In this study, an experimental setup consisting of a sand pack with different configurations of complexity of wormhole patterns was designed. These different configurations have the same wormhole coverage index (WCI) determined by our earlier field-scale history match studies. After saturation of the model with heavy oil, three different solvents (CO2, heptane, and diluents) were introduced into the wormhole structure inside the sand pack (injection phase) and then left for a period of time (soaking phase). Next, the resulted mixture (solvent and oil) was allowed to be produced (production phase). The experiment was aimed to mimic cyclic solvent stimulation at reservoir conditions, and several cycles were run for each experiment. At each cycle, the amount of collected oil and solvent was determined, and the resulting mixture was analyzed through gas chromatography and refractometry. It was observed that some of the injected solvent was trapped inside the sand pack. Therefore, in the final step, a solvent retrieval attempt was made by injecting different low-temperature hot water based on the solvent type. The aim was to observe how much additional solvent could be retrieved from the sand pack at reservoir conditions. Next, the sand-pack experiments were numerically simulated, and effective diffusion coefficients were obtained through history matching. To generate accurate predictions in field-scale simulation, an upscaling procedure from laboratory results of the cyclic solvent injection process was suggested. In this study, we observed that the use of light solvents (CO2) could maintain the sand-pack pressure for a longer period (attributed to foamy oil phenomenon) compared to liquid solvent during the production phase. However, the oil recovery from sole application of light solvent was not as considerable as with the heavier solvents of heptane and diluents. These observations suggest that an improved heavy oil recovery could be achieved using a hybrid application of solvents and hot water in cold heavy oil production with sand (CHOPS) reservoirs. A post-flush hot water was found to play a positive role in solvent retrieval; however, the scale dependency of solvent retrieval behavior needs to be investigated. The findings of this paper can be used in a later study to optimize the field-scale solvent injection schemes considering the economics of the process tested.