@article{mitigatingadaptiveattacksagainstreasonin, title = {Mitigating Adaptive Attacks against Reasoning Models with Activation Consistency Training}, author = {Avidan Shah and Jannik Brinkmann and Rico Angell}, year = {2026}, eprint = {2605.28467}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2605.28467}, }