Revenge of Monosemanticity: Specialized Neurons Improve Data Efficiency in MLPs

A new arXiv preprint (2608.24007v1) shows that multilayer perceptrons (MLPs) develop monosemantic specialized neurons aligned with specific predictive features in clustered regression data, improving data efficiency over global low-dimensional representation methods. The findings challenge the prevailing theory that neural networks learn a single global low-dimensional predictive geometry.

arXiv:2608.24007v1 Announce Type: new Abstract: Understanding how neural networks learn and organize features is central to understanding their behavior. Much existing theory of feature learning has focused on the emergence of a global low-dimensional predictive geometry. We show that this picture is incomplete. In regression problems with clustered data, we demonstrate that multilayer perceptrons MLPs naturally develop monosemantic specialized neurons: individual neurons become strongly aligned with a specific predictive feature relevant to a particular region of the input space. Rather than learning a single global low-dimensional representation, MLPs learn a collection of local low-dimensional representations that can collectively span a high-dimensional space. This specialization provably gives MLPs a data-efficiency advantage over feature-learning methods based on a global low-dimensional representation.