@inproceedings{gupta-etal-2025-advsumm,
    title = "{A}dv{S}umm: Adversarial Training for Bias Mitigation in Text Summarization",
    author = "Gupta, Mukur  and
      Varimalla, Nikhil Reddy  and
      Deas, Nicholas  and
      Subbiah, Melanie  and
      McKeown, Kathleen",
    editor = "Dong, Yue  and
      Xiao, Wen  and
      Zhang, Haopeng  and
      Zhang, Rui  and
      Ernst, Ori  and
      Wang, Lu  and
      Liu, Fei",
    booktitle = "Proceedings of The 5th New Frontiers in Summarization Workshop",
    month = nov,
    year = "2025",
    address = "Hybrid",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-emnlp/2025.newsum-main.12/",
    pages = "172--182",
    ISBN = "979-8-89176-337-1",
    abstract = "Large Language Models (LLMs) have achieved impressive performance in text summarization and are increasingly deployed in real-world applications. However, these systems often inherit associative and framing biases from pre-training data, leading to inappropriate or unfair outputs in downstream tasks. In this work, we present AdvSumm (Adversarial Summarization), a domain-agnostic training framework designed to mitigate bias in text summarization through improved generalization. Inspired by adversarial robustness, AdvSumm introduces a novel Perturber component that applies gradient-guided perturbations at the embedding level of Sequence-to-Sequence models, enhancing the model{'}s robustness to input variations. We empirically demonstrate that AdvSumm effectively reduces different types of bias in summarization{---}specifically, name-nationality bias and political framing bias{---}without compromising summarization quality. Compared to standard transformers and data augmentation techniques like back-translation, AdvSumm achieves stronger bias mitigation performance across benchmark datasets."
}