Local Intrinsic Dimension of Representations Predicts Alignment and Generalization in AI Models and Human Brain

TL;DR

This paper demonstrates that the local intrinsic dimension of neural representations predicts model generalization, alignment with other models and human brain activity, and explains the benefits of scaling in AI systems.

Contribution

It introduces local intrinsic dimension as a key geometric property linking model generalization, alignment, and scaling in AI and neuroscience.

Findings

Lower local intrinsic dimension correlates with better generalization.

Models with stronger alignment to humans have lower local intrinsic dimension.

Scaling reduces local intrinsic dimension, improving model performance.

Abstract

Recent work has found that neural networks with stronger generalization tend to exhibit higher representational alignment with one another across architectures and training paradigms. In this work, we show that models with stronger generalization also align more strongly with human neural activity. Moreover, generalization performance, model--model alignment, and model--brain alignment are all significantly correlated with each other. We further show that these relationships can be explained by a single geometric property of learned representations: the local intrinsic dimension of embeddings. Lower local dimension is consistently associated with stronger model--model alignment, stronger model--brain alignment, and better generalization, whereas global dimension measures fail to capture these effects. Finally, we find that increasing model capacity and training data scale systematically reduces local intrinsic dimension, providing a geometric account of the benefits of scaling. Together, our results identify local intrinsic dimension as a unifying descriptor of representational convergence in artificial and biological systems.