FLeX: Fourier-based Low-rank EXpansion for multilingual transfer

TL;DR

This paper introduces FLeX, a Fourier-based regularization technique combined with LoRA fine-tuning and optimizer strategies to enhance cross-lingual code generation in large language models, achieving significant performance gains.

Contribution

It demonstrates that Fourier-based regularization during LoRA fine-tuning substantially improves multilingual transfer performance in large language models.

Findings

LoRA fine-tuning on high-quality data surpasses broader fine-tuning performance.

Fourier-based regularization improves cross-lingual transfer, increasing pass@1 from 34.2% to 42.1%.

Sophia optimizer converges faster but yields similar final scores as Adam.

Abstract

Cross-lingual code generation is critical in enterprise environments where multiple programming languages coexist. However, fine-tuning large language models (LLMs) individually for each language is computationally prohibitive. This paper investigates whether parameter-efficient fine-tuning methods and optimizer enhancements can improve cross-lingual transfer from Python to languages like Java. We fine-tune the Code Llama 7B model using low-rank adaptation (LoRA) to optimize a small subset of parameters and compare Adam and Sophia optimizers, while exploring a novel Fourier-based regularization technique. Our contributions include: (1)demonstrating that LoRA fine-tuning on a small, high-quality dataset (MBPP) can exceed the pass@1 performance of the more broadly fine-tuned Code Llama-Python-7B model (40.1% vs. 38.4%); (2) showing that while Sophia achieves faster convergence than Adam, final pass@1 scores show marginal differences; and (3) presenting evidence that Fourier-based regularization during fine-tuning significantly improves cross-lingual transfer, achieving 42.1% pass@1 on Java tasks compared to the 34.2% baseline. These findings suggest that combining LoRA, optimized training methods, and frequency-domain regularization can efficiently adapt single-language LLMs to perform well across multiple programming languages.