Multimodal LLMs Vulnerable To Video Jailbreaking
New research indicates that multimodal large language models (MLLMs) exhibit unreliable spatial lexical bias. The study details mechanistic diagnostics of this bias in MLLMs' spatial reasoning capabilities. Authors Chuang Ma, Qianying Liu, and others contributed to the research.
Topics
Developing
- 882d Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore.
- 882d Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
- 882d Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est.
- 882d Sed ut perspiciatis unde omnis iste natus error sit voluptatem accusantium doloremque laudantium.
Sources · 7 independent
Modernity/arxiv
“Mechanistic Diagnostics of Spatial Lexical Bias in Multimodal Large Language Model Spatial Reasoning. Authors: Chuang Ma, Qianying Liu, Tomoyuki Obuchi, Fei Cheng, Wang Yang, Sudong Cai, Shuyuan Zheng, Akiko Aizawa, Sadao Kurohashi Abstract: Multimodal large language models (MLLMs) remain unreliable”
Modernity/arxiv
“Jailbreaking Multimodal Large Language Models using Multi-Clip Video.”
Bluesky Social
“As multimodal large language models (MLLMs) have advanced to process video inputs, concerns have emerged about their poten...”
Unlock the full story
Get a Pro subscription or above to see the live story progression and the full list of independent sources confirming each event as they happen.
Log in to upgrade