VL-Trojan: Multimodal Instruction Backdoor Attacks against Autoregressive Visual Language Models.
Jiawei LiangSiyuan LiangMan LuoAishan LiuDongchen HanEe-Chien ChangXiaochun CaoPublished in: CoRR (2024)
Keyphrases
- language model
- autoregressive
- language modeling
- moving average
- document retrieval
- n gram
- non stationary
- information retrieval
- speech recognition
- random fields
- probabilistic model
- query expansion
- gaussian markov random field
- multi modal
- language modelling
- multimedia
- smoothing methods
- test collection
- statistical language models
- relevance model
- visual information
- retrieval model
- sar images
- language models for information retrieval
- visual features
- low level
- image processing
- gray level
- markov random field
- higher order
- co occurrence
- least squares
- bayesian networks
- feature extraction