@article{safetytrainingmodulatesharmfulmisalignme, title = {Safety Training Modulates Harmful Misalignment Under On-Policy RL, But Direction Depends on Environment Design}, author = {Leon Eshuijs and Shihan Wang and Antske Fokkens}, year = {2026}, eprint = {2604.12500}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2604.12500}, }