Metraipolis › archive › dc_aaa1b9d8556d

Just Align $\bm{x}$: Aligning Predictions, Not Representations — arXiv:2610.00600

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.00600
License (as archived)
CC0-1.0
Archived
2026-10-06T19:11:20.104Z
SHA-256 of the archived text
07bfdb9a21e0eb65de7daa1d991e6f972a20c9eaf686f563aca66ae59288fefb
Stable URL of this page
https://metraipolis.com/documento/dc_aaa1b9d8556d

How to cite

Just Align $\bm{x}$: Aligning Predictions, Not Representations — arXiv:2610.00600 — arXiv (en), https://arxiv.org/abs/2610.00600 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_aaa1b9d8556d (Metraipolis archive dc_aaa1b9d8556d, sha256 07bfdb9a21e0eb65de7daa1d991e6f972a20c9eaf686f563aca66ae59288fefb)
BibTeX
@misc{dcaaa1b9d8556d,
  title = {Just Align $bmx$: Aligning Predictions, Not Representations — arXiv:2610.00600},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.00600},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_aaa1b9d8556d; sha256 07bfdb9a21e0eb65de7daa1d991e6f972a20c9eaf686f563aca66ae59288fefb},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.00600. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Representation alignment has become an effective way to accelerate diffusion training, but its benefits do not transfer reliably to pixel-space clean-image prediction. In JiT, we find that auxiliary feature alignment can improve access to semantic features while reducing access to image variation needed for clean-image prediction, creating a mismatch between the auxiliary objective and the denoising task. This suggests a different principle: auxiliary supervision should improve the prediction target itself rather than impose a separate representation target. We introduce JAx (Just Align x), a prediction-supervision method that aligns clean-image predictions across noise levels. JAx couples a noisier student observation with a cleaner observation through a Markov degradation that preserves the original JiT input distribution. Under this coupling, the oracle prediction from the cleaner state has the same conditional mean as the optimal JiT target, while its conditional target covariance is no greater. Thus, oracle prediction alignment preserves the population JiT objective up to a constant while providing a lower-variance training target. To make this construction practical with an imperfect EMA teacher, JAx combines ground-truth supervision with a reliability-gated coupling band that selects nearby teacher states based on prediction risk. On ImageNet 256x256, JAx consistently improves FID and accelerates convergence across JiT-B/16, L/16, and H/16, without an external encoder or changes to the architecture or sampling procedure. Gradient diagnostics further show reduced minibatch gradient variance, while ablations demonstrate that the gains cannot be explained by time reweighting alone. These results show that prediction-space supervision provides a simple and principled alternative to representation alignment for pixel-space generative models.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-10-02 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2610.00600 — 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 2026-10-06T17:37:36.681Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-10-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).