The Expanded Fluid (EF) viscosity model was recently developed for petroleum applications.While petroleum fluids can be modeled as a single component fluid, process calculations often require that the fluid be treated as a mixture of pseudo-components. For mixtures, the EF model mixing rules require binary interaction parameters, αij, between each component pair in the mixture, which are fitted to experimental data. In this study, a generalized correlation is developed for these interaction parameters as a function of the specific gravity and hydrogen-to-carbon ratio of the mixture components. The proposed correlation was developed based on a dataset which included viscosity and density literature data for pure hydrocarbon pairs at room temperature and atmospheric pressure and new data for pseudo-pairs heavy oil/solvent at temperatures from 21 to 175°C, pressures up to 10 MPa, and solvent contents up to 40 wt%. The correlation was assessed on a distinct test dataset which included viscosity data for pure hydrocarbon mixtures from the literature, new data for deasphalted crude oils diluted with paraffinic and aromatic solvents, and new data for crude oils diluted with similar solvents. The viscosities of the development and test datasets were predicted with an overall average absolute relative deviation (AARD) of 13% and 10%, respectively, compared to 36% and 50% respectively when ideal mixing, αij =0, was assumed. Finally, the correlation was tested on an independent dataset from the literature including viscosity data for crude oils diluted with a variety of solvents. The viscosities of the independent dataset were predicted with an overall average absolute relative deviation (AARD) of 14%, compared to 30% when ideal mixing, αij =0, was assumed. The deviations obtained with the correlated αij are almost as low as those obtained when fitting the data by adjusting the αij.