[Paper Review] A Survey of AI for Materials Science: Foundation Models, LLM Agents, Datasets, and Tools

Minh–Hao Van, Prateek Verma+2 more|ArXiv.org|Jun 25, 2025

Machine Learning in Materials Science3 citations

TL;DR

This survey reviews foundation models, LLM agents, datasets, and tools enabling AI-driven materials science across six task areas, highlighting unimodal, multimodal, and agent-based models and future challenges.

ABSTRACT

Foundation models (FMs) are catalyzing a transformative shift in materials science (MatSci) by enabling scalable, general-purpose, and multimodal AI systems for scientific discovery. Unlike traditional machine learning models, which are typically narrow in scope and require task-specific engineering, FMs offer cross-domain generalization and exhibit emergent capabilities. Their versatility is especially well-suited to materials science, where research challenges span diverse data types and scales. This survey provides a comprehensive overview of foundation models, agentic systems, datasets, and computational tools supporting this growing field. We introduce a task-driven taxonomy encompassing six broad application areas: data extraction, interpretation and Q\&A; atomistic simulation; property prediction; materials structure, design and discovery; process planning, discovery, and optimization; and multiscale modeling. We discuss recent advances in both unimodal and multimodal FMs, as well as emerging large language model (LLM) agents. Furthermore, we review standardized datasets, open-source tools, and autonomous experimental platforms that collectively fuel the development and integration of FMs into research workflows. We assess the early successes of foundation models and identify persistent limitations, including challenges in generalizability, interpretability, data imbalance, safety concerns, and limited multimodal fusion. Finally, we articulate future research directions centered on scalable pretraining, continual learning, data governance, and trustworthiness.

Motivation & Objective

• Motivate the use of foundation models to enable cross-domain, data-rich materials discovery and design.
• Categorize unimodal, multimodal, and agent-based foundation models by task, architecture, and pretraining strategy.
• Summarize datasets, tools, and autonomous platforms that support AI workflows in materials science.

Proposed method

• Provide a task-driven taxonomy spanning six application areas: data extraction, interpretation and Q&A; atomistic simulation; property prediction; materials structure, design and discovery; process planning, discovery and optimization; and multiscale modeling.
• Review unimodal, multimodal, and agent-based foundation models and their cross-domain capabilities.
• Summarize representative models, datasets, and toolchains, and discuss successes, limitations, and future directions.

Experimental results

Research questions

• RQ1 What are the main foundation model architectures and pretraining strategies driving AI in materials science?
• RQ2 How do unimodal, multimodal, and agent-based models perform across core MatSci tasks and material classes?
• RQ3 What datasets and tools support scalable, autonomous AI workflows in materials discovery and design?
• RQ4 What are the persistent limitations and safety concerns hindering deployment of these models in practice?
• RQ5 What future directions will enable scalable pretraining, continual learning, data governance, and trustworthiness in MatSci AI?

Key findings

• GNoME discovered over 2.2 million new stable materials by combining graph neural networks with active-learning-driven DFT validation.
• MatterSim is trained on 17 million DFT-labeled structures and supports universal simulation across all elements and a wide range of temperatures and pressures.
• MACE-MP-0 achieves state-of-the-art accuracy for periodic systems while preserving equivariant inductive biases.
• Multimodal and cross-domain models (e.g., nach0, MultiMat, MatterChat) enable reasoning over structure, text, and spectral data.
• LLM agents (e.g., HoneyComb, MatAgent, ChatMOF, MatPilot) enable autonomous or semi-autonomous tasks across literature review, design, synthesis planning, and experimental workflows.
• A growing ecosystem of toolkits (e.g., Open MatSci ML Toolkit, FORGE) and autonomous platforms (e.g., A-Lab) supports integrated AI-powered materials workflows.