Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Semantic Routing for Enhanced Performance of LLM-Assisted Intent-Based 5G Core Network Management and Orchestration

Dimitrios Michael Manias, Ali Chouman, Abdallah Shami

TL;DR

The paper tackles the reliability and efficiency challenges of LLM-based intent extraction for 5G core MANO. It introduces a semantic routing framework that deterministically maps natural-language intents to predefined routes, enabling end-to-end function calls in the 5G network. By combining seed, variability, and paraphrased prompts, and by evaluating encoder choices and LLM quantization, the authors demonstrate improved accuracy and near-instantaneous responses compared to standalone prompting, even under token/size constraints. The work contributes an open, diverse intent dataset, an end-to-end pipeline, and actionable insights for deploying LLM-assisted MANO in real-world 5G deployments at scale.

Abstract

Large language models (LLMs) are rapidly emerging in Artificial Intelligence (AI) applications, especially in the fields of natural language processing and generative AI. Not limited to text generation applications, these models inherently possess the opportunity to leverage prompt engineering, where the inputs of such models can be appropriately structured to articulate a model's purpose explicitly. A prominent example of this is intent-based networking, an emerging approach for automating and maintaining network operations and management. This paper presents semantic routing to achieve enhanced performance in LLM-assisted intent-based management and orchestration of 5G core networks. This work establishes an end-to-end intent extraction framework and presents a diverse dataset of sample user intents accompanied by a thorough analysis of the effects of encoders and quantization on overall system performance. The results show that using a semantic router improves the accuracy and efficiency of the LLM deployment compared to stand-alone LLMs with prompting architectures.

Semantic Routing for Enhanced Performance of LLM-Assisted Intent-Based 5G Core Network Management and Orchestration

TL;DR

The paper tackles the reliability and efficiency challenges of LLM-based intent extraction for 5G core MANO. It introduces a semantic routing framework that deterministically maps natural-language intents to predefined routes, enabling end-to-end function calls in the 5G network. By combining seed, variability, and paraphrased prompts, and by evaluating encoder choices and LLM quantization, the authors demonstrate improved accuracy and near-instantaneous responses compared to standalone prompting, even under token/size constraints. The work contributes an open, diverse intent dataset, an end-to-end pipeline, and actionable insights for deploying LLM-assisted MANO in real-world 5G deployments at scale.

Abstract

Large language models (LLMs) are rapidly emerging in Artificial Intelligence (AI) applications, especially in the fields of natural language processing and generative AI. Not limited to text generation applications, these models inherently possess the opportunity to leverage prompt engineering, where the inputs of such models can be appropriately structured to articulate a model's purpose explicitly. A prominent example of this is intent-based networking, an emerging approach for automating and maintaining network operations and management. This paper presents semantic routing to achieve enhanced performance in LLM-assisted intent-based management and orchestration of 5G core networks. This work establishes an end-to-end intent extraction framework and presents a diverse dataset of sample user intents accompanied by a thorough analysis of the effects of encoders and quantization on overall system performance. The results show that using a semantic router improves the accuracy and efficiency of the LLM deployment compared to stand-alone LLMs with prompting architectures.
Paper Structure (25 sections, 6 figures, 1 table)

This paper contains 25 sections, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Background
  4. Methodology
  5. System Model
  6. Prompt Generation
  7. Seed Prompts
  8. Variability Prompts
  9. Paraphrased Prompts
  10. Semantic Router Implementation
  11. Threshold Tuning
  12. Experiment Data and Metrics
  13. Experiments
  14. Utterance Experiments
  15. Vocabulary Diversity Experiments
  16. ...and 10 more sections

Figures (6)

  • Figure 1: Semantic routing integration in 5G Core Intent-Based MANO Pipeline
  • Figure 2: Exploring the various prompt transformation methods used.
  • Figure 3: Semantic router with the Hugging Face encoder before and after route threshold training based on number of utterances provided.
  • Figure 4: Semantic router with the Hugging Face encoder before and after route threshold optimization when varying the utterance vocabulary composition.
  • Figure 5: Semantic router with the OpenAI encoder before and after route threshold training based on number of utterances provided.
  • ...and 1 more figures