@article{probebeforeyoutalktowardsblackbox, title = {Probe before You Talk: Towards Black-box Defense against Backdoor Unalignment for Large Language Models}, author = {Biao Yi and Tiansheng Huang and Sishuo Chen and Tong Li and Zheli Liu and Zhixuan Chu and Yiming Li}, year = {2025}, eprint = {2506.16447}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2506.16447v1}, }