Emergent Generalization by Representation Learning in Artificial Neural Networks
Abstract
Dimensionality reduction has proven powerful for identifying neural manifolds, which are low-dimensional structures underlying high-dimensional neural activity. These low-dimensional representations have improved the interpretability of population-level coding. Yet whether such low-dimensional representations are biologically relevant and confer functional advantages in learning systems, or merely reflect neuron-level activity, remains contested in neuroscience. We show that an explicit information bottleneck forcing a recurrent neural network to learn a low-dimensional representation is necessary for rotational and out-of-distribution generalisation in a time-series prediction task. Using information-theoretic measures of causal emergence, we characterise the dynamics of this representation across the memorisation-to-generalisation transition, finding a non-monotonic trajectory which shows an initial decrease, a minimum, and a subsequent rise to a maximum, even as prediction loss falls monotonically. This trajectory scales with task complexity, and the magnitude of emergent structure reliably predicts generalisation performance. Analysis of hippocampal CA1 and medial prefrontal activity in rats learning an alternating maze task reveals analogous non-monotonic emergence dynamics that track behavioural performance. Together, these findings indicate that the ability of neural networks to learn compact, distributed and emergent representations confers a functional advantage for generalisation, supporting a causal role for learned representations in cognition.
Related
- High-dimensional hippocampus representation facilitates learning of low-dimensional abstract representation and supports generalization
- Invariance before structure: two modes of emergent compression in representation learning
- Neural networks reveal emergent properties of collective learning in democratic but not despotic groups
- ReGraph: A Computational Account of Emergent Generalization in the "what" and "where" Dual Visual Streams
- Particle-based representation of phytoplankton modulates emergent coexistence and biogeography in a model ocean
BibTeX
@misc{rajpal2026emergent,
title = {{Emergent Generalization by Representation Learning in Artificial Neural Networks}},
author = {Hardik Rajpal and Dan Goodman},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2607.10430},
eprint = {2607.10430},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2607.10430},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Emergent Generalization by Representation Learning in Artificial Neural Networks
AU - Hardik Rajpal
AU - Dan Goodman
PY - 2026
DA - 2026/11/05
PB - arXiv
DO - 10.48550/arxiv.2607.10430
UR - https://arxiv.org/abs/2607.10430
AB - Dimensionality reduction has proven powerful for identifying neural manifolds, which are low-dimensional structures underlying high-dimensional neural activity. These low-dimensional representations have improved the interpretability of population-level coding. Yet whether such low-dimensional representations are biologically relevant and confer functional advantages in learning systems, or merely reflect neuron-level activity, remains contested in neuroscience. We show that an explicit information bottleneck forcing a recurrent neural network to learn a low-dimensional representation is necessary for rotational and out-of-distribution generalisation in a time-series prediction task. Using information-theoretic measures of causal emergence, we characterise the dynamics of this representation across the memorisation-to-generalisation transition, finding a non-monotonic trajectory which shows an initial decrease, a minimum, and a subsequent rise to a maximum, even as prediction loss falls monotonically. This trajectory scales with task complexity, and the magnitude of emergent structure reliably predicts generalisation performance. Analysis of hippocampal CA1 and medial prefrontal activity in rats learning an alternating maze task reveals analogous non-monotonic emergence dynamics that track behavioural performance. Together, these findings indicate that the ability of neural networks to learn compact, distributed and emergent representations confers a functional advantage for generalisation, supporting a causal role for learned representations in cognition.
ER -