Mathematical Reasoning on MATH (Accuracy, Length)
93.67AccuracyAutoTHINK
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AutoTHINKBackbone=DeepSeek-R1-7B2026.02 | 93.67 | 2,449.16 | |
| StandardpBackbone=DeepSeek-R1-7B2026.02 | 92.33 | 4,261.18 | |
| C3oTBackbone=DeepSeek-R1-7B2026.02 | 92 | 3,805.73 | |
| S3-CoTBackbone=DeepSeek-R1-7B2026.02 | 92 | 2,833.27 | |
| EffReaBackbone=DeepSeek-R1-7B2026.02 | 91.33 | 2,700.98 | |
| LASERDEBackbone=DeepSeek-R1-7B2026.02 | 91.33 | 1,795.42 | |
| S3-CoTscBackbone=DeepSeek-R1-7B2026.02 | 90.67 | 3,016.64 | |
| EfficientpBackbone=DeepSeek-R1-7B2026.02 | 90.33 | 3,206.63 | |
| LC-R1Backbone=DeepSeek-R1-7B2026.02 | 88.67 | 1,374.06 | |
| CoT-ValveBackbone=DeepSeek-R1-7B2026.02 | 80.33 | 1,499.4 | |
| TokenSkipBackbone=Qwen2.5-7B2026.02 | 74.67 | 512.12 | |
| StandardpBackbone=Qwen2.5-7B2026.02 | 72.67 | 559.49 | |
| ShorterBetterBackbone=DeepSeek-R1-7B2026.02 | 72.33 | 585.57 | |
| C3oTBackbone=Qwen2.5-7B2026.02 | 71.33 | 536.77 | |
| CoT-ValveBackbone=Qwen2.5-7B2026.02 | 71 | 619.19 | |
| S3-CoTBackbone=Qwen2.5-7B2026.02 | 70.5 | 426.91 | |
| EfficientpBackbone=Qwen2.5-7B2026.02 | 70 | 300.83 | |
| S3-CoTscBackbone=Qwen2.5-7B2026.02 | 69.67 | 433.43 |