MP-20

The MP-20 dataset is a benchmark containing a diverse set of crystal structures used to evaluate the stability and novelty of generative models in crystal discovery.

Papers using MP-20 (7)

All-atom Diffusion Transformers: Unified generative modelling of molecules and materials2025 · 6 citesTransformer-Enhanced Variational Autoencoder for Crystal Structure Prediction2025 · 2 citesAll that structure matches does not glitter2025 · 1 citesCrys-JEPA: Accelerating Crystal Discovery via Embedding Screening and Generative Refinement2026Composable Crystals: Controllable Materials Discovery via Concept Learning2026Multimodal Crystal Flow: Any-to-Any Modality Generation for Unified Crystal Modeling2026Rao-Blackwell Gradient Estimators for Equivariant Denoising Diffusion2025