Inhibitory message passing improves multi-agent reasoning accuracy and efficiency
Neuro-Inspired Inhibitory Message Passing for Accurate and Token-Efficient Multi-Agent Reasoning
Artificial IntelligenceComputation and LanguageMultiagent Systems
Summary
When many AI agents work together, they share information to solve problems better than alone. But repeating facts can clutter communication and hide important clues. The authors found a way to 'quiet down' repeated information while keeping unique facts, inspired by how the brain works. This method helps groups of agents reason more accurately and use fewer words, across many models and setups.
What this means in practice
- •For distributed ai developers: Implement communication routing that filters repeated information to improve AI agent collaboration accuracy and reduce communication cost.
- •For cloud service architects: Design multi-agent AI systems that scale to many agents while keeping messages concise and reasoning precise.
Authors
Nima H. Siboni, Magdalena Springer, Vahid Rostami
Abstract
Multi-agent systems promise to solve problems by bringing together information beyond the reach of any individual agent. Retrieval and tool use can distribute this information across agents, making communication essential to collaboration. A central challenge is keeping crucial evidence visible in each agent's context as messages accumulate. Repeated information can crowd out rare, decisive evidence. Models of cortical circuits suggest a principle for addressing this imbalance: inhibition can selectively suppress shared inputs while preserving sensitivity to weaker signals. We adapt this principle into inhibitory message passing, a training-free communication router that cancels redundant evidence separately for each evidence sentence and receiver. Each round, it forwards evidence sentences restated in agents' messages, delivering each at most once to receivers whose own evidence lacks it and excluding the senders' reasoning and answers. Against standard message passing, accuracy in peer groups rose by 39-77 percentage points across three benchmarks, and tokens per run fell by 10-46%. Gains held across seven open models spanning four to thirty-two billion parameters and two commercial models, in groups up to sixty-four agents, and across multi-agent architectures including peer-to-peer, orchestrator-worker, and multilevel hierarchical organizations. These findings support a scalable design pattern for more accurate and token-efficient multi-agent reasoning.