Skip to content
#

supermix

Here are 5 public repositories matching this topic...

Tether trains a 2,762-parameter receiver over frozen digit classifiers. It compares ordinary supervised learning, donor dropout, an IIT-style intervention control, and two additional losses: matching relevant intervention effects and suppressing irrelevant or damaged-expert effects. The main comparison is Tether versus identical-data IIT, with sing

  • Updated Oct 24, 2026
  • Python

Add this topic to your repo

To associate your repository with the supermix topic, visit your repo's landing page and select "manage topics."

Learn more