Dafny Code Synthesis on APPS Vericoding-derived (test)
83Pass RateModel union
Evaluation Results
| Method | Links | |
|---|---|---|
| Model unionAttempts per task=52026.05 | 83 | |
| GPT-5 miniAttempts per task=52026.05 | 71.6 | |
| Claude Opus 4.1Attempts per task=52026.05 | 66.6 | |
| Gemini 2.5 FlashAttempts per task=52026.05 | 36.9 |