Diminishing Stereotype Bias in Image Generation Model using Reinforcemenlent Learning Feedback
This work addresses gender bias in AI image generation, which is a critical fairness issue for users and society, though it is incremental as it builds on existing methods like stable diffusion and RLAIF.
The study tackled gender bias in image generation models by using Reinforcement Learning from Artificial Intelligence Feedback (RLAIF) with a Denoising Diffusion Policy Optimization (DDPO) pipeline, achieving effective bias mitigation without compromising image quality or requiring extra data.
This study addresses gender bias in image generation models using Reinforcement Learning from Artificial Intelligence Feedback (RLAIF) with a novel Denoising Diffusion Policy Optimization (DDPO) pipeline. By employing a pretrained stable diffusion model and a highly accurate gender classification Transformer, the research introduces two reward functions: Rshift for shifting gender imbalances, and Rbalance for achieving and maintaining gender balance. Experiments demonstrate the effectiveness of this approach in mitigating bias without compromising image quality or requiring additional data or prompt modifications. While focusing on gender bias, this work establishes a foundation for addressing various forms of bias in AI systems, emphasizing the need for responsible AI development. Future research directions include extending the methodology to other bias types, enhancing the RLAIF pipeline's robustness, and exploring multi-prompt fine-tuning to further advance fairness and inclusivity in AI.