Epistemic Gain, Aleatoric Cost: Uncertainty Decomposition in Multi-Agent Debate for Math Reasoning
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866912934967377920 |
|---|---|
| author | Qiao, Dan Chen, Binbin Cai, Fengyu Chen, Jianlong Li, Wenhao Jiang, Fuxin Chen, Zuzhi Zha, Hongyuan Zhang, Tieying Wang, Baoxiang |
| author_facet | Qiao, Dan Chen, Binbin Cai, Fengyu Chen, Jianlong Li, Wenhao Jiang, Fuxin Chen, Zuzhi Zha, Hongyuan Zhang, Tieying Wang, Baoxiang |
| contents | Multi-Agent Debate (MAD) has shown promise in leveraging collective intelligence to improve reasoning and reduce hallucinations, yet it remains unclear how information exchange shapes the underlying ability. Empirically, MAD exhibits paradoxical phenomena, such as accuracy improvement accompanied by substantial increase in token entropy, and remarkable divergence between homogeneous and heterogeneous model combinations. In this paper, we propose a Bayesian uncertainty analysis framework for MAD, which decomposes total predictive uncertainty into epistemic uncertainty reducible by debate context and aleatoric uncertainty induced by internal model noise. Across multiple model configurations, we find that effective debate hinges on achieving high epistemic gain under controlled aleatoric cost. Building on this insight, we design an uncertainty-guided multi-agent reinforcement learning (MARL) algorithm that explicitly optimizes aleatoric noise reduction and epistemic information utilization. Experiments show that our training significantly improves post-debate accuracy and stability, and enhances individual reasoning beyond single-agent RL, providing a unified Bayesian uncertainty perspective for understanding and improving MAD. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_01221 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Epistemic Gain, Aleatoric Cost: Uncertainty Decomposition in Multi-Agent Debate for Math Reasoning Qiao, Dan Chen, Binbin Cai, Fengyu Chen, Jianlong Li, Wenhao Jiang, Fuxin Chen, Zuzhi Zha, Hongyuan Zhang, Tieying Wang, Baoxiang Multiagent Systems Multi-Agent Debate (MAD) has shown promise in leveraging collective intelligence to improve reasoning and reduce hallucinations, yet it remains unclear how information exchange shapes the underlying ability. Empirically, MAD exhibits paradoxical phenomena, such as accuracy improvement accompanied by substantial increase in token entropy, and remarkable divergence between homogeneous and heterogeneous model combinations. In this paper, we propose a Bayesian uncertainty analysis framework for MAD, which decomposes total predictive uncertainty into epistemic uncertainty reducible by debate context and aleatoric uncertainty induced by internal model noise. Across multiple model configurations, we find that effective debate hinges on achieving high epistemic gain under controlled aleatoric cost. Building on this insight, we design an uncertainty-guided multi-agent reinforcement learning (MARL) algorithm that explicitly optimizes aleatoric noise reduction and epistemic information utilization. Experiments show that our training significantly improves post-debate accuracy and stability, and enhances individual reasoning beyond single-agent RL, providing a unified Bayesian uncertainty perspective for understanding and improving MAD. |
| title | Epistemic Gain, Aleatoric Cost: Uncertainty Decomposition in Multi-Agent Debate for Math Reasoning |
| topic | Multiagent Systems |
| url | https://arxiv.org/abs/2603.01221 |