Code Generation on HumanEval-X C++
94.5Compilation Success RateVanilla
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VanillaModel=DeepSeek-Coder 6.7B2026.05 | 94.5 | 59.8 | 80.5 | |
| RepromptModel=DeepSeek-Coder 6.7B2026.05 | 90.9 | 64 | 82.9 | |
| VanillaModel=CodeLlama 7B2026.05 | 90.2 | 30.5 | 54.9 | |
| RepromptModel=CodeLlama 7B2026.05 | 81.7 | 31.1 | 55.5 | |
| CDCModel=Dream-Coder 7B2026.05 | 79.2 | 65.2 | 83.5 | |
| CFG-CDModel=Dream 7B2026.05 | 70.7 | 11 | 60.4 | |
| VanillaModel=Dream-Coder 7B2026.05 | 67.1 | 34.1 | 55.5 | |
| CFG-CDModel=Dream-Coder 7B2026.05 | 61.6 | 37.8 | 53.7 | |
| CDCModel=DiffuCoder 7B2026.05 | 50 | 37.2 | 68.3 | |
| CFG-CDModel=DiffuCoder 7B2026.05 | 49.4 | 25 | 50 | |
| CDCModel=Dream 7B2026.05 | 44.5 | 20.7 | 64.6 | |
| VanillaModel=DiffuCoder 7B2026.05 | 43.3 | 24.4 | 56.1 | |
| VanillaModel=Dream 7B2026.05 | 40.2 | 10.4 | 51.8 |