PulseAugur
EN
LIVE 12:01:51

AI Accelerates Materials Discovery with New Crystal Generation Techniques

By PulseAugur Editorial · [4 sources] ·

Three new research papers introduce advanced AI techniques for accelerating materials discovery. UNATE utilizes unsupervised atomic embeddings to improve crystal property prediction, showing significant gains with limited data. Crys-JEPA develops an energy-aware latent space for crystals, enabling more efficient screening and refinement of generated materials. Composable Crystals introduces a concept-based framework for controllable materials discovery, allowing for guided generation of novel and stable crystal structures. AI

IMPACT These advancements in AI-driven materials discovery could significantly speed up the development of new materials with desired properties.

RANK_REASON The cluster consists of three academic papers detailing novel AI methodologies for materials science research.

Read on arXiv cs.LG →

AI-generated summary · Google Gemini · from 4 sources. How we write summaries →

AI Accelerates Materials Discovery with New Crystal Generation Techniques

COVERAGE [4]

  1. arXiv cs.LG TIER_1 English(EN) · Laura Sol\`a-Garcia, \`Alex Sol\'e, Javier Ruiz-Hidalgo ·

    UNATE: UNsupervised ATomic Embedding for crystal structures property prediction

    arXiv:2605.25866v1 Announce Type: new Abstract: Accurately predicting crystal properties is critical for accelerating materials discovery, but it is often limited by scarce labeled data and costly theoretical calculations. To alleviate this, we propose UNATE (Unsupervised Atomic …

  2. arXiv cs.LG TIER_1 English(EN) · Nian Liu, Nikita Kazeev, Stephen Gregory Dale, Artem Maevskiy, Yuwei Zeng, Ryoji Kubo, Pengru Huang, Thomas Laurent, Yann LeCun, Kostya S. Novoselov, Xavier Bresson ·

    Crys-JEPA: Accelerating Crystal Discovery via Embedding Screening and Generative Refinement

    arXiv:2605.14759v2 Announce Type: replace Abstract: De novo crystal generation seeks to discover materials that are not merely realistic, but also stable and novel. However, most existing generative models are trained to maximize the likelihood of observed crystals, which encoura…

  3. arXiv cs.LG TIER_1 English(EN) · Nian Liu, Yuwei Zeng, Ryoji Kubo, Nikita Kazeev, Stephen Gregory Dale, Artem Maevskiy, Pengru Huang, Thomas Laurent, Kostya S. Novoselov, Xavier Bresson ·

    Composable Crystals: Controllable Materials Discovery via Concept Learning

    arXiv:2605.14769v2 Announce Type: replace Abstract: De novo crystal generation, a central task in materials discovery, aims to generate crystals that are simultaneously valid, stable, unique, and novel. Existing methods mainly rely on black-box stochastic sampling, providing limi…

  4. arXiv cs.LG TIER_1 English(EN) · Javier Ruiz-Hidalgo ·

    UNATE: UNsupervised ATomic Embedding for crystal structures property prediction

    Accurately predicting crystal properties is critical for accelerating materials discovery, but it is often limited by scarce labeled data and costly theoretical calculations. To alleviate this, we propose UNATE (Unsupervised Atomic Embedding), a framework that leverages structura…

RELATED ENTITIES

RELATED TOPICS