $-0.02$ dex for open clusters. As an example application, we confirm the presence of metal-poor stars on thin-disc-like orbits, including stars with [Fe/H] $< -2$, showing that the catalogue can be used to study rare Galactic populations on a consistent multi-survey scale.