ELTEX

This is the official repository for the paper "ELTEX: A Framework for Domain-Driven Synthetic Data Generation". ELTEX is a framework that generates high-quality synthetic training data by systematically extracting domain-specific knowledge, enabling resource-efficient models to achieve GPT-4-competitive performance in specialized domains.

ELTEX Sheets: Build Synthetic Datasets in Google Sheets

We've developed a full-featured Google Sheets add-on that brings our framework directly to users! Currently used internally and coming soon for public release! Stay tuned for updates

Key Features

• No-Code Data Generation: Generate and manage domain-specific datasets directly in Google Sheets
• User-Friendly: Designed for students, teachers, and domain experts with minimal programming experience
• Full Workflow Support: From parameters → indicators → synthetic data → QA → export, all in your spreadsheet

🤗 Datasets

• 12k Synthetic Social Media Messages for Early Cyberattack Detection on Blockchain

Attack Type Distribution

Attack Vectors	Seed Examples
Social Engineering & Phishing	Credential theft, wallet phishing
Smart Contract Exploits	Token claim vulnerabilities, flash loans
Exchange Security Breaches	Hot wallet compromises, key theft
DeFi Protocol Attacks	Liquidity pool manipulation, bridge exploits

For more about Cyberattack Vectors, read Attack Vectors Wiki by DNI

---

We're actively working on expanding our research with additional domain-specific datasets!

Citation

@misc{razmyslovich2025eltexframeworkdomaindrivensynthetic,
      title={ELTEX: A Framework for Domain-Driven Synthetic Data Generation}, 
      author={Arina Razmyslovich and Kseniia Murasheva and Sofia Sedlova and Julien Capitaine and Eugene Dmitriev},
      year={2025},
      eprint={2503.15055},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2503.15055}, 
}