Performance Comparison of IBN orchestration using LLM and SLMs

Wai Lwin Phone; Brahim El Boudani; Tasos Dagiuklas; Saptarshi Ghosh

cs.NIFeb 27, 2026

Performance Comparison of IBN orchestration using LLM and SLMs

Authors: Wai Lwin Phone, Brahim El Boudani, Tasos Dagiuklas, Saptarshi Ghosh

Organizations: Dept. of Computer Science & Digital Technologies, London South Bank University, London, UK · Digital Catapult, London, UK

Abstract

The evolution of both 5G and 6G networks is driving the advancement of fully autonomous network management, placing Intent-Based Networking at the centre of this transformation. This paper introduces a novel framework for 5G and 6G IBN orchestration that leverages a stateful, hierarchical multi-agent architecture to achieve full automation using both SLMs and LLMs. Both models have been evaluated for translation accuracy using metrics such as BLEU, METEOR, and ROUGE-L, as well as computational complexity. Experimental results show that both models exhibit similar accuracy. However, result shows that SLMs can improve the overall completion speed of the IBN lifecycle by 20%.

Figures & tables

Fig. 1: Proposed IBN Orchestration Architecture.

Model	Agent Type	Role	Topology Options	5G Service Types	Routing Strategy
TinyLlama-1.1B (SLM)	Junior Agent 1	Topology & Service Analysis	Full-mesh, Partial-mesh, Hub-and-spoke	eMBB, URLLC, MMTC	–
	Junior Agent 2	Topology & Service Analysis	Full-mesh, Partial-mesh, Hub-and-spoke	eMBB, URLLC, MMTC	–
	Senior Agent	Generate topology & weights	–	–	–
	Policy Agent	Decide routing strategy	–	–	OSPF or Dual
SmolLM-1.7B (SLM)	Junior Agent 1	Topology & Service Analysis	Full-mesh, Partial-mesh, Hub-and-spoke	eMBB, URLLC, MMTC	–
	Junior Agent 2	Topology & Service Analysis	Full-mesh, Partial-mesh, Hub-and-spoke	eMBB, URLLC, MMTC	–

TABLE I: Scenarios (LLMs and SLMs and network topologies)

Agent Type	Model	$\mu$	$\sigma$
Senior Agent	TinyLlama-1.1B(SLM)	0.8884	0.08
	mistral-small3.2:24b	0.8240	0.09
	gpt-5-nano	0.7621	0.10
Policy Agent	TinyLlama-1.1B(SLM)	0.8080	0.11
	gpt-5-nano	0.7886	0.14
	mistral-small3.2:24b	0.7226	0.14

TABLE II: BLEU-2 score comparison across agents and models (

N=500

per group). BLEU-2 is computed at the corpus level with equal unigram and bigram weights (no smoothing).

Agent Type	Model	$\mu$	$\sigma$
Senior Agent	TinyLlama-1.1B(SLM)	0.8839	0.06
	mistral-small3.2:24b	0.8748	0.07
	gpt-5-nano	0.8349	0.06
Policy Agent	TinyLlama-1.1B(SLM)	0.7077	0.11
	gpt-5-nano	0.5800	0.23
	mistral-small3.2:24b	0.4111	0.33

TABLE III: Meteor score comparison across agents and models (

N=500

per group). METEOR: SENTENCE-LEVEL AVERAGE USING NLTK’S IMPLEMENTATION.

Agent Type	Model	$\mu$	$\sigma$
Senior Agent	TinyLlama-1.1B(SLM)	0.8810	0.09
	mistral-small3.2:24b	0.8578	0.08
	gpt-5-nano	0.7769	0.12
Policy Agent	TinyLlama-1.1B(SLM)	0.6987	0.17
	gpt-5-nano	0.6600	0.18
	mistral-small3.2:24b	0.5422	0.25

TABLE IV: ROUGE-L score comparison across agents and models (

N=500

per group), with standard deviation. ROUGE-L: Average F-measure computed with stemming enabled.

Agent Type	Iter. 1 (%)	Iter. 2 (%)	Iter. 3 (%)
Intent UI agent	13	16	18
Junior agents (Both)	7	7	7
Senior agent	51	52	52
Policy agent	29	25	23

TABLE V: Agent Time Distribution Across Iterations (SLM) (%)

Agent Type	Iter. 1 (%)	Iter. 2 (%)	Iter. 3 (%)
Intent UI agent	14	18	21
Junior agents (Both)	8	9	8
Senior agent	53	50	51
Policy agent	25	23	20

TABLE VI: Agent Time Distribution Across Iterations (gpt-5-nano) (%)

Agent Type	Iter. 1 (%)	Iter. 2 (%)	Iter. 3 (%)
Intent UI agent	11	12	14
Junior agents (Both)	13	13	12
Senior agent	54	55	53
Policy agent	22	20	21

TABLE VII: Agent Time Distribution Across Iterations (mistral-small3.2:24b) (%)

Explore similar work

LLM-Powered Agentic AI for 5G/6G Networks: A Tutorial and Survey on Architectures, Protocols, and Standardization
Jul 17, 2026Mazene Ameur, Abdelkader Mekrache, Bouziane Brik +1Multi-Agent LLM Systems Agent Communication Protocols
EURECOM, France · Simula Metropolitan, Norway · University of Sharjah, UAE
Advanced LLM-Enhanced Intent-Based 5G Network Management using Dynamic Semantic Routes
Aug 23, 2026Thomas Benton Townsend, Dimitrios Michael ManiasIntent-Based Networking LLM Routing
The Department of Computer Science and Engineering, Mississippi State University
From Agentic to Autogenic Network Management for AI-Native 6G and Beyond: A Standards Perspective
Jul 7, 2026Petar Djukic, Sudipta Acharya, Takai Eddine Kennouche +16G Networks Wireless Communications
Bell Labs Research, 600 March Rd, Kanata, ON K2K 2E6 · University of Ottawa, 350 Legget Drive, Kanata, ON, Canada, K2K 2W7 · Nokia Technology Standards, Nokia France, 12 rue Jean Bart Massy 91300

Jul 17, 2026J/K move · Enter open · S save

LLM-Powered Agentic AI for 5G/6G Networks: A Tutorial and Survey on Architectures, Protocols, and Standardization

Mazene Ameur, Abdelkader Mekrache, Bouziane Brik, Adlen Ksentini

EURECOM, France · Simula Metropolitan, Norway · University of Sharjah, UAE

Multi-Agent LLM Systems Agent Communication Protocols 6G Networks Agentic AI

Open paper AskPDF arXiv

Agentic Artificial Intelligence (AI), enabled by Large Language Models, marks a shift from rule-based automation toward autonomous, goal-driven control of Next-Generation Networks (NGNs). Existing surveys treat the two domains in isolation, leaving protocol integration, evaluation, and standardization alignment underexplored. To address this gap, a two-part tutorial-and-survey is presented. Part I formalises the control, management, and AI-native planes of 5G and 6G. It then covers the foundations of agentic systems: reasoning, planning, tool use, multi-agent coordination, and evaluation. Part II maps agentic capabilities onto 5G/6G control surfaces, standardization, and major 6G initiatives. Finally, it identifies open challenges shaping autonomous telecommunications.

Abstract

Explore similar work

LLM-Powered Agentic AI for 5G/6G Networks: A Tutorial and Survey on Architectures, Protocols, and Standardization

Advanced LLM-Enhanced Intent-Based 5G Network Management using Dynamic Semantic Routes

From Agentic to Autogenic Network Management for AI-Native 6G and Beyond: A Standards Perspective