Practical image generation requires both human preference alignment and few-step efficiency, yet combining reinforcement learning (RL) with distillation is often multi-stage and may erode RL gains We...
AdvFidelity unifies preference alignment and few-step image generation by embedding artifact-sensitive fidelity supervision directly into reinforcement learning. It trains a discriminator to rank high-quality references above content-matched reconstructions, enabling online adaptation of fidelity rewards without explicit distillation. The method achieves alignment, task performance, and generation quality comparable to full-step LeapAlign while using only 4 or 8 steps and no classifier-free guidance.
By Sicong Kang, Fangao Zeng, Yuhan Li, Hao Zhou, Mengfei Xu, Wenxiang Shang, Wei Li, Bingbing Ni, Pipei Huang
arXiv:2608. 09226v1 Announce Type: cross Abstract: Efficient text-to-image generation requires both reinforcement-learning (RL)-based reward alignment and few-step distillation, yet these procedures are typically performed sequentially, increasing training cost and risking the loss of reward gains during compression.
By Yuhan Li, Fangao Zeng, Sicong Kang, Mengfei Xu, Hao Zhou, Wei Li, Pipei Huang, Bingbing Ni
arXiv:2608.29804v1 Announce Type: new
Abstract: Virtual try-on (VTON) requires not only realistic generation but also faithful preservation of garment characteristics. However, existing evaluation me...
By Kaidong Zhang, Yukang Ding, Xiaoyu Liu, Ying Chen
arXiv:2606. 19162v1 Announce Type: new Abstract: Score- and flow-matching models often rely on preference-based reinforcement learning for two purposes: aligning with subjective preferences and, surprisingly, recovering properties such as visual realism and coherent object structure that matching-based training is intended to learn from the data itself.
By Nicolas Beltran-Velez, Felix Friedrich, Zhang Xiaofeng, Reyhane Askari-Hemmat, Xiaochuang Han, Adriana Romero-Soriano, Michal Drozdzal
Efficient text-to-image generation requires both reinforcement-learning (RL)-based reward alignment and few-step distillation, yet these procedures are typically performed sequentially, increasing training cost and risking the loss of reward gains during compression. We instead take an RL-native perspective: diffusion RL already generates reward-scored finite-step trajectories, whose intermediate states provide a natural source of distillation supervision rather than a disposable byproduct of sampling.