Injecting Undetectable Backdoors in Deep Learning and Language Models.
Alkis KalavasisAmin KarbasiArgyris OikonomouKaterina SotirakiGrigoris VelegkasManolis ZampetakisPublished in: CoRR (2024)
Keyphrases
- deep learning
- language model
- language modeling
- unsupervised learning
- n gram
- document retrieval
- information retrieval
- probabilistic model
- machine learning
- language modelling
- unsupervised feature learning
- statistical language models
- test collection
- retrieval model
- query expansion
- mental models
- smoothing methods
- knn
- weakly supervised
- computer vision
- information extraction
- relevance model
- pattern recognition
- multiscale