RBF-Solver: A Multistep Sampler for Diffusion Probabilistic Models via Radial Basis Functions

TL;DR

RBF-Solver introduces a flexible, learnable multistep diffusion sampler using Gaussian RBFs that improves sampling efficiency and image quality in diffusion models, outperforming polynomial-based methods especially at higher function evaluations.

Contribution

It proposes RBF-Solver, a novel multistep diffusion sampler with learnable Gaussian RBFs, enabling explicit trajectory optimization and superior performance over existing polynomial-based samplers.

Findings

Outperforms polynomial-based samplers at high NFE (NFE >= 15).

Achieves state-of-the-art FID scores on CIFAR-10 with fewer evaluations.

Reduces FID significantly in low-NFE conditional generation on ImageNet.

Abstract

Diffusion probabilistic models (DPMs) are widely adopted for their outstanding generative fidelity, yet their sampling is computationally demanding. Polynomial-based multistep samplers mitigate this cost by accelerating inference; however, despite their theoretical accuracy guarantees, they generate the sampling trajectory according to a predefined scheme, providing no flexibility for further optimization. To address this limitation, we propose RBF-Solver, a multistep diffusion sampler that interpolates model evaluations with Gaussian radial basis functions (RBFs). By leveraging learnable shape parameters in Gaussian RBFs, RBF-Solver explicitly follows optimal sampling trajectories. At first order, it reduces to the Euler method (DDIM). At second order or higher, as the shape parameters approach infinity, RBF-Solver converges to the Adams method, ensuring its compatibility with existing samplers. Owing to the locality of Gaussian RBFs, RBF-Solver maintains high image fidelity even at fourth order or higher, where previous samplers deteriorate. For unconditional generation, RBF-Solver consistently outperforms polynomial-based samplers in the high-NFE regime (NFE >= 15). On CIFAR-10 with the Score-SDE model, it achieves an FID of 2.87 with 15 function evaluations and further improves to 2.48 with 40 function evaluations. For conditional ImageNet 256 x 256 generation with the Guided Diffusion model at a guidance scale 8.0, substantial gains are achieved in the low-NFE range (5-10), yielding a 16.12-33.73% reduction in FID relative to polynomial-based samplers.