Residual Reward Models: Leveraging Prior Knowledge for Efficient Preference-based Reinforcement Learning in Robotics

arXiv:2507.00611v2 Announce Type: replace-cross Abstract: Preference-based Reinforcement Learning (PbRL) provides a promising alternative to heuristic reward design in complex robotic environments. However, PbRL often suffers from poor sample efficiency, requiring extensive and costly human feedback, which limits its real-world applicability. Prior work has proposed learning a reward model from demonstrations and fine-tuning it using preferences. However, when the model is a neural network, transitioning between different loss functions across training phases often leads to unstable optimization and performance degradation. In this paper, we propose a method to effectively leverage prior knowledge with a Residual Reward Model (RRM). An RRM assumes that the true reward of the environment can be split into a sum of two parts: a prior reward and a learned reward. The prior reward is a term available before training, such as an engineering heuristic ``best guess'', a language-generated reward, or a reward function learned from inverse reinforcement learning, and the learned reward is then trained with preferences as a residual offset. Experimental results in Meta-World and DM-Control show that RRMs substantially improve the sample efficiency of common PbRL methods across various prior reward types. Furthermore, we demonstrate the practical efficacy of our method on a physical Franka Panda robot, accelerating policy learning and achieving high success rates in fewer steps than baselines.

Abstract:Preference-based Reinforcement Learning (PbRL) provides a promising alternative to heuristic reward design in complex robotic environments. However, PbRL often suffers from poor sample efficiency, requiring extensive and costly human feedback, which limits its real-world applicability. Prior work has proposed learning a reward model from demonstrations and fine-tuning it using preferences. However, when the model is a neural network, transitioning between different loss functions across training phases often leads to unstable optimization and performance degradation. In this paper, we propose a method to effectively leverage prior knowledge with a Residual Reward Model (RRM). An RRM assumes that the true reward of the environment can be split into a sum of two parts: a prior reward and a learned reward. The prior reward is a term available before training, such as an engineering heuristic ``best guess'', a language-generated reward, or a reward function learned from inverse reinforcement learning, and the learned reward is then trained with preferences as a residual offset. Experimental results in Meta-World and DM-Control show that RRMs substantially improve the sample efficiency of common PbRL methods across various prior reward types. Furthermore, we demonstrate the practical efficacy of our method on a physical Franka Panda robot, accelerating policy learning and achieving high success rates in fewer steps than baselines.

Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)

arXiv:2507.00611 [cs.LG]

(or arXiv:2507.00611v2 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2507.00611 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

aiqbench analysis

Residual Reward Models: Leveraging Prior Knowledge for Efficient Preference-based Reinforcement Learning in Robotics

Related coverage