<p>The surface chemical composition of the Moon holds key insights into its geological evolution and resource potential. With future missions targeting the lunar south pole, there is a pressing need for high-resolution chemical maps of this region. Here we integrate spectral data with the latest Chang’E-6 samples, which are the first returned from the lunar farside, to refine global estimates of major oxides. A classical equation was applied to produce updated maps of FeO, TiO₂, and optical maturity, while a deep learning model was constructed to invert more complex oxides across the lunar surface. The addition of Chang’E-6 samples further corrected the TiO₂ content of lunar surface. The machine learning approach captures nonlinear links between reflectance and geochemistry, enabling robust mapping of oxides. These results reveal basaltic signatures in the south polar region, suggesting volcanic activity, and provide a framework for investigating the Moon’s thermal evolution and resource potential.</p>