@article{safetyarithmeticaframeworkfortesttime, title = {Safety Arithmetic: A Framework for Test-time Safety Alignment of Language Models by Steering Parameters and Activations}, author = {Rima Hazra and Sayan Layek and Somnath Banerjee and Soujanya Poria}, year = {2024}, eprint = {2406.11801}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2406.11801v2}, }