Understanding sulfur-nitrogen partitioning is essential for predicting secondary aerosol formation and mixing-state evolution, yet the mechanisms governing its cross-phase coupling remain poorly constrained. Here, we integrate single-particle mass spectrometry (SPA-MS) with air-pollutant and meteorological observations from two regional emission-control periods. We define three sulfur-to-nitrogen ratio metrics in the gas (gSNR), particle (pSNR), and number-based (nSNR) domains, and use causal inference and interpretable machine learning to identify their linkages and environmental drivers. The results reveal a stepwise propagation from precursor composition to particle chemistry and then to population mixing-state evolution. Although gSNR sets the first-order constraint on sulfur-nitrogen partitioning, the aerosol response is strongly particle-type dependent, with more pronounced sulfate enrichment in black-carbon-containing and organic-rich particles than in BC-free particles. Relative humidity (RH) emerges as the primary regulator of this coupling by modulating aerosol liquid water and phase transitions. Under dry conditions, the three SNR metrics diverge and aerosols remain largely externally mixed; under humid conditions, the metrics converge and aerosols evolve toward a more internally mixed state. Our results support the inclusion of RH- and particle-type-dependent parameterizations of cross-phase coupling and chemical heterogeneity in air-quality models.