RIDD: Plug-and-Play Interpretable Responsible Text-to-Image Generation

CVPR 2025

RIDD introduces a novel plug-and-play framework for interpretable and responsible text-to-image (T2I) generation. Our approach enables fine-grained control over generated images by dynamically adjusting attributes such as age, gender, and race while maintaining fidelity and fairness in AI-generated content.

---

Introduction

The rapid advancement in diffusion models has enabled high-quality text-to-image synthesis. However, interpretability, fairness, and responsible generation remain key challenges. RIDD addresses these challenges by:

• Providing fine-grained control over attributes like age, gender, and race.
• Using knowledge distillation and concept whitening for interpretable changes.
• Ensuring responsible AI by mitigating biases in generated images.

---

Demo

🔗 Live Demo: Coming Soon
🔗 Project Page: (Plug and Play Control)

---

Code will be made available soon.

BibTeX

If you use RIDD in your research, please cite:

@inproceedings{azam2025ridd,
  author    = {Basim Azam and Naveed Akhtar},
  title     = {Plug-and-Play Interpretable Responsible Text-to-Image Generation via Dual-Space Multi-facet Concept Control},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year      = {2025},
  url       = {https://arxiv.org/abs/XXXX.XXXXX},
}

Acknowledments

We acknowledge contributions from

• The University of Melbourne
• OpenAI and Stable Diffusion
• Github Repos ()