@inproceedings{farn-etal-2025-safeguard,
    title = "Safeguard Fine-Tuned {LLM}s Through Pre- and Post-Tuning Model Merging",
    author = "Farn, Hua  and
      Su, Hsuan  and
      H. Kumar, Shachi  and
      Sahay, Saurav  and
      Chen, Shang-Tse  and
      Lee, Hung-yi",
    editor = "Christodoulopoulos, Christos  and
      Chakraborty, Tanmoy  and
      Rose, Carolyn  and
      Peng, Violet",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/name-variant-enfa-fane/2025.findings-emnlp.901/",
    doi = "10.18653/v1/2025.findings-emnlp.901",
    pages = "16589--16602",
    ISBN = "979-8-89176-335-7",
    abstract = "Fine-tuning large language models (LLMs) for downstream tasks often leads to catastrophic forgetting, notably degrading the safety of originally aligned models. While some existing methods attempt to restore safety by incorporating additional safety data, the quality of such data typically falls short of that used in the original alignment process. Moreover, these high-quality safety datasets are generally inaccessible, making it difficult to fully recover the model{'}s original safety. We ask: How can we preserve safety while improving downstream task performance without additional safety data? We show that simply merging the weights of pre- and post-fine-tuned models effectively mitigates safety degradation while enhancing performance. Experiments across different downstream tasks and models validate the method{'}s practicality and effectiveness."
}