English

Debiased Model-based Representations for Sample-efficient Continuous Control

Machine Learning 2026-05-13 v1 Artificial Intelligence

Abstract

Model-based representations recently stand out as a promising framework that embeds latent dynamics information into the representations for downstream off-policy actor-critic learning. It implicitly combines the advantages of both model-free and model-based approaches while avoiding the training costs associated with model-based methods. Nevertheless, existing model-based representation methods can fail to capture sufficient information about relevant variables and can overfit to early experiences in the replay buffer. These incur biases in representation and actor-critic learning, leading to inferior performance. To address this, we propose Debiased model-based Representations for Q-learning, tagged DR.Q algorithm. DR.Q explicitly maximizes the mutual information between the representations of the current state-action pair and the next state besides minimizing their deviations, and samples transitions with faded prioritized experience replay. We evaluate DR.Q on numerous continuous control benchmarks with a single set of hyperparameters, and the results demonstrate that DR.Q can match or surpass recent strong baselines, sometimes outperforming them by a large margin. Our code is available at https://github.com/dmksjfl/DR.Q.

Keywords

Cite

@article{arxiv.2605.11711,
  title  = {Debiased Model-based Representations for Sample-efficient Continuous Control},
  author = {Jiafei Lyu and Zichuan Lin and Scott Fujimoto and Kai Yang and Yangkun Chen and Saiyong Yang and Zongqing Lu and Deheng Ye},
  journal= {arXiv preprint arXiv:2605.11711},
  year   = {2026}
}

Comments

ICML 2026