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 2502.14848cs.CL
收藏

GATE:跨多样任务的基于图的自适应工具演化

GATE: Graph-based Adaptive Tool Evolution Across Diverse Tasks

Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);School of Artificial Intelligence, University of Chinese Academy of Sciences(中国科学院大学人工智能学院);Microsoft Research Asia(微软亚洲研究院);Nanjing University of Posts and Telecommunications(南京邮电大学);Ricoh Software Research Center (Beijing)(理光软件研究中心(北京))

Jianwen Luo, Yiming Huang, Jinxiang Meng, Fangyu Lei, Shizhu He, Xiao Liu, Shanshan Jiang, Bin Dong, Jun Zhao, Kang Liu

AI总结:

本文提出 GATE 框架,通过层次化工具图在多场景中动态构建和演化可复用工具,并在 Minecraft、智能体与代码生成任务上显著超越现有方法。

摘要:

大语言模型(LLM)在工具制造方面展现出巨大潜力,但现有框架往往难以高效构建可靠的工具集,并且局限于单任务设置。为应对这些挑战,我们提出 GATE(Graph-based Adaptive Tool Evolution,基于图的自适应工具演化),这是一个自适应框架,可在多个场景中动态构建并演化由可复用工具组成的层次化图。我们在开放式任务(Minecraft)、基于智能体的任务(TextCraft、DABench)以及代码生成任务(MATH、Date、TabMWP)上评估 GATE。结果表明,与此前的 SOTA 相比,GATE 在 Minecraft 中实现了最高 4.3 倍的里程碑完成速度提升;在代码生成任务上较现有工具制造方法平均提升 9.23%,在智能体任务上平均提升 10.03%。GATE 展示了自适应演化的能力,在保持高效率的同时平衡工具数量、复杂性与功能性。代码和数据可在 https://github.com/ayanami2003/GATE 获取。

英文摘要:

Large Language Models (LLMs) have shown great promise in tool-making, yet existing frameworks often struggle to efficiently construct reliable toolsets and are limited to single-task settings. To address these challenges, we propose GATE (Graph-based Adaptive Tool Evolution), an adaptive framework that dynamically constructs and evolves a hierarchical graph of reusable tools across multiple scenarios. We evaluate GATE on open-ended tasks (Minecraft), agent-based tasks (TextCraft, DABench), and code generation tasks (MATH, Date, TabMWP). Our results show that GATE achieves up to 4.3x faster milestone completion in Minecraft compared to the previous SOTA, and provides an average improvement of 9.23% over existing tool-making methods in code generation tasks and 10.03% in agent tasks. GATE demonstrates the power of adaptive evolution, balancing tool quantity, complexity, and functionality while maintaining high efficiency. Code and data are available at \url{https://github.com/ayanami2003/GATE}.

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。

↑