Leveraging Perturbation Robustness to Enhance Out-of-Distribution Detection
Wenxi Chen, Raymond A. Yeh, Shaoshuai Mou, Yan Gu · 2025
Read Leveraging Perturbation Robustness to Enhance Out-of-Distribution Detection by Wenxi Chen, Raymond A. Yeh, Shaoshuai Mou, Yan Gu (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-2503-18784.txt.
Description
Wenxi Chen, Raymond Yeh, Shaoshuai Mou, and Yan Gu improve detection of out of distribution inputs in a 2025 paper. Vision models must flag images that differ from their training data to stay safe in open settings. Their post hoc method rests on a simple insight: odd inputs lose confidence faster than normal ones when slightly perturbed.
The method searches downhill from each input for a low score nearby, widening the gap between normal and odd samples. It needs no change to the model itself. On a standard benchmark it led post hoc methods for small models. On one adversarially trained model, false alarms at a fixed catch rate fell by more than ten percent.
Catalog record
| CREATOR | Wenxi Chen, Raymond A. Yeh, Shaoshuai Mou, Yan Gu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-18784