Visual-RolePlay: Universal Jailbreak Attack on MultiModal Large Language Models via Role-playing Image Characte.
Siyuan MaWeidi LuoYu WangXiaogeng LiuMuhao ChenBo LiChaowei XiaoPublished in: CoRR (2024)
Keyphrases
- language model
- image data
- role playing
- language modeling
- low level
- image content
- image retrieval
- image features
- image classification
- n gram
- image collections
- probabilistic model
- visual features
- query expansion
- image regions
- document retrieval
- image representation
- visual data
- speech recognition
- smoothing methods
- multimedia
- retrieval model
- visual information
- image database
- key frames
- multi modal
- document ranking
- language modelling
- information retrieval
- statistical language models
- language models for information retrieval