利用知识蒸馏压缩多语言神经机器翻译模型的实证研究（英文版）

英文标题：An Empirical Study of Leveraging Knowledge Distillation for Compressing Multilingual Neural Machine Translation Models中文摘要：本文探讨了如何通过知识蒸馏来压缩 MNMT 模型，发现这是一项具有挑战性的任务，并提出了一些设计思考和优化方案。英文摘要：Knowledge distillation (KD) is a well-known method for compressing neuralmodels. However, works focusing on distilling knowledge from large multilingualneural machine translation (MNMT) models into smaller ones are practicallynonexistent, despite the popularity and superiority of MNMT. This paper bridgesthis