Leveraging Fine-Tuned Retrieval-Augmented Generation with Long-Context Support: For 3GPP Standards

TL;DR

This paper introduces a fine-tuned retrieval-augmented generation system based on Phi-2 SLM, enhancing telecom standards processing by adaptive chunking, re-ranking, context expansion, and efficient fine-tuning, outperforming larger models like GPT-4.

Contribution

It presents a novel RAG system for telecom standards using Phi-2 SLM with techniques like semantic chunking, re-ranking, SelfExtend, and LoRA for improved performance and efficiency.

Findings

Outperforms existing QA methods in telecom domain

Exceeds GPT-4 performance despite smaller size

Demonstrates effective context expansion and fine-tuning techniques

Abstract

Recent studies show that large language models (LLMs) struggle with technical standards in telecommunications. We propose a fine-tuned retrieval-augmented generation (RAG) system based on the Phi-2 small language model (SLM) to serve as an oracle for communication networks. Our developed system leverages forward-looking semantic chunking to adaptively determine parsing breakpoints based on embedding similarity, enabling effective processing of diverse document formats. To handle the challenge of multiple similar contexts in technical standards, we employ a re-ranking algorithm to prioritize the most relevant retrieved chunks. Recognizing the limitations of Phi-2's small context window, we implement a recent technique, namely SelfExtend, to expand the context window during inference, which not only boosts the performance but also can accommodate a wider range of user queries and design requirements from customers to specialized technicians. For fine-tuning, we utilize the low-rank adaptation (LoRA) technique to enhance computational efficiency during training and enable effective fine-tuning on small datasets. Our comprehensive experiments demonstrate substantial improvements over existing question-answering approaches in the telecom domain, achieving performance that exceeds larger language models such as GPT-4 (which is about 880 times larger in size). This work presents a novel approach to leveraging SLMs for communication networks, offering a balance of efficiency and performance. This work can serve as a foundation towards agentic language models for networks.