Reproduce “Curriculum unlearning in humans and neural networks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jirko Rubruck, Alice Zhang, Andrew Saxe, Christopher Summerfield
The authors test whether presenting rule-following examples before exceptions helps learners overcome an incomplete categorization rule. In a bug-classification task, people and a two-layer neural network learned exceptions better with a simple-to-complex sequence than with no curriculum. The network benefit was larger after one-dimensional than XOR pretraining, but representational changes were measured only in the network, not people.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.