Code on HumanEval (pass@1)
67.07Pass@1CodePref
Evaluation Results
| Method | Links | |
|---|---|---|
| CodePrefTraining Method=DPO, Dataset=CodePref2025.11 | 67.07 | |
| UltraMixTraining Method=DPO, Dataset=UM-190k2025.11 | 66.96 | |
| UltraMixTraining Method=DPO, Dataset=UM-187k2025.11 | 66.84 | |
| UltraMixTraining Method=DPO, Dataset=UM-170k2025.11 | 65.72 | |
| TuluDPOTraining Method=DPO, Dataset=TuluDPO2025.11 | 65.25 | |
| ORPOTraining Method=ORPO, Dataset=ORPO2025.11 | 63.61 | |
| HelpSteerTraining Method=DPO, Dataset=HelpSteer2025.11 | 60.01 | |
| UltraFBTraining Method=DPO, Dataset=UltraFB2025.11 | 57.97 | |
| SmolLM-3-3B-SFTTraining Method=SFT2025.11 | 52.2 |