Metraipolis › archive › dc_0ea5018693c4

Difficulty-Calibrated Interpolation Paths for Conditional Flow Matching — arXiv:2608.21286

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.21286
License (as archived)
CC0-1.0
Archived
2026-10-18T10:56:25.385Z
SHA-256 of the archived text
f8e2570053b31910cd0e7b01d7de6ffcb67a99530c1eeb05841a8ea721a717fb
Stable URL of this page
https://metraipolis.com/documento/dc_0ea5018693c4

How to cite

Difficulty-Calibrated Interpolation Paths for Conditional Flow Matching — arXiv:2608.21286 — arXiv (en), https://arxiv.org/abs/2608.21286 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0ea5018693c4 (Metraipolis archive dc_0ea5018693c4, sha256 f8e2570053b31910cd0e7b01d7de6ffcb67a99530c1eeb05841a8ea721a717fb)
BibTeX
@misc{dc0ea5018693c4,
  title = {Difficulty-Calibrated Interpolation Paths for Conditional Flow Matching — arXiv:2608.21286},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.21286},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0ea5018693c4; sha256 f8e2570053b31910cd0e7b01d7de6ffcb67a99530c1eeb05841a8ea721a717fb},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.21286. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Conditional Flow Matching trains generative models by regressing a network onto the velocity of a prescribed noise-to-data interpolation path. The interpolation schedule that shapes this path is known to affect convergence and sample quality, yet it is invariably fixed in advance, independent of both the data and the model. We show that the regression difficulty of Conditional Flow Matching varies systematically along the path, and we propose Difficulty-Calibrated Flow Matching, which derives the schedule from the model itself: a short pilot run with the linear path records the per-time loss, and the schedule is set to the quantile function of this difficulty profile, so the trajectory lingers where the velocity is hardest to learn. The method has a single hyperparameter, leaves the training objective and its gradient equivalence intact, composes with classifier-free guidance, and adds about two percent training overhead. In controlled experiments on CIFAR-10, MNIST, and Fashion-MNIST with an identical compact U-Net, the calibrated path attains the best FID on CIFAR-10 at full sampling budget and clearly outperforms all fixed schedules in the large-batch, few-update regime, precisely the setting where compute is scarcest.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-08-24 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2608.21286 — 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 2026-10-18T10:48:13.608Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-08-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).