kr-cen/BAGEL-MICo
🥯 BAGEL-MICo • BAGEL-Variant finetuned on Multi-Image Composition Dataset (MICo-150K)
   
<!-- <p align="left"> <a href="https://mico-150k.github.io/"> <img src="https://img.shields.io/badge/MICo-Website-0A66C2?logo=safari&logoColor=white" style="display: inline-block; vertical-align: middle;" alt="MICo Website" /> </a> <a href="https://www.arxiv.org/abs/2512.07348"> <img src="https://img.shields.io/badge/MICo-Paper-red?logo=arxiv&logoColor=red" style="display: inline-block; vertical-align: middle;" alt="MICo Paper on arXiv" /> </a> <a href="https://github.com/A113N-W3I/MICo-150K" target="_blank" style="margin: 2px;"> <img alt="Github" src="https://img.shields.io/badge/MICo-Bench-536af5?color=536af5&logo=github" style="display: inline-block; vertical-align: middle;" alt="MICo Bench" /> </a>
</p> -->
🎨 Demo




✍️ Citation
If you find this work useful, please cite:
~~~ @article{wei2025mico, title={MICo-150K: A Comprehensive Dataset Advancing Multi-Image Composition}, author={Wei, Xinyu and Cen, Kangrui and Wei, Hongyang and Guo, Zhen and Li, Bairui and Wang, Zeqing and Zhang, Jinrui and Zhang, Lei}, journal={arXiv preprint arXiv:2512.07348}, year={2025} } ~~~
License
BAGEL-MICo is licensed same as BAGEL: BAGEL is licensed under the Apache 2.0 license. It is finetuned from Qwen2.5-7B-Instruct and siglip-so400m-14-384-flash-attn2 model, and uses the FLUX.1-schnell VAE model, all under Apache 2.0.
