arXiv 2502.189902025-02-26cs.CL

GenTool：通过从零到一和从弱到强模拟增强语言模型中的工具泛化能力

GenTool: Enhancing Tool Generalization in Language Models through Zero-to-One and Weak-to-Strong Simulation

School of Informatics, University of Edinburgh（爱丁堡大学信息学院）；Microsoft Corporation（微软公司）

Jie He, Jennifer Neville, Mengting Wan, Longqi Yang, Hui Liu, Xiaofeng Xu, Xia Song, Jeff Z. Pan, Pei Zhou

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

GenTool通过合成数据模拟从零到一和从弱到强两种工具泛化场景，采用两阶段微调优化工具排序与选择，显著提升1B至8B参数LLMs的工具使用能力，性能超越GPT-4o。

摘要：

大型语言模型（LLMs）通过集成外部工具可以增强其作为AI助手的能力，从而访问更广泛的信息。尽管最近的LLMs通常在监督微调（SFT）阶段使用工具使用示例进行微调，但它们能否发展出稳健的工具使用技能并有效泛化到未见过的查询和工具，仍然存在疑问。在本工作中，我们提出了GenTool，一种新颖的训练框架，旨在让LLMs为工具使用中的多样化泛化挑战做好准备。我们的方法解决了实际应用中至关重要的两个基本维度：从零到一泛化，使模型能够在最初缺乏合适工具时，在工具可用时采纳并使用它；以及从弱到强泛化，使模型能够利用现有工具的增强版本来解决查询。为实现这一目标，我们开发了模拟这两个工具使用维度的合成训练数据，并引入了一种两阶段微调方法：先优化工具排序，再优化工具选择。通过在四个泛化场景中的广泛实验，我们证明了我们的方法显著增强了参数规模从1B到8B的LLMs的工具使用能力，其性能超越了GPT-4o。此外，我们的分析还为LLMs在工具泛化中遇到的挑战提供了有价值的见解。

英文摘要：

Large Language Models (LLMs) can enhance their capabilities as AI assistants by integrating external tools, allowing them to access a wider range of information. While recent LLMs are typically fine-tuned with tool usage examples during supervised fine-tuning (SFT), questions remain about their ability to develop robust tool-usage skills and can effectively generalize to unseen queries and tools. In this work, we present GenTool, a novel training framework that prepares LLMs for diverse generalization challenges in tool utilization. Our approach addresses two fundamental dimensions critical for real-world applications: Zero-to-One Generalization, enabling the model to address queries initially lacking a suitable tool by adopting and utilizing one when it becomes available, and Weak-to-Strong Generalization, allowing models to leverage enhanced versions of existing tools to solve queries. To achieve this, we develop synthetic training data simulating these two dimensions of tool usage and introduce a two-stage fine-tuning approach: optimizing tool ranking, then refining tool selection. Through extensive experiments across four generalization scenarios, we demonstrate that our method significantly enhances the tool-usage capabilities of LLMs ranging from 1B to 8B parameters, achieving performance that surpasses GPT-4o. Furthermore, our analysis also provides valuable insights into the challenges LLMs encounter in tool generalization.

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。