Metraipolis › archive › dc_db875e4d34f5
Embedding Prediction Helps Image Generation — arXiv:2610.02203
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.02203
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T19:15:51.406Z
- SHA-256 of the archived text
8a725226bd077a988918a9fb8196da5f3db8be0b278262d0bba2a4c08b46d8ea- Stable URL of this page
https://metraipolis.com/documento/dc_db875e4d34f5
How to cite
Embedding Prediction Helps Image Generation — arXiv:2610.02203 — arXiv (en), https://arxiv.org/abs/2610.02203 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_db875e4d34f5 (Metraipolis archive dc_db875e4d34f5, sha256 8a725226bd077a988918a9fb8196da5f3db8be0b278262d0bba2a4c08b46d8ea)
BibTeX
@misc{dcdb875e4d34f5,
title = {Embedding Prediction Helps Image Generation — arXiv:2610.02203},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.02203},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_db875e4d34f5; sha256 8a725226bd077a988918a9fb8196da5f3db8be0b278262d0bba2a4c08b46d8ea},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.02203. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
In diffusion transformers, a class label or a text prompt is embedded once, and the same condition is reused at every denoising step. We ask whether predicted embeddings can serve as this condition instead. Next-Embedding Predictive Autoregression (NEPA) trains a Transformer to predict the next continuous embedding in a sequence. In generation, the clean image follows the noisy image, so its embeddings are the next embeddings after the condition and the noisy image. We train a NEPA model to predict them all at once with Multi-Embedding Prediction, and in Embedding Conditioned Generation, a DiT generator is conditioned on these predictions, recomputed at every denoising step, so the conditioning signal adapts to the current noisy state. Experiments on class-conditional ImageNet $256\times256$ study the condition of the generator, the design of Multi-Embedding Prediction, and the scaling of both models. The NEPA model adds a second network to every sampling step; with it, and combined with REPA, our final model, NEPA-DiT-XL, reaches an FID of 1.32 using about a third of the training compute of REPA. 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) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.02203 — 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 2026-10-06T17:37:36.899Z 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).