mlsanityData to 14 Oct 2026

CrystalDiT: A Diffusion Transformer for Crystal Generation

arXiv 2508.16614 ↗First posted 13 Aug 2025cs.LG cond-mat.mtrl-scicited by 6

On the map: Diffusion models → Flow matching, time series and science

Abstract (arXiv)

We present CrystalDiT, a diffusion transformer for crystal structure generation that achieves state-of-the-art performance by challenging the trend of architectural complexity. Instead of intricate, multi-stream designs, CrystalDiT employs a unified transformer that imposes a powerful inductive bias: treating lattice and atomic properties as a single, interdependent system. Combined with a periodic table-based atomic representation and a balanced training strategy, our approach achieves 8.78% SUN (Stable, Unique, Novel) rate on MP-20, substantially outperforming recent methods including FlowMM (4.21%) and MatterGen (3.66%). Notably, CrystalDiT generates 63.28% unique and novel structures while maintaining comparable stability rates, demonstrating that architectural simplicity can be more effective than complexity for materials discovery. Our results suggest that in data-limited scientific domains, carefully designed simple architectures outperform sophisticated alternatives that are prone to overfitting.

Key facts. Cited by 6 papers in the map (data through 14 Oct 2026). Placed in Flow matching, time series and science, one of 14 directions in Diffusion models. Builds on 13 of its 41 references that are in the map. First posted on arXiv 13 Aug 2025.

Builds on

13 of 41 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025All-atom Diffusion Transformers: Unified generative modelling of molecules and materials5 Mar
  2. 2024FlowLLM: Flow Matching for Material Generation with Large Language Models as Base Distributions30 Oct

Other references · 11

  1. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  2. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,101
  3. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,719
  4. 2024Scaling Diffusion Language Models via Adaptation from Autoregressive Modelscited by 274
  5. 2021Crystal Diffusion Variational Autoencoder for Periodic Material Generationcited by 270
  6. 2024FlowMM: Generating Materials with Riemannian Flow Matchingcited by 109
  7. 2025SymmCD: Symmetry-Preserving Crystal Generation with Diffusion Modelscited by 51
  8. 2025Wyckoff Transformer: Generation of Symmetric Crystalscited by 27
  9. 2025NTPP: Generative Speech Language Modeling for Dual-Channel Spoken Dialogue via Next-Token-Pair Predictioncited by 20
  10. 2025MSDformer: Multi-scale Discrete Transformer For Time Series Generationcited by 3
  11. 2023ETDock: A Novel Equivariant Transformer for Protein-Ligand Dockingcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 6.

6 papers.

  1. 2026Crystalite: A Lightweight Transformer for Efficient Crystal Modelingcited by 10
  2. 2026Multimodal Crystal Flow: Any-to-Any Modality Generation for Unified Crystal Modelingcited by 7
  3. 2026Crys-JEPA: Accelerating Crystal Discovery via Embedding Screening and Generative Refinementcited by 4
  4. 2026Fourier Transformers for Latent Crystallographic Diffusion and Generative Modelingcited by 1
  5. 2026Generating Symmetric Materials using Latent Flow Matchingcited by 0
  6. 2026GLASS: Global Latent Aggregation with Slot-based Set Decoding for Scalable All-Atom Crystal Generationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Generative Inverse Design of Crystal Structures via Diffusion Models with Transformers
  2. 2024Generative Design of Crystal Structures by Point Cloud Representations and Diffusion Model
  3. 2025LLM Meets Diffusion: A Hybrid Framework for Crystal Material Generation
  4. 2025CrystalGRW: Generative Modeling of Crystal Structures with Targeted Properties via Geodesic Random Walks
  5. 2025Kinetic Langevin Diffusion for Crystalline Materials Generation
  6. 2023Scalable Diffusion for Materials Generation
  7. 2025WyckoffDiff -- A Generative Diffusion Model for Crystal Symmetry
  8. 2026OxiGen: Oxidation-State-Aware Crystal Generation