diallm-gemma-gspo-ind

DiaLLM — Gemma 3-4B-it — en-IN — GSPO

Built with Gemma.

Part of DiaLLM: An Investigation into the Robustness-Generation Gap in English Dialect Adaptation (EMNLP 2026 Main).

• Base model: Gemma 3-4B-it
• Target variety: en-IN (Indian English)
• Adaptation thread: explicit (variety-targeted)
• Alignment method: GSPO

Fine-tuned from jordanpainter/diallm-gemma-sft-ind via GSPO with the target-variety preference pairs.

Code, checkpoints, preference datasets, linguistic-analysis toolkit: https://github.com/surrey-nlp/diallm

Paper: https://arxiv.org/abs/2607.07669

This model is a fine-tuned version of jordanpainter/diallm-gemma-sft-ind, trained using TRL.

Training procedure

This model was trained with GSPO.

Framework versions

• TRL: 0.28.0
• Transformers: 4.57.6
• Pytorch: 2.5.1+cu121
• Datasets: 4.5.0
• Tokenizers: 0.22.2

Citation

@article{painter2026diallm,
  title     = {DiaLLM: An Investigation into the Robustness-Generation Gap in English Dialect Adaptation},
  author    = {Painter, Jordan and Srirag, Dipankar and Kappiyath, Adarsh and Kanojia, Diptesh and Joshi, Aditya and Yin, Lu},
  year      = {2026},
  eprint    = {2607.07669},
  archivePrefix = {arXiv}
}

@article{zheng2025groupsequencepolicyoptimization,
    title  = {{Group Sequence Policy Optimization}},
    author = {Chujie Zheng and Shixuan Liu and Mingze Li and Xiong-Hui Chen and Bowen Yu and Chang Gao and Kai Dang and Yuqiong Liu and Rui Men and An Yang and Jingren Zhou and Junyang Lin},
    year   = 2025,
    eprint = {arXiv:2507.18071}
}

@software{vonwerra2020trl,
  title   = {{TRL: Transformers Reinforcement Learning}},
  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
  license = {Apache-2.0},
  url     = {https://github.com/huggingface/trl},
  year    = {2020}
}