ReAcTree: Hierarchical LLM Agent Trees with Control Flow for Long-Horizon Task Planning

2025-11-04

AI Summary

To address decision coupling and historical information overload in large language model (LLM) agents tackling complex, long-horizon tasks—stemming from single-trajectory modeling—this paper proposes ReAcTree. The framework constructs a dynamic hierarchical agent tree that recursively decomposes global goals into subgoal nodes and incorporates control-flow nodes to orchestrate execution policies. It further introduces a dual-memory system—comprising episodic memory and working memory—to enable cross-node experience retrieval and observation sharing. This design effectively decouples temporal dependencies in long-horizon planning, enhancing both planning robustness and execution efficiency. Evaluated on WAH-NL and ALFRED benchmarks, ReAcTree significantly outperforms baselines including ReAct. On WAH-NL, it achieves a 61% task completion rate using Qwen2.5-72B, nearly doubling ReAct’s 31%.

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
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Recent advancements in large language models (LLMs) have enabled significant progress in decision-making and task planning for embodied autonomous agents. However, most existing methods still struggle with complex, long-horizon tasks because they rely on a monolithic trajectory that entangles all past decisions and observations, attempting to solve the entire task in a single unified process. To address this limitation, we propose ReAcTree, a hierarchical task-planning method that decomposes a complex goal into more manageable subgoals within a dynamically constructed agent tree. Each subgoal is handled by an LLM agent node capable of reasoning, acting, and further expanding the tree, while control flow nodes coordinate the execution strategies of agent nodes. In addition, we integrate two complementary memory systems: each agent node retrieves goal-specific, subgoal-level examples from episodic memory and shares environment-specific observations through working memory. Experiments on the WAH-NL and ALFRED datasets demonstrate that ReAcTree consistently outperforms strong task-planning baselines such as ReAct across diverse LLMs. Notably, on WAH-NL, ReAcTree achieves a 61% goal success rate with Qwen 2.5 72B, nearly doubling ReAct's 31%.

Problem

• Hierarchical decomposition of complex long-horizon tasks
• Dynamic agent tree construction with control flow coordination
• Integration of episodic and working memory systems

Innovation

• Hierarchical agent tree decomposes complex goals
• Control flow nodes coordinate agent execution strategies
• Episodic and working memory systems enhance reasoning