作者位次 #3/3ACL 2026Findings· Findings of ACL
DPL o RA : A Dual-Pruning Framework based on ILP Optimization and Progressive Pruning for Parameter-Efficient L o RA Fine-Tuning
Changjun Park · Sejong Yoon · Jaekwang Kim
摘要
We propose DPLoRA (Dual-Pruning Low-Rank Adaptation), a framework that optimizes rank allocation via two stages: (1) an initial pruning stage (OPLoRA; Optimal Pruning LoRA) that uses Integer Linear Programming (ILP) to determine optimal layer-wise ranks without manual tuning; and (2) a progressiv...