一作ICLR 2026Poster· Social Aspects / Trustworthy Machine Learning
Reinforcement Unlearning via Group Relative Policy Optimization
Efstratios Zaradoukas · Bardh Prenkaj · Gjergji Kasneci
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Efstratios Zaradoukas · Bardh Prenkaj · Gjergji Kasneci
Yuxiao Li · Alina Fastowski · Efstratios Zaradoukas · Bardh Prenkaj · Gjergji Kasneci
Activation steering has emerged as a powerful tool to shape LLM behaviour without the need for weight updates. While its inherent brittleness and unreliability are well-documented, its safety implications remain underexplored. In this work, we present a systematic safety audit of steering vectors...