NEFMind: Parameter-Efficient Fine-Tuning of Open-Source LLMs for Telecom APIs Automation

2025-08-12

AI Summary

To address the escalating complexity of service discovery and management in 5G service-based architectures—driven by the rapid proliferation of network functions (NFs) and their associated APIs—this paper proposes NEFMind, a lightweight, domain-specialized framework. NEFMind leverages the open-source small language model Phi-2, enhanced via NEF API specification-driven synthetic data generation and quantized low-rank adaptation (QLoRA) fine-tuning. It further introduces a multi-dimensional evaluation protocol combining GPT-4 Ref Score and BERTScore for robust performance assessment. Evaluated on real-world 5G NF APIs, NEFMind achieves 98–100% accuracy in API call identification using only a 1.3B-parameter model, reduces communication overhead by 85%, and matches GPT-4’s functional performance. Its core contribution lies in establishing an efficient paradigm for adapting compact LMs to telecom-specific API semantics—enabling high deployment efficiency, low-latency inference, and scalable integration within resource-constrained 5G network environments.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Generation
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

The use of Service-Based Architecture in modern telecommunications has exponentially increased Network Functions (NFs) and Application Programming Interfaces (APIs), creating substantial operational complexities in service discovery and management. We introduce extit{NEFMind}, a framework leveraging parameter-efficient fine-tuning of open-source Large Language Models (LLMs) to address these challenges. It integrates three core components: synthetic dataset generation from Network Exposure Function (NEF) API specifications, model optimization through Quantized-Low-Rank Adaptation, and performance evaluation via GPT-4 Ref Score and BertScore metrics. Targeting 5G Service-Based Architecture APIs, our approach achieves 85% reduction in communication overhead compared to manual discovery methods. Experimental validation using the open-source Phi-2 model demonstrates exceptional API call identification performance at 98-100% accuracy. The fine-tuned Phi-2 model delivers performance comparable to significantly larger models like GPT-4 while maintaining computational efficiency for telecommunications infrastructure deployment. These findings validate domain-specific, parameter-efficient LLM strategies for managing complex API ecosystems in next-generation telecommunications networks.

Problem

• Automating telecom API discovery and management complexities
• Reducing communication overhead in 5G service-based architectures
• Optimizing LLMs for efficient telecom infrastructure deployment

Innovation

• Parameter-efficient fine-tuning of open-source LLMs
• Synthetic dataset generation from NEF API specs
• Quantized-Low-Rank Adaptation for model optimization