Completing the 2x2 factorial: HRM's hierarchy with full BPTT reveals gradient-architecture interaction
-
Updated
Jun 16, 2026 - Jupyter Notebook
#
gradient-methods
Here are 2 public repositories matching this topic...
Disentangling gradient quality from architecture in recursive reasoning. Controlled experiment: 1-step gradient approximation is the sole bottleneck in HRM vs TRM performance gap.
-
Updated
Jun 16, 2026 - Jupyter Notebook
Improve this page
Add a description, image, and links to the gradient-methods topic page so that developers can more easily learn about it.
Add this topic to your repo
To associate your repository with the gradient-methods topic, visit your repo's landing page and select "manage topics."