Fetching the paper…

Universal scaling laws in the gradient descent training of neural networks · Around