@inproceedings{garcia-gilabert-etal-2025-salamandra,
    title = "From {SALAMANDRA} to {SALAMANDRATA}: {BSC} Submission for {WMT}25 General Machine Translation Shared Task",
    author = "Garcia Gilabert, Javier  and
      Liao, Xixian  and
      Da Dalt, Severino  and
      Bohman, Ella  and
      Mash, Audrey  and
      De Luca Fornaciari, Francesca  and
      Baucells, Irene  and
      Llop, Joan  and
      Claramunt, Miguel  and
      Escolano, Carlos  and
      Melero, Maite",
    editor = "Haddow, Barry  and
      Kocmi, Tom  and
      Koehn, Philipp  and
      Monz, Christof",
    booktitle = "Proceedings of the Tenth Conference on Machine Translation",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-emnlp/2025.wmt-1.37/",
    pages = "614--637",
    ISBN = "979-8-89176-341-8",
    abstract = "In this paper, we present the SalamandraTA family of models, an improved iteration of Salamandra LLMs (Gonzalez-Agirre et al., 2025) specifically trained to achieve strong performance in translation-related tasks for 38 European languages. SalamandraTA comes in two scales: 2B and 7B parameters. For both versions, we applied the same training recipe with a first step of continual pre-training on parallel data, and a second step of supervised fine-tuning on high-quality instructions.The BSC submission to the WMT25 General Machine Translation shared task is based on the 7B variant of SalamandraTA. We first extended the model vocabulary to support the additional non-European languages included in the task. This was followed by a second phase of continual pretraining and supervised fine-tuning, carefully designed to optimize performance across all translation directions for this year{'}s shared task. For decoding, we employed two quality-aware strategies: Minimum Bayes Risk Decoding and Translation Reranking using Comet and Comet-kiwi.We publicly release both the 2B and 7B versions of SalamandraTA, along with the newer SalamandraTA-v2 model, on Hugging Face."
}