This is the code for EMNLP2022 paper Adaptive Token-level Cross-lingual Feature Mixing for Multilingual Neural Machine Translation, we will release the code in the near future.