MPO: Boosting LLM Agents with Meta Plan Optimization
This addresses planning inefficiencies for LLM agents in interactive tasks, offering a plug-and-play solution for improved efficiency and generalization, though it appears incremental as it builds on existing agent frameworks.
The paper tackles planning hallucinations and retraining needs in LLM-based agents by proposing the Meta Plan Optimization (MPO) framework, which uses high-level guidance and feedback to enhance planning, resulting in significant performance improvements over baselines in experiments on two tasks.
Recent advancements in large language models (LLMs) have enabled LLM-based agents to successfully tackle interactive planning tasks. However, despite their successes, existing approaches often suffer from planning hallucinations and require retraining for each new agent. To address these challenges, we propose the Meta Plan Optimization (MPO) framework, , which enhances agent planning capabilities by directly incorporating explicit guidance. Unlike previous methods that rely on complex knowledge, which either require significant human effort or lack quality assurance, MPO leverages high-level general guidance through meta plans to assist agent planning and enables continuous optimization of the meta plans based on feedback from the agent's task execution. Our experiments conducted on two representative tasks demonstrate that MPO significantly outperforms existing baselines. Moreover, our analysis indicates that MPO provides a plug-and-play solution that enhances both task completion efficiency and generalization capabilities in previous unseen scenarios.