Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

NextG-GPT: Leveraging GenAI for Advancing Wireless Networks and Communication Research

Ahmad M. Nazar, Mohamed Y. Selim, Daji Qiao, Hongwei Zhang

TL;DR

NextG-GPT introduces a domain-specific, RAG-augmented LLM framework for advancing wireless network research. By integrating telecom-focused datasets with an end-to-end workflow (data preprocessing, chunking, embeddings, semantic search, and context-aware generation) and evaluating multiple open-source LLMs, it demonstrates substantial improvements in answer relevancy, contextual recall, and factual correctness over vanilla models. The system is validated in the ARA Wireless Living Lab, enabling real-time experimentation, network optimization, and intelligent debugging, with potential for autonomous, self-optimizing operations in future multi-modal deployments. These results highlight the value of domain-tuned knowledge bases and retrieval-augmented generation in delivering precise, standards-aligned guidance for O-RAN, 5G/6G research, and telecom experimentation workflows.

Abstract

Artificial intelligence (AI) and wireless networking advancements have created new opportunities to enhance network efficiency and performance. In this paper, we introduce Next-Generation GPT (NextG-GPT), an innovative framework that integrates retrieval-augmented generation (RAG) and large language models (LLMs) within the wireless systems' domain. By leveraging state-of-the-art LLMs alongside a domain-specific knowledge base, NextG-GPT provides context-aware real-time support for researchers, optimizing wireless network operations. Through a comprehensive evaluation of LLMs, including Mistral-7B, Mixtral-8x7B, LLaMa3.1-8B, and LLaMa3.1-70B, we demonstrate significant improvements in answer relevance, contextual accuracy, and overall correctness. In particular, LLaMa3.1-70B achieves a correctness score of 86.2% and an answer relevancy rating of 90.6%. By incorporating diverse datasets such as ORAN-13K-Bench, TeleQnA, TSpec-LLM, and Spec5G, we improve NextG-GPT's knowledge base, generating precise and contextually aligned responses. This work establishes a new benchmark in AI-driven support for next-generation wireless network research, paving the way for future innovations in intelligent communication systems.

NextG-GPT: Leveraging GenAI for Advancing Wireless Networks and Communication Research

TL;DR

NextG-GPT introduces a domain-specific, RAG-augmented LLM framework for advancing wireless network research. By integrating telecom-focused datasets with an end-to-end workflow (data preprocessing, chunking, embeddings, semantic search, and context-aware generation) and evaluating multiple open-source LLMs, it demonstrates substantial improvements in answer relevancy, contextual recall, and factual correctness over vanilla models. The system is validated in the ARA Wireless Living Lab, enabling real-time experimentation, network optimization, and intelligent debugging, with potential for autonomous, self-optimizing operations in future multi-modal deployments. These results highlight the value of domain-tuned knowledge bases and retrieval-augmented generation in delivering precise, standards-aligned guidance for O-RAN, 5G/6G research, and telecom experimentation workflows.

Abstract

Artificial intelligence (AI) and wireless networking advancements have created new opportunities to enhance network efficiency and performance. In this paper, we introduce Next-Generation GPT (NextG-GPT), an innovative framework that integrates retrieval-augmented generation (RAG) and large language models (LLMs) within the wireless systems' domain. By leveraging state-of-the-art LLMs alongside a domain-specific knowledge base, NextG-GPT provides context-aware real-time support for researchers, optimizing wireless network operations. Through a comprehensive evaluation of LLMs, including Mistral-7B, Mixtral-8x7B, LLaMa3.1-8B, and LLaMa3.1-70B, we demonstrate significant improvements in answer relevance, contextual accuracy, and overall correctness. In particular, LLaMa3.1-70B achieves a correctness score of 86.2% and an answer relevancy rating of 90.6%. By incorporating diverse datasets such as ORAN-13K-Bench, TeleQnA, TSpec-LLM, and Spec5G, we improve NextG-GPT's knowledge base, generating precise and contextually aligned responses. This work establishes a new benchmark in AI-driven support for next-generation wireless network research, paving the way for future innovations in intelligent communication systems.

Paper Structure

This paper contains 42 sections, 4 equations, 3 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. RAG-LLM Assistants in Wireless System Development
  4. LLM Applications in Telecommunications
  5. Innovation of NextG-GPT
  6. Implementation of NextG-GPT
  7. Datasets
  8. Embeddings
  9. LLMs
  10. NextG-GPT Workflow
  11. Data Preprocessing
  12. Data Chunking
  13. Embedding Data
  14. Knowledge Base Creation
  15. Semantic Search
  16. ...and 27 more sections

Figures (3)

  • Figure 1: NextG-GPT workflow where Steps 1-4 involve RAG initialization; on prompting, Steps 5-6 show the semantic search and result ranking mechanism, and Step 7 shows the response generation where the user prompt and appropriate contexts are provided to the LLM to generate a response.
  • Figure 2: Evaluation Metrics of NextG-GPT where A) shows answer relevancy scores, B) shows context recall scores, C) includes vanilla LLM and RAG-LLM answer correctness scores, and D) shows answer faithfulness.
  • Figure 3: Comparison of Vanilla LLaMa and NextG-GPT responses to an ARA-specific O-RAN experiment setup query. NextG-GPT provides accurate, structured instructions with validated 3GPP and O-RAN references, while Vanilla LLaMa gives a generic and partially incorrect response, as shown in the red text.