Towards Modular LLMs by Building and Reusing a Library of LoRAs

TL;DR

This paper introduces a library of task-specific adapters for large language models, along with a novel routing method, Arrow, enabling zero-shot task generalization and improved transfer learning across multiple tasks.

Contribution

It proposes a new clustering method, MBC, for building adapter libraries and a zero-shot routing mechanism, Arrow, for dynamic adapter selection without retraining.

Findings

MBC effectively groups tasks based on adapter similarity.

Arrow enables zero-shot adapter routing for new inputs.

Adapters and routing outperform traditional joint training on various tasks.

Abstract

The growing number of parameter-efficient adaptations of a base large language model (LLM) calls for studying whether we can reuse such trained adapters to improve performance for new tasks. We study how to best build a library of adapters given multi-task data and devise techniques for both zero-shot and supervised task generalization through routing in such library. We benchmark existing approaches to build this library and introduce model-based clustering, MBC, a method that groups tasks based on the similarity of their adapter parameters, indirectly optimizing for transfer across the multi-task dataset. To re-use the library, we present a novel zero-shot routing mechanism, Arrow, which enables dynamic selection of the most relevant adapters for new inputs without the need for retraining. We experiment with several LLMs, such as Phi-2 and Mistral, on a wide array of held-out tasks, verifying that MBC-based adapters and Arrow routing lead to superior generalization to new tasks. We make steps towards creating modular, adaptable LLMs that can match or outperform traditional joint training.