MS student, McGill University, McGill University
1 paper at NeurIPS 2025
We show that deep neural networks across architectures and training conditions all instantiate the same abstract algorithm for modular addition.
