arXiv stat.ML
· Papers
Breaking Data Symmetry is Needed For Generalization in Feature Learning Kernels
arXiv:2604.00316v2 Announce Type: replace Abstract: Grokking occurs when a model achieves high training accuracy but generalization to unseen test points happens long after that. This phenomenon was initially observed on a class of algebraic problems, such as learning modular arithmetic (Power et al., 2022). We study g