Code Generation on LiveCodeBench (Avg@8)
27.78Avg@8CodeScaler
Evaluation results
| Method | Links | |
|---|---|---|
| CodeScalerTraining Dataset=KodCode, Backbone Model=Qwen3-14B, Training Protocol=CodeScaler2026.02 | 27.78 | |
| CodeScalerTraining Dataset=rStarCoder, Backbone Model=Qwen3-14B, Training Protocol=CodeScaler2026.02 | 27.5 | |
| RLVR (Binary)Training Dataset=rStarCoder, Backbone Model=Qwen3-14B, Training Protocol=RLVR2026.02 | 27.24 | |
| RLVR (Binary)Training Dataset=KodCode, Backbone Model=Qwen3-14B, Training Protocol=RLVR2026.02 | 26.12 | |
| CodeScalerTraining Dataset=rStarCoder, Backbone Model=Qwen3-8B, Training Protocol=CodeScaler2026.02 | 24.5 | |
| RLVR (Binary)Training Dataset=rStarCoder, Backbone Model=Qwen3-8B, Training Protocol=RLVR2026.02 | 23.3 | |
| CodeScalerTraining Dataset=KodCode, Backbone Model=Qwen3-8B, Training Protocol=CodeScaler2026.02 | 23.11 | |
| RLVR (Binary)Training Dataset=KodCode, Backbone Model=Qwen3-8B, Training Protocol=RLVR2026.02 | 21.4 | |
| Qwen3-14B-BaseTraining Dataset=KodCode, Backbone Model=Qwen3-14B, Training Protocol=Base2026.02 | 21.37 | |
| Qwen3-14B-BaseTraining Dataset=rStarCoder, Backbone Model=Qwen3-14B, Training Protocol=Base2026.02 | 21.37 | |
| Qwen3-8B-BaseTraining Dataset=KodCode, Backbone Model=Qwen3-8B, Training Protocol=Base2026.02 | 13.75 | |
| Qwen3-8B-BaseTraining Dataset=rStarCoder, Backbone Model=Qwen3-8B, Training Protocol=Base2026.02 | 13.75 |