Backdoor Activation Attack: Attack Large Language Models using Activation Steering for Safety-Alignment.
Haoran WangKai ShuPublished in: CoRR (2023)
Keyphrases
- language model
- language modeling
- document retrieval
- query expansion
- language modelling
- probabilistic model
- n gram
- speech recognition
- test collection
- retrieval model
- information retrieval
- statistical language models
- context sensitive
- document ranking
- pseudo relevance feedback
- smoothing methods
- document level
- document length
- language models for information retrieval