通过奖励倾斜分布匹配强化少步生成器

Reinforcing Few-step Generators via Reward-Tilted Distribution Matching

AI 导读

本文提出奖励倾斜分布匹配蒸馏（RTDMD），这是一个将分布匹配蒸馏与奖励引导强化学习统一应用于少步流生成器的两阶段框架。该方法通过最小化到奖励倾斜教师分布的KL散度，自然分解为分布匹配项与奖励最大化项。第一阶段引入环境一致分布匹配蒸馏（AC-DMD），在子区间进行分布匹配，并通过一致性正则化辅助分数模型追踪生成器分布。第二阶段联合优化两项，并推导混合策略梯度及步子集GRPO（SubGRPO）以降低方差。在SD3、SD3.5和FLUX.2上的实验表明，RTDMD仅用4步推理即可在偏好、美学和组合指标上达到新的 state-of-the-art。

推荐理由

这篇直接把分布匹配蒸馏和奖励建模拧在一起，在 SD3/3.5/FLUX.2 上用 4 步推理就压了之前所有文生图对齐方法，做图像生成训练和偏好对齐的该看。

正文 · 原文

Recent advances in few-step diffusion distillation have enabled efficient image generation, yet aligning these models with human preferences remains challenging. We propose Reward-Tilted Distribution Matching Distillation (RTDMD), a two-stage framework that unifies distribution matching distillation with reward-guided reinforcement learning for few-step flow generators. We show that minimizing the KL divergence to a reward-tilted teacher distribution naturally decomposes into a distribution matching term and a reward maximization term. In the first stage, we introduce Ambient-Consistent Distribution Matching Distillation (AC-DMD), which performs subinterval-wise distribution matching and augments the fake score objective with a consistency regularizer to help the fake score model track the shifting generator distribution under limited updates. In the second stage, we jointly optimize both terms: for the reward maximization term, we derive a hybrid policy gradient that combines a GRPO-style estimator for the stochastic intermediate transitions with direct reward backpropagation through the deterministic final step, and further introduce step-subset GRPO (SubGRPO) to reduce variance. Experiments on SD3, SD3.5, and FLUX.2 demonstrate that RTDMD establishes new state-of-the-art results across preference, aesthetic, and compositional metrics with only 4 inference steps, outperforming previous few-step text-to-image generation methods. Code and models are available at https://github.com/Harahan/RTDMD.