In recent years, the introduction of Large Language Model (LLM) has revolutionised the approach to document analysis and management, making it possible to process increasingly relevant and contextual transactions. However, the use of these technologies brings with it significant challenges in terms of privacy, security and data control.
Against this backdrop is thehe new study AI4Cyberwhich presents the development of a system capable of querying business documents via theRAG approach (Retrieval Augmented Generation), capable of combine the power of LLMs with the ability to extract information from relevant textual sources.
The solution presented in the study employs Large Language Model (LLM) running locally, with the aim of offering a private AI assistant that guarantees the confidentiality, integrity and availability of the data processed.
The model that showed a better trade-off between efficiency and accuracy is Phi4, chosen after several comparative tests with Llama 3.1 and Mistral. An intuitive user interface and the ability to manage not only structured and unstructured documents, but also scanned documents thanks to the integration of theOCRmaking the system versatile and easily adaptable to companies' operational needs.
If you wish to learn more, here is the link to our studio complete.
In addition, you can subscribe to the specific mailing list Cyber Studios by Tinexta Defence, to receive updates on upcoming research:
