Broken Identifier

arXiv:2604.27342 · detector doi_compliance · incontrovertible · 2026-05-19 19:22:52.130363+00:00

Paper page Integrity report arXiv Try DOI

Evidence text

Institute of Industrial Science, the University of Tokyo, Tokyo, Japan S1. Data Distribution To ensure balanced configurational coverage across the full structural phase space of bilayer h- BN, training data for both models were assembled from seven structural categories: AA′, AA′ sliding, AB, AB sliding, BA, moiré superlattices, and monolayer configurations. The MACE fine- tuning dataset comprises 2,808 structures sampled from AIMD trajectories and static sliding pathways, while the EGCNN training dataset comprises 4,045 structures with DFT -computed BEC tensors. The notably larger representation of moiré structures reflects the structural complexity and atomic diversity of the ∑7, ∑13, and ∑19 commensurate superlattices. For both datasets, a stratified 8:2 train–test split was applied independently within each structural category to prevent dataset imbalance from biasing model generalization. The following Fig.S1 represents the distribution of the data used for fine-tunning MACE model and training the EGCNN model. 28 Fig.S1 Structural composition of the training datasets. (a) Distribution of the 2,808 structures used for fine-tuning the MACE machine learning potential, comprising seven structural subsets: monolayer h -BN, AA′, AA′ sliding ( AA′ 1-10), AB, AB sliding (AB_1 -10), BA, moiré superlattices (∑7 commensurabilities), and monolayer configurations. (b) Distribution of the 4,045 structures with DFT-computed BECs used for training the EGCNN model, comprising six struct

Evidence payload

{
  "doi_as_printed": "10.1103/lymz-nlbf",
  "raw_excerpt": "Institute of Industrial Science, the University of Tokyo, Tokyo, Japan S1. Data Distribution To ensure balanced configurational coverage across the full structural phase space of bilayer h- BN, training data for both models were assembled from seven structural categories: AA\u2032, AA\u2032 sliding, AB, AB sliding, BA, moir\u00e9 superlattices, and monolayer configurations. The MACE fine- tuning dataset comprise",
  "ref_index": 7,
  "verdict_class": "incontrovertible"
}