An image is worth 1000 lies: adversarial transferability across prompts on vision-language models
Different from traditional task-specific vision models, recent large VLMs can readily adapt to different vision tasks by simply using different textual instructions, i.e., prompts. However, a well-known concern about traditional task-specific vision models is that they can be misled by imperceptible...
Main Authors: | Luo, H, Gu, J, Liu, F, Torr, P |
---|---|
Format: | Conference item |
Language: | English |
Published: |
OpenReview
2024
|
Similar Items
-
Robust Testing of AI Language Model Resiliency with Novel Adversarial Prompts
by: Brendan Hannon, et al.
Published: (2024-02-01) -
CPT: Colorful Prompt Tuning for pre-trained vision-language models
by: Yuan Yao, et al.
Published: (2024-01-01) -
A picture's worth 1,000 words : a workbook for visual communications /
by: 428491 Westcott, Jean, et al.
Published: (1997) -
A picture is worth 1000 words: textual analysis of routine 20-week scan
by: Alsharid, M, et al.
Published: (2022) -
A Picture is Worth 1000 Words: Teaching Science Communication with Graphical Abstract Assignments
by: Swati Agrawal, et al.
Published: (2023-04-01)