Mathematical Reasoning on GSM8K (AvgMath/Avg Scores)
90.45GSM8K ScoreEqual
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| EqualBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.07 | 90.45 | 73.62 | 63.8 | |
| RegMixBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.07 | 89.61 | 65.21 | 61.4 | |
| DMOBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.07 | 89.61 | 72 | 63.47 | |
| CAUSALMIXBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.07 | 88.86 | 74.72 | 66.66 | |
| DoReMiBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.07 | 88.55 | 65.39 | 62 | |
| ODMBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.07 | 88.32 | 64.74 | 58.77 | |
| GridBackbone=Qwen3-4B, Data Scale=# 20K, Training Mixture Context=AM-Thinking-v1-Distilled-Code&Math2026.07 | 87.34 | 74.27 | 64.74 |