LoRA Methods

LoRA Rank Sensitivity: How r Affects Training

r=4 vs r=64. We tested 8 rank values across 46 runs, identical conditions. The winner wasn’t what the literature suggests. More rank isn’t always better — past a threshold it actively hurts.

⏲
In progress
Full write-up coming soon
Data tables are ready. Full analysis publishing soon.
Built on this research

Everything in this article is measured, not promised. See it running:

Talk to NOESIS liveSee the 33x benchmarkBack the research