Directional Alignment Mitigates Reward Hacking in Reinforcement Learning for Language Models 文章

ArXiv CS.CL2026-05-26NEWSen作者: Wenlong Deng, Jiaji Huang, Kaan Ozkara, Yushu Li, Christos Thrampoulidis, Xiaoxiao Li, Youngsuk Park
查看原文 →
关系图谱

Directional Alignment Mitigates Reward Hacking in Reinforcement Learning for Language Models · 相关技术

相关技术

language modelreinforcement learningreward hackingtrusted-direction projection