Code Generation on MultiPL-E (Language Scores)
76.5Average ScoreQwen2.5-Coder-7B-Instruct
Evaluation results
| Method | Links | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Qwen2.5-Coder-7B-InstructSize=7B2024.11 | 76.5 | 87.8 | 76.5 | 75.6 | 80.3 | 81.8 | 83.2 | 78.3 | 48.7 | — | — | — | — | — | — | — | — | — | |
| DS-Coder-V2-Lite-InstructSize=16B2024.11 | 73.2 | 81.1 | 76.6 | 75.8 | 76.6 | 80.5 | 77.6 | 74.5 | 43 | — | — | — | — | — | — | — | — | — | |
| Yi-Coder-9B-ChatSize=9B2024.11 | 71.8 | 85.4 | 76 | 67.7 | 76.6 | 72.3 | 78.9 | 72.1 | 45.6 | — | — | — | — | — | — | — | — | — | |
| CodeQwen1.5-7B-ChatSize=7B2024.11 | 71.6 | 83.5 | 70.9 | 72 | 75.9 | 76.7 | 77.6 | 73.9 | 41.8 | — | — | — | — | — | — | — | — | — | |
| Stable-DiffCoder-8B-BaseSize=8B2026.01 | 71.2 | 80.5 | 74.1 | 69.4 | 70.3 | 74.8 | 73.1 | 74.4 | 53.2 | — | — | — | — | — | — | — | — | — | |
| OpenCoder-8B-InstructSize=8B2024.11 | 71 | 83.5 | 72.2 | 61.5 | 75.9 | 78 | 79.5 | 73.3 | 44.3 | — | — | — | — | — | — | — | — | — | |
| Seed-Coder-8B-BaseSize=8B2026.01 | 67.6 | 77.4 | 72.8 | 69.6 | 53.8 | 77.4 | 77.6 | 63.9 | 48.1 | — | — | — | — | — | — | — | — | — | |
| DS-Coder-6.7B-InstructSize=6.7B2024.11 | 66.1 | 78.6 | 68.4 | 63.4 | 72.8 | 67.2 | 72.7 | 68.9 | 36.7 | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-Coder-14BSize=14B2026.01 | 62.3 | 83.5 | 46.8 | 69.6 | 63.3 | 69.2 | 61.5 | 64.6 | 39.9 | — | — | — | — | — | — | — | — | — | |
| OpenCoder-8B-BaseSize=8B2026.01 | 61 | 66.5 | 63.9 | 63.4 | 54.3 | 68.6 | 65.8 | 61.5 | 44.3 | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-Coder-7BSize=7B2026.01 | 58.8 | 72 | 53.2 | 62.1 | 60.8 | 64.2 | 60.3 | 59 | 38.6 | — | — | — | — | — | — | — | — | — | |
| OpenCoder-1.5B-InstructSize=1.5B2024.11 | 57.5 | 72.5 | 64.6 | 50.9 | 61.4 | 63.5 | 62.1 | 55.3 | 29.7 | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-Coder-1.5B-InstructSize=1.5B2024.11 | 56.7 | 71.2 | 55.7 | 50.9 | 64.6 | 61 | 62.1 | 59 | 29.1 | — | — | — | — | — | — | — | — | — | |
| Yi-Coder-1.5B-ChatSize=1.5B2024.11 | 51.9 | 67.7 | 51.9 | 49.1 | 57.6 | 57.9 | 59.6 | 52.2 | 19 | — | — | — | — | — | — | — | — | — | |
| DeepSeek-Coder-33B-BaseSize=33B2026.01 | 50.3 | 56.1 | 51.9 | 58.4 | 51.3 | 52.8 | 55.3 | 44.1 | 32.3 | — | — | — | — | — | — | — | — | — | |
| DS-Coder-1.3B-InstructSize=1.3B2024.11 | 48.4 | 65.2 | 51.9 | 45.3 | 55.1 | 59.7 | 52.2 | 45.3 | 12.7 | — | — | — | — | — | — | — | — | — | |
| CodeQwen1.5-7BSize=7B2026.01 | 48.4 | 51.8 | 42.4 | 52.2 | 55.7 | 52.2 | 49.7 | 46.6 | 36.7 | — | — | — | — | — | — | — | — | — | |
| CodeLlama-70B-BaseSize=70B2026.01 | 48.2 | 52.4 | 44.7 | 49.7 | 46.7 | 52.2 | 56.5 | 46.6 | 31.6 | — | — | — | — | — | — | — | — | — | |
| CodeGemma-7B-ItSize=7B2024.11 | 46.5 | 59.8 | 48.1 | 46.6 | 51.9 | 54.7 | 54 | 46.6 | 10.1 | — | — | — | — | — | — | — | — | — | |
| Llama-3.1-70B-BaseSize=70B2026.01 | 46.5 | 54.9 | 41.1 | 41 | 44.2 | 57.9 | 55.3 | 48.4 | 29.1 | — | — | — | — | — | — | — | — | — | |
| DeepSeek-Coder-6.7B-BaseSize=6.7B2026.01 | 44.7 | 49.4 | 43 | 50.3 | 50 | 49.7 | 48.4 | 38.5 | 28.5 | — | — | — | — | — | — | — | — | — | |
| StarCoder2-15BSize=15B2026.01 | 41.7 | 46.3 | 46.2 | 47.2 | 53.2 | 42.1 | 43.5 | 39.1 | 15.8 | — | — | — | — | — | — | — | — | — | |
| DeepSeek-Coder-V2-Lite-BaseSize=2.4B/16B2026.01 | 40.4 | 40.9 | 34.8 | 45.9 | 41.7 | 48.4 | 44.7 | 47.2 | 19.6 | — | — | — | — | — | — | — | — | — | |
| DFTBackbone=Qwen-Math-7B2026.05 | 35.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| StarCoder2-7BSize=7B2026.01 | 34.3 | 35.4 | 38 | 40.4 | 46.2 | 34 | 36 | 30.4 | 13.9 | — | — | — | — | — | — | — | — | — | |
| InfoSFTBackbone=Qwen-Math-7B2026.05 | 34.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CodeLlama-7B-InstructSize=7B2024.11 | 33 | 45.7 | 32.2 | 28.6 | 32.9 | 39 | 43.5 | 31.7 | 10.1 | — | — | — | — | — | — | — | — | — | |
| DFTBackbone=Llama-3.1-8B2026.05 | 32.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InfoSFTBackbone=Llama-3.1-8B2026.05 | 32.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFTBackbone=Llama-3.1-8B2026.05 | 32 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| baseBackbone=Llama-3.1-8B2026.05 | 31 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| baseBackbone=Qwen-Math-7B2026.05 | 30.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DFTBackbone=Qwen-Math-1.5B2026.05 | 29.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InfoSFTBackbone=Qwen-Math-1.5B2026.05 | 28.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFTBackbone=Qwen-Math-7B2026.05 | 28.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFTBackbone=Qwen-Math-1.5B2026.05 | 28 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| baseBackbone=Qwen-Math-1.5B2026.05 | 26 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GRIP 16BModel Size=16B, Total Budget=300B tokens2026.02 | 25.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GRIP 8BModel Size=8B, Total Budget=300B tokens2026.02 | 23.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 6. GRIP (Full)Model=GRIP-8B, Total Budget=100B tokens2026.02 | 19.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 4. + Loss ReplayModel=GRIP-8B, Total Budget=100B tokens2026.02 | 16.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 5. + DiversityModel=GRIP-8B, Total Budget=100B tokens2026.02 | 16 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Random 16BModel Size=16B, Total Budget=300B tokens2026.02 | 14.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3. + Static ReplayModel=GRIP-8B, Total Budget=100B tokens2026.02 | 14.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Random 8BModel Size=8B, Total Budget=300B tokens2026.02 | 13.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 2. + Static BudgetModel=GRIP-8B, Total Budget=100B tokens2026.02 | 13.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 1. RandomModel=GRIP-8B, Total Budget=100B tokens2026.02 | 11.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-3BModel Backbone=Qwen2.5-3B, Fine-tuning Strategy=None2026.04 | — | — | — | — | — | — | — | — | — | 43.29 | 40.99 | 37.34 | 37.89 | 47.17 | 43.04 | 24.68 | 45.96 | 40.05 | |
| Qwen2.5-3B w/ CADFTModel Backbone=Qwen2.5-3B, Fine-tuning Strategy=CADFT2026.04 | — | — | — | — | — | — | — | — | — | 48.2 | 46.5 | 43.6 | 47.8 | 44.3 | 45.1 | 30.2 | 46.8 | 44.14 | |
| Qwen2.5-3B w/ DFTModel Backbone=Qwen2.5-3B, Fine-tuning Strategy=DFT2026.04 | — | — | — | — | — | — | — | — | — | 45.73 | 44.72 | 41.77 | 45.34 | 42.14 | 43.04 | 27.85 | 44.1 | 41.84 | |
| Qwen2.5-3B w/ SFTModel Backbone=Qwen2.5-3B, Fine-tuning Strategy=SFT2026.04 | — | — | — | — | — | — | — | — | — | 42.07 | 42.24 | 37.97 | 37.27 | 43.4 | 41.77 | 20.25 | 47.83 | 39.1 | |
| Qwen2.5-Coder-3BModel Backbone=Qwen2.5-Coder-3B, Fine-tuning Strategy=None2026.04 | — | — | — | — | — | — | — | — | — | 51.83 | 53.42 | 46.2 | 47.2 | 54.09 | 55.06 | 25.32 | 54.04 | 48.39 | |
| Qwen2.5-Coder-3B w/ CADFTModel Backbone=Qwen2.5-Coder-3B, Fine-tuning Strategy=CADFT2026.04 | — | — | — | — | — | — | — | — | — | 60.2 | 57 | 53.8 | 61.2 | 61 | 63.5 | 34.5 | 56 | 56.67 | |
| Qwen2.5-Coder-3B w/ DFTModel Backbone=Qwen2.5-Coder-3B, Fine-tuning Strategy=DFT2026.04 | — | — | — | — | — | — | — | — | — | 57.32 | 54.66 | 51.27 | 58.39 | 58.49 | 60.76 | 31.01 | 53.42 | 53.16 | |
| Qwen2.5-Coder-3B w/ SFTModel Backbone=Qwen2.5-Coder-3B, Fine-tuning Strategy=SFT2026.04 | — | — | — | — | — | — | — | — | — | 51.22 | 51.55 | 48.1 | 54.66 | 59.12 | 51.27 | 34.18 | 54.04 | 50.52 | |
| Qwen2.5-Coder-7BModel Backbone=Qwen2.5-Coder-7B, Fine-tuning Strategy=None2026.04 | — | — | — | — | — | — | — | — | — | 63.41 | 63.98 | 53.16 | 59.01 | 62.89 | 59.49 | 39.24 | 60.87 | 57.76 | |
| Qwen2.5-Coder-7B w/ CADFTModel Backbone=Qwen2.5-Coder-7B, Fine-tuning Strategy=CADFT2026.04 | — | — | — | — | — | — | — | — | — | 70.5 | 70 | 56.8 | 63 | 72.8 | 67.8 | 52 | 66 | 65.24 | |
| Qwen2.5-Coder-7B w/ DFTModel Backbone=Qwen2.5-Coder-7B, Fine-tuning Strategy=DFT2026.04 | — | — | — | — | — | — | — | — | — | 67.68 | 67.7 | 54.43 | 60.87 | 70.44 | 65.19 | 48.73 | 63.35 | 62.3 | |
| Qwen2.5-Coder-7B w/ SFTModel Backbone=Qwen2.5-Coder-7B, Fine-tuning Strategy=SFT2026.04 | — | — | — | — | — | — | — | — | — | 54.88 | 64.6 | 51.27 | 62.11 | 68.55 | 60.76 | 33.54 | 65.22 | 57.62 |