SEER: Backdoor Detection for Vision-Language Models through Searching Target Text and Image Trigger Jointly.
Liuwan ZhuRui NingJiang LiChunsheng XinHongyi WuPublished in: AAAI (2024)
Keyphrases
- language model
- language modeling
- n gram
- image features
- information retrieval
- image classification
- image content
- document retrieval
- image representation
- statistical language models
- text mining
- test collection
- image regions
- probabilistic model
- document level
- document ranking
- image retrieval
- visual features
- low level
- query expansion
- natural language processing
- retrieval effectiveness
- context sensitive
- document analysis
- smoothing methods