Agentic AI-Native Networks Proposed for 6G

A new paper on arXiv proposes a paradigm shift toward Agentic AI-Native 6G networks, where Large Language Model (LLM)-based agents act as bounded reasoning entities within a semantic control plane. The authors argue that existing optimization-centric approaches lack reasoning capability. They introduce a four-layer architecture integrating deterministic infrastructure, semantic abstraction, hierarchical reasoning, and a distributed multi-agent fabric across device, edge, and core domains. A proof-of-concept framework was developed and empirically tested using domain-specific data.