Design and challenges of open large language model frameworks (Open LLM): a systematic literature mapping

Abstract

Analyzing the frameworks of open large language models (OLLM) is essential to understanding how the management of these artificial intelligence (AI) models can be regulated. This study aims to analyze the evidence published from 2019 to 2024 regarding OLLM frameworks that integrate AI. Systematic mapping was the method for reviewing 227 articles published in the Scopus and Web of Science (WoS) databases. Inclusion, exclusion, and quality criteria filtered the papers to obtain the maximum relevant information. The analysis and classification of articles related to open LLM frameworks and models yielded significant findings per our research questions. The challenges identified were a) improving customization and accuracy through open LLMs, b) latency and efficiency challenges, c) the importance of reliability and security, and d) complex operational management (LLMOps). This review provides a framework for identifying the topic's state of the art and current and emerging research trends.