Mathematical Reasoning on MATH-500 standard (test)
75.8AccuracyOriginal
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OriginalModel=Qwen2.5-14B, Compression Strength=None2026.01 | 75.8 | 583.66 | 1 | 12.99 | |
| CtrlCoTModel=Qwen2.5-14B, Compression Strength=Low-2026.01 | 75.4 | 457.45 | 0.78 | 16.48 | |
| CtrlCoTModel=Qwen2.5-14B, Compression Strength=Low--2026.01 | 74.8 | 481.94 | 0.83 | 15.52 | |
| TruncationModel=Qwen2.5-14B, Compression Strength=Low--2026.01 | 73.6 | 570.48 | 0.98 | 12.9 | |
| TokenSkipModel=Qwen2.5-14B, Compression Strength=Low--2026.01 | 73.4 | 496.55 | 0.85 | 14.78 | |
| TokenSkipModel=Qwen2.5-14B, Compression Strength=Low-2026.01 | 72.2 | 474.31 | 0.81 | 15.22 | |
| OriginalModel=Qwen2.5-7B, Compression Strength=None2026.01 | 71.2 | 574.58 | 1 | 7.7 | |
| CtrlCoTModel=Qwen2.5-7B, Compression Strength=Low--2026.01 | 70.6 | 449.46 | 0.78 | 15.71 | |
| TokenSkipModel=Qwen2.5-7B, Compression Strength=Low--2026.01 | 69.6 | 519.78 | 0.9 | 13.39 | |
| TruncationModel=Qwen2.5-14B, Compression Strength=Low-2026.01 | 69.4 | 552.18 | 0.95 | 12.57 | |
| TruncationModel=Qwen2.5-7B, Compression Strength=Low--2026.01 | 69 | 563.32 | 0.98 | 12.25 | |
| CtrlCoTModel=Qwen2.5-7B, Compression Strength=Low-2026.01 | 68.8 | 428.68 | 0.75 | 16.05 | |
| TokenSkipModel=Qwen2.5-7B, Compression Strength=Low-2026.01 | 67.6 | 491.76 | 0.86 | 13.75 | |
| TruncationModel=Qwen2.5-7B, Compression Strength=Low-2026.01 | 65.8 | 544.86 | 0.95 | 12.08 |