Layer Swapping for Zero-Shot Cross-Lingual Transfer in Large Language Models

TL;DR

This paper introduces a simple layer swapping technique to enhance cross-lingual mathematical reasoning in large language models without additional training, outperforming existing methods on multilingual math benchmarks.

Contribution

It proposes a novel model merging method that combines language and math expertise by swapping layers, enabling effective zero-shot cross-lingual transfer in LLMs.

Findings

Outperforms other merging methods on MGSM benchmark by 10% across four languages

Layer swapping is simple, inexpensive, and based on interpretative analysis

Enables modular and post hoc transfer of reasoning capabilities

Abstract

Model merging, such as model souping, is the practice of combining different models with the same architecture together without further training. In this work, we present a model merging methodology that addresses the difficulty of fine-tuning Large Language Models (LLMs) for target tasks in non-English languages, where task-specific data is often unavailable. We focus on mathematical reasoning and without in-language math data, facilitate cross-lingual transfer by composing language and math capabilities. Starting from the same pretrained model, we fine-tune separate "experts" on math instruction data in English and on generic instruction data in the target language. We then replace the top and bottom transformer layers of the math expert directly with layers from the language expert, which consequently enhances math performance in the target language. The resulting merged models outperform the individual experts and other merging methods on the math benchmark, MGSM, by 10% across four major languages where math instruction data is scarce. In addition, this layer swapping is simple, inexpensive, and intuitive, as it is based on an interpretative analysis of the most important parameter changes during the fine-tuning of each expert. The ability to successfully re-compose LLMs for cross-lingual transfer in this manner opens up future possibilities to combine model expertise, create modular solutions, and transfer reasoning capabilities across languages all post hoc.