Mathematical Reasoning on Mathematical Reasoning Aggregate (Pass@1)
61.5Average Accuracy (Pass@1)Qwen3-8B+E2C-(SFT+RL)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-8B+E2C-(SFT+RL)Model Series=Qwen3 8B, Training Stage=E2C-(SFT+RL)2025.09 | 61.5 | 1,476.41 | |
| Qwen3-8B+GRPO (Baseline)Model Series=Qwen3 8B, Training Stage=GRPO2025.09 | 60 | 1,429.46 | |
| Qwen3-4B+E2C-(SFT+RL)Model Series=Qwen3 4B, Training Stage=E2C-(SFT+RL)2025.09 | 58.4 | 1,456.34 | |
| Qwen3-8B+E2C-SFT+GRPOModel Series=Qwen3 8B, Training Stage=E2C-SFT+GRPO2025.09 | 57.7 | 1,309.62 | |
| Qwen3-4B+GRPO (Baseline)Model Series=Qwen3 4B, Training Stage=GRPO2025.09 | 56.5 | 1,263.15 | |
| Qwen3-4B+E2C-SFT+GRPOModel Series=Qwen3 4B, Training Stage=E2C-SFT+GRPO2025.09 | 55.2 | 1,324.18 | |
| Qwen3-8B+E2C-SFTModel Series=Qwen3 8B, Training Stage=E2C-SFT2025.09 | 46.8 | 1,162.89 | |
| Qwen3-8B+Prompt (Zero-shot)Model Series=Qwen3 8B, Evaluation Mode=Zero-shot2025.09 | 46.6 | 1,142.38 |