<ul><li>Material experts annotated nine distinct categories from 75 research paper abstracts as training data for LLMs.</li><li>The central node 'Material' in the knowledge graph is connected to nodes describing its nomenclature, composition, attributes, and application.</li><li>Normalization process was employed to ensure uniform representation of entities and facilitate effective training of LLMs.</li><li>The study focuses on energy materials, with 150,000 abstracts of peer-reviewed research articles in battery, solar cell, and catalyst fields.</li></ul>

This AI Reads Science Papers Like a Pro, Even When Humans Can’t Agree on the Words

Discover more