Preference-Enhanced Reinforcement Learning for Pluralistic Image Inpainting

Existing image inpainting frameworks rely on strictly supervised training paradigms, often suffering from an over-reliance on ground-truth reconstruction, which leads to conservative outputs with misaligned creativity and limited diversity. To this end, we propose the first framework to explore Group Relative Policy Optimization (GRPO) and Direct Preference Optimization (DPO) for text-guided image inpainting, formulating an efficient online reinforcement learning pipeline that enables flexible, human-aligned aesthetic control via a preference scoring model. Crucially, by decoupling the rigid one-to-one correspondence between text prompts and masked images, our method enables the model to explore diverse, controllable, and high-quality solutions beyond a single target. Furthermore, to balance semantic consistency with physical naturalness at mask boundaries, we introduce a scale-aware dynamic reward mechanism that adaptively emphasizes boundary gradient coherence for small occlusions while prioritizing visual aesthetics in large-scale generation. Extensive experiments demonstrate that our approach consistently produces higher-quality results across different backbone architectures such as Stable Diffusion and FLUX, significantly enhancing the generative capacity of base models.

Existing AI image inpainting tools only produce single, conservative results because they are trained to copy one specific reference image. We developed a new method that lets models learn autonomously through human preference feedback. We also designed an intelligent scoring system that ensures the edited area blends naturally with the original while maintaining visual beauty. Experiments show our method enables mainstream AI editing models to generate more diverse and human-pleasing results.