Mathematical Reasoning on AMC (Accuracy, Token count)
97.5AccuracyVanilla
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VanillaBackbone=Qwen3-14B2026.04 | 97.5 | 6,562 | |
| NoWAITBackbone=Qwen3-14B2026.04 | 97.5 | 5,035 | |
| DEERBackbone=Qwen3-14B2026.04 | 97.5 | 4,514 | |
| DTSRBackbone=Qwen3-14B2026.04 | 97.5 | 3,975 | |
| VanillaBackbone=Qwen3-32B2026.04 | 95 | 6,807 | |
| NoWAITBackbone=Qwen3-32B2026.04 | 95 | 4,894 | |
| DTSRBackbone=Qwen3-32B2026.04 | 95 | 4,617 | |
| VanillaBackbone=Qwen3-8B2026.04 | 92.5 | 7,232 | |
| DEERBackbone=Qwen3-32B2026.04 | 92.5 | 4,966 | |
| NoWAITBackbone=Qwen3-8B2026.04 | 90 | 5,448 | |
| DTSRBackbone=Qwen3-8B2026.04 | 90 | 4,507 | |
| DEERBackbone=Qwen3-8B2026.04 | 87.5 | 5,155 | |
| NoThinkingBackbone=Qwen3-14B2026.04 | 77.5 | 2,065 | |
| NoThinkingBackbone=Qwen3-8B2026.04 | 72.5 | 2,264 | |
| NoThinkingBackbone=Qwen3-32B2026.04 | 70 | 1,738 |