ROSE Doesn't Do That: Boosting the Safety of Instruction-Tuned Large Language Models with Reverse Prompt Contrastive Decoding.
Qihuang ZhongLiang DingJuhua LiuBo DuDacheng TaoPublished in: ACL (Findings) (2024)
Keyphrases
- language model
- language modeling
- n gram
- document retrieval
- speech recognition
- information retrieval
- probabilistic model
- retrieval model
- language modelling
- query expansion
- test collection
- smoothing methods
- statistical language models
- vector space model
- language models for information retrieval
- query terms
- query specific
- multimedia
- learning algorithm
- translation model
- document ranking
- feature selection
- ad hoc information retrieval
- context sensitive
- knn
- term dependencies
- classification accuracy
- document length