SAGA: Learning Signal-Aligned Distributions for Improved Text-to-Image Generation
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (6) · 4293-4301
DOI: 10.1609/aaai.v40i6.42426 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Paul Grimal und 4 weiteren Autor:innen.
Abstract (Originalsprache)
State-of-the-art text-to-image models produce visually impressive results but often struggle with precise alignment to text prompts, leading to missing critical elements or unintended blending of distinct concepts. We propose a novel approach that learns a high-success-rate distribution conditioned on a target prompt, ensuring that generated images faithfully reflect the corresponding prompts. Our method explicitly models the signal component during the denoising process, offering fine-grained control that mitigates over-optimization and out-of-distribution artifacts. Moreover, our framework is training-free and seamlessly integrates with both existing diffusion and flow matching architectures. It also supports additional conditioning modalities -- such as bounding boxes -- for enhanced spatial alignment. Extensive experiments demonstrate that our approach outperforms current state-of-the-art methods.
Zitieren
APA
Grimal, P., Soumm, M., Borgne, H. L., Ferret, O., & Sugimoto, A. (2026). SAGA: Learning Signal-Aligned Distributions for Improved Text-to-Image Generation. Proceedings of the AAAI Conference on Artificial Intelligence, 40(6), 4293-4301. https://doi.org/10.1609/aaai.v40i6.42426
Vancouver
Grimal P, Soumm M, Borgne HL, Ferret O, Sugimoto A. SAGA: Learning Signal-Aligned Distributions for Improved Text-to-Image Generation. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(6):4293-4301. doi:10.1609/aaai.v40i6.42426
BibTeX
@article{Grimal2026,
author = {Grimal, Paul and Soumm, Michael and Borgne, Hervé Le and Ferret, Olivier and Sugimoto, Akihiro},
title = {SAGA: Learning Signal-Aligned Distributions for Improved Text-to-Image Generation},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {6},
pages = {4293--4301},
doi = {10.1609/aaai.v40i6.42426}
}
EUVIMED