@inproceedings{li-etal-2025-logits,
    title = "Logits-Based Finetuning",
    author = "Li, Jingyao  and
      Yang, Senqiao  and
      Wu, Sitong  and
      Shi, Han  and
      Zheng, Chuanyang  and
      Xu, Hong  and
      Jia, Jiaya",
    editor = "Christodoulopoulos, Christos  and
      Chakraborty, Tanmoy  and
      Rose, Carolyn  and
      Peng, Violet",
    booktitle = "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-emnlp/2025.emnlp-main.745/",
    pages = "14763--14775",
    ISBN = "979-8-89176-332-6",
    abstract = "In recent years, developing compact and efficient large language models (LLMs) has emerged as a thriving area of research. However, traditional Supervised Fine-Tuning (SFT), which relies on singular ground truth labels, often fails to capture token-level dependencies and linguistic diversity. To address these limitations, we propose a logits-based fine-tuning framework that integrates the strengths of supervised learning and knowledge distillation. Our approach constructs enriched training targets by combining teacher logits with ground truth labels, preserving both correctness and linguistic diversity. This ensures more reliable and effective training. To validate our approach, we constructed a large-scale 1.2M logits dataset and trained a series of science-focused models. Experimental results demonstrate that our method achieves significant improvements over current SOTA, with accuracy gains of 18{\%} on Mawps and 22.7{\%} on TabMWP. Across nine widely used mathematical benchmarks, our method consistently outperforms prior SFT models, achieving an average improvement of 7.28{\%}. All code and datasets will be open-sourced."
}