💡Model Description

Official model repository for our ACL 2026 Main Conference paper "Language on Demand, Knowledge at Core: Composing LLMs with Encoder-Decoder Translation Models for Extensible Multilinguality".

✨XBridge-base

XBridge-base is trained with stage 1 (cross-model alignment) using trilingual translation data, composing LLaMA3-8B with NLLB-200-1.3B. Training is conducted on 10 languages:

Bn, De, En, Es, Fr, Ja, Ru, Sw, Th, Zh

Despite being trained on a limited set of languages, we observe in our analysis that stage 1 learns a language-agnostic cross-model alignment, which generalizes well beyond the seen languages.

✨XBridge-SFT

XBridge-SFT further extends XBridge-base by training stage 2 (encoder-side adaptation) and stage 3 (decoder-side adaptation) for multilingual mathematical reasoning dataset. We evaluate XBridge-SFT on MGSM benchmark in our paper.

See our paper and github repository for more details!

🌍BayLing-MLingual

XBridge serves as the research foundation of BayLing-MLingual. BayLing-MLingual extends XBridge from a research setting to practical multilingual question answering across 50 languages and 2500 cross-lingual pairs. Try our BayLing-MLingual for general QA among 50 languages!

👉Code: https://github.com/BayLing-Models/BayLing-MLingual

👉Model: https://huggingface.co/BayLing-Models/BayLing-MLingual

📚Citation

If you find this model or our work useful, please cite:

@misc{bu2026languagedemandknowledgecore,
      title={Language on Demand, Knowledge at Core: Composing LLMs with Encoder-Decoder Translation Models for Extensible Multilinguality}, 
      author={Mengyu Bu and Yang Feng},
      year={2026},
      eprint={2603.17512},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2603.17512}, 
}