Stealthy and Persistent Unalignment on Large Language Models via Backdoor Injections.
Yuanpu CaoBochuan CaoJinghui ChenPublished in: NAACL-HLT (2024)
Keyphrases
- language model
- language modeling
- probabilistic model
- n gram
- information retrieval
- document retrieval
- speech recognition
- retrieval model
- language modelling
- statistical language models
- query expansion
- smoothing methods
- test collection
- query terms
- translation model
- ad hoc information retrieval
- language models for information retrieval
- vector space model
- document ranking
- decision trees
- document length
- relevance model
- query specific
- term dependencies
- passage retrieval
- word error rate
- information extraction