Code Generation on MathQA Python Filtered (dev)
20.7PASS@1GPT-Neo 2.7B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-Neo 2.7Bmethod=self-sampling FCS + PCS2022.05 | 20.7 | 36.2 | |
| GPT-Neo 125Mmethod=self-sampling FCS + PCS2022.05 | 11.7 | 28.2 | |
| Codex Davincifew-shot learning=true2022.05 | 5 | 40 |