arXiv:2607. 17572v1 Announce Type: new Abstract: Group Relative Policy Optimization (GRPO) is a powerful reinforcement learning algorithm for aligning generative models with human preferences.
arXiv:2607. 17572v2 Announce Type: replace Abstract: Group Relative Policy Optimization (GRPO) is a powerful reinforcement learning algorithm for aligning generative models with human preferences.
Group Relative Policy Optimization (GRPO) is a powerful reinforcement learning algorithm for aligning generative models with human preferences. While successful in large language models~\cite{shao2024deepseekmathpushinglimitsmathematical}, its extension to diffusion and flow matching models introduces a severe computational bottleneck: gradients must be back-propagated through the high-capacity DiT backbone at \emph{every} timestep of the sampling trajectory, making high-resolution text-to-image (T2I) training prohibitively expensive.
arXiv:2609.37974v1 Announce Type: cross Abstract: Masked diffusion models (MDMs) generate text by unmasking several tokens per step, but they are trained and sampled under different conditions. The m...
arXiv:2604. 18194v2 Announce Type: replace Abstract: Single-step generators promise high-fidelity synthesis at a fraction of the inference and training cost of ordinary differential equation (ODE)-based flow models, a central concern when compute is limited.
arXiv:2606. 08602v1 Announce Type: cross Abstract: We present an online reinforcement learning (RL) algorithm for fine-tuning flow-matching policies in continuous-control problems.
arXiv:2606. 11075v1 Announce Type: new Abstract: Aligning text-to-image flow matching models with human preferences via direct reward backpropagation is sample-efficient but hampered by two well-known pathologies: activations cannot be stored across the full sampling trajectory at modern model scale, and chained Jacobian products across steps inflate the reward gradient as it travels back to early indices.
arXiv:2509.25050v2 Announce Type: replace Abstract: Reinforcement Learning (RL) has emerged as a central paradigm for advancing Large Language Models (LLMs), where both pre-training and RL post-train...
arXiv:2606. 30376v1 Announce Type: new Abstract: Aligning generative flow models on continuous spaces via online reinforcement learning is constrained by intractable trajectory likelihoods.
LeanGRPO eliminates redundant recomputation in diffusion reinforcement learning by reusing computation graphs and activations from rollout for policy updates, or by backpropagating provisional gradients and correcting them later. It introduces two training schedules—LeanGRPO‑Retain and LeanGRPO‑Reweight—that target different model scales and input sizes. Experiments on FlowGRPO/DanceGRPO with FLUX.1‑dev and Wan show up to a 1.83× end‑to‑end speedup while preserving the original optimization objective.
GeoSPRINT is a training‑free framework that constructs non‑uniform sampling schedules for diffusion model inference by detecting geometrically redundant steps in denoising trajectories. It uses a hyperplanarity test in latent space, implemented via QR factorization, to allocate more steps to high‑curvature regions, and introduces the trajectory projection score α_traj as a model‑free diagnostic for flow quality. Across CIFAR‑10, LSUN Church, and Stable Diffusion v1.5, GeoSPRINT consistently outperforms uniform DDIM schedules at matched NFE budgets, improving FID scores by up to 1.93 points.
arXiv:2610.01896v1 Announce Type: cross Abstract: Asynchronous reinforcement learning (RL) improves the efficiency of large language model post-training but introduces stale rollouts generated by ear...
Accelerating Video Diffusion via Training-Free Trajectory Routing (TRACK) introduces a heterogeneous denoising strategy that switches between large and small diffusion models at selected steps, determined by a calibration process that measures disagreement between model predictions. By routing quality-sensitive steps to the large model and low-disagreement steps to the small model, TRACK achieves significant speedups—up to 2.73×—across several video diffusion benchmarks while maintaining comparable quality and diversity. The method requires no retraining, architectural changes, or online dual-model evaluation, making it a practical acceleration paradigm for video diffusion.