Mathematical Reasoning on GSM8K Qwen2.5-Instruct (test)
93.7AccuracyASCoT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ASCoTScale=14B, Ratio=0.72025.08 | 93.7 | 221.17 | 0.71 | |
| ASCoTScale=14B, Ratio=0.92025.08 | 93.6 | 269.76 | 0.86 | |
| ASCoTScale=14B, Ratio=0.82025.08 | 93.5 | 246.57 | 0.79 | |
| ASCoTScale=14B, Ratio=1.02025.08 | 93.3 | 312.3 | 0.99 | |
| OriginalScale=14B2025.08 | 93.1 | 313.11 | — | |
| ASCoTScale=14B, Ratio=0.62025.08 | 92.8 | 177.4 | 0.57 | |
| ASCoTScale=7B, Ratio=1.02025.08 | 91.8 | 294.74 | 0.99 | |
| ASCoTScale=14B, Ratio=0.52025.08 | 91.6 | 158.21 | 0.51 | |
| OriginalScale=7B2025.08 | 91.4 | 297.83 | — | |
| ASCoTScale=7B, Ratio=0.92025.08 | 91.3 | 251.71 | 0.85 | |
| ASCoTScale=7B, Ratio=0.82025.08 | 90.4 | 234.95 | 0.79 | |
| ASCoTScale=7B, Ratio=0.72025.08 | 90.3 | 216.73 | 0.73 | |
| ASCoTScale=7B, Ratio=0.62025.08 | 88.2 | 178.28 | 0.6 | |
| ASCoTScale=7B, Ratio=0.52025.08 | 86.3 | 150.98 | 0.51 | |
| OriginalScale=3B2025.08 | 83.7 | 314.87 | — | |
| ASCoTScale=3B, Ratio=1.02025.08 | 83.5 | 318.32 | 1.01 | |
| ASCoTScale=3B, Ratio=0.92025.08 | 83.2 | 262.31 | 0.83 | |
| ASCoTScale=3B, Ratio=0.82025.08 | 82 | 250.75 | 0.79 | |
| ASCoTScale=3B, Ratio=0.72025.08 | 80.1 | 233.15 | 0.74 | |
| ASCoTScale=3B, Ratio=0.62025.08 | 77.6 | 199.44 | 0.63 | |
| ASCoTScale=3B, Ratio=0.52025.08 | 74.9 | 170.87 | 0.54 |