Acquiring Clean Language Models from Backdoor Poisoned Datasets by Downscaling Frequency Space.
Zongru WuZhuosheng ZhangPengzhou ChengGongshen LiuPublished in: ACL (1) (2024)
Keyphrases
- language model
- language modeling
- document retrieval
- language modelling
- n gram
- speech recognition
- information retrieval
- statistical language models
- probabilistic model
- query expansion
- test collection
- retrieval model
- context sensitive
- query terms
- vector space model
- relevance model
- smoothing methods
- okapi bm
- ad hoc information retrieval
- document ranking
- low dimensional
- word error rate
- co occurrence