Geometry over Density: Few-Shot Cross-Domain OOD Detection
Shawn Li, You Qin, Jiate Li, Charith Peris, Lisa Bauer, Roger Zimmermann, Yue Zhao · 2026
Read Geometry over Density: Few-Shot Cross-Domain OOD Detection by Shawn Li, You Qin, Jiate Li, Charith Peris, Lisa Bauer, Roger Zimmermann, Yue Zhao (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-2605-03410.txt.
Description
Shawn Li, You Qin, Jiate Li, Charith Peris, Lisa Bauer, Roger Zimmermann, and Yue Zhao propose train-once, deploy-anywhere detection of out-of-distribution inputs. Their method uses diffusion geometry instead of density alone. Standard detectors are tied to one training domain. Real deployment faces new domain pairs, with only a handful of known-good samples.
Their UFCOD method reads two energy features from diffusion score paths. The features are path energy and dynamics energy, forming a Sobolev-style signature. Each new task needs about a hundred unlabeled in-domain samples, and no retraining. Across twelve cross-domain benchmarks, accuracy was strong, rivaling methods trained on vastly more data.
Catalog record
| CREATOR | Shawn Li, You Qin, Jiate Li, Charith Peris, Lisa Bauer, Roger Zimmermann, Yue Zhao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.03410