/AI1d ago

CVPR Paper Shows Newer Text-to-Image Models Yield Worse Synthetic Data

19451.2K
#1824Luca Ambrogioni
Stanislav Frolov
Original postLuca Ambrogioni#1824
Stanislav Frolov@stfrolov

We will be presenting our CVPR paper When Pretty Isn’t Useful in the afternoon today.

We benchmark modern t2i models and find that newer models make progressively worse synthetic training data and we investigate why! Come visit us at poster #46.

https://bill2462.github.io/When-Pretty-Isn-t-Useful-page/

6:46 AM · Jun 7, 2026 · 1.2K Views
Sentiment
Sentiment building, check back later.
Cluster Engagement
Posts from X
Most Activity
Most Activity
No ranked X posts are available for this story yet.