← 研踪 PaperTrailarXiv:2611.25998 · v1
Reinforcing Few-Step Diffusion Models via Shadow Posterior Matching
首次提交 2026-11-19 20:18
版本更新 2026-11-19 20:18
雷达观测 2026-11-20 16:53
摘要
Few-step diffusion models are widely deployed for their efficiency. However, reinforcing them with black-box rewards, as is routine for multi-step models, faces two structural challenges. First, a few-step generator is a sampler rather than a score model, so its velocity is not the forward velocity of the generated distribution. Second, it must be anchored to the undistilled teacher, which shares no transitions with it. Existing few-step RL methods handle the latter with a marginal KL to the teacher, but each pays for the former in its reward credit. We propose Shadow Posterior Matching (SPM), which renoises generated samples into shadow inputs and obtains the credit from their shadow posterior. From the KL-regularized reward objective, we derive a regression loss that turns reward values into a reward direction while anchoring the generator's outputs to the teacher and aligning the shadow inputs with the deployed inputs. Applied to 4-step generators, SPM achieves higher rewards with less reward hacking than prior few-step RL methods.
作者备注:Code available at https://github.com/shadow-posterior/code
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-19 20:18 · 最新观测版本
Reinforcing Few-Step Diffusion Models via Shadow Posterior Matching
Few-step diffusion models are widely deployed for their efficiency. However, reinforcing them with black-box rewards, as is routine for multi-step models, faces two structural challenges. First, a few-step generator is a sampler rather than a score model, so its velocity is not the forward velocity of the generated distribution. Second, it must be anchored to the undistilled teacher, which shares no transitions with it. Existing few-step RL methods handle the latter with a marginal KL to the teacher, but each pays for the former in its reward credit. We propose Shadow Posterior Matching (SPM), which renoises generated samples into shadow inputs and obtains the credit from their shadow posterior. From the KL-regularized reward objective, we derive a regression loss that turns reward values into a reward direction while anchoring the generator's outputs to the teacher and aligning the shadow inputs with the deployed inputs. Applied to 4-step generators, SPM achieves higher rewards with less reward hacking than prior few-step RL methods.
查看此版本解读 →查看此版本原文 ↗