Safety
Adaptive Moments are Surprisingly Effective for Plug-and-Play Diffusion Sampling
arXiv:2603.16797v2 Announce Type: replace-cross Abstract: Guided diffusion sampling relies on approximating often intractable likelihood scores, which introduces significant noise into the sampling dy
arXiv:2603.16797v2 Announce Type: replace-cross Abstract: Guided diffusion sampling relies on approximating often intractable likelihood scores, which introduces significant noise into the sampling dynamics. We propose using adaptive moment estimation to stabilize these noisy likelihood scores during sampling. Despite its simplicity, our approach achieves state-of-the-art results on image restoration and class-conditional generation tasks, outperforming more complicated methods, which are often computationally more expensive. We provide empirical analysis of our method on both synthetic and real data, demonstrating that mitigating gradient noise through adaptive moments offers an effective way to improve alignment.
Related
- Denoising, Fast and Slow: Difficulty-Aware Adaptive Sampling for Image Generation
- Guiding Distribution Matching Distillation with Gradient-Based Reinforcement Learning
- DepthMaster: Taming Diffusion Models for Monocular Depth Estimation
Source: arXiv cs.CV | 2026-04-24