Auditory Neural Codes Exhibit Hyperbolic Geometric Signatures

Auditory processing begins by decomposing sounds into frequency components. We investigated how sensory neurons encode this information across auditory regions using hyperbolic embeddings. High-dimensional neural recordings from the primary auditory cortex and two subcortical regions in rats exposed to complex tone-cloud stimuli revealed distinct coding geometries. Multi-feature receptive field analysis based on the maximum entropy principle showed a stronger dominance of quadratic over linear features in the cortex compared to subcortical areas. Geometric analyses further demonstrated that neural responses are organized within low-dimensional hyperbolic manifolds rather than linear Euclidean spaces. Finally, we identified a compact neural code that retains nearly the same stimulus information as the full binary response set while enhancing interpretability. These results suggest that auditory neurons are more effectively characterized through packet-based codes, which may represent a fundamental unit of auditory information processing.