Newest papers

Text-to-Image Generative Modeling Without Language Pretraining

Keshigeyan Chandrasegaran, Yue Zhao, Armin W. Thomas, Ehsan Adeli, Jiajun Wu, Juan Carlos Niebles, Li Fei-Fei, Michael Poli · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26863 · cited by 0 works (OpenAlex) · open-access PDF

Chinese

现代文本到图像(T2I)系统使用互联网规模的语言预训练来生成文本条件的图像。我们询问语言预训练是否需要强大的T2I生成。我们介绍了Halcyon,一个早期合并的多式自动回归模型家族,完全从零开始在GPIC上预训练,一个100M图像文本对的数据集。Halcyon-0.52B比LlamaGen-XL/8(48.0 vs. 63.3 gFD)更出色,以182$\times$更少的预训练计算。此外,Halcyon-1.6B比JiT-T2I-XXL/8++(353 vs. 41.0F gD)更出色,具有275$\$小时更少的预训练计算。控制实验表明,文本预训练不会改善T2I

Machine translation (Meta M2M100); the English original follows.

Abstract

Modern text-to-image (T2I) systems use internet-scale language pretraining for text-conditioned image generation. We ask whether language pretraining is necessary for strong T2I generation. We introduce Halcyon, a family of early-fusion multimodal autoregressive models pretrained entirely from scratch on GPIC, a dataset of 100M image-text pairs. Halcyon-0.52B outperforms LlamaGen-XL/8 (48.0 vs. 63.3 gFD) with 18.2$\times$ less pretraining compute. Further, Halcyon-1.6B outperforms JiT-T2I-XXL/8++ (35.3 vs. 41.0 gFD) with 275$\times$ less pretraining compute. Controlled experiments show that text-only pretraining does not improve T2I generation. Halcyon also learns image captioning by replacing caption dropout with I2T supervision, maintaining competitive T2I generation quality under the same pretraining budget. We show that strong T2I systems can be trained from scratch without internet-scale language pretraining. Code and models are available at https://cs.stanford.edu/~keshik/halcyon

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Closed captioning, Computer science, Artificial intelligence, Scratch, Natural language processing, Language model

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange