Mingzhe Yu, Osamu Tatebe: RRLoRA: Refactorized Low-Rank Adaptation With Learning-Rate Restarts for Efficient Fine-Tuning. IEEE Access 14: 84922-84942 (2026)