arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.08646cs.LG
收藏

梯度预处理实现高效可靠的奖励引导生成

Gradient Preconditioning for Efficient and Reliable Reward-Guided Generation

KAIST(韩国科学技术院)

Jisung Hwang, Minhyuk Sung

AI总结:

提出一种梯度预处理方法,通过将奖励梯度投影到白高斯噪声可行集上,实现一步生成模型在奖励引导生成中的高效性和可靠性,防止奖励黑客攻击并加速优化。

摘要:

我们提出了一种梯度预处理方法,使得使用一步生成模型的奖励引导生成既高效又可靠。测试时噪声优化可以从预训练生成模型中解锁显著更好的奖励引导生成,但它容易导致奖励黑客攻击,从而降低质量,并且通常对于实际使用来说太慢。我们通过将奖励梯度投影到一个精心设计的白高斯噪声可行集上来预处理奖励梯度,该可行集是一个具有块状范数约束的紧凑谱集,紧密捕捉白高斯噪声的统计特性和空间不相关性。这种预处理将每个梯度更新重塑为噪声对齐的方向,驱动更快更有效的奖励上升,同时防止奖励黑客攻击。该投影具有闭式解,并且与FFT的$O(N \log N)$复杂度相匹配,在实践中增加了可忽略的开销。在FLUX上使用四个奖励模型的实验中,我们的方法仅使用最先进的基于正则化方法所需挂钟时间的30%就达到了可比的审美分数。

英文摘要:

We propose a gradient preconditioning method that makes reward-guided generation with one-step generative models both efficient and reliable. Test-time noise optimization can unlock substantially better reward-guided generations from pretrained generative models, but it is prone to reward hacking that degrades quality and is often too slow for practical use. We precondition reward gradients by projecting them onto a carefully designed white Gaussian noise feasible set, a compact spectral set with blockwise norm constraints that tightly captures the statistics and spatial uncorrelatedness of white Gaussian noise. This preconditioning reshapes each gradient update into a noise-aligned direction, driving faster and more effective reward ascent while preventing reward hacking. The projection is closed-form and matches the $O(N \log N)$ complexity of FFT, adding negligible overhead in practice. In experiments on FLUX with four reward models, our approach reaches a comparable Aesthetic Score using only 30% of the wall-clock time required by the state-of-the-art regularization-based method.

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。

↑