Cerebellar granule cells separate behavioral contexts through geometry-preserving affine transformations of cortical neural manifolds rather than high-dimensional expansion, rotating low-rank cortical dynamics apart to enable context-specific motor outputs while preserving the intrinsic manifold structure required for rapid learning—a separation that strengthens with dual-task expertise and computationally optimizes the tradeoff between generalization and interference.

To learn effectively, animals must generalize across yet distinguish between related contexts. Generalization relies on low-dimensional neural manifolds found throughout neocortex, which accelerate learning by constraining neural activity to task-relevant axes. Conversely, context separation is attributed to neural expansion layers that can project information into high-dimensional feature spaces, most famously cerebellar granule cells (GrCs). To investigate the generalization-separation tradeoff, we simultaneously imaged key nodes in the universal cortico-cerebellar pathway—premotor layer 5 pyramidal tract (L5PT) and GrCs—during parallel learning of two distinct skills with shared temporal structure. Rather than expanding the cortical representations, GrCs retained their low-rank encoding of each task. Across contexts, however, despite stable cortico-cerebellar coupling, L5PT activity patterns generalized while GrC patterns temporally remapped. Mechanistically, GrCs used affine transformations that rotated the cortical manifolds apart but preserved their intrinsic low-dimensional geometry. Moreover, GrCs decorrelated cortical trajectories most strongly in expert animals. This reveals a fundamental architectural division of labor: the cortex generates invariant dynamic primitives for smooth generalization, while the cerebellum reconfigures them to drive context-specific output.