TAPS: Tool-Augmented Personalisation via Structured Tagging

TL;DR

This paper introduces TAPS, a novel method that enhances large language models' ability to personalize tool use in goal-oriented dialogues by leveraging structured tagging and uncertainty detection, achieving state-of-the-art results.

Contribution

The paper presents TAPS, a new approach that effectively integrates user preferences into tool-augmented language models, addressing a key gap in personalization capabilities.

Findings

TAPS significantly improves personalization in tool use for LLMs.

Achieves state-of-the-art performance on the NLSI task.

Enhances the integration of user preferences in goal-oriented dialogue agents.

Abstract

Recent advancements in tool-augmented large language models have enabled them to interact with external tools, enhancing their ability to perform complex user tasks. However, existing approaches overlook the role of personalisation in guiding tool use. This work investigates how user preferences can be effectively integrated into goal-oriented dialogue agents. Through extensive analysis, we identify key weaknesses in the ability of LLMs to personalise tool use. To this end, we introduce TAPS, a novel solution that enhances personalised tool use by leveraging a structured tagging tool and an uncertainty-based tool detector. TAPS significantly improves the ability of LLMs to incorporate user preferences, achieving the new state-of-the-art for open source models on the NLSI task.