@inproceedings{shin-etal-2026-dynamixsft,
    title = "{D}ynamix{SFT}: Dynamic Mixture Optimization of Instruction Tuning Collections",
    author = "Shin, Haebin  and
      Ji, Lei  and
      Liu, Xiao  and
      Yu, Zhiwei  and
      Yoo, Hyunwoo  and
      Chen, Qi  and
      Gong, Yeyun",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.findings-acl.1972/",
    pages = "39590--39603",
    ISBN = "979-8-89176-395-1",
    abstract = {As numerous instruction-tuning datasets continue to emerge, dynamically balancing and optimizing their mixtures has become a criticalchallenge. To address this, we propose DynamixSFT, a dynamic and automated method for instruction-tuning dataset mixture optimization. We formulate the problem as a multi-armed bandit setup and introduce a Prior-scaled Boltzmann Exploration that softly anchors the updated sampling distribution to the original dataset proportions, thereby preserving the inherent diversity and coverage of the collection. Sampling probabilities are updated using a lightweight 1-Step Look-ahead Reward, reflecting how much the dataset contributes to improving the model{'}s performance at its current state. We demonstrate that DynamixSFT effectively optimizes the T{\"U}LU-2-mixture andT{\"U}LU-3-mixture collections across 10 benchmarks, while introducing minimal computational overhead over naive sampling. Furthermore, we provide a comprehensive analysis and visualizations to offer deeper insights into the adaptive dynamics of our method.}
}