Self-Adapting Group of Experts for Multi-Agent Reasoning

2026-09-28

AI Summary

This study addresses the limitation of fixed prompts in multi-agent systems, which struggle to accommodate diverse task requirements. To this end, we propose SAGE, a framework introducing a novel training-free adaptive strategy transfer mechanism. Specifically, SAGE identifies optimal strategy donors through answer consistency, prefix consistency, and reciprocal peer review, enabling cross-agent reasoning capability sharing while preserving original role configurations. Furthermore, it incorporates a dynamic sparse directed acyclic graph to optimize collaborative information routing. Extensive experiments demonstrate that SAGE significantly outperforms existing baseline methods in average accuracy across various backbone models and reasoning benchmarks.

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.

Problem

• Multi-agent systems
• Reasoning strategy adaptation
• Large language models
• System prompts

Innovation

• Multi-Agent Reasoning
• Self-Adapting Group of Experts
• Training-free Framework
• Strategy Transfer
• Sparse Directed Acyclic Graph