Code Generation on HumanEval+ and MBPP+
75HumanEval+ ScoreQwen3-1.7B w/ OPD+ReNIO
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-1.7B w/ OPD+ReNIOTeacher Model=Qwen3-8B2026.06 | 75 | — | 63.56 | 69.28 | |
| Qwen3-1.7B w/ OPDTeacher Model=Qwen3-8B2026.06 | 74.4 | — | 62.43 | 68.42 | |
| Qwen3-1.7BTeacher Model=Qwen3-8B2026.06 | 73.63 | — | 57.21 | 65.42 | |
| DS-R1-Qwen-1.5B w/ OPD+ReNIOTeacher Model=DS-R1-Qwen-7B2026.06 | 50.91 | — | 41.2 | 46.06 | |
| DS-R1-Qwen-1.5B w/ OPDTeacher Model=DS-R1-Qwen-7B2026.06 | 50.15 | — | 41.14 | 45.65 | |
| DS-R1-Qwen-1.5BTeacher Model=DS-R1-Qwen-7B2026.06 | 49.09 | — | 41.07 | 45.08 | |
| DASDBackbone=Qwen3-8B2026.05 | — | 73.7 | — | — | |
| GRPOBackbone=Qwen3-8B2026.05 | — | 72.3 | — | — | |
| OPSDBackbone=Qwen3-8B2026.05 | — | 69.9 | — | — | |
| Qwen3-8BPost-training=Base2026.05 | — | 72.1 | — | — |