In environments rich with complex multisensory stimuli, flexible and effective behaviors rely on our ability to transfer learned associations across sensory modalities. Here, we explored the intertwined cortical representations of whisker and visual sensations in mice, focusing on their role in cross-modal transfer learning. Mice trained to differentiate between the stimulations of two vertically-aligned whiskers could seamlessly transfer this learned behavior to distinguish between two visual stimuli based on their vertical positioning. Generalization was achieved only when the spatial reward contingency remained congruent across the modality switch. Employing multi-scale calcium imaging across the dorsal cortex revealed two associative domains within the ventral and dorsal streams for visuo-tactile integration. These areas exhibited multimodal spatial congruency, functionally and anatomically, supported by topographically organized feedforward and feedback projections to and from primary sensory regions. Single-cell responses in these domains demonstrated the hallmark properties of an amodal representation of sensory space, exhibiting congruent visual and tactile positional tuning that enabled effective cross-modal generalization when stimulus positions were decoded with a classifier. By employing a loss-of-function approach and suppressing synaptic transmission in the dorsal stream, we observed a disruption in transfer learning, highlighting the critical role of the dorsal multisensory region. Conversely, using a gain-of-function approach through optogenetic activation of neurons in the dorsal region that encode the task's visual positions, we tested the sufficiency of this region for transfer behavior. Our results revealed a crucial cortical pathway for visuo-tactile integration and cross-modal transfer learning, essential for goal-directed behaviors.