PMMT: Preference Alignment in Multilingual Machine Translation via LLM Distillation

TL;DR

This paper introduces PMMT, a novel approach using Large Language Models to generate multilingual corpora aligned with human preferences and distill these preferences into smaller MT models, improving translation quality and efficiency.

Contribution

It presents a new method for generating preference-aligned multilingual data and distilling human preferences into compact MT models, advancing personalized translation.

Findings

Outperforms existing methods on preference-aligned translation tasks

Achieves competitive results on WMT and Flores benchmarks

Efficiently supports large-scale online translation services

Abstract

Translation is important for cross-language communication, and many efforts have been made to improve its accuracy. However, less investment is conducted in aligning translations with human preferences, such as translation tones or styles. In this paper, a new method is proposed to effectively generate large-scale multilingual parallel corpora with specific translation preferences using Large Language Models (LLMs). Meanwhile, an automatic pipeline is designed to distill human preferences into smaller Machine Translation (MT) models for efficiently and economically supporting large-scale calls in online services. Experiments indicate that the proposed method takes the lead in translation tasks with aligned human preferences by a large margin. Meanwhile, on popular public benchmarks like WMT and Flores, on which our models were not trained, the proposed method also shows a competitive performance compared to SOTA works.