# Deep Generative Crystal Structure Prediction: A Benchmark Study and a Controlled Test of Prototype Dependence

arXiv: https://arxiv.org/abs/2609.26502 · HTML: https://mlsanity.com/paper/2609.26502/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 39 references that are in the map.
- First posted on arXiv 22 Sep 2026.

## Builds on (most cited first, 12 of 12)

- [Crystal Diffusion Variational Autoencoder for Periodic Material Generation](https://mlsanity.com/paper/2110.06197/) (2021)
- [Crystal Structure Prediction by Joint Equivariant Diffusion](https://mlsanity.com/paper/2309.04475/) (2023)
- [SymmCD: Symmetry-Preserving Crystal Generation with Diffusion Models](https://mlsanity.com/paper/2502.03638/) (2025)
- [Open Materials Generation with Stochastic Interpolants](https://mlsanity.com/paper/2502.02582/) (2025)
- [LeMat-GenBench: A Unified Evaluation Framework for Crystal Generative Models](https://mlsanity.com/paper/2512.04562/) (2025)
- [Periodic Materials Generation using Text-Guided Joint Diffusion Model](https://mlsanity.com/paper/2503.00522/) (2025)
- [Vector Field Oriented Diffusion Model for Crystal Material Generation](https://mlsanity.com/paper/2401.05402/) (2023)
- [Equivariant Diffusion for Crystal Structure Prediction](https://mlsanity.com/paper/2512.07289/) (2025)
- [CrystalGRW: Generative Modeling of Crystal Structures with Targeted Properties via Geodesic Random Walks](https://mlsanity.com/paper/2501.08998/) (2025)
- [Substitution-Based Analysis of Structural Novelty for Generative Models of Materials](https://mlsanity.com/paper/2606.23166/) (2026)
- [Introducing physics-informed generative models for targeting structural novelty in the exploration of chemical space](https://mlsanity.com/paper/2510.23181/) (2025)
- [Transport Novelty Distance: A Distributional Metric for Evaluating Material Generative Models](https://mlsanity.com/paper/2512.09514/) (2025)

## Abstract (arXiv)

Deep generative models are widely reported to enable de novo crystal structure prediction (CSP), but their capability has not been measured consistently against template-based methods. We evaluate 12 representative generative CSP models, spanning latent-variable, diffusion, flow-matching, autoregressive, and manifold random-walk architectures, against TCSP 2.0 on 180 test structures and a leakage-controlled subset of 46. All methods use identical structure-matching, symmetry, and consensus criteria. Template retrieval is the strongest single method, reaching 68.3% top-1 success; symmetry-aware EquiCSP (66.4%) and Uni-3DAR (62.9%) form the next tier. However, comparison with TCSP 2.0 shows that most structures correctly predicted by generative models are also correctly predicted by template substitution. Thus, the set of structures uniquely reachable by generation is small, limiting its practical advantage for discovering structures outside existing prototype libraries. To test the source of this performance, we removed entire stoichiometric prototype families from the training set and retrained the strongest generative model. Accuracy declined by 50-78% across four families, establishing that performance is substantially prototype-dependent. A small minority of structures survived removal of their prototype family, demonstrating a real but limited retrieval-independent predictive capacity. Present generative CSP models therefore function largely as implicit, softer-edged prototype libraries rather than genuinely de novo predictors. Enlarging this residual capacity, rather than aggregate match rate alone, is the central open problem.