Learning and Reusing Policy Decompositions for Hierarchical Generalized Planning with LLM Agents
For LLM-based agents, this work improves task accuracy and efficiency by integrating classical planning concepts, showing strong gains on unseen tasks.
The paper presents HCL-GP, a method combining generalized planning and hierarchical decomposition for LLM agents, achieving 98.2% accuracy on normal tasks and 97.8% on challenge tasks in the AppWorld benchmark, improving 15.8 points over static synthesis on challenging scenarios.
We present a dynamic policy-learning approach that combines generalized planning and hierarchical task decomposition for LLM-based agents. Our method, Hierarchical Component Learning for Generalized Policies (HCL-GP ), learns parameterized policies that generalize across task instances and automatically extracts reusable components from successful executions, organizing them into a component library for compositional policy generation. We address three challenges: (1) learning components through automated decomposition, (2) generalizing components to maximize reuse, and (3) efficient retrieval via semantic search. Evaluated on the AppWorld benchmark, our approach achieves 98.2% accuracy on normal tasks and 97.8% on challenge tasks with unseen applications, improving 15.8 points over static synthesis on challenging scenarios. For open-source models, dynamic reuse enables 62.5% success versus near-zero without reuse. This demonstrates that classical planning concepts can be effectively integrated with LLM agents for improved accuracy and efficiency.