Exploring Multi-Agent Debate for Zero-Shot Stance Detection: A Novel Approach

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(9), P. 4612 - 4612

Published: April 22, 2025

Zero-shot stance detection aims to identify the expressed in social media text aimed at specific targets without relying on annotated data. However, due insufficient contextual information and inherent ambiguity of language, this task faces numerous challenges low-resource scenarios. This work proposes a novel zero-shot method based multi-agent debate (ZSMD) address aforementioned challenges. Specifically, we construct two debater agents representing supporting opposing stances. A knowledge enhancement module supplements original tweet target with relevant background knowledge, providing richer support for argument generation. Subsequently, engage over predetermined number rounds, employing rebuttal strategies such as factual verification, logical analysis, sentiment analysis. If no consensus is reached within specified referee agent synthesizes process input deliver final determination. We evaluate ZSMD benchmark datasets, SemEval-2016 Task 6 P-Stance, compare it against strong baselines MB-Cal COLA. The experimental results show that not only achieves higher accuracy than these baselines, but also provides deeper insights into subtle differences opinion expression, highlighting potential structured argumentation settings.

Language: Английский

MGTAB: A Multi-Relational Graph-Based Twitter Account Detection Benchmark

Neurocomputing, Journal Year: 2025, Volume and Issue: unknown, P. 130490 - 130490

Published: May 1, 2025

Language: Английский

LLM Chain Ensembles for Scalable and Accurate Data Annotation

2021 IEEE International Conference on Big Data (Big Data), Journal Year: 2024, Volume and Issue: unknown, P. 2110 - 2118

Published: Dec. 15, 2024

Language: Английский