LLM Agents Orchestrate 6G Network Slice Configurations

A January 10, 2026 arXiv paper presents a hierarchical multi-agent framework using LLM-based agents to translate natural-language intents into executable 6G network slice configurations. The system employs an orchestrator plus RAN and Core specialist agents using ReAct-style reasoning and structured network state, outperforming rule-based systems and direct LLM prompting across benchmark scenarios. Results highlight applicability to O-RAN deployments and the need for careful prompt engineering.