[10p-N304-1~5] 18.4 Methods, Theory, Foundations, Others

1:30 PM - 1:45 PM JST(4:30 AM - 4:45 AM UTC)

[10p-N304-1] Literature retrieval and answer evaluation for NASICON-type solid electrolyte research using large language models

〇Toshiharu Ohnuma¹, Gaku Yamazaki¹, Hiroshi Murata¹, Takeshi Kobayashi¹ (1.CREIPI)

Keywords:

Retrieval-augmented generation (RAG),Large Language Model,NASICON-type solid electrolyte

To address challenges in literature surveys on NASICON-type solid electrolytes, a question-answering system based on retrieval-augmented generation (RAG) was developed by combining a structured database with large language models (LLMs). Retrieval performance and answer quality were comprehensively evaluated, clarifying the effects of database design and differences among LLM families on response characteristics.