Adam: Splitting

Based on your interest in "Splitting Adam," you are likely referring to research surrounding the widely used in machine learning. There isn't one single paper with that exact title, but several "interesting" papers analyze splitting the algorithm's components or its behavior in complex ways: 1. The Sign, Magnitude and Variance of Stochastic Gradients

It proposes Coupled Adam to fix this specific side effect. Splitting Adam

By testing these separately, researchers found that "Stochastic Sign Descent" can actually outperform standard Adam on specific datasets like MNIST and CIFAR10. 2. Adaptive Multilevel Splitting (ADAM) Based on your interest in "Splitting Adam," you

Published in 2025, this paper "splits" the problem of in LLM embeddings. It argues that Adam's second moment actually causes

It argues that Adam's second moment actually causes word representations to become narrow and directional (anisotropic).

It's often applied to power grid reliability or particle transport. 3. Adam Reduces a Unique Form of Sharpness

It isolates the stochastic direction (the sign of the gradient) from the adaptive step size (the relative variance).