Abstract: Chinese-Vietnamese cross-lingual query expansion aims to add terms and concepts matching the semantics of Chinese queries and convert them into Vietnamese. To deal with low-resource language in a reasonable computational costs, this paper proposes a Chinese-Vietnamese cross-lingual query expansion method based on retrieval augmentation and knowledge distillation. By leveraging knowledge distillation and retrieval augmentation, the method injects the chain-of-thought generation capabilities and external knowledge from large language models into multilingual pre-trained models with fewer parameters. Experimental results on MLQA, XQuAD, and a constructed Chinese-Vietnamese dataset demonstrate that this method outperforms baseline models, with improvements of 3.4%, 1.6%, 2.9%, and 3.4% in MAP, Recall, NDCG, and MRR metrics, respectively.