Diffusion Theory as a Scalpel: Detecting and Purifying Poisonous Dimensions in Pre-trained Language Models Caused by Backdoor or Bias.
Zhiyuan ZhangDeli ChenHao ZhouFandong MengJie ZhouXu SunPublished in: ACL (Findings) (2023)
Keyphrases
- language model
- pre trained
- language modeling
- n gram
- probabilistic model
- document retrieval
- statistical language models
- language modelling
- query expansion
- information retrieval
- retrieval model
- speech recognition
- relevance model
- document ranking
- test collection
- training data
- language models for information retrieval
- data sets
- neural network
- smoothing methods
- bayesian networks
- clustering algorithm
- learning algorithm
- machine learning
- spoken term detection