As firm as their foundations: can open-sourced foundation models be used to create adversarial examples for downstream tasks?
Foundation models pre-trained on web-scale vision-language data, such as CLIP, are widely used as cornerstones of powerful machine learning systems. While pre-training offers clear advantages for downstream learning, it also endows downstream models with shared adversarial vulnerabilities that can b...
Main Authors: | Hu, A, Gu, J, Pinto, F, Kamnitsas, K, Torr, P |
---|---|
格式: | Internet publication |
语言: | English |
出版: |
2024
|
相似书籍
-
As firm as their foundations: creating transferable adversarial examples across downstream tasks with CLIP
由: Hu, A, et al.
出版: (2024) -
Open Sourcing as a Profit-Maximizing Strategy for Downstream Firms
由: Gambardella, Alfonso, et al.
出版: (2021) -
OSGeo - Open Source Geospatial Foundation
由: Margherita Di Leo, et al.
出版: (2012-09-01) -
OSGeo - Open Source Geospatial Foundation
由: Margherita Di Leo, et al.
出版: (2012-09-01) -
Causality-driven feature selection and domain adaptation for enhancing chemical foundation models in downstream tasks
由: Eduardo Soares, et al.
出版: (2025-01-01)