@InProceedings{10.1007/978-3-032-37189-8_33,
author="Fang, Yan
and Lan, Mengcheng
and Huang, Zilong
and Lei, Weixian
and Zhao, Yunqing
and Zhong, Yujie
and Yu, Yingchen
and She, Qi
and Zhao, Yao
and Wei, Yunchao",
editor="Favaro, Paolo
and Kukelova, Zuzana
and Maki, Atsuto
and Rohrbach, Anna
and Schindler, Konrad
and Tombari, Federico",
title="Let ViT Speak: Generative Language-Image Pre-training",
booktitle="Computer Vision -- ECCV 2026",
year="2026",
publisher="Springer Nature Switzerland",
address="Cham",
pages="588--608",
abstract="In this paper, we present Generative Language-Image Pre-training (GenLIP), a simplified generative pretraining approach for Vision Transformers (ViTs) tailored to multimodal large language models (MLLMs). To better align ViTs with the autoregressive nature of LLMs, GenLIP trains a ViT to predict language tokens directly from visual tokens using a standard language modeling objective--without contrastive batch construction or an additional text decoder. GenLIP offers three key advantages: (1) Simplicity: a single transformer jointly models visual and linguistic tokens; (2) Scalability: it scales efficiently with both data and model size; (3) Performance: it achieves competitive or superior results across diverse multimodal benchmarks. With only 1/5 of the seen samples, GenLIP matches or surpasses strong baselines such as SigLIP2. With continued pretraining on multi-resolution images at native aspect ratios, GenLIP further excels at detail-sensitive tasks such as OCR, chart understanding, and visual question answering, making it a strong foundation for vision encoders in MLLMs.",
isbn="978-3-032-37189-8"
}