Mathematical Reasoning on AMC (Acc, #Tok, ∆%)
89.2AccuracyVanilla
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| VanillaReasoner=Qwen3-8B2026.06 | 89.2 | 10,583 | — | |
| ACTSπSFTReasoner=Qwen3-8B2026.06 | 87.2 | 8,371 | 20.9 | |
| DEERReasoner=Qwen3-8B2026.06 | 86.8 | 8,766 | 17.2 | |
| ACTSReasoner=Qwen3-8B2026.06 | 86.7 | 8,492 | 19.8 | |
| VanillaReasoner=DeepSeek-R1-Distill-Qwen-7B2026.06 | 83.1 | 7,863 | — | |
| CoDReasoner=Qwen3-8B2026.06 | 82.2 | 8,422 | 20.4 | |
| ACTSReasoner=DeepSeek-R1-Distill-Qwen-7B2026.06 | 79.8 | 4,486 | 42.9 | |
| BudgetGuidanceReasoner=DeepSeek-R1-Distill-Qwen-7B2026.06 | 78.8 | 6,600 | 16.1 | |
| ACTSπSFTReasoner=DeepSeek-R1-Distill-Qwen-7B2026.06 | 77.4 | 4,937 | 37.2 | |
| DEERReasoner=DeepSeek-R1-Distill-Qwen-7B2026.06 | 71.3 | 4,891 | 37.8 | |
| BudgetGuidanceReasoner=Qwen3-8B2026.06 | 71.3 | 4,555 | 57 | |
| CoDReasoner=DeepSeek-R1-Distill-Qwen-7B2026.06 | 67 | 4,431 | 43.6 | |
| NoThinkReasoner=Qwen3-8B2026.06 | 65.3 | 2,939 | 72.2 | |
| ACTSReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.06 | 65.1 | 5,143 | 52.5 | |
| VanillaReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.06 | 63.1 | 10,835 | — | |
| ACTSπSFTReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.06 | 63.1 | 5,362 | 50.5 | |
| NoThinkReasoner=DeepSeek-R1-Distill-Qwen-7B2026.06 | 54.5 | 1,600 | 79.6 | |
| DEERReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.06 | 54 | 5,169 | 52.3 | |
| CoDReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.06 | 53.5 | 7,421 | 31.5 | |
| NoThinkReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.06 | 49.2 | 2,394 | 77.9 |