作者位次 #5/10ACL 2026Findings· Findings of ACL
FARSS : Fisher-Optimized Adaptive Low-Rank and Singular-Vector Selection for Knowledge-Preserving Fine-Tuning
Renxing Chen · Ziwei Xiang · Peisong Wang · Hongjian Fang · Meng Li · Fanhu Zeng · Yanan Zhu · Peipei Yang · Xu-yao Zhang · Jian Cheng
摘要
Parameter-efficient fine-tuning (PEFT) has become a prevalent approach for adapting large language models (LLMs). However, low-rank adaptation methods face an inherent trade-off: improving target task performance can compromise pre-trained world knowledge, while aggressively constraining updates...