一作ACL 2026Findings· Findings of ACL
Federated L o RA Fine-Tuning with Pipelined Error-Mitigated Aggregation and Matrix-Wise Freezing
Haoran Wang · Xiong Wang · Yuqing Li · Jing Chen · Junyi Zhang · Nan Yan · Kun He · Wei Wang
摘要
Federated low-rank adaptation (LoRA) enables multiple clients to collaboratively fine-tune large language models (LLMs) without disclosing their raw data. However, existing works often experience performance degradation due to biased model aggregation and are hindered by significant communication...