Evaluations with different pretrained language models on dev set.
| System . | Dev . |
|---|---|
| RYANSQL | 43.4% |
| RYANSQL + BERT-base | 51.4% |
| RYANSQL + BERT-large | 66.6% |
| RYANSQL + RoBERTa | 65.7% |
| RYANSQL + ELECTRA | 63.6% |
| System . | Dev . |
|---|---|
| RYANSQL | 43.4% |
| RYANSQL + BERT-base | 51.4% |
| RYANSQL + BERT-large | 66.6% |
| RYANSQL + RoBERTa | 65.7% |
| RYANSQL + ELECTRA | 63.6% |