system-one-gemma

Reimplements a System One decision model on Gemma 3 270M with a LoRA of 2.6M trainable parameters, scoring all options in one forward pass; trains in a free Colab and ships weights in the repo.

Notes

64.4% held-out accuracy, ECE 0.047, ~50 ms per decision, ~15 minutes of training on a T4. Weights CC-BY-NC.

source-backed — Public repository, docs or live artifact. About this label

Destinations

Filed under

• #replica
• #gemma
• #github

In collections

• Local & replica models →