@inproceedings{han-etal-2026-guardrails, title = "Guardrails without Shackles: Mitigating Over-Defense in Inference-Time Detection of Fixed-Target Backdoors for {LLM}s", author = "Han, Bingxu and Wang, Fang and Zhang, Bo and Xia, Xingyou and Zhang, Jun and Cao, Yunbo and Luo, Zhunchen and Bai, Xiaoying and Geng, Guotong and Wang, Mingqiang", editor = "Lee, Hung-Yi and Dev, Sunipa and Li, Jessy and Elliott, Desmond", booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026", month = oct, year = "2026", address = "Budapest, Hungary", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2026.findings-emnlp.1998/", pages = "42584--42607", ISBN = "979-8-89176-462-0" }