Mathematical Reasoning on GSM8K (test) (Top@1, Time, Tokens)
93.33Top-1 AccuracyFull Reasoning
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Full ReasoningBase Model=DeepSeek-R1-Qwen2.5-14b2026.01 | 93.33 | 35.24 | 493 | |
| SyncThinkBase Model=DeepSeek-R1-Qwen2.5-14b2026.01 | 93.03 | 31.59 | 442 | |
| Answer ConvergenceBase Model=DeepSeek-R1-Qwen2.5-14b2026.01 | 92.23 | 33.49 | 381 | |
| Reasoning RatioBase Model=DeepSeek-R1-Qwen2.5-14b, r=0.752026.01 | 92.12 | 31.17 | 459 | |
| No ReasoningBase Model=DeepSeek-R1-Qwen2.5-14b2026.01 | 92.04 | 18.07 | 251 | |
| Reasoning RatioBase Model=DeepSeek-R1-Qwen2.5-14b, r=0.52026.01 | 91.62 | 27.09 | 379 | |
| Reasoning RatioBase Model=DeepSeek-R1-Qwen2.5-14b, r=0.252026.01 | 91.15 | 22.59 | 319 | |
| No ReasoningBase Model=DeepSeek-R1-Qwen2.5-7b2026.01 | 88.09 | 11.39 | 266 | |
| Answer ConvergenceBase Model=DeepSeek-R1-Qwen2.5-7b2026.01 | 87.82 | 15.28 | 278 | |
| Full ReasoningBase Model=DeepSeek-R1-Qwen2.5-7b2026.01 | 87.57 | 18.79 | 440 | |
| SyncThinkBase Model=DeepSeek-R1-Qwen2.5-7b2026.01 | 87.49 | 14.39 | 383 | |
| Reasoning RatioBase Model=DeepSeek-R1-Qwen2.5-7b, r=0.752026.01 | 86.43 | 17.05 | 417 | |
| Reasoning RatioBase Model=DeepSeek-R1-Qwen2.5-7b, r=0.52026.01 | 85.97 | 15.22 | 359 | |
| Reasoning RatioBase Model=DeepSeek-R1-Qwen2.5-7b, r=0.252026.01 | 85.52 | 13.45 | 317 | |
| SFT ModelBase Model=DeepSeek-R1-Qwen2.5-14b2026.01 | 84.43 | 156.58 | 2,191 | |
| No ReasoningBase Model=DeepSeek-R1-LLaMA3.1-8b2026.01 | 80.51 | 12.72 | 267 | |
| Full ReasoningBase Model=DeepSeek-R1-LLaMA3.1-8b2026.01 | 78.47 | 23.66 | 498 | |
| SyncThinkBase Model=DeepSeek-R1-LLaMA3.1-8b2026.01 | 78.32 | 19.85 | 419 | |
| Answer ConvergenceBase Model=DeepSeek-R1-LLaMA3.1-8b2026.01 | 78.18 | 15.33 | 385 | |
| Reasoning RatioBase Model=DeepSeek-R1-LLaMA3.1-8b, r=0.752026.01 | 77.45 | 21.16 | 486 | |
| SFT ModelBase Model=DeepSeek-R1-Qwen2.5-7b2026.01 | 77.41 | 4.53 | 106 | |
| Reasoning RatioBase Model=DeepSeek-R1-LLaMA3.1-8b, r=0.52026.01 | 77.04 | 18.49 | 403 | |
| Reasoning RatioBase Model=DeepSeek-R1-LLaMA3.1-8b, r=0.252026.01 | 76.63 | 15.52 | 340 | |
| SFT ModelBase Model=DeepSeek-R1-LLaMA3.1-8b2026.01 | 63.99 | 3.99 | 84 |