Simulation-free and finite-time diffusion model
cs.LG
Submitted: 2026-08-04
Updated: 2026-08-31
Code: https://github.com/kentarokaba/sfft-diffusion
License: http://creativecommons.org/licenses/by/4.0/
The gist: The performance of generative diffusion models is determined by the choice of the reference diffusion process connecting the empirical and prior distributions.
Terminology
Abstract
The performance of generative diffusion models is determined by the choice of the reference diffusion process connecting the empirical and prior distributions. Conventional approaches typically trade off simulation-free training against finite-time generation. We propose a framework for designing the reference process that achieves both simultaneously. The key idea is to prescribe tractable time-dependent conditional distributions and then construct the reference process realizing them as its marginals. This framework reveals that score matching is not fundamental to diffusion-model training but instead emerges naturally through reversal of the reference process. We further show that conditional flow matching arises as the small-noise limit of the proposed framework.
Sources
- A Survey on Audio Diffusion Models: Text To Speech Synthesis and Enhancement in Generative AI
- Sampling, Diffusions, and Stochastic Localization
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks