Metraipolis › archive › dc_8cb0937cd421

Gaussian Flow-Matching Schedules: Implications for Sampling and Training — arXiv:2609.25839

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.25839
License (as archived): CC0-1.0
Archived: 2026-10-08T15:07:36.503Z
SHA-256 of the archived text: b33463c40da6d974233c9c68fd7978af5f3b60fc79cd17612833a61bb593b4d5
Stable URL of this page: https://metraipolis.com/documento/dc_8cb0937cd421

How to cite

Gaussian Flow-Matching Schedules: Implications for Sampling and Training — arXiv:2609.25839 — arXiv (en), https://arxiv.org/abs/2609.25839 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8cb0937cd421 (Metraipolis archive dc_8cb0937cd421, sha256 b33463c40da6d974233c9c68fd7978af5f3b60fc79cd17612833a61bb593b4d5)

BibTeX

@misc{dc8cb0937cd421,
  title = {Gaussian Flow-Matching Schedules: Implications for Sampling and Training — arXiv:2609.25839},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.25839},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8cb0937cd421; sha256 b33463c40da6d974233c9c68fd7978af5f3b60fc79cd17612833a61bb593b4d5},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.25839. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Flow-matching schedules affect both sampling dynamics and the variance of the regression target. For centered commuting Gaussians, we show that a direction-dependent schedule decomposes into two independent design choices: a variance path, which fully determines the intermediate laws and probability flow, and a factorization, which leaves this flow unchanged while controlling irreducible regression variance. On the sampling side, we analyze finite-step Euler accuracy and derive a necessary drift bound for exact N -step sampling, connecting the geodesic and the logarithmic path. On the training side, for any fixed path, we derive closed-form factorizations that either minimize time-averaged regression variance or make it constant along the path.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-23 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.25839 — 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 2026-10-08T14:37:29.436Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).