Code Completion on APC Hard Completion 1.0 (test)
100HCRQwen2.5-Coder-1.5B-BASE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-Coder-1.5B-BASEBackbone=Qwen2.5-Coder, Parameters=1.5B, Training=BASE2026.04 | 100 | 41.88 | 77.36 | |
| Qwen2.5-Coder-1.5B-SFTBackbone=Qwen2.5-Coder, Parameters=1.5B, Training=SFT2026.04 | 100 | 44.55 | 74.96 | |
| Qwen2.5-Coder-1.5B-GRPOBackbone=Qwen2.5-Coder, Parameters=1.5B, Training=GRPO2026.04 | 100 | 41.02 | 70.9 | |
| Qwen2.5-Coder-3B-BASEBackbone=Qwen2.5-Coder, Parameters=3B, Training=BASE2026.04 | 100 | 46.74 | 83.19 | |
| Qwen2.5-Coder-3B-SFTBackbone=Qwen2.5-Coder, Parameters=3B, Training=SFT2026.04 | 100 | 48.57 | 82.73 | |
| Qwen2.5-Coder-3B-GRPOBackbone=Qwen2.5-Coder, Parameters=3B, Training=GRPO2026.04 | 100 | 46.49 | 77.97 | |
| Qwen2.5-Coder-7B-BASEBackbone=Qwen2.5-Coder, Parameters=7B, Training=BASE2026.04 | 100 | 24.29 | 54.68 | |
| Qwen2.5-Coder-7B-SFTBackbone=Qwen2.5-Coder, Parameters=7B, Training=SFT2026.04 | 100 | 49.24 | 83.27 | |
| Qwen2.5-Coder-7B-GRPOBackbone=Qwen2.5-Coder, Parameters=7B, Training=GRPO2026.04 | 100 | 51.08 | 86.07 | |
| Qwen2.5-Coder-14B-BASEBackbone=Qwen2.5-Coder, Parameters=14B, Training=BASE2026.04 | 100 | 30.35 | 57.77 | |
| Qwen2.5-Coder-14B-SFTBackbone=Qwen2.5-Coder, Parameters=14B, Training=SFT2026.04 | 100 | 51.32 | 84.21 | |
| Qwen2.5-Coder-14B-GRPOBackbone=Qwen2.5-Coder, Parameters=14B, Training=GRPO2026.04 | 100 | 47.04 | 79.75 | |
| CodeGemma-2B-BASEBackbone=CodeGemma, Parameters=2B, Training=BASE2026.04 | 100 | 43.24 | 80.56 | |
| CodeGemma-2B-SFTBackbone=CodeGemma, Parameters=2B, Training=SFT2026.04 | 100 | 32.67 | 40.31 | |
| CodeGemma-2B-GRPOBackbone=CodeGemma, Parameters=2B, Training=GRPO2026.04 | 100 | 43.1 | 74.81 | |
| CodeGemma-7B-BASEBackbone=CodeGemma, Parameters=7B, Training=BASE2026.04 | 100 | 40.51 | 77.7 | |
| CodeGemma-7B-SFTBackbone=CodeGemma, Parameters=7B, Training=SFT2026.04 | 100 | 34.21 | 66.94 | |
| CodeGemma-7B-GRPOBackbone=CodeGemma, Parameters=7B, Training=GRPO2026.04 | 100 | 40.1 | 74.98 | |
| CodeLlama-7B-BASEBackbone=CodeLlama, Parameters=7B, Training=BASE2026.04 | 100 | 11.54 | 42.78 | |
| CodeLlama-7B-SFTBackbone=CodeLlama, Parameters=7B, Training=SFT2026.04 | 100 | 37.1 | 70.14 | |
| CodeLlama-7B-GRPOBackbone=CodeLlama, Parameters=7B, Training=GRPO2026.04 | 100 | 36.83 | 70.48 | |
| CodeLlama-13B-BASEBackbone=CodeLlama, Parameters=13B, Training=BASE2026.04 | 100 | 17.07 | 48.99 | |
| CodeLlama-13B-SFTBackbone=CodeLlama, Parameters=13B, Training=SFT2026.04 | 100 | 31.18 | 63.76 | |
| CodeLlama-13B-GRPOBackbone=CodeLlama, Parameters=13B, Training=GRPO2026.04 | 100 | 35.72 | 68.49 | |
| StarCoder2-3B-BASEBackbone=StarCoder2, Parameters=3B, Training=BASE2026.04 | 100 | 38.46 | 75.47 | |
| StarCoder2-3B-SFTBackbone=StarCoder2, Parameters=3B, Training=SFT2026.04 | 100 | 42.2 | 77.15 | |
| StarCoder2-3B-GRPOBackbone=StarCoder2, Parameters=3B, Training=GRPO2026.04 | 100 | 43.4 | 79.23 | |
| StarCoder2-7B-BASEBackbone=StarCoder2, Parameters=7B, Training=BASE2026.04 | 100 | 42.9 | 79.78 | |
| StarCoder2-7B-SFTBackbone=StarCoder2, Parameters=7B, Training=SFT2026.04 | 100 | 45.67 | 81.65 | |
| StarCoder2-7B-GRPOBackbone=StarCoder2, Parameters=7B, Training=GRPO2026.04 | 100 | 38.98 | 74.16 | |
| StarCoder2-15B-BASEBackbone=StarCoder2, Parameters=15B, Training=BASE2026.04 | 100 | 40.53 | 78.43 | |
| StarCoder2-15B-SFTBackbone=StarCoder2, Parameters=15B, Training=SFT2026.04 | 100 | 44.04 | 79.51 | |
| StarCoder2-15B-GRPOBackbone=StarCoder2, Parameters=15B, Training=GRPO2026.04 | 100 | 44.05 | 78.69 |