StackPlanner: A Centralized Hierarchical Multi-Agent System with Task-Experience Memory Management

2026-01-09

arXiv.org

AI Summary

This work addresses the challenges of long-horizon multi-agent collaboration in centralized systems, where ineffective memory management often leads to context bloat, error accumulation, and poor cross-task generalization. To overcome these limitations, we propose StackPlanner, a hierarchical framework that decouples high-level coordination from low-level task execution. StackPlanner introduces, for the first time, task-level active memory control combined with a reinforcement learning–based structured experience memory mechanism, enabling efficient retrieval and reuse of reusable collaborative experiences. Experimental results across multiple multi-agent benchmark tasks demonstrate that our approach significantly enhances the stability, efficiency, and generalization capability of long-horizon collaborative performance.

Technology Category

• Multiagent Systems: Multiagent Planning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Multi-agent systems based on large language models, particularly centralized architectures, have recently shown strong potential for complex and knowledge-intensive tasks. However, central agents often suffer from unstable long-horizon collaboration due to the lack of memory management, leading to context bloat, error accumulation, and poor cross-task generalization. To address both task-level memory inefficiency and the inability to reuse coordination experience, we propose StackPlanner, a hierarchical multi-agent framework with explicit memory control. StackPlanner addresses these challenges by decoupling high-level coordination from subtask execution with active task-level memory control, and by learning to retrieve and exploit reusable coordination experience via structured experience memory and reinforcement learning. Experiments on multiple deep-search and agent system benchmarks demonstrate the effectiveness of our approach in enabling reliable long-horizon multi-agent collaboration.

Problem

• multi-agent systems
• memory management
• long-horizon collaboration
• coordination experience
• centralized architecture

Innovation

• hierarchical multi-agent system
• memory management
• experience reuse
• reinforcement learning
• centralized coordination