让多个AI助手协作推理时，记忆越用越强，突破传统方法的性能瓶颈。
ReM-MoA: Reasoning Memory Sustains Mixture-of-Agents Scaling
- 引入分级记忆存储与评分机制，持续积累并优化推理过程
- 在五项推理任务中，深度扩展时表现持续提升，超越现有方法
- 适合需要长链逻辑推理的复杂任务，如数学证明与代码生成
混合专家（MoA）架构通过将多个大语言模型代理组织为分层推理流水线，实现推理阶段的可扩展性。然而，现有MoA变体在深度增加时无法持续保持性能优势，出现退化、早期饱和或停滞。本文提出ReM-MoA，一种增强记忆的MoA框架，通过两个机制维持扩展性：(1) 采用比较式评审代理持续存储并排序各层的推理轨迹的排名推理记忆；(2) 设计精选多样化记忆路由策略，使不同代理接触不同组合的成功与失败轨迹，在保留探索多样性的同时传播高质量推理。此外，我们引入可选的多领域评审器蒸馏流程，通过前沿模型监督提升排序质量。在涵盖数学、形式逻辑、代码、知识与常识的五个推理基准上，ReM-MoA在深度与宽度扩展下均持续优于先前的MoA变体，且优势随深度增大而扩大，证实结构化跨层推理记忆是实现可扩展多代理推理的关键缺失机制。
原文摘要 · Abstract (English)
Mixture-of-Agents (MoA) architectures improve inference-time scaling by organizing multiple LLM agents into layered reasoning pipelines. However, existing MoA variants fail to sustain gains as depth increases, exhibiting degradation, early plateauing, or saturation. We propose ReM-MoA, a memory-augmented MoA framework that sustains scaling through two mechanisms: (1) a Ranked Reasoning Memory that persistently stores and ranks reasoning traces from all layers using a comparative Reviewer Agent, and (2) a Curated Diversified Memory Routing scheme that exposes different agents to distinct combinations of successful and failed traces, preserving exploration diversity while propagating high-quality reasoning. We further introduce an optional multi-domain Reviewer distillation pipeline that improves ranking quality through frontier-model supervision. Across five reasoning benchmarks spanning math, formal logic, code, knowledge, and commonsense, ReM-MoA consistently outperforms prior MoA variants across both depth and width scaling, and its advantage widens with depth, establishing structured cross-layer reasoning memory as a key missing mechanism for scalable multi-agent inference.
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。