Associate Professor, Macquarie University
1 paper at NeurIPS 2025
The paper introduces a novel two-stage sketching framework for federated LoRA that enables differentially private and effective updating of low-rank matrices in large language models.
