@inproceedings{zheng-etal-2025-asymmetric,
    title = "Asymmetric Conflict and Synergy in Post-training for {LLM}-based Multilingual Machine Translation",
    author = "Zheng, Tong  and
      Wen, Yan  and
      Bao, Huiwen  and
      Guo, Junfeng  and
      Huang, Heng",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingestion-acl-25/2025.findings-acl.944/",
    pages = "18362--18383",
    ISBN = "979-8-89176-256-5",
    abstract = "The emergence of Large Language Models (LLMs) has advanced the multilingual machine translation (MMT), yet the Curse of Multilinguality (CoM) remains a major challenge. Existing work in LLM-based MMT typically mitigates this issue via scaling up training and computation budget, which raises a critical question: Is scaling up the training and computation budget truly necessary for high-quality MMT, or can a deeper understanding of CoM provide a more efficient solution? To explore this problem, we analyze the linguistic conflicts and synergy, the underlying mechanism of CoM during post-training phase. We identify an asymmetric phenomenon in linguistic conflicts and synergy: the dominance of conflicts and synergy varies in different translation directions, leading to sub-optimal adaptation in existing post-training methods. We further find that a significant bottleneck in MMT appears to lie in post-training rather than multilingual pre-training, suggesting the need for more effective adaptation strategies. Building on these new insights, we propose a direction-aware training approach, combined with group-wise model merging, to address asymmetry in linguistic conflicts and synergy explicitly. Leveraging this strategy, our method fine-tunes X-ALMA-13B-Pretrain{---}trained only with multilingual pre-training{---}achieving comparable performance to XALMA-13B (only SFT) while using only 20B pretraining tokens and 17B parameters{---}5.5{\texttimes} fewer pretraining-tokens and 1.7x fewer model size{---}with just 0.85 COMET drop on Flores-200 testsets of 50 languages."
}