← 研踪 PaperTrailarXiv:2502.18770 · v9
Reward Shaping to Mitigate Reward Hacking in RLHF
首次提交 2025-02-26 10:57
版本更新 2026-11-11 00:12
雷达观测 2026-11-11 13:07
摘要
Reinforcement learning from human feedback (RLHF) is widely used to align large language models with human preferences, but it remains vulnerable to \emph{reward hacking}, whereby a policy exploits flaws in the reward function at the expense of the intended behavior. Reward shaping can stabilize training and partially mitigate reward hacking, yet its design principles remain underexplored. Our systematic study of common shaping methods suggests that shaped rewards should be bounded and should rise rapidly before gradually saturating. We introduce \emph{Preference as Reward} (PAR), which applies a sigmoid to centered rewards to produce a signal that reflects the reward model's latent preferences. We show that PAR has a variance-reduction property that stabilizes training and substantially widens the window for effective early stopping. Using Proximal Policy Optimization (PPO) and Gemma2-2B, PAR consistently outperforms the other shaping methods evaluated and demonstrates high data efficiency and robustness. Comparisons with unshaped rewards across four base models and four reinforcement learning algorithms show that PAR is effective across models and particularly beneficial for actor--critic methods when value estimates become unstable. The code is available at https://github.com/PorUna-byte/PAR.
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v9 · 2026-11-11 00:12 · 最新观测版本
Reward Shaping to Mitigate Reward Hacking in RLHF
Reinforcement learning from human feedback (RLHF) is widely used to align large language models with human preferences, but it remains vulnerable to \emph{reward hacking}, whereby a policy exploits flaws in the reward function at the expense of the intended behavior. Reward shaping can stabilize training and partially mitigate reward hacking, yet its design principles remain underexplored. Our systematic study of common shaping methods suggests that shaped rewards should be bounded and should rise rapidly before gradually saturating. We introduce \emph{Preference as Reward} (PAR), which applies a sigmoid to centered rewards to produce a signal that reflects the reward model's latent preferences. We show that PAR has a variance-reduction property that stabilizes training and substantially widens the window for effective early stopping. Using Proximal Policy Optimization (PPO) and Gemma2-2B, PAR consistently outperforms the other shaping methods evaluated and demonstrates high data efficiency and robustness. Comparisons with unshaped rewards across four base models and four reinforcement learning algorithms show that PAR is effective across models and particularly beneficial for actor--critic methods when value estimates become unstable. The code is available at https://github.com/PorUna-byte/PAR.
查看此版本解读 →查看此版本原文 ↗