Enhancing Code Translation in Language Models with Few-Shot Learning via Retrieval-Augmented Generation

TL;DR

This paper presents a retrieval-augmented few-shot learning method to improve code translation accuracy in large language models by dynamically leveraging relevant code examples from existing repositories.

Contribution

It introduces a novel Retrieval-Augmented Generation approach for code translation that enhances model performance without extensive retraining, outperforming traditional zero-shot methods.

Findings

Significant improvement in translation quality, especially for complex tasks.

Effective across diverse datasets and models, including open and commercial LLMs.

Robustness demonstrated with various shot numbers and embedding models.

Abstract

The advent of large language models (LLMs) has significantly advanced the field of code translation, enabling automated translation between programming languages. However, these models often struggle with complex translation tasks due to inadequate contextual understanding. This paper introduces a novel approach that enhances code translation through Few-Shot Learning, augmented with retrieval-based techniques. By leveraging a repository of existing code translations, we dynamically retrieve the most relevant examples to guide the model in translating new code segments. Our method, based on Retrieval-Augmented Generation (RAG), substantially improves translation quality by providing contextual examples from which the model can learn in real-time. We selected RAG over traditional fine-tuning methods due to its ability to utilize existing codebases or a locally stored corpus of code, which allows for dynamic adaptation to diverse translation tasks without extensive retraining. Extensive experiments on diverse datasets with open LLM models such as Starcoder, Llama3-70B Instruct, CodeLlama-34B Instruct, Granite-34B Code Instruct, and Mixtral-8x22B, as well as commercial LLM models like GPT-3.5 Turbo and GPT-4o, demonstrate our approach's superiority over traditional zero-shot methods, especially in translating between Fortran and CPP. We also explored varying numbers of shots i.e. examples provided during inference, specifically 1, 2, and 3 shots and different embedding models for RAG, including Nomic-Embed, Starencoder, and CodeBERT, to assess the robustness and effectiveness of our approach.