DUAL-REFLECT: Enhancing Large Language Models for Reflective Translation through Dual Learning Feedback Mechanisms

TL;DR

DUAL-REFLECT is a novel framework that improves large language models' translation quality by using dual learning to provide effective feedback, especially benefiting low-resource language pairs.

Contribution

It introduces a dual learning feedback mechanism to enhance LLMs' self-reflection capabilities for better translation accuracy.

Findings

Improves translation accuracy across multiple tasks.

Effectively eliminates ambiguities in translations.

Enhances performance in low-resource language pairs.

Abstract

Recently, large language models (LLMs) enhanced by self-reflection have achieved promising performance on machine translation. The key idea is guiding LLMs to generate translation with human-like feedback. However, existing self-reflection methods lack effective feedback information, limiting the translation performance. To address this, we introduce a DUAL-REFLECT framework, leveraging the dual learning of translation tasks to provide effective feedback, thereby enhancing the models' self-reflective abilities and improving translation performance. The application of this method across various translation tasks has proven its effectiveness in improving translation accuracy and eliminating ambiguities, especially in translation tasks with low-resource language pairs.