{
  "abstract": "We propose generative marginal learning (GML), a framework for equilibrium sampling and local observable estimation in molecular simulations. By training a conditional diffusion model on large-system simulations, GML learns the marginal distribution of a retained region, enabling configuration generation without explicitly simulating the surrounding environment or imposing periodic boundary conditions on the retained region. We derive a marginal effective potential and the associated conditional mean-force identity, which provide the basis for both denoising and force-alignment training. As an application, we estimate the spatially averaged Irving--Kirkwood stress near a vacancy defect. Numerical results demonstrate that GML substantially improves pair structure and local stress accuracy compared with small periodic molecular dynamics simulations, mitigating finite-size effects in calculating local molecular observables.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2610.25105",
  "author_total": 1,
  "authors": [
    {
      "name": "Cheng Yuan"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/8dt9gr5m77854.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2610.25105",
  "fulltext_url": null,
  "id": "8dt9gr5m77854",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2610.25105",
  "language": null,
  "licence": "CC-BY-NC-ND-4.0",
  "licence_label": "CC BY-NC-ND 4.0",
  "licence_source": "article",
  "licence_url": "http://creativecommons.org/licenses/by-nc-nd/4.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cond-mat.stat-mech"
    ],
    [
      "arxiv",
      "physics.comp-ph"
    ],
    [
      "openalex",
      "2505|Machine Learning in Materials Science"
    ]
  ],
  "openalex": "W7223030027",
  "pdf_url": "https://arxiv.org/pdf/2610.25105",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-18",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/8dt9gr5m77854.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:30:52Z",
      "native_id": "2610.25105",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2610.25105"
    }
  ],
  "subjects": [
    {
      "field": "physics",
      "origin": "arxiv:physics.comp-ph",
      "subfield": "computational"
    },
    {
      "field": "physics",
      "origin": "arxiv:cond-mat.stat-mech",
      "subfield": "statistical-nonlinear"
    }
  ],
  "title": "Generative marginal learning for finite-size molecular simulation with application to stress calculation",
  "title_text": "Generative marginal learning for finite-size molecular simulation with application to stress calculation",
  "type": "preprint",
  "updated_at": "2026-11-09T00:20:37Z",
  "url": "https://research.draconitic.com/paper/8dt9gr5m77854",
  "venue": "arXiv",
  "versions": []
}