LoRA-One: One-Step Full Gradient Could Suffice for Fine-Tuning Large Language Models, Provably and Efficiently.

Yuanhe Zhang, Fanghui Liu, Yudong Chen

Browse the full ICML paper archive.