Continuous Embedding Attacks via Clipped Inputs in Jailbreaking Large Language Models.
Zihao XuYi LiuGelei DengKailong WangYuekang LiLing ShiStjepan PicekPublished in: CoRR (2024)
Keyphrases
- language model
- language modeling
- probabilistic model
- n gram
- watermarking algorithm
- document retrieval
- speech recognition
- information retrieval
- retrieval model
- language modelling
- watermarking technique
- statistical language models
- smoothing methods
- watermarking scheme
- query expansion
- test collection
- context sensitive
- query terms
- vector space
- language models for information retrieval
- ad hoc information retrieval
- relevance model
- language model for information retrieval
- machine learning
- vector space model
- query specific
- visual features
- document length
- word error rate
- document ranking
- okapi bm
- continuous variables