Saved in:
Bibliographic Details
Main Authors: Razmyslovich, Arina, Murasheva, Kseniia, Sedlova, Sofia, Capitaine, Julien, Dmitriev, Eugene
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.15055
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912328325267456
author Razmyslovich, Arina
Murasheva, Kseniia
Sedlova, Sofia
Capitaine, Julien
Dmitriev, Eugene
author_facet Razmyslovich, Arina
Murasheva, Kseniia
Sedlova, Sofia
Capitaine, Julien
Dmitriev, Eugene
contents We introduce Efficient LLM Token Extraction (ELTEX), a framework addressing the critical challenge of LLM domain specialization by systematically extracting and integrating domain indicators throughout synthetic data generation. Unlike approaches relying on implicit knowledge transfer, ELTEX explicitly leverages domain signals to maintain specialized knowledge integrity. In our cybersecurity case study, ELTEX-enhanced data enables a fine-tuned Gemma-2B model to achieve performance competitive with GPT-4o on blockchain cyberattack classification while reducing computational requirements. Our Google Sheets implementation makes ELTEX accessible to non-technical users. Our contributions include: (1) the ELTEX framework; (2) Google Sheets Add-on implementation; (3) empirical validation showing how ELTEX bridges performance gaps between small and large models; and (4) a synthetic dataset of 11,448 texts for blockchain cyberattack detection.
format Preprint
id arxiv_https___arxiv_org_abs_2503_15055
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ELTEX: A Framework for Domain-Driven Synthetic Data Generation
Razmyslovich, Arina
Murasheva, Kseniia
Sedlova, Sofia
Capitaine, Julien
Dmitriev, Eugene
Computation and Language
We introduce Efficient LLM Token Extraction (ELTEX), a framework addressing the critical challenge of LLM domain specialization by systematically extracting and integrating domain indicators throughout synthetic data generation. Unlike approaches relying on implicit knowledge transfer, ELTEX explicitly leverages domain signals to maintain specialized knowledge integrity. In our cybersecurity case study, ELTEX-enhanced data enables a fine-tuned Gemma-2B model to achieve performance competitive with GPT-4o on blockchain cyberattack classification while reducing computational requirements. Our Google Sheets implementation makes ELTEX accessible to non-technical users. Our contributions include: (1) the ELTEX framework; (2) Google Sheets Add-on implementation; (3) empirical validation showing how ELTEX bridges performance gaps between small and large models; and (4) a synthetic dataset of 11,448 texts for blockchain cyberattack detection.
title ELTEX: A Framework for Domain-Driven Synthetic Data Generation
topic Computation and Language
url https://arxiv.org/abs/2503.15055