Code Generation on Code domain benchmarks
91.5HumanEvalOmniThought-0528
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| OmniThought-0528Backbone=Qwen3-8B-Base, Size=365k2025.12 | 91.5 | 86.8 | 29.4 | 64.6 | 68.1 | 0.043 | |
| AM-Thinking (code)Backbone=Qwen3-8B-Base, Size=324k2025.12 | 91.5 | 89.5 | 43 | 79.3 | 75.8 | 0.073 | |
| ODA-Mixture-500kBackbone=Qwen3-8B-Base, Size=506k2025.12 | 91.5 | 83.3 | 41 | 76.8 | 73 | 0.041 | |
| OpenCodeReasoningBackbone=Qwen2.5-7B-Base, Size=752k2025.12 | 88.4 | 76.7 | 42.7 | 79.3 | 71.7 | 0.029 | |
| ODA-Mixture-100kBackbone=Qwen3-8B-Base, Size=101k2025.12 | 87.2 | 85.6 | 38.7 | 81.1 | 73.2 | 0.206 | |
| rStar-CoderBackbone=Qwen2.5-7B-Base, Size=990k2025.12 | 86.6 | 77.4 | 46.2 | 78.7 | 72.2 | 0.022 | |
| MegaScienceBackbone=Qwen3-8B-Base, Size=1.25M2025.12 | 83.5 | 69.7 | 21.9 | 75 | 62.5 | 0.008 | |
| Qwen3-8B-BaseBackbone=Qwen3-8B-Base2025.12 | 82.9 | 75.5 | 16.9 | 34.2 | 52.4 | — | |
| MiroMind-M1-SFTBackbone=Qwen3-8B-Base, Size=719k2025.12 | 82.3 | 79.8 | 21.2 | 71.3 | 63.6 | 0.016 | |
| EpiCoder-funcBackbone=Qwen2.5-7B-Base, Size=380k2025.12 | 82.1 | 73.4 | 15.9 | 76.4 | 62 | 0.031 | |
| ODA-Mixture-100kBackbone=Qwen2.5-7B-Base, Size=101k2025.12 | 81.1 | 75.9 | 28 | 72.6 | 64.4 | 0.141 | |
| ODA-Mixture-500kBackbone=Qwen2.5-7B-Base, Size=506k2025.12 | 80.5 | 78.6 | 33.3 | 74.4 | 66.7 | 0.033 | |
| OpenThoughts3Backbone=Qwen2.5-7B-Base, Size=1.2M2025.12 | 78.7 | 79.4 | 31.9 | 78.1 | 67 | 0.014 | |
| AM-Thinking (code)Backbone=Qwen2.5-7B-Base, Size=324k2025.12 | 78.1 | 79 | 40.9 | 76.8 | 68.7 | 0.057 | |
| Qwen2.5-7B-BaseBackbone=Qwen2.5-7B-Base2025.12 | 77.4 | 71.6 | 8.2 | 43.3 | 50.1 | — | |
| OpenThoughts-114kBackbone=Qwen3-8B-Base, Size=114k2025.12 | 72 | 75.5 | 31.5 | 75 | 63.5 | 0.097 |