{"total":20,"items":[{"citing_arxiv_id":"2606.23214","ref_index":109,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Universal Interatomic Potentials as Configuration-Space Generators for One-Shot and Iterative Fine-Tuning of Ab Initio-Accurate Material-Specific Models","primary_cat":"cond-mat.mtrl-sci","submitted_at":"2026-06-22T12:01:35+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Universal MLIPs serve as configuration generators whose DFT-relabeled subsamples enable one-shot or iterative training of material-specific MLIPs that recover accurate reactive energy profiles with 600-2000 DFT calculations.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.28905","ref_index":18,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"DFT Accuracy on Crystal Structure Prediction with Machine Learning Interatomic Potentials","primary_cat":"physics.chem-ph","submitted_at":"2026-05-27T16:31:39+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"CSP-MACE-Å matches PBE DFT with Neumann-Perrin and B86bPBE-XDM DFT performance on two CSP benchmark sets while running much faster and benefiting from harmonic free-energy reranking.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.19124","ref_index":270,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Atomistic Modeling of Chemical Disorder in Materials: Bridging Classical Methods and AI-Assisted Approaches","primary_cat":"cond-mat.mtrl-sci","submitted_at":"2026-05-18T21:22:47+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":3.0,"formal_verification":"none","one_line_summary":"A review of classical and AI-assisted methods for modeling chemical disorder in atomistic simulations of alloys and complex materials.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.18404","ref_index":21,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"JanusPipe: Efficient Pipeline Parallel Training for Machine Learning Interatomic Potentials","primary_cat":"cs.DC","submitted_at":"2026-05-18T13:45:58+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"JanusPipe introduces SymFold and WaveK to enable efficient 3D-parallel training for conservative MLIPs, reporting 1.51x and 1.45x average throughput gains over 1F1B and Hanayo baselines on 32 GPUs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.17147","ref_index":73,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Spatial statistics for screening molecular structures","primary_cat":"cond-mat.mtrl-sci","submitted_at":"2026-05-16T20:42:23+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Spatial statistics on voxelized structures using FFT correlations and PCA yield low-dimensional convex features that support accurate predictions with as few as 10 training samples.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"Finally, PCA projects these high-dimensional statistics onto a low-dimensional, convex manifold that serves as input to lean CNNs or GPR models. The high-dimensional nature of the computed two-point correlations necessitates a robust dimen- sionality reduction strategy (see Figure 4 to construct a tractable input space for predictive modeling. Principal Component Analysis (PCA) [73] is systematically applied to the ensemble of two-point statistics, projecting the high-dimensional spatial data onto an orthogonal, low-dimensional feature space [52] that retains the vast majority of the structural variance. This chemically agnostic, principal component manifold serves as a universal structural descriptor that seamlessly integrates with highly"},{"citing_arxiv_id":"2605.13788","ref_index":32,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Force-Aware Neural Tangent Kernels for Scalable and Robust Active Learning of MLIPs","primary_cat":"cs.LG","submitted_at":"2026-05-13T17:08:37+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Force-aware Neural Tangent Kernels combined with chunked acquisition provide scalable and distribution-robust active learning for MLIPs, outperforming baselines on OC20 and remaining competitive on other benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.10458","ref_index":21,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"QT-Net: Rethinking Evaluation of AI Models in Atomic Chemical Space","primary_cat":"cs.LG","submitted_at":"2026-05-11T12:29:10+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"QT-Net predicts atomic electron populations and multipoles via a new SOAP-cluster held-out test, improving molecular property prediction and recovering QM9 dipole moments from per-atom outputs.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"framework, molecular dipole moments computed from QT-Net's per-atom outputs recover the ground-truth values reported in QM9. We release all code and data, including a JAX implementation of QT-Net, to support the broader use of learned QTA properties as inductive biases for atomic-scale molecular machine learning. Preprint. arXiv:2605.10458v1 [cs.LG] 11 May 2026 1 Introduction Molecular property prediction[ 21, 31], accelerated simulations[ 20, 22], andde novomolecular design[40, 15] are among the main drivers of new artificial intelligence (AI) for science methods. However, out-of-distribution (OOD) generalization and broad sampling are among the most pressing challenges in the field. [ 11, 35, 2] Molecular representations built from first principles atomic"},{"citing_arxiv_id":"2605.09394","ref_index":18,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Systematic Fine-Tuning of MACE Interatomic Potentials for Catalysis","primary_cat":"physics.chem-ph","submitted_at":"2026-05-10T07:43:42+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Fine-tuned MACE MLIPs achieve lower mean absolute errors on catalytic reaction energies and barriers than from-scratch models, with a large fine-tuned model performing best on both metallic and oxide systems including out-of-distribution cases.","context_count":1,"top_context_role":"dataset","top_context_polarity":"use_dataset","context_text":"might be a limitation, as performance remains heavily dependent on data availability and quality, which may restrict their generalizability when determining bothE r and Ea Wood, et al. [15] have used a mixture of experts model to train a family of universal models (UMA) on a combined dataset comprised of OC22 (oxide electrocatalysts) [16], OC25 (solid-liquid interfaces) [17], OMOL (molecules) [18], OpenDac [19] and OMAT (bulk inorganic materials)[20]and is able to estimateE r within 0.1 eV of DFT, for 74 % ofE r reported in AdsorbML task. In another effort, Batatia, et al. pretrained a base model to OMAT dataset [20], then fine-tuned it using a multi-head replay with 10% of the OMAT to avoid catastrophic forgetting, and 6 other heads that consists of"},{"citing_arxiv_id":"2605.08885","ref_index":33,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Compact SO(3) Equivariant Atomistic Foundation Models via Structural Pruning","primary_cat":"cs.LG","submitted_at":"2026-05-09T11:07:03+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Structural pruning of SO(3) equivariant atomistic models from large checkpoints yields 1.5-4x fewer parameters and 2.5-4x less pre-training compute than small models trained from scratch, while outperforming them on most Matbench Discovery metrics and downstream tasks.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"molecules and peptides for training machine learning potentials.Scientific Data, 10(1):11, 2023. [32] Luis Barroso-Luque, Muhammed Shuaibi, Xiang Fu, Brandon M Wood, Misko Dzamba, Meng Gao, Ammar Rizvi, C Lawrence Zitnick, and Zachary W Ulissi. Open materials 2024 (omat24) inorganic materials dataset and models.arXiv preprint arXiv:2410.12771, 2024. [33] Daniel S Levine, Muhammed Shuaibi, Evan Walter Clark Spotte-Smith, Michael G Taylor, Muhammad R Hasyim, Kyle Michel, Ilyes Batatia, Gábor Csányi, Misko Dzamba, Peter Eastman, et al. The open molecules 2025 (omol25) dataset, evaluations, and models.arXiv preprint arXiv:2505.08762, 2025. [34] Vahe Gharakhanyan, Luis Barroso-Luque, Yi Yang, Muhammed Shuaibi, Kyle Michel, Daniel S Levine,"},{"citing_arxiv_id":"2605.03964","ref_index":29,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Pretrained Model Representations as Acquisition Signals for Active Learning of MLIPs","primary_cat":"cs.LG","submitted_at":"2026-05-05T16:48:23+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Kernels from pretrained MLIP latent spaces outperform standard acquisition methods in active learning for reactive chemistry, reducing required labels by 38% for energy error and 28% for force error.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.21069","ref_index":90,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Accelerating point defect simulations using data-driven and machine learning approaches","primary_cat":"cond-mat.mtrl-sci","submitted_at":"2026-04-22T20:29:41+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":2.0,"formal_verification":"none","one_line_summary":"Machine learning models trained on quantum mechanical data can predict defect properties in solids with high accuracy but at much lower computational cost than traditional methods.","context_count":1,"top_context_role":"background","top_context_polarity":"unclear","context_text":"lation with equivariant interatomic potentials,\"Nature Reviews Physics, vol. 5, pp. 437-438, July 2023. [89] I. Batatia, S. Batzner, D. P. Kov' acs, A. Musaelian, G. N. C. Simm, R. Drautz, C. Ortner, B. Kozinsky, and G. Cs' anyi, \"The design space of E(3)-equivariant atom-centred interatomic potentials,\"Nature Machine Intelligence, vol. 7, pp. 56-67, Jan. 2025. [90] D. S. Levine, M. Shuaibi, E. W. C. Spotte-Smith, M. G. Taylor, M. R. Hasyim, K. Michel, I. Batatia, G. Cs' anyi, M. Dzamba, P. Eastman, N. C. Frey, X. Fu, V. Gharakhanyan, A. S. Krishnapriyan, J. A. Rackers, S. Raja, A. Rizvi, A. S. Rosen, Z. Ulissi, S. Vargas, C. L. Zitnick, S. M. Blau, and B. M. Wood, \"The Open Molecules 2025 (OMol25) Dataset, Evaluations, and Models,\" May 2025."},{"citing_arxiv_id":"2604.20458","ref_index":36,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Surrogate Functionals for Machine-Learned Orbital-Free Density Functional Theory","primary_cat":"cs.LG","submitted_at":"2026-04-22T11:36:18+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Surrogate functionals for OF-DFT are defined and trained solely to make fixed density optimization reach the ground-state density, achieving competitive accuracy on QM9 and QMugs without O(N^3) orthonormalization.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.12607","ref_index":56,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Hierarchical generative modeling for the design of multi-component systems","primary_cat":"physics.comp-ph","submitted_at":"2026-04-14T11:32:08+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A closed-loop framework jointly optimizes molecular composition and geometry in multi-component systems, demonstrated by a 30% reduction in activation barrier for a Claisen rearrangement via post-hoc validation.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.11827","ref_index":86,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Inverse Design of Inorganic Compounds with Generative AI","primary_cat":"physics.chem-ph","submitted_at":"2026-04-11T16:37:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":2.0,"formal_verification":"none","one_line_summary":"A review of generative AI for inverse design of inorganic compounds, analyzing adaptations for their complexity in composition, geometry, symmetry, and electronic structure, with discussion of future benchmarks and synthesizability metrics.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2603.21942","ref_index":6,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Suiren-1.0 Technical Report: A Family of Molecular Foundation Models","primary_cat":"physics.chem-ph","submitted_at":"2026-03-23T12:59:43+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Suiren-1.0 is a family of three molecular foundation models (Base, Dimer, ConfAvg) pre-trained on 70M+ DFT samples and distilled to achieve claimed state-of-the-art performance on quantum property prediction tasks from 2D inputs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2602.17709","ref_index":2,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"UBio-MolFM: A Universal Molecular Foundation Model for Bio-Systems","primary_cat":"physics.chem-ph","submitted_at":"2026-02-13T04:38:28+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"UBio-MolFM achieves ab initio-level fidelity on large out-of-distribution biomolecular systems using a new multi-fidelity dataset, E2Former-V2 architecture, and three-stage curriculum learning.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2602.04849","ref_index":71,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"El Agente Estructural: An Artificially Intelligent Molecular Editor","primary_cat":"physics.chem-ph","submitted_at":"2026-02-04T18:38:48+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"El Agente Estructural is a new multimodal agent that performs natural-language-driven 3D molecular geometry editing and generation using integrated domain tools and vision-language models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2601.16331","ref_index":5,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Accuracy and Efficiency Benchmarks of Pretrained Machine Learning Potentials for Molecular Simulations","primary_cat":"physics.chem-ph","submitted_at":"2026-01-22T21:32:16+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Benchmarks of 15 MLIPs show parameter count and training set size correlate with accuracy, architecture drives speed and memory, and explicit Coulomb terms provide no benefit.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2601.10791","ref_index":26,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"OmniMol: Transferring Particle Physics Knowledge to Molecular Dynamics with Point-Edge Transformers","primary_cat":"physics.chem-ph","submitted_at":"2026-01-15T19:00:04+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"OmniMol transfers a billion-jet pre-trained PET foundation model from HEP to molecular dynamics via an interaction-matrix attention bias, delivering strong performance on the oMol dataset with minimal fine-tuning and fast inference.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2512.05717","ref_index":27,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Comparing the latent features of universal machine-learning interatomic potentials","primary_cat":"physics.chem-ph","submitted_at":"2025-12-05T13:45:01+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Different uMLIPs encode chemical space in distinct ways, with high cross-model feature reconstruction errors, and fine-tuning preserves strong pre-training bias in the latent features.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}