让AI辩论更智能：按需启动、自动停止、过滤异常，提升推理准确率。
ARMOR-MAD: Adaptive Routing for Heterogeneous Multi-Agent Debate in Large Language Model Reasoning
- 根据答案一致性决定是否辩论，节省计算资源。
- 辩论提前收敛时自动停止，效率提升明显。
- 识别并弱化异常答案，提高最终决策质量。
多智能体辩论（MAD）可提升大模型推理能力，但固定辩论流程常导致计算浪费，并放大相似智能体的共现错误。本文提出无需训练的异构辩论框架ARMOR-MAD，将辩论视为条件计算。该框架包含三个组件：预辩论一致性路由（PAR）判断初始回答是否需要辩论；早期共识终止评估器（EASE）在达成共识后提前停止辩论；语义离群检测（SOD）在聚合时降低异常答案权重。在MATH Level 5、GSM8K、MMLU和MMLU-Pro上，ARMOR-MAD均优于相同模型池的固定轮次异构辩论，准确率分别达到65.5%、96.5%、90.0%和81.5%。结果表明，真实模型异质性与基于一致性的控制对提升MAD的准确性与效率至关重要。
原文摘要 · Abstract (English)
Multi-agent debate (MAD) can improve large language model reasoning, but fixed debate pipelines often waste computation and can amplify correlated errors among similar agents. We propose ARMOR-MAD, a training-free heterogeneous MAD framework that treats debate as conditional computation. ARMOR-MAD combines three components: Pre-debate Agreement Routing (PAR) decides whether independently generated Round-0 answers require debate; Early Agreement Stopping Evaluator (EASE) stops debate after convergence; and Semantic Outlier Detection (SOD) down-weights abnormal final answers during aggregation. Across MATH Level 5, GSM8K, MMLU, and MMLU-Pro, ARMOR-MAD consistently improves over fixed-round heterogeneous debate with the same model pool, reaching 65.5\%, 96.5\%, 90.0\%, and 81.5\% accuracy, respectively. The results suggest that genuine model heterogeneity and agreement-based control are both important for making MAD more accurate and efficient.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。