arXiv：cs.AI· Seungdong Yoa, Ye Seul Sim, Suhee Yoon, Sanghyu Yoon, Dongmin Kim, Soonyoung Lee, Bumsoo Kim, Junhyun Lee·· 16 小时前
LASE：置信度引导的自适应多智能体辩论框架，以接近单智能体成本实现多智能体推理性能
Efficient Multi-Agent Reasoning via Confidence-Guided Adaptive Debate
AI 导读
研究者提出 LASE（Leader-Adaptive Structured Engagement），一种以 leader 为中心的多智能体辩论框架，仅在非中性场景下选择性触发交互，默认使用简单聚合。
正文
Abstract:Multi-agent debate has shown promise for improving the reasoning of large language models, yet recent theory suggests its benefits are highly regime-dependent. While interaction can amplify informative signals under corrective conditions, symmetric debate dynamics are neutral in expectation, often making majority voting preferable. We reconcile these views by arguing that debate is effective only when invoked at the right time and with appropriate structure. Based on this insight, we propose LASE: Leader-Adaptive Structured Engagement, a leader-centric multi-agent debate framework that selectively engages interaction only in non-neutral regimes. LASE introduces an asymmetric leader-supporter architecture that enables directed information flow and selective signal amplification, while defaulting to simple aggregation otherwise. Experiments across diverse reasoning benchmarks show that LASE achieves multi-agent-level performance with near single-agent token cost, substantially improving efficiency over static debate and voting baselines.
| Comments: | Accepted at ICML 2026 |
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.16343 [cs.AI] |
| (or arXiv:2610.16343v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.16343 arXiv-issued DOI via DataCite (pending registration) |
|
| Journal reference: | Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:149412-149429, 2026 |
Submission history
From: Seungdong Yoa [view email]
[v1]
Mon, 12 Oct 2026 08:02:41 UTC (879 KB)
来源：arXiv：cs.AI · arxiv.org