| title | DiFA: Inference-Time Forward-Process Alignment for Diffusion Models |
|---|---|
| authors | Shigui Li, Delu Zeng |
| date | 2026-07-20 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2607.17972 |
| affiliation | South China University of Technology |
| upvotes | 7 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | The prevailing inference framework for diffusion models formulates generation fundamentally as a problem of numerical integration. This perspective casts the model as an exact estimator, neglecting the inherent statistical uncertainty of the denoising process. In this work, we propose Forward-Process Aligned Diffusion prediction (DiFA), a training-free framework that reframes inference-time data prediction refinement as a sequential state estimation problem. Rather than reusing past outputs solely for numerical integration, DiFA treats iterative data predictions along the reverse trajectory as correlated observations to build a forward-aligned temporal consensus. Inspired by Kalman filtering, this consensus aggregates historical predictions according to structural consistency and noise-level compatibility. To counteract the over-smoothing tendency of temporal consensus, we introduce a deviation guidance mechanism to adaptively preserve residual details. Empirically, DiFA yields significant improvements on CIFAR-10 and ImageNet across the evaluated metrics, including FID, IS, and FD-DINOv2, demonstrating that aligning inference with the forward statistical structure substantially improves generative fidelity. |
cite
@misc{li2026difa,
title={DiFA: Inference-Time Forward-Process Alignment for Diffusion Models},
author={Shigui Li and Delu Zeng},
year={2026},
eprint={2607.17972},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2607.17972}
}