InfoPrompt: Information-Theoretic Soft Prompt Tuning for Natural Language Understanding.
Junda WuTong YuRui WangZhao SongRuiyi ZhangHandong ZhaoChaochao LuShuai LiRicardo HenaoPublished in: NeurIPS (2023)
Keyphrases
- information theoretic
- natural language understanding
- information theory
- mutual information
- knowledge representation
- text understanding
- natural language
- semantic analysis
- theoretic framework
- natural language processing
- language understanding
- jensen shannon divergence
- information theoretic measures
- information bottleneck
- dialogue system
- semantic representations
- log likelihood
- kullback leibler divergence
- minimum description length
- relative entropy
- spoken dialog systems
- image registration
- image processing
- machine learning