[论文解读] Tool Learning with Foundation Models

Yujia Qin, Shengding Hu+39 more|arXiv (Cornell University)|Apr 17, 2023

Mobile Crowdsensing and Crowdsourcing被引用 29

一句话总结

本文提出了一个面向基础模型的工具学习通用框架，综述了背景与现有工作，并通过使用 18 tools 的实验进行验证，突出显示挑战与未来方向。

ABSTRACT

Humans possess an extraordinary ability to create and utilize tools, allowing them to overcome physical limitations and explore new frontiers. With the advent of foundation models, AI systems have the potential to be equally adept in tool use as humans. This paradigm, i.e., tool learning with foundation models, combines the strengths of specialized tools and foundation models to achieve enhanced accuracy, efficiency, and automation in problem-solving. Despite its immense potential, there is still a lack of a comprehensive understanding of key challenges, opportunities, and future endeavors in this field. To this end, we present a systematic investigation of tool learning in this paper. We first introduce the background of tool learning, including its cognitive origins, the paradigm shift of foundation models, and the complementary roles of tools and models. Then we recapitulate existing tool learning research into tool-augmented and tool-oriented learning. We formulate a general tool learning framework: starting from understanding the user instruction, models should learn to decompose a complex task into several subtasks, dynamically adjust their plan through reasoning, and effectively conquer each sub-task by selecting appropriate tools. We also discuss how to train models for improved tool-use capabilities and facilitate the generalization in tool learning. Considering the lack of a systematic tool learning evaluation in prior works, we experiment with 18 representative tools and show the potential of current foundation models in skillfully utilizing tools. Finally, we discuss several open problems that require further investigation for tool learning. In general, we hope this paper could inspire future research in integrating tools with foundation models.

研究动机与目标

• 介绍工具使用与基础模型的认知与范式背景。
• 提出一个将工具、环境、控制器和感知器整合在一起的通用工具学习框架。
• 回顾现有的工具学习研究并识别核心问题与解决方案。
• 通过 18 tools 的实验，展示基础模型使用多样化工具的潜力。
• 讨论在安全、可扩展和个性化的工具学习方面的开放问题与未来方向。

提出的方法

• 用四个要素定义一个统一的工具学习框架：工具集、环境、控制器（基础模型）和感知器。
• 描述从用户意图到可执行计划与工具执行的一般过程。
• 概述训练策略：通过演示学习和通过反馈学习。
• 通过标准化接口实现对多工具交互的通用工具学习的讨论。
• 用 18 个代表性工具进行实验，以评估当前基础模型利用工具的能力。

实验结果

研究问题

• RQ1 如何构建基础模型以学习并协调跨越多种工具的使用？
• RQ2 哪些训练策略能使基础模型实现稳健且可泛化的工具使用？
• RQ3 最先进的基础模型在实际任务中有效利用广泛工具的程度有多大？
• RQ4 在将工具学习与基础模型结合部署时，关键挑战有哪些（安全性、个性化、工具创建）？
• RQ5 统一接口如何促进将工具使用技能迁移到新工具和新情境？

主要发现

• 基础模型（如 ChatGPT）可以通过简单提示有效地使用工具来完成任务。
• 一个通用的工具学习框架可以统一工具、环境和模型之间的交互。
• 来自示范和来自反馈的训练策略是提升工具使用能力的核心。
• 对 18 个工具的实验展示了当前基础模型在工具操作方面的潜力和局限。
• 本文指出了包括安全性、工具创建、个性化以及在复杂系统中部署等主要开放问题。