Simplicity Prevails: The Emergence of Generalizable AIGI Detection in Visual Foundation Models
Yue Zhou, Xinan He, Kaiqing Lin, Bing Fan, Feng Ding, Bin Li · 2026
Read Simplicity Prevails: The Emergence of Generalizable AIGI Detection in Visual Foundation Models by Yue Zhou, Xinan He, Kaiqing Lin, Bing Fan, Feng Ding, Bin Li (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-2602-01738.txt.
Description
Yue Zhou and colleagues show simple beats complex in detecting generated images: specialized detectors collapse on wild data. A linear model on frozen vision foundation model signs sets a state of the art. It matches detectors and beats them in the wild, boosting scores by over thirty percent. The authors trace that power to training data with fake content. Vision language models hold forgery as a concept, while self taught models acquire crime signs. Limits: recapture and sending degrade scores, and models stay blind to VAE rebuilds and editing.
Catalog record
| CREATOR | Yue Zhou, Xinan He, Kaiqing Lin, Bing Fan, Feng Ding, Bin Li |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51959 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.01738