New Method Rebalances LoRA Fine-Tuning Without Extra Training

Original: Learn the Directions, Normalize the Gains: Post-Training Normalization for LoRA

Source: arxiv.org ↗