AI RESEARCH
Improving Visual Representation Alignment Generation with GRPO
arXiv CS.AI
•
ArXi:2606.00583v1 Announce Type: cross Recent diffusion transformers have nstrated strong image synthesis capabilities but remain inefficient to train due to weak alignment between generative and discriminative representations. While representation alignment frameworks such as REPA improve convergence by aligning noisy denoising features with pretrained visual encoders, their externally supervised alignment loss is static and lacks adaptivity during