Prompting and Evaluating Large Language Models for Proactive Dialogues: Clarification, Target-guided, and Non-collaboration

TL;DR

This paper analyzes the capabilities of large language models in proactive dialogue scenarios, introduces a new prompting scheme to enhance goal planning, and discusses empirical results to guide future research in this area.

Contribution

It proposes the Proactive Chain-of-Thought prompting scheme to improve LLMs' goal planning in proactive dialogues and provides a comprehensive analysis of their limitations and potentials.

Findings

Proactive Chain-of-Thought prompting enhances goal planning in LLMs.

LLMs still face challenges in handling ambiguous queries and refusals.

Empirical results suggest directions for future proactive dialogue system development.

Abstract

Conversational systems based on Large Language Models (LLMs), such as ChatGPT, show exceptional proficiency in context understanding and response generation. However, despite their impressive capabilities, they still possess limitations, such as providing randomly-guessed answers to ambiguous queries or failing to refuse users' requests, both of which are considered aspects of a conversational agent's proactivity. This raises the question of whether LLM-based conversational systems are equipped to handle proactive dialogue problems. In this work, we conduct a comprehensive analysis of LLM-based conversational systems, specifically focusing on three aspects of proactive dialogue systems: clarification, target-guided, and non-collaborative dialogues. To trigger the proactivity of LLMs, we propose the Proactive Chain-of-Thought prompting scheme, which augments LLMs with the goal planning capability over descriptive reasoning chains. Empirical findings are discussed to promote future studies on LLM-based proactive dialogue systems.