During the molten salt electrorefining of spent nuclear fuel, multiple phases such as oxide, solid metal, liquid metal, and molten salt often co-exist. Computational modeling can be a useful tool for understanding the reaction mechanism across the multiple phases. The new model has been developed and applied to a lab-scale electrorefining with liquid metal anode and solid cathode LiCl-KCl molten salt. The benchmark study predicts anodic dissolution and cathodic deposition of U and Pu with minor disagreements. In particular, the on-set of Pu deposition on the surface of the solid cathode is well estimated, which is important for the quality of U ingot and the safeguards of process. The underestimation of U deposition (similar to 6%) and the overestimation of Pu dissolution (similar to 7%) at the end of simulation are explained by unconsidered reaction species such as Np and Am from the liquid Cd anode, which over-estimates the dissolution of Pu from the anode compared to the measured data. The sensitivity study also reveals the transition behaviors of electrochemical reactions for U and Pu on the solid cathode are changed by diffusion boundary layer thickness, transfer coefficients, and the difference of electrochemical potentials more sensitively than those of the liquid metal anode. For this specific experiment case, the thinner diffusion boundary layer improves the prediction of cathodic reactions particularly at the end of electrorefining. (C) 2017 Elsevier B.V. All rights reserved.