Enhancing Large Language Models for Telecom Networks Using Retrieval-Augmented Generation
Bibliographic record
Abstract
This paper presents a comprehensive approach for fine-tuning large language models (LLMs) for domain-specific tasks in the telecommunications field. We utilize a dataset with 1,827 multiple-choice questions (MCQs) from 3GPP standard documents. A publicly available LLM named "Phi-2" is used to answer the MCQs correctly. We develop a Retrieval-Augmented Generation (RAG) pipeline to improve Phi-2 model's performance. The RAG pipeline comprises document segmentation, synthetic question-answer (QA) generation, custom fine-tuning of the embedding model, and incremental fine-tuning of Phi-2. Our experiments show that accuracy greatly increased by combining all the above-mentioned steps in the RAG pipeline. The proposed approach outperforms the baseline Phi-2 model by 45.20% in terms of accuracy. This study identifies the limitations of instruction fine-tuning in specialized fields and explores the possibility of using sophisticated data processing with fine-tuned models to improve performance even more.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.006 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.003 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".