{
  "abstract": "Generative models make molecular crystal structure prediction fast, but their samples still exhibit geometric and packing violations. Physics can be introduced during training, post-training, or inference, yet these choices are rarely compared with the generator and physical signal held fixed. We introduce CrystAF, an all-atom crystal flow-map generation model, and use it with the UMA interatomic potential to systematically study where physics should enter. Post-training learns physical preferences directly into CrystAF, improving molecular validity and crystal packing while leaving sampling unchanged: physics is paid for once during training rather than repeatedly at deployment. In contrast, UMA relaxation is effective at repairing local clashes but makes generation 6--26$\\times$ slower, while learning from relaxed targets provides little benefit. These routes are complementary rather than competing. Physics-informed post-training first shifts the generated distribution toward more physically reasonable structures, after which inexpensive inference-time corrections further remove clashes and restore stereochemistry that the generator cannot represent. Importantly, the same post-training strategy also improves the multi-step all-atom Clari-M and rigid-body MolCrystalFlow generators, demonstrating transfer across architectures and representations. Together, our results suggest a simple principle: learn reusable physical alignment into the generator, and reserve inference-time physics for residual constraints that are better corrected than learned.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2609.36398",
  "author_total": 3,
  "authors": [
    {
      "name": "Haocheng Tang"
    },
    {
      "name": "Junmei Wang"
    },
    {
      "name": "Wengong Jin"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/ajynfpdmy5y5a.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2609.36398",
  "fulltext_url": null,
  "id": "ajynfpdmy5y5a",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2609.36398",
  "language": null,
  "licence": "LicenseRef-arXiv-nonexclusive-1.0",
  "licence_label": "arXiv licence",
  "licence_source": "article",
  "licence_url": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cs.AI"
    ],
    [
      "arxiv",
      "cs.CE"
    ],
    [
      "arxiv",
      "q-bio.BM"
    ],
    [
      "openalex",
      "2505|Machine Learning in Materials Science"
    ]
  ],
  "openalex": "W7214926900",
  "pdf_url": "https://arxiv.org/pdf/2609.36398",
  "pmcid": null,
  "pmid": null,
  "published": "2026-09-28",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/ajynfpdmy5y5a.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:25:36Z",
      "native_id": "2609.36398",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2609.36398"
    }
  ],
  "subjects": [
    {
      "field": "biochemistry",
      "origin": "arxiv:q-bio.BM",
      "subfield": "biophysics"
    }
  ],
  "title": "Where Should Physics Enter a Molecular Crystal Generator?",
  "title_text": "Where Should Physics Enter a Molecular Crystal Generator?",
  "type": "preprint",
  "updated_at": "2026-10-30T12:25:36Z",
  "url": "https://research.draconitic.com/paper/ajynfpdmy5y5a",
  "venue": "arXiv",
  "versions": []
}