AI RESEARCH
Gradient-Based LoRA Rank Allocation Under GRPO: An Empirical Study
arXiv CS.CL
•
ArXi:2605.07366v1 Announce Type: new Adaptive rank allocation for LoRA, allocating parameters to important layers and fewer to unimportant ones, consistently improves efficiency under supervised fine-tuning (SFT). We investigate whether this success transfers to reinforcement learning, specifically Group Relative Policy Optimization (GRPO). Using gradient-magnitude profiling on Qwen 2.5 1.5B with GSM8K, we find that it does not: proportional rank allocation degrades accuracy by 4.5 points compared to uniform allocation (70.0% vs. 74.5%), despite using identical parameter budgets.