Does Combining Parameter-efficient Modules Improve Few-shot Transfer Accuracy?

arXiv:2402.15414v1 Announce Type: new
Abstract: Parameter-efficient fine-tuning stands as the standard for efficiently fine-tuning large language and vision models on downstream tasks. Specifically, the efficiency of low-rank adaptation has facilitated the creation and sharing of hundreds of custom LoRA modules, each trained on distinct data from various downstream tasks. In this paper, we explore the composability of LoRA modules, examining if combining these pre-trained modules enhances generalization to unseen downstream tasks. Our investigation involves evaluating two approaches: (a) uniform composition, …

abstract accuracy arxiv cs.cv cs.lg data efficiency few-shot fine-tuning language large language lora low low-rank adaptation modules paper standard tasks transfer type vision vision models