Curriculum unlearning in humans and neural networks
Abstract
During learning, humans often benefit from curricula that teach simple rules before exceptions. However, we lack a computational explanation for this phenomenon. Here, using a canonical behavioural task and modelling with neural networks, we develop and test a theory for why simple-to-complex curricula succeed. We show that during learning, both humans and neural networks show simplicity biases, a tendency to learn simple rules that are underfit to the data. In neural networks this bias is accompanied by compression of neural representations. Simple-to-complex curricula help both types of learners overcome this compression, developing representations (and policy) with appropriate dimensionality for the task. We test and confirm a further prediction from this theory: learners that have been pre-exposed to simple tasks and with more compressed initial representations benefit more from curricula. These findings suggest that curricula help with ``unlearning'' of simple rules.
Related
- Geometric and algebraic inference in humans and neural networks
- Neural encoding of grasp and object properties in the posterior parietal and motor cortices of the cortical grasping network in tetraplegic humans
- Walking delays auditory processing in humans alongside a global shift in aperiodic neural activity
- A rapid, machine-learning readout of cochlear neural degeneration that transfers from animals to humans
- Constricted pupil states index neural replay during wakeful rest in humans
BibTeX
@misc{rubruck2026curriculum,
title = {{Curriculum unlearning in humans and neural networks}},
author = {Jirko Rubruck and Alice Zhang and Andrew Saxe and Christopher Summerfield},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.04565},
eprint = {2611.04565},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.04565},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Curriculum unlearning in humans and neural networks
AU - Jirko Rubruck
AU - Alice Zhang
AU - Andrew Saxe
AU - Christopher Summerfield
PY - 2026
DA - 2026/11/03
PB - arXiv
DO - 10.48550/arxiv.2611.04565
UR - https://arxiv.org/abs/2611.04565
AB - During learning, humans often benefit from curricula that teach simple rules before exceptions. However, we lack a computational explanation for this phenomenon. Here, using a canonical behavioural task and modelling with neural networks, we develop and test a theory for why simple-to-complex curricula succeed. We show that during learning, both humans and neural networks show simplicity biases, a tendency to learn simple rules that are underfit to the data. In neural networks this bias is accompanied by compression of neural representations. Simple-to-complex curricula help both types of learners overcome this compression, developing representations (and policy) with appropriate dimensionality for the task. We test and confirm a further prediction from this theory: learners that have been pre-exposed to simple tasks and with more compressed initial representations benefit more from curricula. These findings suggest that curricula help with ``unlearning'' of simple rules.
ER -