Test-Driven Code Reranking on HumanEval+
75Pass@1Oracle
Evaluation Results
| Method | Links | |
|---|---|---|
| OracleSamples (N)=202026.03 | 75 | |
| OracleSamples (N)=102026.03 | 70.12 | |
| MIST-RLSamples (N)=202026.03 | 62.8 | |
| CODERM-8BSamples (N)=202026.03 | 61.59 | |
| QWEN3-14BSamples (N)=202026.03 | 55.49 | |
| MIST-RLSamples (N)=102026.03 | 48.78 | |
| CODERM-8BSamples (N)=102026.03 | 45.73 | |
| QWEN3-14BSamples (N)=102026.03 | 44.51 | |
| RANDOM SELECTIONSamples (N)=102026.03 | 22.5 | |
| RANDOM SELECTIONSamples (N)=202026.03 | 22.5 |