Qwen2.5-1.5B Continually Pretrained on Mongolian (Traditional Mongolian Script)

This model is a continual pretraining (CPT) checkpoint built by further pretraining Qwen2.5 1.5B on the Mongolian (Traditional Mongolian Script) portion of the MC^2 Corpus.

The model is intended to improve Mongolian (Traditional Mongolian Script) language modeling and to support research on low-resource language adaptation.

Training details and methodology are described in: "Efficient Low-Resource Language Adaptation via Multi-Source Dynamic Logit Fusion" (ACL 2026).

Training Data

Corpus: Mongolian subset of MC^2 Corpus
Language: Mongolian (mn, Traditional Mongolian Script)
Training paradigm: Continual pretraining (CPT) starting from Qwen2.5-1.5B

Intended Use

This checkpoint is released primarily for research purposes. Researchers are welcome to use this CPT checkpoint as a base model for future work, particularly in model merging and logit fusion.

Citation

If you use this model, please cite:

@article{zhang2026efficient,
  title={Efficient Low-Resource Language Adaptation via Multi-Source Dynamic Logit Fusion},
  author={Zhang, Chen and Lin, Jiuheng and Liao, Zhiyuan and Feng, Yansong},
  journal={arXiv preprint arXiv:2604.18106},
  year={2026}
}