1 paper
Yangming Zhang, Ling Yi, Datong Li +4
Low-rank adaptation (LoRA) often underperforms full fine-tuning. We investigate this performance gap through a tangent-space lens: factorized updates are confined to the tangent sp…