arXiv:2607. 22929v1 Announce Type: new Abstract: A short fine-tuning run can undo the safety guards of an open-weight model---retraining a refusal-trained assistant to aid weapons development or produce hate speech.
By Domenic Rosati, Ali Dadsetan, Hong Huang, Xijie Zeng, Hassan Chowdhry, Subhabrata Majumdar, Hassan Sajjad, Frank Rudzicz
arXiv:2607. 13175v1 Announce Type: cross Abstract: Safe reinforcement learning typically enforces safety by bounding expected cumulative costs, a criterion that often fails to detect rare but catastrophic tail events.
By Yassine Chemingui, Chenhua Fan, Honghao Wei, Janardhan Rao Doppa
The paper introduces new evaluation metrics for safe reinforcement learning that go beyond average safety guarantees by examining how often and how severely safety bounds are violated, consistency across tasks and bounds, and the relationship between training-time and final policy behavior. It also proposes a safety tier system for categorizing algorithms and presents empirical safety evaluations on multiple navigation tasks. The authors recommend reporting aggregate metrics, distributional data, and task‑specific results together, and provide an open‑source suite, SafeRLEval, to facilitate reliable safety assessment.
By Lindsay Spoor, Aske Plaat, Thomas Moerland
arXiv:2609.38938v1 Announce Type: new
Abstract: Reinforcement learning with human feedback (RLHF) learns from human comparisons, which can be corrupted or deliberately manipulated. This paper studies...
By Xinyi Ni, Lifeng Lai
arXiv:2605. 00155v3 Announce Type: replace Abstract: Reinforcement learning from human feedback (RLHF) is a central post-training tool for aligning large language models, but its training reward is only a learned proxy for true human utility.
By Yikai Wang, Shang Liu, Jose Blanchet
arXiv:2604. 26360v2 Announce Type: replace-cross Abstract: Reinforcement learning from human feedback (RLHF) systems face a compounding alignment challenge: not only are learned reward models uncertain about unseen state-action pairs, but the human preference annotations they are trained on are themselves inconsistent, context-dependent, and noisy.
By Disha Singha