Fetching the paper…

Gradient Descent on Two-layer Nets: Margin Maximization and Simplicity Bias · Around