Refusal in Language Models Is Mediated by a Single Direction.
Andy ArditiOscar ObesoAaquib SyedDaniel PalekaNina RimskyWes GurneeNeel NandaPublished in: CoRR (2024)
Keyphrases
- language model
- language modeling
- probabilistic model
- statistical language models
- n gram
- document retrieval
- information retrieval
- retrieval model
- language modelling
- query expansion
- speech recognition
- context sensitive
- test collection
- relevance model
- language model for information retrieval
- co occurrence
- language models for information retrieval
- query terms
- machine learning
- retrieval effectiveness
- document ranking
- active learning
- document length
- search engine