danish-foundation-models/gemma-3-1b-cpt-dynaword-matched-v1

Model Card for gemma-3-1b-cpt-dynaword-full-v1

This model was trained as a part the Dynaword paper. It was trained as a part of a set of experiments to show the relative improvement of training on the Danish Dynaword.

Citation

If you use this work please cite the paper:

@misc{enevoldsen2025dynawordoneshotcontinuouslydeveloped,
      title={Dynaword: From One-shot to Continuously Developed Datasets}, 
      author={Kenneth Enevoldsen and Kristian Nørgaard Jensen and Jan Kostkan and Balázs Szabó and Márton Kardos and Kirten Vad and Johan Heinsen and Andrea Blasi Núñez and Gianluca Barmina and Jacob Nielsen and Rasmus Larsen and Peter Vahlstrup and Per Møldrup Dalum and Desmond Elliott and Lukas Galke and Peter Schneider-Kamp and Kristoffer Nielbo},
      year={2025},
      eprint={2508.02271},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2508.02271}, 
}