Bird-SR: Bidirectional Reward-Guided Diffusion for Real-World Image Super-Resolution

arXiv:2602.07069v2 Announce Type: replace-cross Abstract: Powered by multimodal text-to-image priors, diffusion-based super-resolution excels at synthesizing intricate details; however, models trained on synthetic low-resolution (LR) and high-resolution (HR) image pairs often degrade when applied to real-world LR images due to significant distribution shifts. We propose Bird-SR, a bidirectional reward-guided diffusion framework that formulates super-resolution as trajectory-level preference opti