Metraipolis › archive › dc_b29c4a67493f

Finite-Step Geometry of Marginal-Preserving Velocity Fields in Generative Flows — arXiv:2610.37936

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.37936
License (as archived)
CC0-1.0
Archived
2026-11-01T02:23:18.292Z
SHA-256 of the archived text
a13a9ddae3c3fbed72f0641d9a7d8a2b1e52bd369a614d8a0e06db1f6a740984
Stable URL of this page
https://metraipolis.com/documento/dc_b29c4a67493f

How to cite

Finite-Step Geometry of Marginal-Preserving Velocity Fields in Generative Flows — arXiv:2610.37936 — arXiv (en), https://arxiv.org/abs/2610.37936 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b29c4a67493f (Metraipolis archive dc_b29c4a67493f, sha256 a13a9ddae3c3fbed72f0641d9a7d8a2b1e52bd369a614d8a0e06db1f6a740984)
BibTeX
@misc{dcb29c4a67493f,
  title = {Finite-Step Geometry of Marginal-Preserving Velocity Fields in Generative Flows — arXiv:2610.37936},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.37936},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b29c4a67493f; sha256 a13a9ddae3c3fbed72f0641d9a7d8a2b1e52bd369a614d8a0e06db1f6a740984},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.37936. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Generative flows can admit different velocity fields that induce the same continuous marginal evolution but behave differently after numerical discretization. We study this finite-step non-equivalence while keeping the marginal path and numerical method fixed. Projected modified equations characterize the component of local integration error that changes the marginal law and give sufficient conditions for higher marginal convergence order. For affine fixed-shape distributions, we characterize exact finite-step reachability and derive spectral obstructions and sharp nearest-reachable Wasserstein errors. For diffusion probability-flow ODEs, we construct skew-score perturbations that preserve the exact marginal path under exact scores and quantify their leakage under score error. Controlled Gaussian and mixture experiments recover the predicted order, reachability, and obstruction phenomena.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Machine Learning (stat.ML).
Pagina del lavoro: https://arxiv.org/abs/2610.37936 — 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 2026-11-01T02:12:44.089Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).