Multiagent system adapts expert roles to improve problem solving
Self-Adapting Group of Experts for Multi-Agent Reasoning
Artificial IntelligenceComputation and LanguageMultiagent Systems
Summary
Multi-agent systems use groups of AI experts who talk to each other to solve problems, but usually each expert sticks to its fixed way of thinking. The authors studied if experts can choose better thinking strategies based on early answers and then share those strategies with other experts. They created SAGE, a method that lets these AI agents change their reasoning roles by comparing their answers and reviewing each other without retraining or seeing the problem again. This approach improved their ability to solve reasoning tasks in tests using different AI models.
What this means in practice
- •For software development teams: Improve AI assistant collaboration by enabling dynamic role adaptation to solve complex coding and debugging tasks more accurately.
- •For customer support teams: Enhance multi-agent chatbots with adaptable reasoning strategies for better problem resolution across diverse customer queries.
Authors
Mohammad Atif Quamar, Nurbek Tastan, Karthik Nandakumar, Junpei Komiyama
Abstract
Multi-agent systems bring together language model agents with different roles to propose, review, and refine solutions. Each agent's response depends on its model's capabilities, the reasoning strategy defined by its system prompt, and the information in its input context. Existing frameworks often adapt communication by changing this context while leaving individual prompts fixed, even when a problem calls for different skills. We study whether agents' initial responses can identify a strategy better suited to the current problem and guide its transfer to other agents. To address this, we introduce SAGE (Self-Adapting Group of Experts), a training-free framework that uses answer agreement, prefix consistency, and reciprocal peer review to select a strategy donor. SAGE transfers the selected donor's reasoning strategy to the other agents while preserving their original roles. This transfer uses only the agents' original system prompts, without access to the problem or generated solutions. After strategy adaptation, agents exchange responses through a dynamic, sparse directed acyclic graph that routes information from higher-scoring agents to lower-scoring agents. Experiments across multiple agent backbones and reasoning benchmarks show that SAGE achieves higher average accuracy than the evaluated baselines. Our code is available at https://github.com/atifquamar07/sage.