Abstract
Recent advances in distillation and flow-map models have enabled deterministic one- or few-step generation for high-quality data, facilitating a new branch of reward alignment approaches that directly optimize the initial noise from a Gaussian distribution. However, most existing initial-noise optimization methods rely on first-order gradient information, which is either inapplicable or suffers from instability and inefficiency in black-box reward scenarios. Here, we introduce ZeNOVA, a stable and efficient initial noise alignment method in a gradient-free manner. Specifically, we address existing algorithms' major challenge in black-box scenarios through annealed soft-value guidance, manifold-constrained hyperspherical Langevin dynamics, and Metropolis-Hastings jumping. Extensive experiments on image and video generative models show that ZeNOVA outperforms all evaluated zeroth-order baselines by optimizing the initial noise toward higher rewards substantially more stably while exploiting the geometry of the Gaussian prior, demonstrating its practical applicability to various black-box reward alignment.
Keywords
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Chang, J., & Ye, J. C. (2026). Manifold-Constrained Initial Noise Optimization for Efficient Generative Model Alignment. https://omanscience.com/en/articles/manifold-constrained-initial-noise-optimization-for-efficient-generative-model-alignment
MLA 9
Chang, Jinho, and Jong Chul Ye. "Manifold-Constrained Initial Noise Optimization for Efficient Generative Model Alignment." https://omanscience.com/en/articles/manifold-constrained-initial-noise-optimization-for-efficient-generative-model-alignment.
Chicago (author–date)
Chang, Jinho, and Jong Chul Ye. 2026. "Manifold-Constrained Initial Noise Optimization for Efficient Generative Model Alignment." https://omanscience.com/en/articles/manifold-constrained-initial-noise-optimization-for-efficient-generative-model-alignment.
Harvard
Chang, J. and Ye, J. C. (2026) 'Manifold-Constrained Initial Noise Optimization for Efficient Generative Model Alignment', Available at: https://omanscience.com/en/articles/manifold-constrained-initial-noise-optimization-for-efficient-generative-model-alignment.
Vancouver
Chang J, Ye JC. Manifold-Constrained Initial Noise Optimization for Efficient Generative Model Alignment. https://omanscience.com/en/articles/manifold-constrained-initial-noise-optimization-for-efficient-generative-model-alignment
IEEE
J. Chang, and J. C. Ye, "Manifold-Constrained Initial Noise Optimization for Efficient Generative Model Alignment," https://omanscience.com/en/articles/manifold-constrained-initial-noise-optimization-for-efficient-generative-model-alignment.