@inproceedings{duan-etal-2026-projecting,
    title = "Projecting Out the Malice: A Global Subspace Approach to {LLM} Detoxification",
    author = "Duan, Zenghao  and
      Yin, Zhiyi  and
      Shi, Zhichao  and
      Pang, Liang  and
      Jing, Shaoling  and
      Huang, Zihe  and
      Wu, Jiayi  and
      Yan, Yu  and
      Deng, Jingcheng  and
      Shen, Huawei  and
      Cheng, Xueqi",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-long.1652/",
    pages = "35697--35719",
    ISBN = "979-8-89176-390-6",
    abstract = "Large language models (LLMs) exhibit exceptional performance but pose inherent risks of generating toxic content, restricting their safe deployment. While traditional methods (e.g., alignment) adjust output preferences, they fail to eliminate underlying toxic regions in parameters, leaving models vulnerable to adversarial attacks. Prior mechanistic studies characterize toxic regions as ``toxic vectors'' or ``layer-wise subspaces'', yet our analysis identifies critical limitations: i) Removed toxic vectors can be reconstructed via linear combinations of non-toxic vectors, demanding targeting of entire toxic subspace; ii) Contrastive objective over limited samples inject noise into layer-wise subspaces, hindering stable extraction. These highlight the challenge of identifying robust toxic subspace and removing them. Therefore, we propose GLOSS (GLobal tOxic Subspace Suppression), a lightweight method that mitigates toxicity by identifying and eliminating this global subspace from FFN parameters. Experiments on LLMs (e.g., Qwen3) show GLOSS achieves SOTA detoxification while preserving general capabilities without requiring large-scale retraining."
}