Thrilled to share our work “On Fairness of Unified Multimodal Large Language Model for Image Generation”

[pkulium]

Thrilled to share our new paper “On Fairness of Unified Multimodal Large Language Model for Image Generation”
arXiv
. We benchmark leading U-MLLMs, uncover significant demographic biases in image generation, and propose a balanced preference model that mitigates bias while preserving semantic fidelity
arXiv

Read it on arXiv: https://arxiv.org/abs/2502.03429!

[sshan-zhao]

Thrilled to share our new paper “On Fairness of Unified Multimodal Large Language Model for Image Generation” arXiv . We benchmark leading U-MLLMs, uncover significant demographic biases in image generation, and propose a balanced preference model that mitigates bias while preserving semantic fidelity arXiv

Read it on arXiv: https://arxiv.org/abs/2502.03429!

Great work! We welcome and greatly appreciate any support for our project and survey aimed at advancing the development of unified multimodal models. We will provide regular updates on our project and survey, incorporating new findings and highlighting any significant work we may have overlooked.