Mathematical Reasoning on AIME24 (Accuracy, Token consumption)
80AccuracyCoT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CoTBackbone=QwQ-32B, Reasoning framework=CoT2026.03 | 80 | 0.13 | |
| CoT-SCBackbone=QwQ-32B, Reasoning framework=CoT-SC2026.03 | 80 | 0.85 | |
| MAD-M2 (O)Backbone=QwQ-32B, Masking Strategy=Objective2026.03 | 80 | 0.67 | |
| MADBackbone=QwQ-32B, Reasoning framework=MAD2026.03 | 76.7 | 1 | |
| MAD-M2 (S)Backbone=QwQ-32B, Masking Strategy=Subjective2026.03 | 76.7 | 1.28 | |
| CoT-SCBackbone=Qwen2.5-Math-7B-Instruct, Reasoning framework=CoT-SC2026.03 | 23.3 | 0.53 | |
| MADBackbone=Qwen2.5-7B-Instruct, Reasoning framework=MAD2026.03 | 13.3 | 1 | |
| MAD-M2 (S)Backbone=Qwen2.5-7B-Instruct, Masking Strategy=Subjective2026.03 | 13.3 | 1.13 | |
| CoTBackbone=Qwen2.5-Math-7B-Instruct, Reasoning framework=CoT2026.03 | 13.3 | 0.08 | |
| MAD-M2 (O)Backbone=Qwen2.5-Math-7B-Instruct, Masking Strategy=Objective2026.03 | 13.3 | 0.67 | |
| CoT-SCBackbone=Qwen2.5-7B-Instruct, Reasoning framework=CoT-SC2026.03 | 10 | 0.43 | |
| MAD-M2 (O)Backbone=Qwen2.5-7B-Instruct, Masking Strategy=Objective2026.03 | 6.7 | 0.68 | |
| MADBackbone=Qwen2.5-Math-7B-Instruct, Reasoning framework=MAD2026.03 | 6.7 | 1 | |
| MAD-M2 (S)Backbone=Qwen2.5-Math-7B-Instruct, Masking Strategy=Subjective2026.03 | 6.7 | 1.37 | |
| CoTBackbone=Qwen2.5-7B-Instruct, Reasoning framework=CoT2026.03 | 3.3 | 0.07 | |
| CoT-SCBackbone=DeepSeek-Math-7B-Instruct, Reasoning framework=CoT-SC2026.03 | 3.3 | 0.44 | |
| CoTBackbone=DeepSeek-Math-7B-Instruct, Reasoning framework=CoT2026.03 | 0 | 0.07 | |
| MADBackbone=DeepSeek-Math-7B-Instruct, Reasoning framework=MAD2026.03 | 0 | 1 | |
| MAD-M2 (S)Backbone=DeepSeek-Math-7B-Instruct, Masking Strategy=Subjective2026.03 | 0 | 1.32 | |
| MAD-M2 (O)Backbone=DeepSeek-Math-7B-Instruct, Masking Strategy=Objective2026.03 | 0 | 0.67 |