The UniMate dataset is a mechanical metamaterial benchmark dataset released by the UNIMATE team in 2025. The related paper results are “UNIMATE: A Unified Model for Mechanical Metamaterial Generation, Property Prediction, and Condition Confirmation", has been accepted by ICML 2025. This dataset is based on the topological structure of Lumpe & Stankovic (2021). It is the first high-quality dataset that uniformly includes three-dimensional structure, density information and mechanical properties. It aims to solve the problem that existing metamaterial datasets lack multi-modal and multi-task coverage, and is suitable for tasks such as structure generation, property prediction, and multi-modal modeling.

The UniMate dataset contains 15,000 samples, each of which contains three-dimensional topological structure, density information and its corresponding homogenized mechanical properties, covering low-density (ρ=0.1) to medium-density (ρ=0.5) scenarios, and the topological structure satisfies cubic symmetry and periodicity.