Multimodal Scientific Learning Beyond Diffusions and Flows
Leonardo Ferreira Guilhoto, Akshat Kaushal, Paris Perdikaris · 2026
Read Multimodal Scientific Learning Beyond Diffusions and Flows by Leonardo Ferreira Guilhoto, Akshat Kaushal, Paris Perdikaris (2026) 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-2602-00960.txt.
Description
Leonardo Ferreira Guilhoto and colleagues argue that mixture density nets, long overlooked, suit science problems with many possible answers. Diffusion and flow models, though expressive, tend to need data and compute, and fit poorly with the solution shapes of physics problems. As density estimators, mixture nets spread probability across distinct answer branches at once, which helps when data are scarce. Tests on inverse, multistable, and chaotic tasks show these nets beat rivals on generalization, clarity, and efficiency. The authors frame them as a lean route to multimodal uncertainty in science.
Catalog record
| CREATOR | Leonardo Ferreira Guilhoto, Akshat Kaushal, Paris Perdikaris |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 80527 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.00960