@inproceedings{siu-etal-2025-cosmic,
    title = "{COSMIC}: Generalized Refusal Direction Identification in {LLM} Activations",
    author = "Siu, Vincent  and
      Crispino, Nicholas  and
      Yu, Zihao  and
      Pan, Sam  and
      Wang, Zhun  and
      Liu, Yang  and
      Song, Dawn  and
      Wang, Chenguang",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/display_plenaries/2025.findings-acl.1310/",
    pages = "25534--25553",
    ISBN = "979-8-89176-256-5",
    abstract = "Large Language Models encode behaviors like refusal within their activation space, but identifying these behaviors remains challenging. Existing methods depend on predefined refusal templates detectable in output tokens or manual review. We introduce **COSMIC** (Cosine Similarity Metrics for Inversion of Concepts), an automated framework for direction selection that optimally identifies steering directions and target layers using cosine similarity, entirely independent of output text. COSMIC achieves steering effectiveness comparable to prior work without any prior knowledge or assumptions of a model{'}s refusal behavior such as the use of certain refusal tokens. Additionally, COSMIC successfully identifies refusal directions in adversarial scenarios and models with weak safety alignment, demonstrating its robustness across diverse settings."
}