Crossmark

Adaptive ensemble attack: breaking Large Multimodal Models via dynamic caption selection and weighted gradients

Published Online: 2026-05-19

Published Print: 2026-06

Authors

Pandey, Sudhir Kumar

Mi, Jian-Xun

Ahmad, Israr

Pathan, Muhammad Salman
Funding

Funding for this research was provided by:

Dublin City University
License Information

Text and Data Mining valid from 2026-05-19

Version of Record valid from 2026-05-19
More Information

Article History

Received: 17 December 2025

Accepted: 30 March 2026

First Online: 19 May 2026

Declarations

:

: The authors declare no conflict of interest.

Document is current