The relative importance of being Gaussian
F. Alberto Grünbaum and Tongdi Xu · 2025
Read The relative importance of being Gaussian by F. Alberto Grünbaum and Tongdi Xu (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-2507-08059.txt.
Description
Alberto Grünbaum and Tongdi Xu asked in 2025 how much of a champion denoising method's magic is simply that its noise is Gaussian. Diffusion models clean images superbly, and their derivations lean on Gaussian facts: the bell curve is fixed by its mean and spread, and two Gaussians sum to a Gaussian. The note swaps the noise and changes nothing else.
Run the same algorithm on uniform noise, on beta-shaped noise, or on a mix of two Gaussians of very different widths, the authors propose, and see what survives. The point is diagnostic rather than helpful: to learn which steps of the method are truly Gaussian-dependent and which are robust. The head spells the second name slightly differently from the catalog; surnames are used here. Readers get a stress test of a fashionable algorithm's favorite assumption.
Catalog record
| CREATOR | F. Alberto Grünbaum and Tongdi Xu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 7 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-08059