An Empirical Study of Sampling Hyperparameters in Diffusion-Based Super-Resolution
Yudhistira Arief Wibowo · 2025
Read An Empirical Study of Sampling Hyperparameters in Diffusion-Based Super-Resolution by Yudhistira Arief Wibowo (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-17675.txt.
Description
Yudhistira Arief Wibowo runs a tuning study for diffusion models that upscale a single photo. The models dream up fine detail from a blurry low-resolution input, steered by conditioning methods that nudge each step toward matching the original. Those methods bring knobs, and nobody had ranked which knobs actually matter for picture quality.
One knob towers over the rest. The conditioning step size changed results far more than the number of diffusion steps, a surprise since step count gets most of the attention and compute. Step sizes between 2.0 and 3.0 won across the FFHQ face benchmark tests. For practitioners, the message is practical: spend tuning time on step size first, and save the compute that extra steps would burn.
Catalog record
| CREATOR | Yudhistira Arief Wibowo |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-17675