Source: arXiv — 2026-07-17
Summary
Researchers Mazene Ameur, Abdelkader Mekrache, Bouziane Brik, and Adlen Ksentini published a two-part tutorial-and-survey formalizing how LLM-powered agentic AI — reasoning, planning, tool use, and multi-agent coordination — maps onto the existing structure of 5G and 6G networks. Part I lays out the control, management, and emerging AI-native planes of next-generation networks alongside the foundations of agentic systems; Part II maps specific agentic capabilities onto real 5G/6G control surfaces and existing telecom standardization efforts. The paper's stated gap is that prior surveys treat agentic AI and telecom networking as separate literatures, leaving protocol integration and standards alignment largely unaddressed.
Key Takeaways
- The paper's organizing move is treating a next-gen network as having three planes — control, management, and a newer AI-native plane — and asking, for each one, what an LLM agent's reasoning/planning/tool-use loop would actually plug into.
- It's explicitly framed as a tutorial-and-survey rather than a new system or benchmark: Part I builds shared vocabulary (agent foundations plus network plane architecture), Part II does the mapping onto standardization bodies and existing 6G initiatives.
- The stated contribution is closing a literature gap — agentic-AI surveys and 5G/6G surveys have mostly stayed in separate lanes, so protocol-level and standards-level integration between the two has been under-specified until now.
- As with most review papers, the value here is structural: a shared reference frame for where "agentic AI" concepts (multi-agent coordination, tool use, autonomous planning) would sit inside standards-driven telecom architecture, rather than a specific new technique with measured results.
- Worth a skim for anyone working at the intersection of telecom infrastructure and agent frameworks, since it explicitly ties agent concepts to standardization bodies rather than leaving that connection implicit.