DARN: Dynamic Adaptive Regularization Networks for Efficient and Robust Foundation Model Adaptation
Dhenenjay Yadav, Rohan Sawai · 2025
Read DARN: Dynamic Adaptive Regularization Networks for Efficient and Robust Foundation Model Adaptation by Dhenenjay Yadav, Rohan Sawai (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-2511-04766.txt.
Description
Dhenenjay Yadav and his co-author present DARN, a decoder for adapting foundation models to satellite imagery. Foundation models give strong features, but satellite scenes vary wildly in content and difficulty. Fixed regularization in the decoder ignores that variety.
DARN first predicts each sample's task complexity with a light predictor. Dropout rates then shift per sample across a set range based on that estimate. Dynamic capacity gating also adjusts how many channels stay active. Together, the three parts tune the decoder's capacity to each image, aiming for robust adaptation across mixed geospatial scenes.
Catalog record
| CREATOR | Dhenenjay Yadav, Rohan Sawai |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-04766