MoRe Fine-Tuning with 10x Fewer Parameters

TL;DR

MoRe introduces a neural architecture search-based framework for parameter-efficient fine-tuning that surpasses existing methods like LoRA in performance and parameter efficiency across various tasks and models.

Contribution

MoRe provides a novel, theoretically more expressive and empirically more efficient framework for adapter architecture search in PEFT, reducing parameter count significantly.

Findings

MoRe outperforms state-of-the-art PEFT methods on multiple tasks.

MoRe uses only 5% of LoRA's parameters while maintaining or improving performance.

MoRe is more expressive than LoRA according to theoretical analysis.

Abstract

Parameter-efficient fine-tuning (PEFT) techniques have unlocked the potential to cheaply and easily specialize large pretrained models. However, the most prominent approaches, like low-rank adapters (LoRA), depend on heuristics or rules-of-thumb for their architectural choices -- potentially limiting their performance for new models and architectures. This limitation suggests that techniques from neural architecture search could be used to obtain optimal adapter architectures, but these are often expensive and difficult to implement. We address this challenge with Monarch Rectangular Fine-tuning (MoRe), a simple framework to search over adapter architectures that relies on the Monarch matrix class. Theoretically, we show that MoRe is more expressive than LoRA. Empirically, our approach is more parameter-efficient and performant than state-of-the-art PEFTs on a range of tasks and models, with as few as 5\% of LoRA's parameters.