Dafny Code Synthesis on Dafny Hard Subset Quality-filtered (val)
31.1Pass RateMulti-turn RLVR
Evaluation Results
| Method | Links | |
|---|---|---|
| Multi-turn RLVRTurns=Multi-turn, Filtering=Filtered, Evaluation protocol=verify@≤ 42026.05 | 31.1 | |
| Filtered RLVRTurns=Single-turn, Filtering=Filtered2026.05 | 9.7 |