MergeDistill: Merging Pre-trained Language Models using Distillation.
Simran KhanujaMelvin JohnsonPartha P. TalukdarPublished in: CoRR (2021)
Keyphrases
- multimedia
- language model
- pre trained
- language modeling
- training data
- n gram
- training examples
- speech recognition
- document retrieval
- information retrieval
- language modelling
- probabilistic model
- retrieval model
- query expansion
- test collection
- statistical language models
- smoothing methods
- language models for information retrieval
- document ranking
- control signals
- relevance model
- query specific
- neural network