Capture LoRA's low-rank update in the smallest possible form: a frozen base scalar plus a trainable product of two adapter scalars.
Required APIdef lora_scalar_update(base, left, right):
return adapted_weightThink through the mechanism first if you want the extra reasoning step. It never blocks the editor.
lora_scalar_update(10, 2, 3)the adapter changes behavior without rewriting the base
lora_scalar_update(10, 0, 3)zero trainable update preserves the pretrained value
2 hidden edge tests run after the visible contract passes.
Run Tests to see the contract verdicts here.