Metraipolis › archive › dc_7d75f1be3ff0

A Language-Informed Loss for Visual Generation — arXiv:2610.23653

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.23653
License (as archived)
CC0-1.0
Archived
2026-10-19T09:21:30.526Z
SHA-256 of the archived text
d13ebeb1053651c21c3178081650debb717f7cf017526704d762a7f37a50d1a2
Stable URL of this page
https://metraipolis.com/documento/dc_7d75f1be3ff0

How to cite

A Language-Informed Loss for Visual Generation — arXiv:2610.23653 — arXiv (en), https://arxiv.org/abs/2610.23653 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7d75f1be3ff0 (Metraipolis archive dc_7d75f1be3ff0, sha256 d13ebeb1053651c21c3178081650debb717f7cf017526704d762a7f37a50d1a2)
BibTeX
@misc{dc7d75f1be3ff0,
  title = {A Language-Informed Loss for Visual Generation — arXiv:2610.23653},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.23653},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7d75f1be3ff0; sha256 d13ebeb1053651c21c3178081650debb717f7cf017526704d762a7f37a50d1a2},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.23653. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

End-to-end visual generation with pixel-space loss has recently achieved competitive performance without the need for learned image tokenizers. In this work, we study whether language-informed supervision can provide a richer training signal than pixel loss for visual generation. We introduce LILIE (pronounced "lily"), a Language-Informed and Lossless Image Encoder. LILIE is an invertible image transform trained such that a vision-language model can recover captions from noised features. The encoder uses a normalizing flow transformer trained jointly with a vision-language model via next token prediction. We then use the encoder to extract a simple language-informed loss for text-to-image generation. We train JiT models with LILIE loss, reducing gFD by 14 points (66.84 $\to$ 53.01) against a pixel-only baseline on the GPIC dataset. Our code, encoders, and generative models are available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.23653 — 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 2026-10-19T09:17:41.727Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).