Learning from human feedback has been shown to improve text-to-image models. These techniques first learn a reward function that captures what humans care about in the task and then improve the models based on the learned reward function. Even though relatively simple approaches (e.g., rejection sampling based on reward scores) have been investigated, fine-tuning text-to-image models with the reward function remains challenging. In this work, we propose using online reinforcement learning (RL) to fine-tune text-to-image models. We focus on diffusion models, defining the fine-tuning task as an RL problem, and updating the pre-trained text-to-image diffusion models using policy gradient to maximize the feedback-trained reward. Our approach, coined DPOK, integrates policy optimization with KL regularization. We conduct an analysis of KL regularization for both RL fine-tuning and supervised fine-tuning. In our experiments, we show that DPOK is generally superior to supervised fine-tuning with respect to both image-text alignment and image quality. Our code is available at https://github.com/google-research/google-research/tree/master/dpok.

翻译：利用人类反馈来学习已被证明能够改善文本到图像模型。这类技术首先学习一个奖励函数，该函数捕捉人类在任务中所关注的内容，然后基于所学到的奖励函数对模型进行改进。尽管一些相对简单的方法（例如基于奖励分数的拒绝采样）已被研究，但基于奖励函数微调文本到图像模型仍然具有挑战性。在这项工作中，我们提出使用在线强化学习（RL）对文本到图像模型进行微调。我们聚焦于扩散模型，将微调任务定义为一个RL问题，并通过策略梯度更新预训练的文本到图像扩散模型，以最大化基于反馈训练的奖励。我们提出的方法称为DPOK，它将策略优化与KL正则化相结合。我们对RL微调和监督微调中的KL正则化进行了分析。实验表明，DPOK在图像-文本对齐和图像质量方面均普遍优于监督微调。我们的代码可在https://github.com/google-research/google-research/tree/master/dpok获取。