arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.04248cs.AIcs.CL
收藏

经验-MCTS:通过双经验蒙特卡洛树搜索实现连续智能体进化

Empirical-MCTS: Continuous Agent Evolution via Dual-Experience Monte Carlo Tree Search

JianChengXingYun Technology Co., Ltd.(健成星云科技有限公司)

Hao Lu, Haoyuan Huang, Yulin Zhou, Chen Li, Ningxin Zhu

AI总结:

经验-MCTS通过双环框架实现连续智能体进化,结合局部探索与全局记忆优化,提升复杂推理任务性能。

摘要:

在推理时间扩展策略,特别是蒙特卡洛树搜索(MCTS),显著增强了大语言模型(LLMs)的推理能力。然而,当前方法主要处于无状态,每次问题实例后丢弃成功的推理模式,无法模仿人类问题解决中经验积累的特征。为弥合这一差距,我们引入了经验-MCTS,一种双环框架,将无状态搜索转变为连续、非参数化的学习过程。该框架通过两种新机制统一局部探索与全局记忆优化:配对经验进化元提示(PE-EMP)和记忆优化代理。PE-EMP在局部搜索中作为反思优化器,利用配对反馈动态合成适应性标准并实时进化元提示(系统提示)。同时,记忆优化代理管理一个全局存储库作为动态策略先验,利用原子操作在问题间提炼高质量见解。在复杂推理基准测试中,包括AIME25、ARC-AGI-2和MathArena Apex的广泛评估显示,经验-MCTS显著优于无状态MCTS策略和独立经验驱动代理。这些结果强调了将结构化搜索与经验积累结合对于掌握复杂、开放性推理任务的重要性。

英文摘要:

Inference-time scaling strategies, particularly Monte Carlo Tree Search (MCTS), have significantly enhanced the reasoning capabilities of Large Language Models (LLMs). However, current approaches remain predominantly stateless, discarding successful reasoning patterns after each problem instance and failing to mimic the empirical accumulation of wisdom characteristic of human problem-solving. To bridge this gap, we introduce Empirical-MCTS, a dual-loop framework that transforms stateless search into a continuous, non-parametric learning process. The framework unifies local exploration with global memory optimization through two novel mechanisms: Pairwise-Experience-Evolutionary Meta-Prompting (PE-EMP) and a Memory Optimization Agent. PE-EMP functions as a reflexive optimizer within the local search, utilizing pairwise feedback to dynamically synthesize adaptive criteria and evolve meta-prompts (system prompts) in real-time. Simultaneously, the Memory Optimization Agent manages a global repository as a dynamic policy prior, employing atomic operations to distill high-quality insights across problems. Extensive evaluations on complex reasoning benchmarks, including AIME25, ARC-AGI-2, and MathArena Apex, demonstrate that Empirical-MCTS significantly outperforms both stateless MCTS strategies and standalone experience-driven agents. These results underscore the critical necessity of coupling structured search with empirical accumulation for mastering complex, open-ended reasoning tasks.

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。

↑