| title | From Corpora to Co-Evolving Capabilities: Capability-Centric Data Design for Generalist Image Generation |
|---|---|
| authors | Xingjian Wang, Zhao Wang, Taihang Hu, Jun Zheng, Qing Jin, Qinye Zhou, Zhengtao Wu, Yongchao Du, Zuan Gao, Chao Lin, Yefeng Shen, Xiaoli Xu, Zhengze Xu, Hao Yan, Yuhang Yu, Mingzhou Zhang, Mengting Chen |
| date | 2026-08-18 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2608.18076 |
| affiliation | alibaba |
| upvotes | 13 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | Large-scale image generation has benefited from advances in data scale, quality, rebalancing, and recaptioning, yet conventional pipelines typically optimize task-specific datasets in isolation. A central challenge is not only how to curate each task-specific corpus, but also how to organize heterogeneous supervision according to the dependencies among generative capabilities. We present a capability-driven data infrastructure that couples capability-specific supervision construction with capability-aligned curriculum scheduling. Its three specialized yet interoperable data engines build complementary relational supervision for text-image grounding, inter-image transformation, and image-knowledge association, while caption experts align T2I and editing supervision across tasks and granularities. A multi-stage curriculum jointly evolves task composition, visual-concept distribution, data quality, and image resolution along the dependency order of capability acquisition, with capability-aware evaluation closing the loop through targeted retrieval, expert construction, and gap-aware resampling. At scale, the framework curates a 440M-image T2I corpus, 120M editing pairs, and over 27M image-entity pairs. With this infrastructure, we train multimodal diffusion models at two scales from scratch, with 3B and 6B sizes respectively. We conduct quantitative evaluation on CPI-Bench, along with qualitative evaluations across diverse text-to-image and editing scenarios. Experimental results present broad visual coverage, versatile rendering, and effective transfer across generative capabilities. |
cite
@misc{wang2026from,
title={From Corpora to Co-Evolving Capabilities: Capability-Centric Data Design for Generalist Image Generation},
author={Xingjian Wang and Zhao Wang and Taihang Hu and Jun Zheng and Qing Jin and Qinye Zhou and Zhengtao Wu and Yongchao Du and Zuan Gao and Chao Lin and Yefeng Shen and Xiaoli Xu and Zhengze Xu and Hao Yan and Yuhang Yu and Mingzhou Zhang and Mengting Chen},
year={2026},
eprint={2608.18076},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2608.18076}
}