METRO: Towards Strategy Induction from Expert Dialogue Transcripts for Non-collaborative Dialogues

TL;DR

METRO introduces a novel method leveraging large language models to automatically induce strategic behavior from expert dialogue transcripts, formalized into a hierarchical Strategy Forest, enabling scalable non-collaborative dialogue agents.

Contribution

It is the first to automatically induce strategy actions and planning logic from raw transcripts using large language models, formalized as a Strategy Forest structure.

Findings

Outperforms existing methods by 9-10% on benchmarks.

Demonstrates robust cross-task transferability.

Reveals the importance of behavioral diversity and foresight.

Abstract

Developing non-collaborative dialogue agents traditionally requires the manual, unscalable codification of expert strategies. We propose \ours, a method that leverages large language models to autonomously induce both strategy actions and planning logic directly from raw transcripts. METRO formalizes expert knowledge into a Strategy Forest, a hierarchical structure that captures both short-term responses (nodes) and long-term strategic foresight (branches). Experimental results across two benchmarks show that METRO demonstrates promising performance, outperforming existing methods by an average of 9%-10%. Our further analysis not only reveals the success behind METRO (strategic behavioral diversity and foresight), but also demonstrates its robust cross-task transferability. This offers new insights into building non-collaborative agents in a cost-effective and scalable way. Our code is available at https://github.com/Humphrey-0125/METRO.