Mathematical Reasoning on MATH (Accuracy & Token Consumption)
82AccuracyCoT-SC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoT-SCBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 82 | 0.44 | |
| CoT-SCBackbone=QwQ-32B2026.03 | 81.6 | 0.61 | |
| CoTBackbone=QwQ-32B2026.03 | 80.8 | 0.1 | |
| MAD-M2 (O)Backbone=Qwen2.5-Math-7B-Instruct, Masking Strategy=Objective2026.03 | 80.2 | 0.62 | |
| MAD-M2 (S)Backbone=QwQ-32B, Masking Strategy=Subjective2026.03 | 79.6 | 1.27 | |
| MADBackbone=QwQ-32B2026.03 | 79.2 | 1 | |
| CoTBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 77.8 | 0.08 | |
| MAD-M2 (O)Backbone=QwQ-32B, Masking Strategy=Objective2026.03 | 75 | 0.67 | |
| MADBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 71.2 | 1 | |
| MAD-M2 (S)Backbone=Qwen2.5-Math-7B-Instruct, Masking Strategy=Subjective2026.03 | 71.2 | 1.41 | |
| CoT-SCBackbone=Qwen2.5-7B-Instruct2026.03 | 58 | 0.45 | |
| MAD-M2 (S)Backbone=Qwen2.5-7B-Instruct, Masking Strategy=Subjective2026.03 | 56.8 | 1.2 | |
| MADBackbone=Qwen2.5-7B-Instruct2026.03 | 55.6 | 1 | |
| MAD-M2 (O)Backbone=Qwen2.5-7B-Instruct, Masking Strategy=Objective2026.03 | 54.2 | 0.67 | |
| CoTBackbone=Qwen2.5-7B-Instruct2026.03 | 49.2 | 0.07 | |
| CoT-SCBackbone=DeepSeek-Math-7B-Instruct2026.03 | 44.4 | 0.47 | |
| MAD-M2 (O)Backbone=DeepSeek-Math-7B-Instruct, Masking Strategy=Objective2026.03 | 39.8 | 0.68 | |
| MADBackbone=DeepSeek-Math-7B-Instruct2026.03 | 38.6 | 1 | |
| MAD-M2 (S)Backbone=DeepSeek-Math-7B-Instruct, Masking Strategy=Subjective2026.03 | 37 | 1.31 | |
| CoTBackbone=DeepSeek-Math-7B-Instruct2026.03 | 34.2 | 0.08 |