Unlocking Few-Step Diffusion for Faithful Previews
Abstract
Sampling latency compounds in diffusion workflows, where users generate and discard many candidates before keeping one. Surprisingly, the poor outputs of standard few-step samplers do not reflect a lack of reconstruction capacity: by optimizing only the initial noise, frozen 3-4-step samplers can closely reproduce their corresponding full-step outputs. Building on this finding, we learn corrections to the initial noise and denoising updates using endpoint supervision, improving correspondence with full-step outputs generated from the same noise and prompt. The resulting previews allow users to screen candidates cheaply and reserve full-step generation for promising ones. Input correction also transfers across sampling budgets without retraining. Experiments show substantial improvements in reference fidelity, including 53-78% lower reconstruction MSE than retrained LD3 on unconditional benchmarks, alongside improved ranking preservation and candidate selection on SD1.5, SDXL, and FLUX.1-dev.
Related
- Unlocking Cross-Scenario Physical Layer Security: A Mixture-of-Experts Framework with Generative Diffusion Models
- A Faithful Action of the Mapping Class Group Mod(Sg) on the (2g−1)-Sphere
- The Burau representation of the braid group is faithful for n = 4
- A fully faithful p-adic Riemann--Hilbert functor for filtered coherent D-modules
- A faithful representation and profinite completion
BibTeX
@misc{jia2026unlocking,
title = {{Unlocking Few-Step Diffusion for Faithful Previews}},
author = {Jing Jia and Sifan Liu and Guanyang Wang},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.34406},
eprint = {2609.34406},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.34406},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Unlocking Few-Step Diffusion for Faithful Previews
AU - Jing Jia
AU - Sifan Liu
AU - Guanyang Wang
PY - 2026
DA - 2026/09/28
PB - arXiv
DO - 10.48550/arxiv.2609.34406
UR - https://arxiv.org/abs/2609.34406
AB - Sampling latency compounds in diffusion workflows, where users generate and discard many candidates before keeping one. Surprisingly, the poor outputs of standard few-step samplers do not reflect a lack of reconstruction capacity: by optimizing only the initial noise, frozen 3-4-step samplers can closely reproduce their corresponding full-step outputs. Building on this finding, we learn corrections to the initial noise and denoising updates using endpoint supervision, improving correspondence with full-step outputs generated from the same noise and prompt. The resulting previews allow users to screen candidates cheaply and reserve full-step generation for promising ones. Input correction also transfers across sampling budgets without retraining. Experiments show substantial improvements in reference fidelity, including 53-78% lower reconstruction MSE than retrained LD3 on unconditional benchmarks, alongside improved ranking preservation and candidate selection on SD1.5, SDXL, and FLUX.1-dev.
ER -