作者位次 #2/3ACL 2026Poster· Natural Language Processing
When Identity Skews Debate: Anonymization for Bias-Reduced Multi-Agent Reasoning
Hyeong Kyu Choi · Jerry Zhu · Sharon Li
摘要
Multi-agent debate (MAD) aims to improve large language model (LLM) reasoning by letting multiple agents exchange answers and then aggregate their opinions. Yet recent studies reveal that agents are not neutral: they are prone to identity-driven sycophancy and self-bias, uncritically adopting a p...