ProvMind: provenance-grounded reasoning for materials synthesis

Abstract

Materials process optimization requires reasoning over routes, conditions, tools and material-flow dependencies, yet most computational formulations flatten synthesis procedures into text or ordered steps. We introduce MatProcBench, a provenance-grounded benchmark constructed from literature-mined MatPROV graphs, to evaluate seven process-reasoning tasks spanning route continuity, step-level variable inference and global material-flow consistency under both same-split and shift-aware evaluation. Its strict dual-distribution-shift (dual-OOD) split simultaneously tests transfer to later publications and to a held-out material class. We further introduce ProvMind, a process-memory reasoning framework that retrieves analogous training processes, converts them into provenance-aware option-level compatibility scores, and uses a language model for constrained final decision making. ProvMind achieves 52.84% accuracy on the dual-OOD split, exceeding the aggregate accuracies of the evaluated prompting, retrieval-augmented and supervised fine-tuning baselines. These results show the potential of provenance-grounded process memory under controlled distribution shift, while complex route and joint-condition tasks remain challenging.