LISA: A Lithium-Ion Solid-State Assistant using large language models for knowledge defragmentation in battery science and beyond

This work presents the development and implementation of a research assistant tool, Lithium-Ion Solid-State Assistant (LISA), based on the Retrieval-Augmented Generation (RAG) architecture. This assistant has been specifically tailored to enhance the retrieval and extraction of information from the domain of solid-state battery research. The system employs sophisticated retrieval techniques to efficiently identify the most pertinent document segments in response to researcher queries. The segments above are subsequently collated into prompts for a Large Language Model (LLM), which generates accurate, contextually enhanced responses to queries about solid-state battery-related subjects. This approach has the potential to markedly improve the accessibility and usability of a range of documentation, from project reports to complex scientific literature. The system provides researchers with a powerful tool to bridge disciplinary gaps, facilitate cross-disciplinary communication, accelerate knowledge discovery, and drive innovation in the field of solid-state batteries. A comprehensive evaluation was conducted to assess the system’s performance, with results indicating its potential to transform scientific research workflows. The system offers a robust open-source framework for future advancements in automated knowledge retrieval, understanding, and management, particularly in supporting the development of new materials.