Reasoning-Enhanced GraphRAG-Based Large and Small AI Model Collaboration: A Unified 6G Network Management Approach

With the rapid development of network technology, the synergistic collaboration between large AI models (LAMs) and specialized small AI models (SAMs) offers a new paradigm for intelligent network management. Empowered by Chain-of-Thought (CoT) and graph-based Retrieval-Augmented Generation (GraphRAG) techniques, we propose a reasoning-enhanced GraphRAG-based LAM-SAM collaboration framework integrating high-level reasoning with domain-specific deployability. By leveraging structured knowledge graphs, the framework enhances decision-making accuracy and effectively mitigates hallucinations during inference. To enable intelligent coordination, we further implement policy alignment via Kullback-Leibler (KL) divergence between LAM and SAM policies. Comprehensive case studies demonstrate the framework's effectiveness in addressing complex network challenges. The framework improves the throughput by approximately 12% in coverage optimization. Moreover, the proposed framework enhances base station health by over 8% in network operations and maintenance without additional cost overhead. Finally, we outline promising future research directions for practical deployment.