LLM4Mat-Bench: Benchmarking Large Language Models for Materials Property Prediction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Rubungo, Andre Niyongabo, Li, Kangming, Hattrick-Simpers, Jason, Dieng, Adji Bousso
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913591868784640
author Rubungo, Andre Niyongabo
Li, Kangming
Hattrick-Simpers, Jason
Dieng, Adji Bousso
author_facet Rubungo, Andre Niyongabo
Li, Kangming
Hattrick-Simpers, Jason
Dieng, Adji Bousso
contents Large language models (LLMs) are increasingly being used in materials science. However, little attention has been given to benchmarking and standardized evaluation for LLM-based materials property prediction, which hinders progress. We present LLM4Mat-Bench, the largest benchmark to date for evaluating the performance of LLMs in predicting the properties of crystalline materials. LLM4Mat-Bench contains about 1.9M crystal structures in total, collected from 10 publicly available materials data sources, and 45 distinct properties. LLM4Mat-Bench features different input modalities: crystal composition, CIF, and crystal text description, with 4.7M, 615.5M, and 3.1B tokens in total for each modality, respectively. We use LLM4Mat-Bench to fine-tune models with different sizes, including LLM-Prop and MatBERT, and provide zero-shot and few-shot prompts to evaluate the property prediction capabilities of LLM-chat-like models, including Llama, Gemma, and Mistral. The results highlight the challenges of general-purpose LLMs in materials science and the need for task-specific predictive models and task-specific instruction-tuned LLMs in materials property prediction.
format Preprint
id arxiv_https___arxiv_org_abs_2411_00177
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LLM4Mat-Bench: Benchmarking Large Language Models for Materials Property Prediction
Rubungo, Andre Niyongabo
Li, Kangming
Hattrick-Simpers, Jason
Dieng, Adji Bousso
Materials Science
Computation and Language
Large language models (LLMs) are increasingly being used in materials science. However, little attention has been given to benchmarking and standardized evaluation for LLM-based materials property prediction, which hinders progress. We present LLM4Mat-Bench, the largest benchmark to date for evaluating the performance of LLMs in predicting the properties of crystalline materials. LLM4Mat-Bench contains about 1.9M crystal structures in total, collected from 10 publicly available materials data sources, and 45 distinct properties. LLM4Mat-Bench features different input modalities: crystal composition, CIF, and crystal text description, with 4.7M, 615.5M, and 3.1B tokens in total for each modality, respectively. We use LLM4Mat-Bench to fine-tune models with different sizes, including LLM-Prop and MatBERT, and provide zero-shot and few-shot prompts to evaluate the property prediction capabilities of LLM-chat-like models, including Llama, Gemma, and Mistral. The results highlight the challenges of general-purpose LLMs in materials science and the need for task-specific predictive models and task-specific instruction-tuned LLMs in materials property prediction.
title LLM4Mat-Bench: Benchmarking Large Language Models for Materials Property Prediction
topic Materials Science
Computation and Language
url https://arxiv.org/abs/2411.00177