Title here
Summary here
LeMaterial
Publications
Publications as part of LeMaterial projects.
2026
4 publications
Atompack
Atompack: A Storage and Distribution Layer for Read-Heavy Atomistic ML Training Datasets
An append-oriented storage format and distribution layer for compact atomistic datasets with fast shuffled reads in machine-learning training workflows.
LeMat-Rho
LeMat-Rho: High-Fidelity Charge Density Dataset for Atomistic Materials Modeling
A high-fidelity dataset of roughly 69,000 inorganic crystal structures computed with r2SCAN, including charge densities, Bader charges, forces, stresses, and energies.
TriForces
TriForces: Augmenting Atomistic GNNs for Transferable Representations
A model-agnostic three-stream framework that separates composition, structure, and interaction information to improve transferable atomistic representations.
Dope2Real
High-throughput ML Screening of Doped Cathode Active Materials
An automated high-throughput screening pipeline combining universal machine-learning interatomic potentials and DFT to evaluate dopants and co-dopants for cathode active materials, with experimental validation of promising candidates.
2025
4 publications
LeMat-GenBench
LeMat-GenBench: A Unified Evaluation Framework for Crystal Generative Models
A unified benchmark and evaluation suite for crystal generative models with standardized metrics and a reproducible leaderboard setup.
LeMat-Bulk
LeMat-Bulk: aggregating, and de-duplicating quantum chemistry materials databases
A unified, de-duplicated large-scale materials dataset across major sources with robust structure hashing for interoperability.
LeMat-Synth
LeMat-Synth: a multi-modal toolbox to curate broad synthesis procedure databases from scientific literature
A multi-modal extraction pipeline and dataset that structures synthesis procedures from 81k materials papers across methods and material classes.
LeMat-Traj
LeMat-Traj: A Scalable and Unified Dataset of Materials Trajectories for Atomistic Modeling
A large-scale, standardized trajectory dataset with over 120M atomic configurations to support transferable atomistic modeling workflows.