CrystalREPA: Transferring Physical Priors from Universal MLIPs to Crystal Generative Models

• URL: http://arxiv.org/abs/2605.08960v1
• Date: Sat, 09 May 2026 13:56:23 GMT
• Title: CrystalREPA: Transferring Physical Priors from Universal MLIPs to Crystal Generative Models
• Abstract summary: We show a substantial representation gap between state-of-the-art crystal generative models and pretrained universal machine learning interatomic potentials.<n>We propose Crystal REPresentation Alignment (CrystalREPA), a plug-and-play framework that aligns the atom-wise hidden states of generative encoders with frozen MLIP representations.<n>CrystalREPA consistently improves the thermodynamic stability, structural validity, and structural fidelity of generated crystals.
• Score: 10.446104330081189
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Crystal generative models mainly learn what stable crystals look like, with little explicit supervision for what makes them stable. We reveal a substantial representation gap between state-of-the-art crystal generative models and pretrained universal machine learning interatomic potentials (MLIPs) via energy probing, and show this gap can be closed by a simple training-time alignment. We propose Crystal REPresentation Alignment (CrystalREPA), a plug-and-play framework that aligns the atom-wise hidden states of generative encoders with frozen MLIP representations through an element-aware contrastive objective, transferring stability-aware atomistic priors with marginal training overhead and no additional inference cost. Across three generative frameworks, ten MLIP teachers, and two benchmark datasets, CrystalREPA consistently improves the thermodynamic stability, structural validity, and structural fidelity of generated crystals. Equally important, we find that an MLIP's transfer effectiveness is poorly predicted by its accuracy on standard leaderboards (e.g., Matbench Discovery) but strongly predicted by the distinguishability of its atom-wise representation space, yielding a practical, accuracy-independent criterion for selecting MLIP teachers for generative transfer.

Related papers

• Symmetry-Breaking De Novo Crystal Generation via Markovian Jump Diffusion [50.136954476803766]
  We propose a novel diffusion-based framework that generates full structure specifications by reversing from the lowest-symmetry priors.<n>Our model, dubbed emphSymmetry-breaking Crystal Diffusion (SbCD), introduces a principled approach to explicitly incorporate inter-space-group transitions into the generative process.<n>In de novo generation experiments on MP20 and MPTS-52, SbCD outperforms its symmetry-preserving counterpart by a substantial margin.
  arXiv  Detail & Related papers  (2026-08-13T16:41:48Z)
• Composable Crystals: Controllable Materials Discovery via Concept Learning [25.01951802810537]
  We introduce a concept-based compositional framework for crystal generation.<n>We train a vector-quantized variational autoencoder to automatically discover a shared set of reusable crystal concepts.<n>By recombining such concepts, our framework enables controllable exploration of novel crystals beyond the training distribution.
  arXiv  Detail & Related papers  (2026-05-14T12:36:23Z)
• CrystalReasoner: Reasoning and RL for Property-Conditioned Crystal Structure Generation [7.340047961454058]
  We propose CrystalReasoner (CrysReas), an end-to-end LLM framework that generates crystal structures from natural language instructions through reasoning and alignment.<n>CrysReas introduces physical priors as thinking tokens, which include crystallographic symmetry, local coordination environments and predicted physical properties before generating atomic coordinates.<n>For property-conditioned tasks, we design task-specific reward functions and train specialized models for discrete constraints and continuous properties.
  arXiv  Detail & Related papers  (2026-05-14T04:08:51Z)
• OXtal: An All-Atom Diffusion Model for Organic Crystal Structure Prediction [63.318434943975255]
  We introduce OXtal, a large-scale 100M parameter all-atom diffusion model that learns the conditional joint distribution over intramolecular conformations and periodic packing.<n>By leveraging a large dataset of 600K experimentally validated crystal structures, OXtal achieves orders-of-improvement over prior ab initio machine learning CSP methods.<n> OXtal attains over 80% packing similarity rate, demonstrating its ability to model both thermodynamic and kinetic regularities of molecular crystallization.
  arXiv  Detail & Related papers  (2025-12-07T20:46:30Z)
• Beyond Structure: Invariant Crystal Property Prediction with Pseudo-Particle Ray Diffraction [23.89478649565297]
  Crystal property prediction is computationally prohibitive to solve exactly for large many-body systems using traditional density functional theory.<n>We introduce PRDNet that leverages unique reciprocal-space diffraction besides graph representations.<n>Extensive experiments are conducted on Materials Project, JARVIS-DFT, and MatBench, demonstrating that the proposed model achieves state-of-the-art performance.
  arXiv  Detail & Related papers  (2025-09-26T02:30:23Z)
• CrystalDiT: A Diffusion Transformer for Crystal Generation [52.45780803467369]
  We present CrystalDiT, a diffusion transformer for crystal structure generation that achieves state-of-the-art performance.<n>CrystalDiT employs a unified transformer that imposes a powerful inductive bias: treating lattice and atomic properties as a single, interdependent system.
  arXiv  Detail & Related papers  (2025-08-13T12:53:32Z)
• Active Learning for Conditional Inverse Design with Crystal Generation and Foundation Atomic Models [2.8408587358426725]
  We present an active learning framework that combines crystal generation models and foundation atomic models to enhance the accuracy and efficiency of inverse design.<n>Our framework is general to accommodate different crystal generation and foundation atomic models, and establishes a scalable approach for AI-driven materials discovery.
  arXiv  Detail & Related papers  (2025-02-24T09:15:38Z)
• CrystalGRW: Generative Modeling of Crystal Structures with Targeted Properties via Geodesic Random Walks [1.2141052067494946]
  We introduce CrystalGRW, a diffusion-based generative model that can predict stable phases validated by density functional theory.<n>CrystalGRW demonstrates the ability to generate realistic crystal structures that are close to their ground states with accuracy comparable to existing models.<n>These features help accelerate materials discovery and inverse design by offering stable, symmetry-consistent crystal candidates for experimental validation.
  arXiv  Detail & Related papers  (2025-01-15T18:26:35Z)
• Scalable Diffusion for Materials Generation [99.71001883652211]
  We develop a unified crystal representation that can represent any crystal structure (UniMat) UniMat can generate high fidelity crystal structures from larger and more complex chemical systems. We propose additional metrics for evaluating generative models of materials.
  arXiv  Detail & Related papers  (2023-10-18T15:49:39Z)
• Latent Conservative Objective Models for Data-Driven Crystal Structure Prediction [62.36797874900395]
  In computational chemistry, crystal structure prediction is an optimization problem. One approach to tackle this problem involves building simulators based on density functional theory (DFT) followed by running search in simulation. We show that our approach, dubbed LCOMs (latent conservative objective models), performs comparably to the best current approaches in terms of success rate of structure prediction.
  arXiv  Detail & Related papers  (2023-10-16T04:35:44Z)
• Crystal-GFN: sampling crystals with desirable properties and constraints [103.79058968784163]
  We introduce Crystal-GFN, a generative model of crystal structures that sequentially samples structural properties of crystalline materials. In this paper, we use as objective the formation energy per atom of a crystal structure predicted by a new proxy machine learning model trained on MatBench. The results demonstrate that Crystal-GFN is able to sample highly diverse crystals with low (median -3.1 eV/atom) predicted formation energy.
  arXiv  Detail & Related papers  (2023-10-07T21:36:55Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.