Pith. sign in

REVIEW 3 cited by

From Molecules to Materials: Pre-training Large Generalizable Models for Atomic Property Prediction

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2310.16802 v2 pith:EQNKHU6K submitted 2023-10-25 cs.LG

From Molecules to Materials: Pre-training Large Generalizable Models for Atomic Property Prediction

classification cs.LG
keywords pre-trainingchemicaldomainsmodelspredictionpropertytaskstraining
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Foundation models have been transformational in machine learning fields such as natural language processing and computer vision. Similar success in atomic property prediction has been limited due to the challenges of training effective models across multiple chemical domains. To address this, we introduce Joint Multi-domain Pre-training (JMP), a supervised pre-training strategy that simultaneously trains on multiple datasets from different chemical domains, treating each dataset as a unique pre-training task within a multi-task framework. Our combined training dataset consists of $\sim$120M systems from OC20, OC22, ANI-1x, and Transition-1x. We evaluate performance and generalization by fine-tuning over a diverse set of downstream tasks and datasets including: QM9, rMD17, MatBench, QMOF, SPICE, and MD22. JMP demonstrates an average improvement of 59% over training from scratch, and matches or sets state-of-the-art on 34 out of 40 tasks. Our work highlights the potential of pre-training strategies that utilize diverse data to advance property prediction across chemical domains, especially for low-data tasks. Please visit https://nima.sh/jmp for further information.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. GFFMERGE: Efficient Merging of Graph Neural Force Fields and Beyond

    cs.LG 2026-06 unverdicted novelty 7.0

    GFFMERGE formulates GNN force field merging as a convex embedding-alignment problem with an analytical solution, recovering near joint-training performance on MD17, MD22, LiPS20 and other benchmarks while delivering 5...

  2. PaMM: Periodic Motif Memory for Atomistic Models with an Explicit Local-Structure Interface

    cs.LG 2026-05 conditional novelty 6.0

    PaMM augments an equivariant atomistic model with explicit pair and triplet motif memory tables, producing modest gains in energy and force MAE on OMAT benchmarks at fixed training budget.

  3. MatterSim: A Deep Learning Atomistic Model Across Elements, Temperatures and Pressures

    cond-mat.mtrl-sci 2024-05 unverdicted novelty 6.0

    MatterSim delivers a single deep learning force field that simulates inorganic materials across elements, 0-5000 K, and up to 1000 GPa with near first-principles accuracy for lattice dynamics, mechanics, and Gibbs fre...