📄 中文摘要
知识蒸馏（KD）作为一种研究领域，近年来在自然语言处理（NLP）中作为压缩工具受到了广泛关注，旨在解决日益庞大的模型所带来的挑战。在机器翻译（MT）中，KD不仅仅是压缩工具，还作为一种通用的知识转移机制，影响监督方式、翻译质量和效率。该综述综合了截至2025年10月1日的105篇关于机器翻译领域知识蒸馏的文献，首先为非专业人士介绍机器翻译和知识蒸馏的基本概念，随后概述了与机器翻译应用相关的标准知识蒸馏方法，最后根据方法论贡献和应用效果对KD4MT文献中的进展进行了分类。
KD4MT：机器翻译中的知识蒸馏综述
出处: KD4MT: A Survey of Knowledge Distillation for Machine Translation
发布: 2026年2月19日
📄 English Summary
KD4MT: A Survey of Knowledge Distillation for Machine Translation
Knowledge Distillation (KD) has emerged as a significant research area in recent years, serving as a compression tool to tackle the challenges posed by increasingly large models in Natural Language Processing (NLP). In the context of Machine Translation (MT), KD operates not only as a compression method but also as a general-purpose knowledge transfer mechanism that influences supervision, translation quality, and efficiency. This survey synthesizes findings from 105 papers on KD for MT (KD4MT) up to October 1, 2025. It begins with an introduction to both MT and KD for non-experts, followed by an overview of standard KD approaches relevant to MT applications. Advances in the KD4MT literature are categorized based on their methodological contributions and application outcomes.