11 October 2023
MatChat model fine-tuned from LLaMA2-7B using materials synthesis data
13,878 synthesis processes extracted from papers used for training
MatChat shows strong reasoning abilities for material synthesis
Both generative and inferential capabilities demonstrated
Web platform developed for easy access to MatChat model
Large language model for materials science
This paper trains a large language model called MatChat to predict inorganic material chemical synthesis pathways. The authors fine-tune the open-source LLaMA2-7B model using a dataset of 13,878 synthesis processes extracted from papers. Experiments show MatChat can reason about and generate high-quality synthesis methods not seen during training.
No comments yet, be the first to start the conversation...
Sign up to comment on this paper