arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.10539cs.LG
收藏

残差强化学习中价值学习为何高效?

What Makes Value Learning Efficient in Residual Reinforcement Learning?

Nanyang Technical University(南洋理工大学);Mila - Quebec AI Institute(魁北克AI研究所);National University of Singapore(新加坡国立大学)

Guozheng Ma, Lu Li, Haoyu Wang, Zixuan Liu, Pierre-Luc Bacon, Dacheng Tao

AI总结:

DAWN通过数据锚定预热和归一化解决残差强化学习中价值学习的冷启动和尺度不匹配问题,提升效率。

摘要:

残差强化学习(RL)通过冻结基础策略并仅学习有限的修正,实现了对表达能力强的预训练策略的稳定在线细化。然而,残差RL中的价值学习面临着独特的挑战,这些挑战仍然缺乏深入理解。在本文中,我们识别出两个关键瓶颈:冷启动病态,即批评者缺乏对基础策略周围价值景观的知识;以及结构尺度不匹配,即残差贡献被基础动作所淹没。通过系统研究,我们揭示了这些瓶颈的机制,发现简单的但原则性的解决方案足够:基础策略转移为隐式预热提供了必要的价值锚点,而批评者归一化有效恢复了表示敏感性以区分价值差异。基于这些见解,我们提出了DAWN(数据锚定预热和归一化),一种针对残差RL中高效价值学习的最小方法。通过解决这些瓶颈,DAWN在多样化的基准、策略架构和观测模态中展示了显著的效率提升。

英文摘要:

Residual reinforcement learning (RL) enables stable online refinement of expressive pretrained policies by freezing the base and learning only bounded corrections. However, value learning in residual RL poses unique challenges that remain poorly understood. In this work, we identify two key bottlenecks: cold start pathology, where the critic lacks knowledge of the value landscape around the base policy, and structural scale mismatch, where the residual contribution is dwarfed by the base action. Through systematic investigation, we uncover the mechanisms underlying these bottlenecks, revealing that simple yet principled solutions suffice: base-policy transitions serve as an essential value anchor for implicit warmup, and critic normalization effectively restores representation sensitivity for discerning value differences. Based on these insights, we propose DAWN (Data-Anchored Warmup and Normalization), a minimal approach targeting efficient value learning in residual RL. By addressing these bottlenecks, DAWN demonstrates substantial efficiency gains across diverse benchmarks, policy architectures, and observation modalities.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑