Fine-tuning Large Language Models for Adaptive Machine Translation

TL;DR

The paper tackles real-time adaptive machine translation by fine-tuning an efficient open-source LLM (Mistral 7B) with a mixture of zero-shot and one-shot prompts in the medical Spanish–English domain. It demonstrates that a relatively small fine-tuning dataset (~20k segments) can enhance both standard translation and inference-time domain adaptation, achieving zero-shot performance comparable to NLLB-3.3B and surpassing ChatGPT, while one-shot gains approach or exceed those of commercial LLMs. The authors establish a practical, self-hosted approach with strong efficiency advantages and provide a thorough information-retrieval pipeline to supply context for adaptive prompts. They also outline future work across domains, languages, and larger models, aiming to extend the gains of adaptive MT to broader settings and data scales.

Abstract

This paper presents the outcomes of fine-tuning Mistral 7B, a general-purpose large language model (LLM), for adaptive machine translation (MT). The fine-tuning process involves utilising a combination of zero-shot and one-shot translation prompts within the medical domain. The primary objective is to enhance real-time adaptive MT capabilities of Mistral 7B, enabling it to adapt translations to the required domain at inference time. The results, particularly for Spanish-to-English MT, showcase the efficacy of the fine-tuned model, demonstrating quality improvements in both zero-shot and one-shot translation scenarios, surpassing Mistral 7B's baseline performance. Notably, the fine-tuned Mistral outperforms ChatGPT "gpt-3.5-turbo" in zero-shot translation while achieving comparable one-shot translation quality. Moreover, the zero-shot translation of the fine-tuned Mistral matches NLLB 3.3B's performance, and its one-shot translation quality surpasses that of NLLB 3.3B. These findings emphasise the significance of fine-tuning efficient LLMs like Mistral 7B to yield high-quality zero-shot translations comparable to task-oriented models like NLLB 3.3B. Additionally, the adaptive gains achieved in one-shot translation are comparable to those of commercial LLMs such as ChatGPT. Our experiments demonstrate that, with a relatively small dataset of 20,000 segments that incorporate a mix of zero-shot and one-shot prompts, fine-tuning significantly enhances Mistral's in-context learning ability, especially for real-time adaptive MT.

Figures (1)

• Figure 1: Zero-shot and one-shot prompts used for fine-tuning Mistral