Metraipolis › archive › dc_2e06bcd2a73b

Where Should Physics Enter a Molecular Crystal Generator? — arXiv:2609.36398

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.36398
License (as archived): CC0-1.0
Archived: 2026-10-07T16:25:46.934Z
SHA-256 of the archived text: 7344efd6e07f615493a4a94b22fae6837e70af959f32edfa55b586e546430360
Stable URL of this page: https://metraipolis.com/documento/dc_2e06bcd2a73b

How to cite

Where Should Physics Enter a Molecular Crystal Generator? — arXiv:2609.36398 — arXiv (en), https://arxiv.org/abs/2609.36398 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2e06bcd2a73b (Metraipolis archive dc_2e06bcd2a73b, sha256 7344efd6e07f615493a4a94b22fae6837e70af959f32edfa55b586e546430360)

BibTeX

@misc{dc2e06bcd2a73b,
  title = {Where Should Physics Enter a Molecular Crystal Generator? — arXiv:2609.36398},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.36398},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2e06bcd2a73b; sha256 7344efd6e07f615493a4a94b22fae6837e70af959f32edfa55b586e546430360},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.36398. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

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.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI).
Categorie arXiv: Biomolecules (q-bio.BM) ; Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE).
Pagina del lavoro: https://arxiv.org/abs/2609.36398 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T15:54:53.832Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).