黑暗森林：少交流，高精度——面向多智能体大语言模型

摘要

多智能体大语言模型（LLM）系统通过整合多个智能体的输出提升推理能力，但交互密集型方法可能引发误差传播和高通信开销。当智能体交换原始响应或推理轨迹时，错误的中期推理可能被采纳并放大，导致形成看似合理却错误的共识；多轮通信还会增加令牌消耗、延迟和推理成本。本文提出一种受控通信的协调框架DarkForest。该框架首先保持智能体独立性，使每个智能体在不查看其他智能体输出的情况下生成答案；接着将原始响应解析为结构化候选记录，将语义等价的候选记录聚类，并利用智能体可靠性、置信度、解析质量、支撑模式可靠性和独立性校正来估算这些聚类上的校准信念分布。协调器仅从该信念状态接收策略允许的证据，实现受控通信。在六个推理基准上的实验表明，DarkForest取得了领先的整体质量，在基准指标上较最强基线最高提升30.7%，相比高通信开销基线将令牌消耗降低高达6.5倍。

English

Multi-agent LLM systems improve reasoning by combining outputs from multiple agents, but interaction-heavy methods can introduce error propagation and high communication overhead. When agents exchange raw responses or reasoning traces, incorrect intermediate reasoning may be adopted and amplified, leading to confident but wrong consensus; multi-round communication also increases token consumption, latency, and inference cost. In this paper, we propose a controlled-communication coordination framework named DarkForest. DarkForest first keeps agents independent, so each agent produces an answer without seeing the others' outputs. It then parses the raw responses into structured candidate records, groups semantically equivalent candidates into clusters, and estimates a calibrated belief distribution over these clusters using agent reliability, confidence, parse quality, support-pattern reliability, and independence corrections. A coordinator receives only policy-permitted evidence from this belief state with controlled communication. Experiments on six reasoning benchmarks show that DarkForest achieves leading overall quality, improves the strongest baseline by up to 30.7\% on benchmark metrics, and reduces token consumption by up to 6.5times compared with communication-heavy baselines.