@article{learningtoundorollbackaugmentedreinforce, title = {Learning to Undo: Rollback-Augmented Reinforcement Learning with Reversibility Signals}, author = {Andrejs Sorstkins and Omer Tariq and Muhammad Bilal}, year = {2025}, eprint = {2510.14503}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2510.14503}, }