Advancing and Benchmarking Personalized Tool Invocation for LLMs

Xu Huang, Yuefeng Huang, Weiwen Liu, Xingshan Zeng, Yasheng Wang, Ruiming Tang, Hong Xie, Defu Lian·May 07, 2025

Summary

PTTool enhances LLMs with personalized tool invocation, outperforming competitors in untrained scenarios. WineTraveler38 registers for a MegaMart account, specifying preferences including language and opting out of marketing emails.

Background
Overview of PTTool and its role in enhancing Large Language Models (LLMs)
Explanation of personalized tool invocation in the context of PTTool
Comparison of PTTool's performance against competitors in untrained scenarios
Method
Data Collection
Description of the data used to train and test PTTool
Explanation of how WineTraveler38's preferences were gathered and integrated into the data set
Data Preprocessing
Steps taken to prepare the data for PTTool's use, including WineTraveler38's preference specifications
How these preferences were incorporated into the data preprocessing phase
Application
WineTraveler38's Registration
Detailed account creation process for WineTraveler38 on MegaMart
Explanation of the preferences WineTraveler38 specified during registration, including language preference and opt-out of marketing emails
PTTool's Role
How PTTool utilizes WineTraveler38's preferences to enhance the user experience
Performance metrics and outcomes of PTTool's application in untrained scenarios
Basic info
papers
computation and language
artificial intelligence
Advanced features
Insights
How does PTTool integrate with LLMs to enhance their functionality?
What are the key features of PTTool that allow it to outperform competitors in untrained scenarios?
What compatibility considerations are there for integrating PTTool with existing LLMs?