Search

Materials artificial intelligence (MAI) has revolutionized the discovery, design, and optimization of new materials by leveraging machine learning algorithms to analyze complex datasets and predict properties with high accuracy. However, the rapid proliferation of MAI tools has raised critical questions about benchmarking practices, which are essential for evaluating model performance, ensuring reproducibility, and addressing ethical concerns. This narrative review examines current benchmarking frameworks in MAI, highlighting what is effectively measured—such as predictive accuracy and computational efficiency—and what is often overlooked —such as data bias, interpretability, fairness, and ethical implications. Drawing on recent advances in frameworks such as JARVIS-Leaderboard and Matbench, the review discusses challenges in data quality, reproducibility, and the integration of explainable AI (XAI) methods. It also explores active learning strategies for optimizing materials discovery under limited data conditions and proposes directions for more inclusive and transparent benchmarking. By synthesizing insights from diverse studies, this review aims to guide future MAI research toward robust, equitable, and ethically sound practices that accelerate innovation while mitigating risks.

The accelerating integration of artificial intelligence into materials discovery offers transformative potential for identifying novel compounds and optimizing properties at unprecedented speeds. Yet, this promise is tempered by persistent challenges in maintaining scientific rigor across computational workflows. This review employs a structured literature synthesis grounded exclusively in 35 peer-reviewed publications from 2017 to 2025, identified through targeted searches across Web of Science, Scopus, and arXiv using strings focused on scientific rigor, reproducibility in materials machine learning, reporting standards in materials informatics, methodological quality in AI-driven science, validation standards for materials AI, benchmarking in materials property prediction, replication in computational materials science, and quality assessment frameworks for AI in materials discovery, with inclusion criteria limited to studies addressing AI-assisted discovery practices and exclusion of purely experimental or non-computational works, following a PRISMA-style screening that yielded the final corpus after removing duplicates and off-topic items. Scientific rigor in this domain is understood as the systematic application of thorough, accurate, and transparent methods that ensure independent verification of AI-generated predictions while upholding honesty in reporting both positive and negative outcomes. Current practices in materials AI demonstrate growing sophistication in model development and data utilization but reveal inconsistent transparency in code and data sharing, limited replication efforts, and reliance on internal validation that falls short of broader scientific benchmarks, even as select studies begin to engage with established checklists and principles. Critical gaps emerge in the absence of tailored materials-AI rigor frameworks, the rarity of external experimental validation, and insufficient community mechanisms for enforcing completeness in reporting, which collectively risk resource misallocation and diminished confidence in AI-driven claims. Targeted recommendations for authors, reviewers, journals, and funders emphasize mandatory code and data deposition, comprehensive hyperparameter disclosure, and cultural shifts toward valuing replication and negative results to bridge these deficiencies and elevate the field’s overall integrity.

Benchmarking has become central to the development of machine learning interatomic potentials (MLIPs), yet the epistemic reliability of reported comparisons remains insufficiently scrutinized. This review synthesizes prevailing practices and demonstrates that current evaluation protocols systematically misrepresent model progress. A coherent taxonomy of methodological failure emerges, spanning opaque data handling, structurally flawed train–test partitioning, restricted metric selection, weak baseline construction, limited reproducibility, and the near absence of extrapolation analysis. Under these conditions, widely cited performance indicators—such as sub-10 meV/atom energy MAE or sub-0.1 eV/Å force RMSE—primarily capture interpolation within constrained training distributions, offering limited insight into generalization, dynamical stability, or deployment viability. A related deficiency lies in the systematic exclusion of physically and computationally salient regimes, including long-range interactions, finite-temperature behavior, low-symmetry and disordered structures, calibrated uncertainty, defect-rich configurations, and explicit cost–accuracy trade-offs. Existing benchmark suites, including Materials Project–derived datasets, QM9 adaptations, COMP6, and bespoke collections, inherit these constraints, reinforcing an evaluative paradigm that privileges narrow optimization over robust, application-relevant performance. Recasting benchmark outcomes as contingent on methodological design rather than intrinsic model capability reveals how evaluation choices implicitly structure model rankings. In response, this work advances a set of directly implementable standards: diversified splitting strategies, distribution-aware multi-metric reporting, transparent baseline inclusion, controlled extrapolation regimes, complete reproducibility artifacts, and normalized cost accounting. Aligning benchmarking practice with these principles is necessary to transition from incremental leaderboard gains toward reliable and transferable interatomic potentials for materials discovery.