Recent context-control learning studies have shown that switch costs are reduced in a particular context predicting a high probability of switching as compared to another context predicting a low probability of switching. These context-specific switch probability effects suggest that control of task sets, through experience, can become associated with a particular context cue and be retrieved subsequently to modulate task-switching efficiency. However, advanced task foreknowledge, mediated by top-down retrieval of a specific task set, also modulates switch costs, as shown in the task-switching literature. Here, we examined how the retrieval of task sets interacts with the retrieval of task control. In Experiments 1 and 2, we provided the task foreknowledge midway through the experiment and examined how it might alter the item-specific switch probability (ISSP) effect and the list-wide switch probability effect (LWSP). Both ISSP and LWSP effects were replicated with no task foreknowledge, demonstrating modulations of switch costs due to context-control learning. These modulations, however, were swallowed once task foreknowledge became available, likely due to the certainty of task foreknowledge overpowering learned probabilistic information. Experiment 3 provided task foreknowledge throughout and compared how it affects the list- versus the item-based control learning. In this case, we detected the LWSP but not the ISSP effect. Experiment 3's findings suggest that, with task foreknowledge, list-based context-control learning does modulate switch costs--likely because the switch probability associations are retrieved proactively. Together, these data suggest that task-switching performance benefits from multiple paths to retrieve information in memory, best if done proactively.