paper-with-me

홈 › Papers

DiffusionNFT: Online Diffusion Reinforcement with Forward Process

2025-09-19 · Kaiwen Zheng, Huayu Chen, Haotian Ye, Haoxiang Wang, Qinsheng Zhang, Kai Jiang, Hang Su, Stefano Ermon, Jun Zhu, Ming-Yu Liu arxiv

Online reinforcement learning (RL) has been central to post-training language models, but its extension to diffusion models remains challenging due to intractable likelihoods. Recent works discretize the reverse sampling process to enable GRPO-style training, yet they inherit fundamental drawbacks, including solver restrictions, forward-reverse inconsistency, and complicated integration with classifier-free guidance (CFG). We introduce Diffusion Negative-aware FineTuning (DiffusionNFT), a new online RL paradigm that optimizes diffusion models directly on the forward process via flow matching. DiffusionNFT contrasts positive and negative generations to define an implicit policy improvement direction, naturally incorporating reinforcement signals into the supervised learning objective. This formulation enables training with arbitrary black-box solvers, eliminates the need for likelihood estimation, and requires only clean images rather than sampling trajectories for policy optimization. DiffusionNFT is up to $25\times$ more efficient than FlowGRPO in head-to-head comparisons, while being CFG-free. For instance, DiffusionNFT improves the GenEval score from 0.24 to 0.98 within 1k steps, while FlowGRPO achieves 0.95 with over 5k steps and additional CFG employment. By leveraging multiple reward models, DiffusionNFT significantly boosts the performance of SD3.5-Medium in every benchmark tested.

📄 PDF Abstract BibTeX arXiv:2509.16117

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

FlowAWR: Online Adaptive Flow Reinforcement via Advantage-Weighted Rectification

2026-06-29 · Zheming Fu, Ruizhe He, Wei Shang, Xiaoxiao Ma 외 arxiv

Aligning generative flow models on continuous spaces via online reinforcement learning is constrained by intractable trajectory likelihoods. Existing density-approximated policy gradient methods rely on stochastic SDE sa…

Reinforcement Learning

MeanFlowNFT: Bringing Forward-Process RL to Average-Velocity Generators

2026-07-16 · Yushi Huang, Xiangxin Zhou, Jun Zhang, Liefeng Bo 외 hf

MeanFlow generators achieve fast few-step sampling by predicting average velocities over time intervals, making them attractive for efficient generation. Reinforcement learning (RL) has become a powerful way to align dif…

Reinforcement LearningVideo Generation

Designing Reinforcement Learning for Diffusion Models: A Unified Path-Space View

2026-08-14 · Yixian Xu, Yuanrui Zhang, Shengjie Luo, Liwei Wang 외 arxiv

Reinforcement learning (RL) post-training provides a direct way to align diffusion models with human preferences and task-specific rewards. However, current RL algorithms for diffusion models remain fragmented: reverse-t…

Reinforcement Learning

V-GRPO: Online Reinforcement Learning for Denoising Generative Models Is Easier than You Think

2026-04-25 · Bingda Tang, Yuhui Zhang, Xiaohan Wang, Jiayuan Mao 외 arxiv

Aligning denoising generative models with human preferences or verifiable rewards remains a key challenge. While policy-gradient online reinforcement learning (RL) offers a principled post-training framework, its direct …

Reinforcement Learning

Flow-Factory: A Unified Framework for Reinforcement Learning in Flow-Matching Models

2026-02-13 · Bowen Ping, Chengyou Jia, Minnan Luo, Hangwei Qian 외 arxiv

Reinforcement learning has emerged as a promising paradigm for aligning diffusion and flow-matching models with human preferences, yet practitioners face fragmented codebases, model-specific implementations, and engineer…

Reinforcement Learning