ReasonGen-R1：通过SFT和RL为自回归图像生成模型赋予CoT
ReasonGen-R1: CoT for Autoregressive Image generation models through SFT and RL
AI总结：
本文提出ReasonGen-R1两阶段框架，通过推理理由数据SFT与GRPO强化学习，让自回归图像生成模型先进行文本思维链推理再生成图像，并在多个基准上达到更优表现。
摘要：
尽管思维链推理和强化学习（RL）已推动自然语言处理（NLP）取得突破，但它们与生成式视觉模型的结合仍未得到充分探索。我们提出ReasonGen-R1，这是一个两阶段框架：首先在新生成的包含书面理由的推理数据集上进行监督微调（SFT），使自回归图像生成器具备显式的基于文本的“思考”技能；随后使用Group Relative Policy Optimization（GRPO）优化其输出。为使模型能够在生成图像前通过文本进行推理，我们自动生成并发布了一个由模型构建的理由语料库，并与视觉提示配对，从而实现对物体布局、风格和场景构图的可控规划。我们的GRPO算法利用来自预训练视觉语言模型的奖励信号评估整体视觉质量，并在每次更新中优化策略。在GenEval、DPG和T2I基准上的评估表明，ReasonGen-R1始终优于强基线模型和先前的最先进模型。更多信息见：aka.ms/reasongen。
英文摘要：
Although chain-of-thought reasoning and reinforcement learning (RL) have driven breakthroughs in NLP, their integration into generative vision models remains underexplored. We introduce ReasonGen-R1, a two-stage framework that first imbues an autoregressive image generator with explicit text-based "thinking" skills via supervised fine-tuning on a newly generated reasoning dataset of written rationales, and then refines its outputs using Group Relative Policy Optimization. To enable the model to reason through text before generating images, We automatically generate and release a corpus of model crafted rationales paired with visual prompts, enabling controlled planning of object layouts, styles, and scene compositions. Our GRPO algorithm uses reward signals from a pretrained vision language model to assess overall visual quality, optimizing the policy in each update. Evaluations on GenEval, DPG, and the T2I benchmark demonstrate that ReasonGen-R1 consistently outperforms strong baselines and prior state-of-the-art models. More: aka.ms/reasongen.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
