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

TL;DR

ReAcTree introduces a hierarchical LLM-based agent tree with control flow and memory systems to improve long-horizon task planning, significantly outperforming existing methods on complex benchmarks.

Contribution

It presents a novel hierarchical planning framework that decomposes complex tasks into subgoals with dynamic agent trees and integrated memory, enhancing decision-making in long-horizon tasks.

Findings

ReAcTree achieves 61% success rate on WAH-NL, nearly doubling ReAct.

The method outperforms strong baselines across diverse LLMs.

Effective use of episodic and working memory improves task performance.

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 struggle with complex, long-horizon tasks because they rely on a monolithic trajectory that entangles all past decisions and observations 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 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 show 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%. The code is available at https://github.com/Choi-JaeWoo/ReAcTree.git.