BISON uses placeholder tokens in a bidirectional Transformer to generate sequences, and fine-tuning BERT with this scheme beats GPT2 on two dialogue tasks.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Attending to Future Tokens For Bidirectional Sequence Generation
BISON uses placeholder tokens in a bidirectional Transformer to generate sequences, and fine-tuning BERT with this scheme beats GPT2 on two dialogue tasks.