MPO: Boosting LLM Agents with Meta Plan Optimization

2025-03-04

AI Summary

Large language model (LLM)-based agents suffer from hallucination, poor generalization, and reliance on task-specific fine-tuning in interactive planning. Method: This paper proposes Meta-Planning Optimization (MPO), a framework that introduces transferable high-level meta-plans as explicit, lightweight guidance—replacing hand-crafted domain knowledge—and enables dynamic, feedback-driven iterative optimization over execution trajectories without retraining. Contribution/Results: MPO achieves zero-shot cross-task transfer, eliminating the need for task-specific adaptation. It is plug-and-play, verifiable, and imposes minimal computational overhead. Evaluated on two representative interactive planning benchmarks, MPO significantly outperforms existing baselines in both task completion efficiency and generalization to unseen scenarios, demonstrating robustness and scalability without sacrificing interpretability or requiring external knowledge injection.

Technology Category

• Multiagent Systems: Multiagent Planning
• Planning, Routing, and Scheduling: Planning with Language Models
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Agentic search

Abstract

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.

Problem

• Enhances LLM agent planning with meta plans
• Reduces planning hallucinations and retraining needs
• Improves task efficiency and generalization in new scenarios

Innovation

• Meta Plan Optimization enhances agent planning.
• MPO uses high-level guidance for continuous optimization.
• Plug-and-play solution improves task efficiency and generalization.