arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2511.02424cs.AI
收藏

ReAcTree:具有控制流的分层LLM代理树用于长时间任务规划

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

ETRI & UST(ETRI与UST);UST

Jae-Woo Choi, Hyungmin Kim, Hyobin Ong, Youngwoo Yoon, Minsu Jang, Dohyung Kim, Jaehong Kim

AI总结:

ReAcTree通过分层代理树和控制流实现复杂任务规划,显著提升长周期任务的完成效率和成功率。

摘要:

近期大型语言模型(LLMs)的进步使嵌入式自主代理的决策和任务规划取得了显著进展。然而,现有大多数方法在处理复杂、长周期任务时存在困难,因为它们依赖于一个单一的轨迹,将所有过去的决策和观察纠缠在一起,以解决整个任务。为了解决这一限制,我们提出了ReAcTree,一种分层任务规划方法,将复杂目标分解为可管理的子目标,以在动态构建的代理树中处理。每个子目标由一个LLM代理节点处理,该节点能够推理、行动并进一步扩展树,同时控制流节点协调代理节点的执行策略。此外,我们集成了两种互补的记忆系统:每个代理节点从事件性记忆中检索目标特定的、子目标级别的示例,并通过工作记忆共享环境特定的观察。在WAH-NL和ALFRED上的实验表明,ReAcTree在各种LLMs上均优于ReAct等强大的任务规划基线。值得注意的是,在WAH-NL上,ReAcTree使用Qwen 2.5 72B实现了61%的目标成功率,几乎将ReAct的31%翻倍。代码可在https://github.com/Choi-JaeWoo/ReAcTree.git获取。

英文摘要:

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.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑