This paper introduces the Hierarchical Diffuser, a simple and effective planning method that combines the advantages of hierarchical and diffusion-based planning. The method addresses the computational challenges and generalization issues of diffusion-based planning by introducing a "jumpy" planning strategy at the higher level, which allows for a larger receptive field at a lower computational cost. The hierarchical approach involves two diffusers: one for high-level subgoal generation and another for low-level subgoal achievement. The high-level diffuser generates subgoals, while the low-level diffuser refines these subgoals into detailed action sequences. This framework enables efficient planning by reducing the computational burden and improving the model's ability to generalize to new tasks. The Hierarchical Diffuser is evaluated on standard offline reinforcement learning benchmarks, demonstrating superior performance and efficiency compared to non-hierarchical diffusion-based methods and other hierarchical planning approaches. The method is particularly effective in tasks with long horizons and sparse rewards, where traditional methods struggle. The paper also explores the generalization capabilities of the model on compositional out-of-distribution tasks, showing that the hierarchical approach enhances the model's ability to adapt to new scenarios. Theoretical analysis is provided, highlighting the trade-offs between the receptive field size and generalization performance. The results show that a larger receptive field improves the model's ability to capture data distribution but may reduce its generalization capability. The Hierarchical Diffuser is also shown to be more computationally efficient than traditional methods, with significant speed improvements in both training and planning phases. The paper concludes that the Hierarchical Diffuser offers a promising approach to planning in complex environments, combining the strengths of hierarchical planning with the flexibility of diffusion-based methods. The method is particularly effective in tasks requiring long-term planning and generalization to new scenarios. The results demonstrate that the Hierarchical Diffuser outperforms existing methods in terms of performance, efficiency, and generalization capabilities.This paper introduces the Hierarchical Diffuser, a simple and effective planning method that combines the advantages of hierarchical and diffusion-based planning. The method addresses the computational challenges and generalization issues of diffusion-based planning by introducing a "jumpy" planning strategy at the higher level, which allows for a larger receptive field at a lower computational cost. The hierarchical approach involves two diffusers: one for high-level subgoal generation and another for low-level subgoal achievement. The high-level diffuser generates subgoals, while the low-level diffuser refines these subgoals into detailed action sequences. This framework enables efficient planning by reducing the computational burden and improving the model's ability to generalize to new tasks. The Hierarchical Diffuser is evaluated on standard offline reinforcement learning benchmarks, demonstrating superior performance and efficiency compared to non-hierarchical diffusion-based methods and other hierarchical planning approaches. The method is particularly effective in tasks with long horizons and sparse rewards, where traditional methods struggle. The paper also explores the generalization capabilities of the model on compositional out-of-distribution tasks, showing that the hierarchical approach enhances the model's ability to adapt to new scenarios. Theoretical analysis is provided, highlighting the trade-offs between the receptive field size and generalization performance. The results show that a larger receptive field improves the model's ability to capture data distribution but may reduce its generalization capability. The Hierarchical Diffuser is also shown to be more computationally efficient than traditional methods, with significant speed improvements in both training and planning phases. The paper concludes that the Hierarchical Diffuser offers a promising approach to planning in complex environments, combining the strengths of hierarchical planning with the flexibility of diffusion-based methods. The method is particularly effective in tasks requiring long-term planning and generalization to new scenarios. The results demonstrate that the Hierarchical Diffuser outperforms existing methods in terms of performance, efficiency, and generalization capabilities.