A KL-regularization Framework for Learning to Plan with Adaptive Priors

📰 ArXiv cs.AI

Learn to plan with adaptive priors using KL-regularization framework for improved sample efficiency in model-based reinforcement learning

advanced Published 23 May 2026

Action Steps

Implement KL-regularization framework to learn adaptive priors for planning
Update the sampling policy jointly with the planner distribution using MPPI planning
Evaluate the framework on high-dimensional continuous control tasks to assess sample efficiency
Compare the performance of the proposed framework with existing methods
Apply the framework to real-world control tasks to demonstrate its effectiveness

Who Needs to Know This

Researchers and engineers working on model-based reinforcement learning and continuous control tasks can benefit from this framework to improve sample efficiency

Key Insight

💡 KL-regularization framework can be used to learn adaptive priors for planning, improving sample efficiency in model-based reinforcement learning

Key Takeaways

Learn to plan with adaptive priors using KL-regularization framework for improved sample efficiency in model-based reinforcement learning

Full Article

Title: A KL-regularization Framework for Learning to Plan with Adaptive Priors

Abstract:
arXiv:2510.04280v2 Announce Type: replace-cross Abstract: Effective exploration remains a central challenge in model-based reinforcement learning (MBRL), particularly in high-dimensional continuous control tasks where sample efficiency is crucial. A prominent line of recent work leverages learned policies as proposal distributions for Model-Predictive Path Integral (MPPI) planning. Initial approaches update the sampling policy independently of the planner distribution, typically maximizing a lea

Read full paper → ← Back to Reads