Routing Policy Optimization (RPO) introduces stochastic exploration directly within a visual generator’s attention mechanism, avoiding costly rollout-based optimization. By sampling key attention blocks at selected denoising levels and perturbing the transition velocity, RPO uses a Gaussian likelihood and a two‑rollout estimator to provide an advantage signal for learning. Experiments on text‑to‑image and text‑to‑video tasks demonstrate improved generation quality and nearly four times higher training efficiency compared to Flow‑GRPO‑style methods.
By Qiuhong Shen, Yiding Yang, Yuxin Zhang, Xinwei Huang, Minxuan Lin, Bo Liu, Hao Kang, Haotian Yang, Yizhi Wang, Chongyang Ma, Xinchao Wang
The paper introduces Objective-aware Trajectory Credit Assignment (OTCA), a framework that refines reinforcement learning for diffusion-based visual generation. OTCA decomposes credit across denoising steps and allocates multiple reward signals adaptively, addressing the coarse, uniform reward assignment of existing GRPO pipelines. Experiments demonstrate that OTCA consistently enhances image and video generation quality across various metrics.
By Rui Li, Ke Hao, Yuanzhi Liang, Haibin Huang, Chi Zhang, Yun Gu, Xuelong Li
arXiv:2609.38329v1 Announce Type: new
Abstract: Group-relative RL methods such as Flow-GRPO post-train image generators by exploring with isotropic Gaussian noise added at every denoising step. This...
By Shuyue Stella Li, Xiaochuang Han, Yulia Tsvetkov, Luke Zettlemoyer
The paper investigates how reinforcement learning can be effectively applied to diffusion models for visual tasks, focusing on the role of likelihood estimation. By systematically separating policy‑gradient objectives, likelihood estimators, and rollout sampling schemes, the authors find that using an evidence lower bound (ELBO) based likelihood estimator computed from the final generated sample is the key factor for stable and efficient RL optimization, outweighing the choice of loss function. Experiments on SD 3.5 Medium across multiple reward benchmarks confirm that this approach improves GenEval scores from 0.24 to 0.95 in 90 GPU hours, outperforming existing methods such as FlowGRPO and the current state‑of‑the‑art without reward hacking.
By Jaemoo Choi, Yuchen Zhu, Wei Guo, Petr Molodyk, Bo Yuan, Jinbin Bai, Yi Xin, Molei Tao, Yongxin Chen
arXiv:2604. 16557v2 Announce Type: replace Abstract: Current post-training methodologies for adapting Large Vision-Language Models (LVLMs) generally fall into two paradigms: Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL).
By Yuming Yan, Kai Tang, Sihong Chen, Ke Xu, Dan Hu, Qun Yu, Pengfei Hu
arXiv:2608.23664v1 Announce Type: cross
Abstract: Reward fine-tuning is becoming an important tool for adapting diffusion models to human preferences and task-specific objectives, but existing method...
By Jaemoo Choi, Wei Guo, Yuchen Zhu, Arash Vahdat, Molei Tao, Julius Berner, Yongxin Chen
Reinforcement learning can align diffusion models with human preferences and task-specific objectives, but endpoint rewards do not specify how an intermediate denoising prediction should change. We in...
DiffusionART introduces a new reinforcement learning framework for diffusion models that aggregates prompt rewards into a shared direction, eliminating consistency forces inherent in prior methods. It employs a trust‑region‑masked second update for drift‑adaptive restoration, removing the need for EMA tuning and differentiable reward models. Experiments on image and video generation show that DiffusionART trains more efficiently and stably, achieving higher mean rewards with significantly fewer GPU hours compared to existing approaches.
By Zhenglin Wan, Jiayu Chen, Zhengfeiyang Zhang, Pengfei Zhou, Wangbo Zhao, Fangyuan Kong, Pengfei Wan, Kun Gai, Yang You
The paper introduces DiffusionOPSD, an on‑policy self‑distillation framework that transforms image‑level reinforcement learning rewards into explicit targets for intermediate denoising predictions in diffusion models. By generating trajectories with a frozen behavior policy and constructing bounded positive and negative targets around query states, the method trains a policy to fit these targets before updating the behavior policy via an exponential moving average. Experiments on SD 3.5‑M and Z‑Image‑Turbo show that DiffusionOPSD achieves the best held‑out scores in 19 of 20 reward‑matched settings, outperforms the strongest competitor by up to 44 % and cuts GPU‑hour usage by 40–63 % compared to DiffusionNFT.
By Wei Zhou, Xiongwei Zhu, Lingdong Kong, Bo Chen, Lei Zhang, Yongyuan Liang, Xiaoxia Hou, Ye Tian, Xian Sun, Yingshuo Wang, Linfeng Li, Shengqiong Wu, Leigang Qu, Feng Li, Wei Liu, Julian McAuley, Tat-Seng Chua
arXiv:2607. 07693v1 Announce Type: cross Abstract: Reinforcement learning from human feedback (RLHF) has emerged as a powerful paradigm for aligning generative models with human preferences.
By Eric Zhu, Abhinav Shrivastava, Soumik Mukhopadhyay
The paper introduces Reflection-Aware GRPO (RA‑GRPO), a reinforcement‑learning framework that aligns diffusion generative models with human preferences. It uses Diffusion Reflection to correct intermediate sampling paths by reversing the diffusion process, and Counterfactual Path Synthesis to embed these corrected trajectories into the policy, avoiding extra inference cost. Experiments on text‑to‑image and text‑to‑video models show RA‑GRPO outperforms existing methods, reducing reward hacking and improving generalization while remaining architecture‑agnostic.
By Junlong Wu, Jiuzhou Lin, Jia Sun, Boheng Zhang, Huaiqing Wang, Dewen Fan, Houde Liu, Qianqian Gan, Fan Yang, Tingting Gao
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.