This is the code repository for Every Language Counts: Learn and Unlearn in Multilingual LLMs. Dataset is avaliable at https://huggingface.co/datasets/TaiMingLu/news-truthful.